e-book

An Engineer’s Guide to
Industrial Robot Designs
A compendium of technical documentation on robotic system designs

ti.com/robotics Q2 | 2020
Table of Contents/Overview

1. Introduction 3. Robot arm and driving system (manipulator)

1.1 An introduction to an industrial robot system. . . . . . . 3 3.1.1 How to protect battery power management
systems from thermal damage. . . . . . . . . . . . . . . . . 45
2. Robot system controller 3.1.2 Protecting your battery isn’t as hard as
2.1 Control panel you think. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
2.1.1 Using Sitara™ processors for Industry 3.1.3 Position feedback-related reference designs
4.0 servo drives. . . . . . . . . . . . . . . . . . . . . . . 9 for robotic systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.2 Servo drives for robotic systems 4. Sensing and vision technologies
2.2.1 The impact of an isolated gate driver. . . . . . . 13 4.1 TI mmWave radar sensors in robotics
2.2.2 Understanding peak source and applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
sink current parameters . . . . . . . . . . . . . . . . . 17 4.2 Intelligence at the edge powers autonomous
2.2.3 Low-side gate drivers with UVLO versus factories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
BJT totem poles . . . . . . . . . . . . . . . . . . . . . . 19 4.3 Use ultrasonic sensing for graceful robots. . . . . . . . 55
2.2.4 An external gate-resistor design guide for 4.4 How sensor data is powering AI in robotics. . . . . . . 57
gate drivers. . . . . . . . . . . . . . . . . . . . . . . . . . 20
4.5 Bringing machine learning to embedded
2.2.5 High-side motor current monitoring for systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
overcurrent protection. . . . . . . . . . . . . . . . . 22
4.6 Robots get wheels to address new
2.2.6 Five benefits of enhanced PWM rejection for challenges and functions. . . . . . . . . . . . . . . . . . . . . . 65
in-line motor control. . . . . . . . . . . . . . . . 24 4.7 Vision and sensing-technology reference
2.2.7 How to protect control systems from thermal designs for robotic systems. . . . . . . . . . . . . . . . . . . . 66
damage. . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.2.8 How high precision in motor-drive control
5. Technologies for robotic tools (end effector)
enables industrial advances. . . . . . . . . . . . 29 5.1 Achieving accuracy in bin picking with
TI DLP® technology-powered structured
2.2.9 Getting the most out of your power stage
light systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
at the full temperature range. . . . . . . . . . . . 33
2.2.10 Anything but discrete: how to simplify a 48-V 6. Industrial communications for robotics
to 60-VDC-fed three-phase inverter design. . . 35 6.1 Selecting the right industrial communications
2.2.11 Selecting amplifiers for shunt-based current standard for sensors. . . . . . . . . . . . . . . . . . . . . . . . . . . 69
sensing in three-phase motor drives. . . . . . . 38 6.2 Enabling robots to achieve new levels of factory
2.3.1 Servo drive-related reference designs for automation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
robotic systems. . . . . . . . . . . . . . . . . . . . . . . 44 6.3 Is your factory smarter than a fifth grader?. . . . . . . . 76
6.4 Industrial communications-related reference
designs for robotic systems. . . . . . . . . . . . . . . . . . . . . 78

An Engineer’s Guide to Industrial Robot Designs 2 2Q 2020 I Texas Instruments

Chapter 1: Introduction

Analog and embedded technologies and reference designs 1.1 An introduction to an industrial robot system
from Texas Instruments (TI) enable engineers to develop
Before describing the different technologies that go into a
intelligent, autonomous and collaborative robots. This
typical robotic system, let’s discuss the different parts of a
e-book is intended to be a one-stop shop for robotics-
robot system, as shown in Figure 1. As you can see, the
related content.
system is split into the different building blocks: the controller
Our technologies let you build numerous types of industrial system, manipulator, teaching pendant, vision and sensors,
robots with precise motor control, differentiated sensing and end effector (robot tools).
technologies and processing at the edge, all with robust
real-time communication. The primary types of robots
discussed in this e-book are collaborative robots (cobots)
and factory robots:
• Cobots work side by side with humans to improve work
quality. A cobot can sense and stop movement, helping
create a safer working environment.
• Factory robots perform automated programmable
movements in manufacturing. In order to create a safer
working environment, mechanical or sensor technologies
can help keep robots from interfering with human activity.
There are six types of industrial robots:
• Vertically articulated.
• Cartesian.
• Cylindrical.
• Polar.
• Selective compliance assembly robot arm (SCARA).
Figure 1. A robotic system featuring a cobot.
• Delta.
All of these robot types offer a different axis configuration of The International Organization for Standardization (ISO)
the manipulator and include electronic content that enables 8373:2012 standard describes each of the building blocks
the robot to manage its tasks. Task management is driven shown in Figure 1 and defines terms used in relation to
by advances in software, sensors and electronics, and has robots and robotic devices operating in both industrial and
enabled the market for industrial robots to evolve over the nonindustrial environments as:
last 50 years. • The controller system. The ISO 8373 standard states,
This e-book is a compilation of technical articles, white “A set of logic control and power functions which allows
papers and application notes featuring TI’s technologies monitoring and control of the mechanical structure of
suited for these industrial robot system building blocks: the robot and communication with the environment
[equipment and users].” This is the robot’s brain, and can
• The robot controller system. include a motion controller, both internal and external
• The manipulator (robotic arm)/driving system. communication systems, and any potential power stages.
• Sensing and vision technologies.
• The end effector (robot tools).

An Engineer’s Guide to Industrial Robot Designs 3 2Q 2020 I Texas Instruments

Chapter 1: Introduction

• The manipulator. The ISO 8373 standard also states • Cartesian. These are also called rectilinear or gantry
“A machine in which the mechanism usually consists of robots. Cartesian robots have three linear axes that use the
a series of segments, jointed or sliding relative to one Cartesian coordinate system (x, y and z). They may have
another, for the purpose of grasping and/or moving objects an attached axis that enables rotational movement. Three
(pieces or tools) usually in several degrees of freedom or prismatic joints facilitate linear motion along the axis.
axes. A manipulator does not include an end effector.” • Cylindrical. This robot has at least one rotary axis at
The manipulator is commonly known as the robotic arm. the base and at least one prismatic axis to connect the
It’s the part of the robot that defines how many axes the links. The rotary axis uses a rotational motion along
robot is implementing to achieve the movement required the axis, while the prismatic axis uses linear motion.
to perform a task. Cylindrical robots operate within a cylindrical-shaped work
• The teaching pendant. Multifunctional portable environment.
equipment used to program and teach an industrial robot. • Polar. Polar robots are also called spherical robots. For
The pendant typically consists of an LCD touch panel, an these types of robots, the manipulator connects to the
enable button and an e-stop button. The teaching pendant base with a twisting axis and a combination of two rotary
connects to the robot controller system. axes and one linear axis. The axes form a polar coordinate
• The robotic end effector. A device connected on the system and create a spherical-shaped work environment.
robot “wrist” or end-of-arm tooling (EOAT). The system • Selective compliance assembly robot arm (SCARA).
controller controls the robotic end effector by using either Commonly used in assembly applications, this selectively
discrete input/output (I/O) for simple tools or industrial compliant manipulator for robotic assembly is primarily
communication protocols for more advanced tools. cylindrical in design. It features two parallel axes that
• Vision and sensors. These parts of the robot have the provide compliance in one selected plane.
ability to scan the surrounding environment and stop (in • Delta. These spider-like robots are built from jointed
the case of an industrial robot) or reduce (in the case of a parallelograms connected to a common base. A delta
cobot) a robot’s speed when humans approach. Vision/ robot has three axes for the parallelograms; for the end
sensing is implemented with light detection and ranging effector, it can have one to three axes. The parallelograms
(LIDAR), a radar-based safety area scanner or 3D cameras. move a single EOAT in a dome-shaped work area. Heavily
In addition to the safety area scanner, cobots sometimes used in the food, pharmaceutical and electronic industries,
wear a sensor-based “safety skin” that stops the robot arm this robot configuration is capable of delicate, precise
when a human touches it or is in proximity. movements.
When designing the building blocks of a robotic system,
there are mechatronics, robot functions and electrical What amount of payload (weight) and reach will
considerations that you need to understand and obtain the robot have?
specifications for before beginning the actual design. If the robot has to move heavy objects, it needs to have
Let’s discuss some typical considerations when defining the enough force on the motors to enable that capability. This
system architecture of a robot. force is generated with electric energy and is provided to the
motor from the power stage. This power requirement is part
What type of task is the robot supposed to do? of deciding whether the robot will be a high- or low-voltage
Different robot types have different advantages depending on system. A high-voltage robot system will require defined
the application. The typical types of industrial robots are: isolation architecture for safe operation.

• Articulated. This robot design features a rotary axis and

can range from simple three-axis structures to 10 or more
joints. The manipulator connects to the base with a twisting
joint. A rotary axis connects the links in the manipulator.
Each axis provides an additional degree of freedom, or
range of motion.

An Engineer’s Guide to Industrial Robot Designs 4 2Q 2020 I Texas Instruments

Chapter 1: Introduction

Will the electronics be centralized into the system

controller?
In a centralized system, the robot controller cabinet includes
most of the electronic modules that control the robot
manipulator. Figure 2 is an example of a centralized robot.

Idustrial robot
AC mains
Robot controller system cabinet sensing module

Servo drive
Circuit braker power supply module Manipulator
FAN Optional axis controller
depending on gear

Holding brake
Robot controller Servo drive

position sensor
Servo drive
Gear
Industrial robot Servo drive
power stage module
PMSM Axis 1
CPU board
Servo drive
Servo drive position sensor
Industrial robot control module Brake resistor
teach pendant (HMI)
Industrial robot
communication module
Industrial robot Servo drive Holding brake

position sensor
IO module

Servo drive
Gear
Servo drive
power stage module
PMSM Axis 2

Servo drive
position sensor
Brake resistor

E-stop

Servo drive
control panel Holding brake
Servo drive

position sensor
Servo drive
Gear
Servo drive Axis N
power stage module
PMSM

Point-to-point or Servo drive

position sensor
daisy-chain Brake resistor
N is typically
between 3 to 7
Position feedback
aggregator

Servo drive
position feedback
Point-to-point or
daisy-chain

End effector (tool)

Figure 2. Example of a centralized robotic system.

An Engineer’s Guide to Industrial Robot Designs 5 2Q 2020 I Texas Instruments

Chapter 1: Introduction

Robot controller Manipulator

system cabinet Holding brake
Servo drive

position sensor
Servo drive
Gear
Servo drive
AC mains power stage module PMSM Axis 1

Servo drive Servo drive

position sensor
control module
Circuit braker
Optional axis controller
depending on gear

Servo drive
Resistor
Brake

power supply module Holding brake

Servo drive

position sensor
Servo drive
Gear
Servo drive
power stage module PMSM Axis 2
FAN
Servo drive Servo drive
position sensor
Robot controller control module
Industrial robot
CPU board
Industrial robot
Industrial robot communication module Holding brake
teach pendant (HMI) Servo drive

position sensor
Servo drive
Gear
Industrial robot Servo drive
power stage module PMSM Axis N
IO module
sensing module
Industrial robot

Servo drive Servo drive

position sensor
control module
N is typically
E-stop between 3 to 7

Servo drive End effector (tool)

control panel
Point-to-point or
daisy-chain

Figure 3. Example of a decentralized robotic system.

In a decentralized system, some of these modules move How will the different subsystems of the robot
to the robot manipulator to support a variety of factors, communicate with each other? What are the
including the form factor of the cabinet, cabling and more. interface requirements?
Figure 3 is an example of a decentralized robot.
To ensure real-time functionality, you will need to define which
When decentralizing electronic content, it is important to control parameters pass between the subsystems, and the
remember that the environments where the electronics repetition rate and latency of the parameters for both the end
are now used are not the same as the environments of a effector and manipulator communication.
centralized system. This environment change necessitates
Figure 4 includes some typical real-time timing values for a
a re-specification of the electronics and typically requires
robot.
redevelopment of part of the system.
Now that the robot can move, it needs to know how to
move, which leads to the next series of questions.

An Engineer’s Guide to Industrial Robot Designs 6 2Q 2020 I Texas Instruments

Chapter 1: Introduction

Industrial ethernet fieldbus

Industrial robot CPU board Servo drive control module
• Robot/motion control • Position and speed control
• Position and speed control • Current control Potential
isolation barrier
Digital Analog
• Typical high-speed or parallel I/F
• Less signals to potentially isolate
PWM: VDC
Position and speed control: Current control: 8 kHz to 32 kHz 48 V to 1200 V
1 kHz to 32 kHz 1 kHz to 16 kHz 8 kHz to 64 kHz (if double update)

IQREF VA, VB, VC

position sensor
Position and Current Axis 1

Servo drive
Motion Gear
speed control PWM
control control PMSM
FOC unit
FOC
Servo drive
position sensor

DC voltage Absolute
IA, IB, IC, VDC encoder
A/D incremental
encoder
resolver
Phase current (3x or 2x)

Rotor angle/speed Rotor angle Encoder

interface
Motor angle feedback
optional axis angle feedback
Axis angle/speed Encoder
interface

Figure 4. Real-time communication timing needs for robot control.

How does the programming interface work? Is the robot a nonadaptive robot or an adaptive
Will the robot operate from the user interface robot?
or through task programming? Will you need an A nonadaptive robot does not receive feedback from the
extra interface to connect the teaching pendant or environment and will execute its task as programmed.
joystick in order to enable operator functionality? Adaptive robots use input and output data to generate
It’s important to answer these questions early in the environment data. With this data, the robot can react to
design process. Figure 4 also includes some of the design environmental changes and stop its task if necessary.
considerations of motion-control timing. And now, for the For adaptive robots, it is important to define the environment
final question. data to which the robot is reacting. The data might be pre-
defined parameters, like material amounts or sizes or shapes
for quality definitions. Or it might be uncontrolled parameters,
like having people move around the robot or malfunctions
that when detected put the robot in a safe state.

An Engineer’s Guide to Industrial Robot Designs 7 2Q 2020 I Texas Instruments

Chapter 1: Introduction

An adaptive robot requires a sensing module. An area Conclusion

safety scanner or safety skin is placed either at the base
A robot is a very complex system with many design
of the robot or attached somewhere on top of the robot.
challenges within mechatronics and functionality, as well
It supervises the surrounding area of the manipulator and
as electrical considerations. You’ll need to solve these
prevents humans or other machines from getting too close
challenges or make some decisions before a working
to the robot; if they do, the robot stops or slows down.
system is possible.
You should follow both worldwide and local safety
With all of the different technical aspects of a robotic
standards when designing robotic systems. Starting the
system, analog and embedded technologies from Texas
design efforts before looking into relevant standards can
Instruments offer many different products and designs
considerably delay time to market. Several organizations
that can help solve robot-related issues and enable the
offer consulting services to help you understand the
development of intelligent, autonomous and collaborative
hardware implications of designing a safe system
robots.
according to safety standards, including Technischer
Überwachungsverein (TÜV) – Rheinland, TÜV – Süd and This e-book is a compilation of technical articles, white
TÜV – Nord. papers and application notes featuring technologies to help
you design your next robotic system.

Authored by: Kristen Mogensen

An Engineer’s Guide to Industrial Robot Designs 8 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

2.1 Control panel Performance

In servo motor-drive applications, motor control is typically
2.1.1 Using Sitara™ processors for Industry 4.0 separated into several control-loop layers: the current/
servo drives torque loop, speed loop, position loop and a higher-level
The manufacturing and automation industries have used motion-control loop. These loops are typically arranged
servo motor control for many years, but the rise of Industry in a cascade, each with their own real-time processing
4.0 and smart factories has accelerated the adoption of requirements. The current or torque loop is the tightest
automated systems, which in turn has led to increased control loop. Each upstream loop runs at a multiple of
demand for smarter servo drives with more functionality and the loop before it and provides input references to the
the ability to control more axes. downstream loops. Figure 1 shows a typical cascaded
Historically, high-end microcontrollers and large field- control topology.
programmable gate arrays (FPGAs) performed the low-level The blocks in Figure 1 lend themselves well to logical
control algorithms and provided peripherals to connect to partitioning across cores within a heterogeneous processor,
the drive output and motor feedback. The requirements or between a processor and a microcontroller. Spreading
for what a servo drive must support are rapidly changing, the various loops among the different cores in a multicore
however, as equipment gets smarter and higher in processor maximizes the processing bandwidth dedicated
performance. Features like network communications, to each loop. When a processor core receives its control-
functional safety, predictive maintenance and programmable loop input data, it can run the algorithm to completion as
logic controllers (PLCs) are being brought into the servo quickly as possible, provide the reference value for the
control board to optimize cost and space by removing downstream loop and then continue providing other services
external boards. This increased level of integration and need until the next set of input data is ready.
for higher performance are leading designers to look to Processors with higher raw performance can finish the
heterogeneous processors, such as Sitara™ processors control processing faster and have more bandwidth available
from TI, to handle the majority – or all – of the processing to provide additional services and features. Fast processing
needs for Industry 4.0 applications. is especially crucial when cycle times approach 31.25 µs in a

Position (1 kHz to 4 kHz → 1 ms to 250 µs)

Speed/velocity (2 kHz to 8 kHz → 500 µs to 125 µs)

Torque/current (8 kHz to 32 kHz → 125 µs to 31.25 µs)

VDC

x3
Industrial
comms VA,VB,VC PWM (x3)
(position/ Position Speed PWM PMSM
loop loop Torque/ unit
speed/ current
torque loop /PWM (x3)
reference) [FOC]
Current
feedback
(x2-3)
Angle
Σ∆ Σ∆ Voltage sensor
interface ADC feedback
IA,IB,IC,VDC

Position
Angle Speed Angle Angle feedback
feedback
interface

Figure 1. Typical servo-motor control-loop technology.

An Engineer’s Guide to Industrial Robot Designs 9 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

32-kHz control loop or when inputs from multiple axes must performs the media access control layer. Manufacturers
be processed nearly simultaneously. who only plan to support a single protocol prefer this
There are a few options for the strict real-time processing distributed architecture, since ASICs are typically optimized
for that specific communication standard. Once the need to
requirements of servo control, including digital signal
support multiple protocols arises, a multichip solution loses
processors (DSPs), FPGAs and standard Arm® processing
its attractiveness for multiple reasons. Each new protocol
cores. Choosing the right processing core can be difficult
requires that you familiarize yourself with a new device (which
because there’s a balance between flexibility and optimizing
adds development effort and cost). Manufacturers must
control algorithms. In the past, optimizing control algorithms
maintain several versions of their boards for each of the
was the number one priority, so DSPs, application-specific
different protocols.
integrated circuits (ASICs) and FPGAs were the clear choice.
Solutions such as Sitara processors have integrated
Now, the need to add Industry 4.0 services to servo drives
multiprotocol support onto the host processor, helping
has resulted in the adoption of standard Arm Cortex®-A
save costs, board space and development effort, while
and Cortex-R cores. Cortex-A cores can achieve very high
also minimizing the latency associated with communication
bandwidths, which is good for rapid processing, but they
between external components and the host. A single
lack the real-time component of the Cortex-R, which is why
platform supporting multiple standards enables you to
Cortex-R is a better fit than the Cortex-A for servo control.
maintain a single board for the different versions of your end
On the other hand, the Cortex-A is much better suited than
product.
the Cortex-R for many other services, such as networking or
predictive maintenance. Fortunately, multicore devices like If you need to future-proof your products, you must also take
Sitara AM6x processors can contain all of the processing into account the need to support time-sensitive networking
elements mentioned here, enabling all necessary elements in (TSN). The platform you choose for industrial communication
a single chip. must be flexible enough to adapt to evolving TSN standards,
or you risk it being outdated once the standards are finally
Industrial communication set. The Sitara AM6x processor family provides a solution
through its flexible programmable real-time unit-industrial
Industry 4.0 brings many new and exciting things to the
communications subsystem (PRU-ICSS), which enables
factory, but the rapid adoption of multiprotocol industrial
gigabit TSN as well as traditional 100-Mbps protocols like
Ethernet is among the most noticeable in the industrial
EtherCAT.
servo drives sector. There are over a dozen different
communication protocols on the market for industrial
Functional safety
Ethernet, field bus and position encoders, each with its own
pros and cons. EtherCAT, PROFINET and EtherNet/Industrial The trend toward autonomous machine decision-making and
Protocol (IP) are the most popular Ethernet-based protocols operation, as well as increased human-machine interaction
in the servo drives market, while Hiperface Digital Servo Link, in potentially dangerous factory environments, means that
EnDat 2.2 and Bidirectional Interface for Serial/Synchronous functional safety is becoming more important for many
C are among the more popular position-encoder protocols. applications in a smart factory, including servo drives. For a
detailed description of functional safety standards and how
Many of these protocols have ASICs that you can attach
Sitara processors play in the industrial environment, read the
to host processors to support that specific communication
white paper, “The state of functional safety in Industry 4.0.”
protocol. In some cases, with a multichip solution, the
protocol’s stack runs on the host processor and the ASIC

An Engineer’s Guide to Industrial Robot Designs 10 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

System partitions of this architecture is that the total time for the field-oriented
control loop to get inputs from the motor and return a current
The cascaded control loops in a servo drive typically span at
is short, because the entire loop runs on the power-stage
least two circuit boards, separated by a reinforced isolation
board.
boundary. This isolation boundary creates what’s referred to
as a “hot side” and a “cold side.” The hot side is closest to Figure 3 also shows a two-chip solution, but this time both
the motor and includes the high-voltage components that SoCs on the control board are on the cold side. The control
supply power to the motor. The cold side is on the other side loop is split between two SoCs: one handles the algorithm
of the isolation and typically holds the control units. processing and the other acts as an aggregator and
provides pulse-width modulators (PWMs) across the isolation
The modular nature of the various control loops in a motor
boundary. The benefit of this architecture is that it enables
drive give you many possibilities when partitioning your
the use of lower-cost power-stage boards but maintains the
system across the isolation boundary. Figures 2, 3 and 4
same performance levels, whereas the partition shown in
show a few possible partitions of a servo drive.
Figure 2 requires a high-speed interface between the two
Figure 2 shows a two-chip solution, with the two system on SoCs.
chips (SoCs) separated by the isolation boundary. The benefit
Reinforced
isolation
boundary
Sitara™ Analog front end

Drive control board Power stage board(s)

(low voltage/low noise) (high voltage/high noise)

VDC

x3
Industrial
comms VA,VB,VC
PWM (x3)
(position/ Position Speed PWM PMSM
speed/ loop loop Torque/ unit
torque current
reference) loop /PWM (x3)
[FOC]
Current
feedback
(x2-3)
Angle
Σ∆ Σ∆ Voltage sensor
interface ADC feedback
IA,IB,IC,VDC

Position
Angle Speed Angle Angle feedback
feedback
interface

Figure 2. A Sitara processor communicating across the isolation boundary to a separate control unit on the hot side of the system.

Reinforced
isolation
boundary
Sitara™ Aggregator (C2000™/FPGA) Analog front end

Drive control board Power stage board(s)

(low voltage/low noise) (high voltage/high noise)

VDC

x3
Industrial
comms Position Torque/ VA,VB,VC
and PWM (x3)
(position/ current PWM PMSM
speed/ speed loop Data unit
torque loop [FOC] aggregator
reference) /PWM (x3)

Current
feedback
(x2-3)
Angle
Σ∆ Σ∆ Voltage sensor
interface ADC feedback
IA,IB,IC,VDC

Position
Angle Speed Angle Angle feedback
feedback
interface

Figure 3. A Sitara processor acting as a servo processor, with different control functions offloaded to a C2000 microcontroller or
FPGA on the cold side of the system.

An Engineer’s Guide to Industrial Robot Designs 11 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

Reinforced
isolation
boundary
Sitara™ C2000™/FPGA Analog front end

Drive control board Power stage board(s)

(low voltage/low noise) (high voltage/high noise)

VDC

x3
Industrial
comms VA,VB,VC
PWM (x3)
(position/ Position Speed PWM PMSM
speed/ loop loop Torque/ unit
torque current
reference) loop /PWM (x3)
[FOC]
Current
feedback
(x2-3)
Angle
Σ∆ Σ∆ Voltage sensor
interface ADC feedback
IA,IB,IC,VDC

Angle Position
Angle Speed Angle feedback
feedback
interface

Figure 4. Using a Sitara™ processor to implement full servo control on the cold side.

In Figure 4, the entire control loop including the PWM and motion The rest of the Sitara family integrates the PRU-ICSS subsystem,
profile generation (typically handled by a PLC) is integrated into along with other cores and peripherals, to enable control and
a single SoC on the cold side. This architecture enables even communications. The AM6x processor family takes integration
more cost savings through integration and eliminates the latency one step further by offering integrated safety features based on
associated with the interface between SoCs. Hercules microcontrollers to enable a single-chip solution for
communications, servo control and some levels of functional safety.
Solutions from TI
The Sitara processor family has SoCs to handle everything from Conclusion
stand-alone industrial communication modules to fully featured Industry 4.0 is introducing new guidelines and system requirements
multiaxis servo drives for the system partitions discussed for servo drives, making it important for designers to select a
in this chapter. Figure 5 shows what is possible with the solution that fits the needs of current and future servo drives.
different processors within the Sitara, Hercules™ and C2000 Devices like Sitara AM65 processors, which include both Cortex-A
microcontroller families. Sitara AMIC processors contain the and Cortex-R cores and support 100-Mbps and 1-Gbps industrial
PRU-ICSS subsystem and have been optimized for stand-alone networking, are capable of supporting existing and future servo
multi-protocol industrial communications modules. drives. TI also offers a variety of products, including other Sitara
processors and C2000 microcontrollers, to serve the changing
needs of the industrial market.
Sitara™ Sitara™ Sitara™ Sitara™ Sitara™ Hercules™ C2000™ F2837x C2000™ F28004x
AMIC110 Processor AM335x Processors AM437x Processors AM57x Processors AM6x Arm® Processors MCU Delfino™ Piccolo™

Servo drive products from TI MCU MCU

C2000 C2000
AMC110
Product No.
AM335x
Product No. AM437x
Product No.
AM57x
Product No.
AM6x
Product No.
Arm Hercules
Product No. Product No. Product No.
Product Group Name Group Name Group Name Group Name Group Name Group Name F2837x
Group Name F28004x
Group Name
processors
Pruct processors
Pruct processors
Pruct processors
Pruct processors
Pruct MCU
Pruct Pruct MCU Pruct MCU
Delfino Piccolo

Safety Integrated functional safety features 4 4 4 4

Time-sensitive networking (TSN) 4

Industry 4.0 services Integrated security features 4 4 4 4 4 4

Predictive maintenance 4 4 4 4 4 4
High-speed serial interface 4 4 4 4
Control
Integrated servo motor control 4 4 4 4
Communications Industrial Ethernet Multi-protocol Multi-protocol Multi-protocol Multi-protocol Multi-protocol EtherCAT®

Figure 5. TI processing products available for servo drives.

An Engineer’s Guide to Industrial Robot Designs 12 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

2.2 Servo drives for robotic systems Why use an isolated gate driver?
Data center application
2.2.1 Impact of an isolated gate driver
The demand and availability of information transaction and
With information exchange and urbanization rising at a global retrieval is ubiquitous. Time is seldom wasted in using a
level, the need for power management is becoming more smart device, for example. Check out what is in our social
critical than ever. media, our messages or email. All this is happening in the
This affects high-efficiency and robust systems that require cloud. This cloud is the workhorse of real-time connectivity
power electronics with sophisticated intelligence in order to across the globe. The cloud is physically located in a data
meet the demands of power consumption. Several day-to- center.
day applications that necessitate these requirements are Information to and from the data center is transmitted
data centers, telecommunications base stations, industrial through a line such as fiber or coax cable, or wireless through
automation, motor drives and grid infrastructure. Each telecommunications base stations. Inside the data center is
has their own power-level requirements, topologies and the power delivery unit, commonly referred to as the power
appropriate choice of power switches. supply. Information is stored in servers, known as cloud
For example, data centers and telecommunications servers. These servers need power to store and retrieve
applications use power metal oxide semiconductors information back and forth to the users. These power supply
(MOSFET) simply because these applications need to switch units are in the range of a few hundred to thousands of
faster while increasing the system power density. On the watts. They operate off the grid, which is the AC line voltage
other hand, industrial automation and motor drives typically in the range of hundreds of volts. Hence, they are referred to
deploy insulated bipolar gate transistors (IGBT) due to their as the high-voltage unit.
high voltage requirement and higher power levels. Grid There are several low-voltage components such as
infrastructure equipment such as solar inverters have the controllers and communication components required to
flexibility to choose any of the power switches, depending on improve server efficiency. Moreover, these servers are sold
the inverter type and power level. based on their efficiency ratings and therefore are becoming
Human machine interface (HMI) is involved in these mandatory to maintain those ratings. In addition, there are
applications. Additionally, intelligent systems such as humans involved in the cloud operation interacting with the
controllers and communication peripherals need protection servers through HMIs. It is important to avoid any breakdown
from the high-power and high-voltage circuitry. This is and leakage of current from the high-voltage unit into the HMI
achieved through isolation. Also, integrating the isolation as it could damage all the low-voltage components, such as
circuitry with the power transfer components helps to reduce controllers and communications.
system size cost. One key trend is to integrate the gate
Isolation is the answer
driver with an isolator (the device that performs the isolation
function) known as the isolated gate driver. This is becoming An isolation device, which is a semiconductor integrated
a key trend that makes these system-level features attractive. circuit (IC), allows data and power to transfer between the
high-voltage and low-voltage units, while preventing any
This section of the e-book is twofold. The first is to
hazardous DC or uncontrolled transient current flowing
understand why an isolated gate driver has become
from the grid. One well-known example is a lightning strike.
so attractive, which is illustrated by walking through an
Isolation is a means to break the ground loops created
application. The second is to identify the requirements of an
in circuits that have high-energy flow. There are several
isolated driver as a function of the power switch.
methods of isolation. Of all, galvanic isolation is the one that
provides protection for very large potential differences.

An Engineer’s Guide to Industrial Robot Designs 13 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

Day by day this demand for power continues to increase. First off, the GATE terminal controls a MOSFET’s ON/OFF
Not only that, people expect to receive this information state. VGS is the voltage between gate and source.
instantaneously. This means that data center capacity is • To turn ON, apply a positive voltage, VGS > threshold level.
growing by the day with increasing data demand, which
means that power delivery systems need to supply more • To turn OFF, reduce VGS < threshold level.
and more power. However, the data center has limited • GATE is a capacitive input with a high impedance.
real estate constraints. Scaling them to a larger size is • It has CGS, CGD as two parasitic capacitances residing
expensive and highly uneconomical. in the MOSFET’s internal structure.
One way to address this demand is to increase power This is where the gate driver comes into play. It acts as a
density and provide isolation robustness. This can be done power amplifier that accepts a low-power input from the
by significantly improving power supply efficiency and controller IC and produces the appropriate high-current
increasing the power transfer rate, also known as switching gate drive for a power MOSFET to turn ON or OFF.
frequency measured in kHz. This improvement helps to
make the power supply units smaller. Isolation robustness Isolated gate driver versus traditional
is realized by integrating the isolator with a key power transformer isolation
component: the high-speed gate driver. This integrated
Depending on where the controller is positioned or placed,
device is known as the isolated gate driver.
isolation is required between the controller and the driver
for high-voltage applications such as data center power
Gate driver functionality
supplies. A traditional method for isolation is using a gate
To further understand the value of such an integrated driver transformer.
solution, first you need to understand how a gate driver
Figure 2a shows where the transformer is pulsed by a
is used. Gate drivers are implemented in a system
simple low-side, non-isolated gate driver to phase leg of a
that operates under switched-mode power where the
bridge topology (Figure 2b). This is referred to as Type A.
power switch operates in an ON and OFF mode, thereby
HV
consuming zero power, ideally under high-switching VCC VCC CIO
Q1
frequencies. Two common power switches used are
PWM1
the power MOSFET and IGBT. Switched-mode power
CByp
Controller

Low
operates in a controller-based, closed-loop power topology. side
Controlling the ON/OFF status happens at the gate of these PWM2
driver

switches in order to regulate voltage and the flow of current Q2

through the switches. Take a power MOSFET, for example. VSS
SGND PGND
Figure 1 illustrates how the gate terminal works. CIO

CGD
D Figure 2a. Transformer isolation drives both high-side, low-side
G switches in a phase leg.
(Gate-to-source voltage) VGS
VDC-Link 400 V
ON ON
Threshold CGS
OFF OFF S
+
-

Figure 1. Control of a gate terminal in a power switch.

Figure 2b. One phase leg of a full-bridge topology.

An Engineer’s Guide to Industrial Robot Designs 14 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

VBIAS DBOOT
VCC HV

HB Q1
VCC
PWM1 HO
HI
Controller LI HS CByp

PWM2 VSS
LO
EN/NC Q2
VSS COM
SGND Isolator PGND

Figure 3. An isolator and a gate driver configuration highlighting a phase leg controller.

Now consider using an isolator IC instead of a transformer Now consider the isolator and the driver integrated into one
located between the controller and a high-side, low-side IC or a multi-chip module as shown in Figure 4. This is
driver (Figure 3). This is referred to as Type B. referred to as Type C. This solution is the isolated gate driver.
Table 1 shows a typical comparison between Type A and Type C gives a propagation delay similar to or better than a
Type B. discrete transformer solution while it also gives a significant
area reduction greater than 50%. Furthermore, Type C can
Specifications Type A Type B
be tailored to deliver common mode transient immunity
TProp =20 ns =100 ns
(CMTI) greater than 100 V/ns, a number significantly higher
Bias power No Yes
than that achievable by the Type A solution. CMTI is a key
CIO ≥10 pF ≤1 pF
parameter that determines the robustness of a gate driver.
Parasitics Large (LLK) Very small
Overshoot Large Small As explained in the data center application, system solutions
Size Bulky Small are becoming smaller in size. This translates into a power
supply with higher power levels and smaller board space.
Table 1. Type A versus type B comparison.
Integrating as many components as possible is vital, which
Note that the size of the isolator and gate driver combination is why this is the trend in power solutions. An isolated gate
is small compared to the transformer isolation. However, driver (Type C) is the answer to this trend. Galvanic isolation
the propagation delay, a key metric for high-power density technology is typically capacitive, optical and inductive.
applications, is significantly higher for Type B. Additionally, the isolation level (such as reinforced, basic and
functional) depends on the application.

VBIAS DBOOT
VCC HV
LDO HB Q1
VCC
PWM1 HO
HI
Controller

LI HS CByp

PWM2 VSS
LO
EN/NC Q2
VSS COM
SGND Isolator PGND

Figure 4. An integrated solution where an isolated gate driver drives a phase leg.

An Engineer’s Guide to Industrial Robot Designs 15 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

Isolated driver as a function of power switch at higher switching frequencies. CMTI is the ability of an
isolator in the gate driver IC to tolerate high-slew-rate voltage
Isolated driver functionalities are very similar to a non-isolated
transients between its two grounds without corrupting signals
gate driver such as:
passing through it. The higher the CMTI is implies that the
• Propagation delay. isolated gate driver can be used in high-switching frequency
• Common-mode transient immunity (CMTI). applications. Also, with the advent of wide bandgap switches
• Rise time/fall time. such as gallium nitride (GaN) and silicon carbide (SiC), CMTI
is becoming perhaps the most important parameter for
• Maximum driver-side supply voltage. isolated gate drivers.
• UVLO. In particular, SiC MOSFET, due to its superior material
• Channel-to-channel delay. properties, has emerged as the disruptive solution in power
• Protection schemes. electronics—resulting in energy efficient, robust and compact
systems in high-voltage, high-power applications. These
• Dead time control and overlap. applications are increasingly becoming of interest with the
• Enable/disable features. advent of electric vehicles and renewable energy power
The importance of a specification parameter is dictated by systems—making the gate driver requirements for SiC very
the application. For example, power supplies used in data critical. TI has a family of isolated gate drivers, UCC217x,
center servers and telecommunications infrastructure operate with fast, integrated sensing for SiC MOSFETs. Utilizing
at high-switching frequencies above 20 kHz. TI’s capacitive isolation technology, the UCC217x family
maximizes insulation barrier lifetimes while providing high
For such applications where a power MOSFET is used, reinforced isolation ratings, fast data speeds and high density
minimizing switching loss is key. This makes parameters such packaging.
as rise time/fall times and propagation delay very important.
Alternatively, applications like motor drives and high-power (> This stems from TI’s capacitive barrier and industry’s highest
5 kW) solar inverters operate at switching frequencies in the dielectric strength insulator, SiO2. Each capacitive barrier is
range of 5 kHz to 20 kHz. For such high-power applications built using TI proprietary technology with strength exceeding
where IGBT is used, you need to have good protection 12.8 kV of isolation surge-voltage protection and a specified
schemes and high driver-side supply voltages to ensure that isolation voltage of 5.7 kV to ensure strengthened system-
your design can tolerate the harsh environments in these level reliability. In addition, fast short-circuit protection and
applications. quick response time enhanced system protection.

One unique parameter of the isolated gate driver is CMTI, Table 2 compares MOSFET and IGBT isolated gate drivers
which is important to consider when operating the system and summarizes the differences explained earlier.

Power Switch MOSFET IGBT

Switching frequencies High (>20 kHz) Low-to-mediaum (5 kHz to 20 kHz)
Number of channels Single-and dual-channels Single-channel
Protection No Yes–Desaturation, Miller clamping
Max VDD (power supply) 20 V 30 V
VDD range 0 V to 20 V -10 V to 20 V
Operating VDD 10 V to 12 V 12 V to 15 V
UVLO 8V 12 V
CMTI 50 to 100 V/ns <50 V/nx
Propagation delay Smaller the better (<50ns) High (not critical)
Rail voltage Up to 650 V >650 V
Typical applications Power supplies: server, data communications, telecommunications, Motor drives (AC machines), UPS, solar
factory automation, onboard and offboard chargers, solar µ-inverters central and string power inverters (>3 kW),
and string inverters (<3 kW), 400-V to 12-V DC/DC, automotive automotive traction inverters

Table 2. Comparison between MOSFET and IGBT isolated gate drivers.

An Engineer’s Guide to Industrial Robot Designs 16 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

Summary Gate drivers do not need to provide constant current

because they only have to source or sink current when
High-power density and robustness are becoming increasingly
switching the gate of power MOSFET or IGBTs.
important in power management applications such as power
Figure 1 shows the turn-on waveforms.
supplies, solar inverters and HEV/EV DC/DC converters.
With increasing power levels, it is important to protect HMI Understanding the IOH and IOL specifications requires looking
and intelligent systems. Therefore, isolated gate drivers are at the pull-up and pull-down structures inside the device.
becoming the preferred solutions for these applications. The output stage of a gate driver typically comes in some
This section compared the value of the isolated gate driver variation of Figure 2. The UCC5320SC is offered in a split-
to the traditional transformer method. Key requirements for output pinout that gives you more control of the rise and fall
this driver are highlighted and explained as a function of the times without adding extra components like Schottky diodes.
power switch and application. TI offers several isolated gate
drivers for these power switches. These include isolated
UVLO2 VCC2
gate drivers such as the UCC21710-Q1, UCC21732-Q1,
UCC21750, UCC21520, ISO5451/5452, UCC5350 and
Level ROH
UCC21220 family for several applications such as power shifting RNMOS
supplies, motor drivers, solar inverters and automotive and OUTH
control
electrification systems to name a few.  logic OUTL
ROL
2.2.2 Understanding peak source and sink
current parameters
VEE2
Gate drivers are often confused as continuous current
sources because of the high-level output current (IOH) and
low-level output current (IOL) specifications in the data Figure 2. Gate driver output stage.
sheet. For example, designers looking at TI’s UCC5320SC
Under a no-load condition, IOH is determined by VCC2 and the
might read the parameters “4.3-A source” and “4.4-A sink”
parallel combination of RNMOS and ROH, while IOL is set by VCC2
and mistakenly believe that these devices are capable of
and ROL. RNMOS helps the pullup structure deliver the peak
providing these currents continuously.
current, with a brief boost in peak-sourcing current during
the Miller plateau region, shown as interval 3 in Figure 1.
VGS This is done by turning on the N-channel MOSFET during
a narrow instant when the output is changing states from
VTH low to high. When driving MOSFETs and IGBTs high, the
external gate resistor RON and the transistor’s internal gate
IG resistance RGFET_Int reduce the peak output current, as shown
in Equation 1:
⎛ VCC2 ⎞
IOH =min ⎜⎜4.3A, ⎟
⎟
VDS ⎝ RNMOS ||ROH +R ON +RGFET _Int ⎠ (1)
Likewise, the peak sink current is limited by the external gate
resistor ROFF in series with ROL and RGFET_Int and is determined
ID by Equation 2
⎛ VCC 2 ⎞
IOL = min ⎜⎜4.4A, ⎟
⎝ ROL + ROFF + RGFET _ Int ⎟⎠ (2)
1 2 3 4

Figure 1. MOSFET turn-on time intervals.

An Engineer’s Guide to Industrial Robot Designs 17 2Q 2020 I Texas Instruments
Chapter 2: Robot system controller

It is possible to demonstrate different techniques to

determine the peak drive current using the UCC5320SC
isolated single-channel gate driver and a 100-nF capacitive
load. The first method calculates the expected peak currents
based on Equations 1 and 2. Use these equations to
estimate the peak drive current when selecting a gate driver
for your system.
In order to simulate driving a MOSFET or IGBT before
installing one onto a printed circuit board, select a load
capacitor equivalent to the switch’s input capacitance, CISS.
Determine the input capacitance by looking up the required
gate charge from the MOSFET or IGBT’s data sheet at the
Figure 4. Voltage across the series sense resistor.
drive voltage condition.
Table 1 presents the results of the three techniques. Even
A second technique uses CISS and the transient voltages (dv/
with a 0.1-Ω sense resistor in series with the capacitor,
dt) of the switching waveform to determine the source or sink
Equation 1 predicts a 4.30-A sourcing current. Equation 3
current. Figure 3 measures the dv/dt using cursors set to a
uses the largest measured dv/dt value in the linear region of
fixed 35-ns interval and swept across the rising edge in order
the gate drive waveform, which gives an estimated 4.53 A. In
to find the peak dv/dt. As a guideline, set the oscilloscope’s
this same linear region, the voltage across the sense resistor
cursors to a time interval, Δt, that is approximately 10%
is measured in Figure 4, with Ohm’s law determining peak IOH
of the rise time to determine the current through the load
at 4.29 A.
capacitor.
Theoretical vs.
measured Method Result
Theoretical Equation 1:
4.30 A
IOH = min (4.30 A, 4.44 A)
Calculated from Equation 3:
4.53 A
measurement IC = 102 nF (4.44 = MV/s)
Calculated from Ohm’s Law:
4.29 A
measurement IOH = 438 mV/102 mΩ

Table 1. Measurement comparison.

The first method is a good starting point when selecting
a gate driver, but what you will obtain is not an actual
measured value. The second method relies on an accurate
measurement of the highest dv/dt by using a fixed Δt
and sweeping it across the waveform. Finally, the voltage
Figure 3. Measuring peak dv/dt across the load capacitor.
measured across the 0.1-Ω sense resistor will give you a
Use the measured peak dv/dt and load capacitor value to value calculated from the measurement of the peak drive
calculate the peak current, along with Equation 3: current using Figure 4 and Ohm’s law. The key to the third
measurement technique is to select a small-valued sense
dv
IC = C resistor to prevent any limitations in the peak output current.
dt (3)
All presented methods are acceptable approximations of a
A third method inserts a 0.1-Ω sense resistor between the gate driver’s peak output current.
capacitor and ground to calculate IOH or IOL. Figure 4 shows
To reiterate, IOH and IOL are not continuous DC values. The
the voltage waveform across the sense resistor, VSENSE, and
peak current charges or discharges CISS in an instant and
its measurement coincides with the highest dv/dt value of the
then reduces in value as the switch begins to turn on.
Vcap waveform.

An Engineer’s Guide to Industrial Robot Designs 18 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

2.2.3 Low-side gate drivers with UVLO versus Why a BJT totem pole offers no protection
BJT totem poles Figure 2 shows a BJT totem-pole configuration to drive the
Gate drivers have increasingly replaced the use of bipolar MOSFET. Figure 2 shows a typical gate-drive circuit achieved
junction transistor (BJT) totem poles to drive power switches using a bypass capacitor and an additional base resistance
in low-side applications. Gate drivers come with many built-in to limit the input current. At power on and power off, before
safety features, eliminating risks associated with the lack of the BJT drive supply settles, the MOSFET can be subject to
protection in a discrete solution. When driving MOSFETs and a combination of high voltage and high current. You could
IGBTs, safety features are important in ensuring a predictable add external UVLO circuitry to this circuit, but this addition
switching, robust gate drive. Let’s compare the UCC27517 results in an increase in component count, board footprint
gate driver and a discrete totem pole, looking at their and bill-of-materials (BOM) cost.
respective performances during undervoltage lockout (UVLO)
conditions. 3.3 V 1 nF 15 V

100 Ω
The importance of UVLO
The UCC27517 gate driver has an important built-in
protection feature that grounds the output of the driver when 50 Ω 50 Ω

the power supply has not reached the UVLO threshold.

Figure 1 shows how different values of VGS affect the Function
generator
MOSFET for a given drain-to-source voltage. The right side of
the red curve is the saturation region, defined by a constant
drain-to-source current, dependent on the gate-to-source
voltage and independent of the drain-to-source voltage. Figure 2. BJT totem-pole schematic.
This saturation region is where power losses can be high, Figure 3 shows the UCC27517 device driving the power
due to a simultaneous presence of high drain current and switch in the same conditions but with a built-in UVLO, which
high drain-to-source voltage. The left side of the red curve is typically 4.2 V with 300-mV typical hysteresis. When the
is the linear region, where the drain current is proportional to supply voltage reaches UVLO, the output of the driver rises
the low RDS(on) of the MOSFET. For applications with a high with VDD until it reaches steady state. This solution also uses
drain current, a drop in the gate-to-source voltage can be fewer components, has a lower footprint and saves BOM cost.
dangerous for the MOSFET. The UCC27517, as well as other
low-side drivers in the UCC family, prevent that drop with 15 V
their built-in UVLO, enabling safe power up.

100 Ω
Drain current vs drain source voltage for several values of VGS
40 VGS = 10 V

35 50 Ω
3.3 V VDD OUT
30 VGS = 9 V
Linear region GND UCC27517
Drain current / A

25
VGS = 8 V IN+ IN-
20 Function
Saturation region generator
15 VGS = 7 V

10
VGS = 6 V
Figure 3. UCC27517 schematic.
5
VGS = 5 V
VGS = 4 V
0
2 4 6 8 10
Drain to source voltage / V 2V / DV

Figure 1. MOSFET 1 to 5 characteristics.

An Engineer’s Guide to Industrial Robot Designs 19 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

Protecting the MOSFET and IBGT using the

UCC27517 with UVLO
At a 3.3-V startup, there is a significant difference in the
thermal behavior of both gate drives. The UCC27517 gate
driver clamps its output, preventing switching as well as the
drain-to-source voltage drop across the field-effect transistor
(FET) at its output. The waveforms in Figure 4 illustrate this
event. Channel 2, VDS_517, captures no voltage drop across
the MOSFET, while channel 4, IDS_517, shows the drain
current grounded during power up. This process occurs
until the supply voltage reaches the UVLO rising threshold.
However, the BJT allows a voltage drop across the MOSFET
captured by channel 1, VDS_BJT, while the drain current Figure 5. Thermal image of the UCC27517-driven MOSFET (left) and BJT-
rises significantly, as shown by channel 3, IDS_BJT. This driven MOSFET (right) at a 3.3-V power up.
rise in current leads to excessive power dissipation and can
Internal UVLO protection is not limited to the UCC27517 gate
potentially damage the MOSFET.
driver; it also extends to other devices in the UCC family of
low-side gate drivers.

2.2.4 An external gate-resistor design guide for

gate drivers
External gate-drive resistors play a crucial part in limiting
noise and ringing in the gate-drive path. Parasitic
inductances and capacitances, high transient voltages
(dv/dt) and transient currents (di/dt), and body-diode
reverse recovery can cause unwanted behavior without an
appropriately sized gate resistor.
Figure 1 depicts common elements in the gate-drive path:
Figure 4. Waveforms of the UCC27517 and BJT totem-pole gate drives at
the internal resistance of the gate driver, external gate
a 3.3-V power up.
resistance and internal gate resistance of the MOSFET or
Figure 5 shows a thermal image of this event. On the left, IGBT. RGATE is the only component that tunes the gate-drive
the UCC27517 driving the MOSFET (with its built-in UVLO) waveform.
prevents overheating at the FET junctions by grounding its
output. The driver’s output stays grounded regardless of its VDRV VDS, off

input during UVLO conditions. On the right, however, the FET ID

D
at the BJT totem-pole output, with no protection, is exposed CGD
to overheating due to increased power dissipation. RHI
RGATE RG,I
G
UVLO is an important feature, allowing smooth power up and CDS
RLO
power down of the MOSFET by ensuring that switching only
CGS
occurs when enough voltage is supplied. The UCC27517
S
takes care of this issue with its internal UVLO by grounding
its output and thereby preventing overheating of the
MOSFET. This feature is essential, because this excessive
power dissipation at the MOSFET junctions can occur during Figure 1. Gate-drive elements.
power up and power off and potentially damage the FET.

An Engineer’s Guide to Industrial Robot Designs 20 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

Figure 2 shows the parasitic inductances and their effect transistor gate resistance from the total series resistance
on the gate drive waveform created by long trace length and then determines the external gate resistor, as expressed by
poor PCB design. Equation 2:
X L ω LS
Q= =
RG RG
(2)
VDRV
The method described here is an iterative process starting
1
with 0 Ω as the external gate resistance and calculating a
new external gate-resistor value based on the ring frequency,
VIN source inductance and input capacitance.
VIN + VGS, Miller
RGATE 2 2 Two isolated single-channel gate drivers in a half-bridge
1
LS configuration provide proof of concept. Two UCC5310MC
3 VIN
VOUT
gate drivers driven from a 15-V supply drive two 100-V
3
FWD recovery and CSD19536KCS MOSFETs with a typical internal gate
LD VGS= current transfer
resistance, RG,I, of 1.4 Ω.
2 3
MOSFET The CSD19536KCS MOSFET’s relatively small internal
gate resistance shows the effects of adding external gate
Figure 2. Switching theory. resistors. External gate resistors may not be required if a
MOSFET’s or IGBT’s internal gate resistance is large enough.
Parasitic inductance and capacitance cause oscillations in
the gate-drive loop and are modeled by resonant circuits. At 0 Ω, there is unwanted ringing on the gate-source
Fortunately, it is possible to damp the otherwise very high waveform. The internal gate resistance of the CSD19536KCS
Q resonance between the input capacitance, CISS (CGD + MOSFET is not enough to dampen the oscillations found in
CGS), and the source inductance, LS, by the series resistive Figure 3.
components of the loop, RG (RG = RHI or LO + RGATE + RG,I).
An optimal gate-resistor selection is key for a high-
performance design. Without optimization, small resistor
values will result in overshoot in the gate-drive voltage
waveform and also result in a faster turn-on speed. Higher
resistor values will also overdamp the oscillation and extend
the switching times without offering much benefit for the
gate-drive design.
It’s best to select a gate resistor that will give your design
a quality factor Q between 0.5 (critically damped) and 1
(underdamped). A quality factor greater than 0.5 will give you
faster turn-on and turn-off if required.
Start by recording the gate-drive ring with no external Figure 3. External gate resistor where RGATE = 0 Ω.
resistance. This is your ring frequency, fR, used in Equation 1:
Using 3.57 MHz as the ring frequency and 9,250 pF as
1 the input capacitance, Equations 1 and 2 can determine
LS =
CISS(2π f R )2 (1) a critically damped resistor value. Don’t forget to subtract
The MOSFET or IGBT data sheet will provide CISS, which will the series resistive elements RG,I and RHI or LO from this
help you calculate LS. calculated value. Figure 4 (next page) demonstrates the
effects of adding a 7-Ω resistor to the gate-drive path, which
Determine when RG is equal to or twice the inductor’s makes the waveform critically damped.
reactance for underdamped or critically damped
performance. Subtracting the internal gate drive and

An Engineer’s Guide to Industrial Robot Designs 21 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

The selection of external gate resistor will affect three things: Another way to decrease ringing from the series RLC circuit
drive current, gate-driver power dissipation, and rise and fall shown in Figure 5 is to minimize the loop inductance
times. Figures 3 and 4 show the gate resistor’s dampening between the source of the high-side transistor and the
effect and its effect on rise and fall times. source of the low-side transistor. Confining the high peak
currents that charge and discharge the transistor gates to
a minimum physical area is essential. You must place the
gate driver as close as possible to the transistors to reduce
parasitics.

RG LS
VDRV CISS

Figure 5. Resonant circuits in a gate-drive design.

The trade-off between fast rise and fall times versus
Figure 4. A critically damped external gate resistor where RGATE = 7 Ω. oscillations is why the external gate-resistor element of a
gate-drive design is so valuable.
If the rise and fall times are too slow after adding an
optimized gate resistor, another option is to calculate your
gate resistor with a Q factor set to 1. This will promote an 2.2.5 High-side motor current monitoring for
underdamped solution, but be careful to prevent overshoot overcurrent protection
or undershoot. If you still see overshoot or undershoot High-power precision motor systems commonly require
conditions, look at the source and sink current of your gate detailed feedback such as speed, torque and position.
driver and find a device with greater peak currents to replace Sending this feedback to the motor-control circuitry helps
it with. This will charge and discharge your FET at a faster control the motor’s operation to be more precise and
rate, but you will need a new optimized gate resistor to efficient. Other motor-control applications such as fixed-
prevent overshooting. motion tasks do not require the same level of system
For alternate device recommendations, see Table 1. complexity in order to carry out their jobs. Ensuring that the
Optimized Performance
motor has not stalled or encountered an unintended object in
Device parameters trade-off the motor’s path, or that a short in the motor’s winding exists
Requires larger valure may be all of the feedback that’s necessary. More complex
Miller clamp feature gate resistor due to motor-control systems implementing dynamic control and
UCC5350MC
available higher source/sink active monitoring can benefit from adding a simple out-of-
current
range detection function because of the faster indication of
Need to design a
out-of-range events.
Split output feature method to prevent
UCC5320SC
available Miller current induced Placing a current-sense amplifier in series with the DC power
turn-on
supply driving the high side of motor-drive circuitry, as shown
True UVLO2 in Figure 1, enables the measurement of overall current to
UVLO2 referenced monitoring at the
UCC5390EC to GND2 feature expense of not having the motor and the detection of out-of-range conditions. To
available split output or Miller detect small leakages, it is also possible to measure the low-
clamp side return current. A difference between the high- and low-
Difficult to layout both side current levels indicates that a leakage path exists within
Configured as a
transistors close to
UCC21220 half-bridge or two the motor or motor-control circuitry.
each output when
low-side drivers
using a dual channel

Table 1. Alternate device recommendations.

An Engineer’s Guide to Industrial Robot Designs 22 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

Many devices are specified under defined conditions (such

RSHUNT as VCM = 12 V and VS = 5 V) which establishes the baseline
for the default common-mode rejection and power-supply
rejection ratio specifications. In this example, operating at a
60-V common-mode voltage creates a change in VCM of 48 V
(60 V to12 V). A 48-V change with a 80-db of common-mode
rejection results in an additional 4.8 mV of offset voltage – in
RSHUNT addition to the specified input offset voltage found in the
device’s data sheet.
Figure 1. Low- and high-side current sensing. Applications employing calibration schemes are less
The DC voltage level varies depending on the motor’s voltage concerned by this additional induced offset voltage. However,
rating, leading to multiple current-measurement solutions to for applications where system calibration cannot account for
accommodate the corresponding voltage levels. For low- this shift in offset, you must select an amplifier with better
voltage motors (~5 V), the selection of circuitry to monitor common-mode voltage rejection.
this current is simple: multiple amplifier types (current sense, The INA240 is a dedicated current-sense amplifier with a
operational, difference, instrumentation) can perform the common-mode input voltage range of -4 V to +80 V and a
current-measurement function to support a common-mode worst case common-mode rejection specification of 120 dB
input voltage range. over the entire input and temperature range of the device.
For higher-voltage motors (24 V and 48 V), the only available A common-mode rejection of 120 dB corresponds to an
options are dedicated current-sense and differential additional 1 µV of input offset voltage induced for every 1-V
amplifiers. As the voltage requirements continue to increase, change in common-mode voltage. The temperature influence
measurement errors will begin to impact the ability to on the amplifier’s ability to reject common-mode voltages
effectively identify out-of-range conditions. One specification is not well documented in many product data sheets, so
that describes an amplifier’s effectiveness at operating at you should evaluate it in addition to the room temperature
high input voltage levels is common-mode rejection. This specification. The INA240 maintains a guaranteed 120-dB
specification directly describes how well an amplifier’s input common-mode rejection specification over the entire -40°C
circuitry can reject the influence of large input voltages. to +125°C temperature range. Figure 2 shows that the
typical common-mode rejection performance for the INA240
Under the best conditions, an amplifier can completely reject
over the entire temperature range is 135 dB (less than 0.2 µV
and cancel out any voltage common to both input pins and
for every 1-V change).
amplify only the differential voltage between them. As the
common-mode voltage increases, however, leakage currents
Common-mode rejection vs temperature
in the amplifier’s input stage result in an additional input
140
Common-mode rejection (dB)

offset voltage. Monitoring larger input range levels will create 135
proportionally larger measurement errors. 130
125
For example, an difference amplifier or current-sense
120
amplifier that has a common-mode rejection specification
115
of 80 dB will have a significant offset voltage introduced 110
in the measurement based on the input voltage level. An 105
80-dB common-mode rejection specification corresponds 100
-50 -25 0 25 50 75 100 125 150
to an additional 100 µV of offset voltage induced into the
Temperature (°C)
measurement for every volt applied to the input.

Figure 2. Common-mode rejection vs. temperature.

An Engineer’s Guide to Industrial Robot Designs 23 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

A system controller has the ability to use the current-sense capability, the INA149 is a high-performance difference
amplifier’s measurement to evaluate overall system operation. amplifier capable of interfacing with common-mode voltages
Comparing the current information to a predefined operating up to ±275 V off of a ±15-V supply and has a guaranteed
threshold enables the detection of out-of-range events. A common-mode rejection of 90 dB (or 31.6 µV for every 1-V
comparator following the high-side current-sense amplifier input change).
can easily detect and provide alerts quickly to the system, Table 1 summarizes these alternate device
enabling corrective actions. recommendations.
Figure 3 illustrates the signal-chain path for monitoring and
Performance
detecting out-of-range excursions when measuring currents Device Optimized parameter trade-off
on a high-voltage rail driving the motor-drive circuitry. The Package: SOT23-6,
output signal proportional to the measured input current LMP8640HV Accuracy
signal bandwidth
is directed to the analog-to-digital converter; the output Calculated from
VCM Range: ±275V CMR, Gain
signal is also sent to the comparator to detect overcurrent measurement
events. The comparator alert will assert if the input current Calculated from Onboard comparator;
VCM: 0V to 36V
level exceeds the predefined threshold connected as the measurement 35µV VOS

comparator’s reference voltage. Table 1. Alternate device recommendations.

Supply
The INA301 is a precision current-sense amplifier with an
onboard comparator that can detect overcurrent events on
VCM = common-mode voltages up to 36 V.
-4V to +80V
RPULL-UP
Controller

RSHUNT
GPIO 2.2.6 Five benefits of enhanced PWM rejection
for in-line motor control
REF
There’s almost always more than one way to solve a
problem. Sometimes the most widely used method doesn’t
yield the biggest benefit. System designers working on
Figure 3. High-side overcurrent detection. motor-control projects use various current-measurement
methods to ensure that motors are running efficiently, and to
A key requirement for overcurrent detection circuitry is
prevent possible damage.
the ability to detect and respond quickly to out-of-range
conditions. A signal bandwidth of 100 kHz and 2 V/µs As Figure 1 depicts, there are three different ways to
enables the INA240 to accurately measure and amplify measure current in three-phase motor-drive systems: low-
the input current signal and send the output to the high- side, DC link and inline. While Figure 1 shows the traditional
speed comparator, which can issue an alert in the event of three-phase pulse-width modulation (PWM) inverter needed
a shorted condition in just a few microseconds. This fast to drive a DC motor with the three pairs of power MOSFETs
response ensures that other critical system components will IGBTs are very common as well), Figure 1 also includes high-
not be damaged by unintended excess current flowing in the side current sensing, which is typically used for gross fault
system. conditions such as a short to ground.

Alternate device recommendations

For applications measuring high voltage that need a
faster signal bandwidth or a smaller package, consider
the LMP8640. For applications requiring higher-voltage

An Engineer’s Guide to Industrial Robot Designs 24 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

Vsupply

High side

Fault detection:
• High side

Motor control choices:

• Low side
• DC link
• Inline
Inline

Low side

DC link

Figure 1. Various current-sensing methods for three-phase motor-drive systems.

Many designers use the first two methods (low side, DC If I could use an analogy, this is like trying to measure the
link and various combinations thereof) because standard liquid in a cup as it floats along the sea during a hurricane.
current-sensing solutions are readily available – typically with No wonder most designers use low-side sensing!
fast response times, high bandwidths, fast output slew rates
and low common-mode input voltages. But just because
products exist that can sense phase current via low side
or DC link doesn’t mean that these solutions represent the
easiest way. The idea behind measuring current is to try to
replicate the current being driven into the motor windings.
This replication effort occurs in software; it can be quite
involved and is never truly exact.
The inline current-sensing method seems to be the most
logical because that is the current you are ultimately trying
to measure, but there is a challenge associated with this
approach. The PWM signals driving the MOSFETs or IGBTs
wreak havoc on the current-sense amplifier. The common-
mode signals at the sense resistor are driven from the
supply voltage to ground with very fast transient switching
characteristics, while the current-sense amplifier is trying to
measure a small differential signal across the sense resistor Figure 2. Measuring phase current amid fast common-mode transients.
itself. Before describing the potential benefits of inline motor
Figure 2 is an oscilloscope shot of the sinusoidal phase current sensing using enhanced PWM rejection, let me
current (red waveform) generated by the PWM inverter. In explain enhanced PWM rejection. Enhanced PWM rejection
this case, the PWM frequency is 100 kHz, sourced by the is active circuitry that forces the output voltage to settle much
LMG5200 gallium nitride half-bridge power stage. The fast more quickly than traditional methods. As the current-sense
switching signals are what the inline current-sense amplifier amplifier detects input common-mode signals with fast
experiences as it measures the phase current. transitions, internal active circuitry minimizes disturbances
that may propagate to the device’s output.

An Engineer’s Guide to Industrial Robot Designs 25 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

An alternate method to reduce these disturbances, also • The possible elimination of galvanic isolation.
known as ringing, is to use high-bandwidth amplifiers (in the Another benefit of enhanced PWM rejection is subtle but
megahertz range) to settle the output as quickly as possible, important. With enhanced PWM rejection, you may be
but that may be an expensive proposition. able to eliminate the use of an isolated current-sensing
Figure 3 shows the output voltage signal for each of the device when galvanic isolation is not part of the system
phases represented without the introduction of noise. The requirements. Isolated devices will decouple the noise
red waveform is a representation of the signal to show that generated as the PWM signals travel through the sense
the power transistors, which are electronically commutated, resistor. This decoupling is no longer necessary with
replicate a sinusoidal waveform to the motor as closely as enhanced PWM rejection.
possible. The current-sense amplifier will experience an input • Algorithm optimization. With enhanced PWM rejection,
common-mode voltage signal from the power-supply rail the need to replicate or calculate the phase current is no
(VBATT = 48 V, for example) to ground. longer an issue because the answer is already provided
directly. Only minimal software is required to run the motor
VBATT efficiently.
• Increased motor efficiency. Motor manufacturers and
motor-drive system designers are always looking for ways
to improve motor efficiency. High AC and DC accuracy, a
fast output response and reduced blanking time enable
Expected waveform
per phase motor operation at the highest efficiency possible. Precise
timing control of a multiphase motor reduces the blanking
Voltage Signal Voltage Signal

time as much as possible, which in turn maximizes motor

Output

efficiency.
VBATT
Common-Mode

The INA240 current-sense amplifier from TI incorporates

Input

OV enhanced PWM rejection, which brings with it many

system-level benefits to your motor design.
Figure 3. Expected voltage waveform with enhanced PWM rejection.

2.2.7 How to protect control systems from

Five core benefits of in-line motor current sensing using thermal damage
enhanced PWM rejection include: In many control systems, the operating temperature is one
• Reduced blanking time. Common-mode PWM transient of many factors that can affect the system’s performance,
suppression reduces ringing at the output of the current- reliability and safety. Understanding the temperature effects
sense amplifier. Having to wait for the voltage signal to on control systems can help system designers anticipate and
settle is a major drawback, especially for systems that prevent thermal damage.
require low duty cycles (≤10%) because the time to take Typically, control systems operate within limited temperature
the current measurement is shortened (commonly known in ranges. But whenever the system operates beyond
the industry as blanking time). its specified temperature range, its behavior becomes
• In-line current sensing. Coupled with a high common- unpredictable. When operating at high temperatures, the
mode input voltage, enhanced PWM rejection facilitates the control system typically experiences a decrease in efficiency,
ability to monitor current inline. The robustness of a current- an increase in heat dissipation and accelerated aging. On
sense amplifier is a necessity due to the harsh environment the other hand, temperatures that are too low can also
to which it’s exposed. Aside from this requirement, the have a negative impact on the system’s safety and function,
amplifier must also have high AC and DC accuracy to as the operating condition experiences effects such as
provide precise current-sensor measurements. condensation. The combination of these effects may lead to
costly failures.

An Engineer’s Guide to Industrial Robot Designs 26 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

A number of discrete or integrated solutions exist that are Signal

Low pass conditioning
designed to protect control systems from thermal damage. filtering
+
ADC Vref1
Generally, these solutions comprise a temperature sensor, a -
comparator and a voltage reference (see Figure 1). Thermistor

VREF
OT + Comparator for
over temp
RBIAS R1 MCU
-
protection
VTRIP +
VTEMP Vref2
+ -
Calibration
point
RTMP61 -
R2 Figure 3. Example discrete implementation of a temperature monitor and
switch.
The specific implementation will depend on these application
Figure 1. Thermistor + comparator for threshold detection. requirements:

This approach provides real-time thermal protection without • Features.

interrupting the control processing system. Figure 2 shows • Cost.
an example of temperature switch behavior. In this example,
• Footprint.
the trip point is set to 60°C with a 10°C hysteresis.
• Power.

OUT • Accuracy.
Other features to look for include hysteresis, trip-point
VS
programmability, trip test, qualifications (like automotive or
Underwriters Laboratories), output type, channel count and
supply voltage range.

Discrete solutions
It is quite common to see a discrete implementation of a
temperature switch using a negative temperature coefficient
(NTC) thermistor, because the use of these devices is
well established. Thermistor solutions are also considered
low cost. Given the demanding requirements of thermal
50°C 60°C T(TRIP)
protection (like guaranteed performance), discrete solutions
Figure 2. Example of temperature switch trip behavior with hysteresis. often prove to be challenging and costly.
Some applications require both thermal protection and Some of the challenges when designing a discrete thermal
monitor functions, and thus you will also need an analog-to- protection solution include accuracy, reliability and efficiency.
digital converter (see Figure 3). For example:
• Because of the nonlinear nature of NTC thermistors,
maintaining a high-accuracy trip point at high or
low temperatures is difficult without using precision
components, which can increase system cost.
• Calibration is not practical in hardware-based switching
applications.

An Engineer’s Guide to Industrial Robot Designs 27 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

• A discrete implementation requires multiple components • Various threshold programming options (resistor, pin
working together, which can decrease system reliability. programmable, factory preset).
• NTC discrete solutions dissipate a significant amount of • Some devices also offer analog output.
power when hot because the NTC resistance decreases A highly integrated sensor lowers solution costs and enables
significantly at high temperatures. redundancy in safety applications.

Integrated circuit solutions: temperature switches/ TI provides a broad portfolio of temperature switches and
thermostats thermistors such as the TMP302, TMP390 and TMP61,
Another thermal protection solution for control systems is an shown in Figure 4. The TMP302 is available in a small-outline
integrated temperature switch/thermostat. Typically, these transistor-563 package (1.6 mm by 1.2 mm). The device offers
devices have a temperature sensor, comparator and voltage low power (15-μA maximum) and ease of use through pin-
reference fully integrated in a single chip. These temperature selectable trip points and hysteresis. The TMP302 can achieve
switches are smart sensors that autonomously make decisions a trip-point accuracy of ±2°C from -40°C to 125°C without any
to provide real-time thermal protection without interrupting calibration.
the control processing system. The key advantages of these Design tips
sensors are: The TMP302 measures the temperature of the device leads.
• The ability to autonomously enable thermal protection Careful PCB layout considerations are essential to accurately
independent of the control unit. measure ambient or board temperatures. As with any physical
• No software needed. board design, environmental factors can significantly affect
system performance. To avoid leakage and corrosion, the
• Guaranteed temperature accuracy for trip point with system must be kept insulated and dry. This is especially true if
hysteresis. the system operates in cold temperatures where condensation
• Simple and cost-effective over- and/or under-temperature can occur. Printed circuit coatings can help ensure that
detection. moisture will not corrode the sensor or its connections.

VS

Bias

TRIPSET0 Temperature
threshold and
TRIPSET1 hysteresis

HYSTSET Alert OUT

Comparator

Temperature
sensor

GROUND

Figure 4. Pin-programmable IC temperature switch block diagram.

An Engineer’s Guide to Industrial Robot Designs 28 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

2.2.8 How high precision in motor-drive control Benefits of motor control

enables industrial advances Controlling motor drives electronically introduces a level
Picture a robot in an automotive manufacturing plant. It picks of precision that leads to lower costs, greater productivity
up an engine block, moves it to a car chassis, places the and new manufacturing capabilities. Drive control ensures
block exactly, releases it and returns to its original position to a stable rotor position relative to the shunts, so that motor
repeat the process. The robot can lift much greater weight output is more predictable and power usage more efficient.
than humans, move items to specific spots more consistently When the load changes on a motor, electronic control
and repeat the same operation without stopping – 24 hours a can instantaneously modify the voltage input and torque
day if necessary. delivered, thus matching the machine’s output force and
power consumption more closely to the application.
Robots like these have become a mainstay of manufacturing
for automobiles and many other industries, and their use Electronic input control also makes it possible to change
is ever-increasing. But robotics would be impossible to drive speeds within the motor itself instead of relying on
operate without precise motor-drive control. At each point expensive gears, belts and pulleys to output different speeds.
in its operation, a multiaxis robot has to employ different Control enables stepper motors to move rotor positions
amounts of force in three dimensions in order to move the in the small increments or microsteps required for robotic
engine. Motors inside the robot supply variable speed and motion control. For all of these reasons, efficient operation
torque (rotational force) at precise points, which the robot’s enables applications to use controlled motors that are better
controller uses to coordinate motion along different axes for scaled for the job, eliminating much of the overhead that
exact positioning. After the robot releases the car engine, the would be otherwise required.
motors scale back the torque while returning the arm to its Operational efficiency goes hand in hand with increased
initial position. productivity. For instance, a conveyor belt on an assembly
Just as the control of motor drives enables advances in line usually operates in tandem with other systems that load
robotics and other areas, motor control itself depends on items on the belt, perform operations on items, or receive
advances in electronics that make precise control possible items from the belt to perform further steps. Belts usually
during real-time operation. High-end power supplies, move with a consistent forward motion, but operation
intelligent motor drivers, high-performance control signal may sometimes require changing speeds, stopping or
processing and exact sensing feedback all function together briefly reversing direction. These motions, together with
to deliver the precise speed and torque that complex continual changes of the number and weight of items on
machines need instantaneously. The result is greater the belt, require controlled motor drives that can adjust the
functionality, higher productivity and increased safety for both output automatically. Coordination among motors may be
equipment and people. required as well, since multiple belts in a factory are often
synchronized in order to keep items moving at an optimal
TI offers a range of integrated solutions that enable the
rate. Controlled motors that keep belts moving predictably in
development of advanced motor-control systems. TI provides
changing conditions not only have a positive impact on the
design engineers with integrated circuit (IC) products,
productivity of the entire plant, but play an essential role in a
including software and tools, that enable them to develop
contemporary manufacturing environment.
the increasingly precise motor-drive control that industries
require. Based on years of engagement with leading motor Precisely controlled motors also enable new manufacturing
manufacturers, TI helps engineers simplify the design of capabilities. Robots offer the most striking example, where
motor-control systems while enhancing the performance of electronically controlled motors provide fine-grade control
their products. of motion, often combined with strength and speed that
go beyond those of humans. While the earlier example of

An Engineer’s Guide to Industrial Robot Designs 29 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

a robot moving an engine block illustrates strength, other In these three power signals, changes in the motor load
examples emphasize precision or speed of movement. impact the current feedback that is sensed, digitized and
Pick-and-place robots, for instance, repetitively perform fine sent to a digital processing unit (like microcontrollers [MCUs],
motion control with tolerances measurable in micrometers, microprocessors [MPUs], processors or field-programmable
and usually far more rapidly than a human. gate arrays [FPGAs]). High-speed digital signal processing
By helping remove human workers from high-speed, algorithms in the digital processing unit determine in real time
repetitive, sometimes dangerous tasks, electronically whether changing conditions make it necessary to adjust the
controlled motors make work environments safer. A new power being delivered. The processing unit sends a control
development within this trend is occurring today, as robots output with PWM to the three-phase inverter in order to
are being designed to operate safely in collaboration with gate power switching and regulate the power outputs to the
human workers. While safety issues are often related to windings, thus supplying greater or lesser torque or speed
system operations and employee procedures, they also from the motor. Additional sensing data may be fed into the
include the internal control electronics, which have to protect controller to track system input voltage and temperature
both equipment and workers from electrical discharges. changes.
Safety is always an important element in the design of All of these components require a high level of performance
industrial machines and the motors that run them. for precise motor control. The switched-mode power supply
(SMPS) to supply the control system must be capable of
Design challenges for precise motor control ultra-high-speed switching, controlled with consistently high
Many types of motors are used for specific tasks, but the resolution. Power-supply design is exacting because of
majority of industrial motors run on three-phase power from the high voltages and frequencies involved and numerous
an AC electrical supply. Figure 1 shows a block diagram passive components required, which introduce mutually
of the representative control electronics for such a system. reactive impedances that are difficult to manage. Fortunately,
AC power input is rectified to DC. A pulse-width modulation new high-frequency materials and integrated SMPS modules
(PWM) switched three-phase inverter creates three high- are making it much easier to design high-performance power
frequency pulsed voltage waveforms that it outputs in supplies in control systems.
separate phases to the three phase windings of the motor.
Isolated IGBT
Rectifier diodes gate drivers IGBT module

DC+

Input power
supply Drive output
M
PWM signals

DC-

Communication Bus, Isolated current

RS-485, CAN, Control module and voltage sense
Ethernet µP

Encoder

AC motor drive

Figure 1. Control for a three-phase AC induction motor.

An Engineer’s Guide to Industrial Robot Designs 30 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

Accurate motor control also requires extremely high-speed on-the-fly adjustments may be necessary to enable fine-
computation in real time, which is best supplied by MCUs grained robotic pick-and-place movements to consistently
with digital signal processor (DSP) capabilities. DSPs are also measure in micrometers, instead of drifting with heat into
capable of performing digital filtering and other functions to less precise movements with tolerances in millimeters. Given
help protect systems from power transients and other signal that manufacturing environments are often highly demanding
flaws, while at the same time reducing the need for analog in terms of temperature, dust, vibration and other stresses,
components that perform these functions. it is even more important to carefully design motor-control
While dedicated logic and general-purpose MCUs might electronics for precise operation that is consistent throughout
be used for low-cost applications where basic control is a wide range of conditions.
sufficient, industrial motors in robots and other advanced
Enabling technologies for motor-drive control
manufacturing equipment require instantaneous response
and accuracy, as well as the programming flexibility and TI offers technology to help with the design of precise
advanced algorithms that digital signal control MCUs offer. motor-control and reliable driver electronics that can operate
effectively in today’s integrated manufacturing environments.
One of the biggest challenges of motor-control systems is
The company’s solutions include isolated and nonisolated
in designing high-resolution current and voltage sensing
switching gate drivers, feedback signal conversion, and high-
feedback. Designs may measure current feedback from only
speed processing for real-time control, as well as auxiliary
one shunt, but a more thorough (if more compute-intensive)
functions such as programmable clock generators and DC/
approach measures feedback from all three shunts. To avoid
DC power supplies. For advanced SMPS and three-phase
the possibility of analog signal loss or interference, designers
inverter designs, TI provides high-frequency gallium-nitride
increasingly digitize feedback signals as close to the sensor
as possible. However, digital feedback signals can have (GaN) gate drivers and modules that include GaN switches
potential problems with timing, especially as clock speeds and gate drivers. For lower-voltage three-phase inverters, TI
increase and sampling rates rise, which brings narrower provides high-performance smart gate drivers, drivers with
timing windows. Different trace lengths for clocks and data built-in FETs and drivers with integrated control, which result
signals can intensify this problem, potentially causing data in simplified but accurate control and very short development
errors if the signals drift as components heat up during times. Products include safety features, such as reinforced
operation. Good design practices using advanced signal isolation that meets industrial specifications, and they are
modulators can minimize these problems; algorithms that tested and qualified for use in harsh industrial environments.
modify variables in keeping with temperature gain can Among the most important recent TI innovations for motor
compensate as well. control is the AMC1306 isolated delta-sigma modulator, a
The more precise the application requirements, the more device that digitizes signals from current and other sensors
carefully the motor has to deal with changes in temperature, and outputs a combined data and clock signal for maximum
voltage input, timing and other factors. For example, a timing efficiency. The AMC1306 integrates TI’s integrated
robot arm that moves an object in a straight line in three- capacitive isolation technology in series to achieve reinforced
dimensional space may change its trajectory when the isolation in a minimal footprint. Delta-sigma analog-to-digital
system is operating at a high temperature, unless the control conversion of the change in sensor output level is followed by
design compensates for these changes with temperature Manchester coding of the clock rate onto the data stream, as
sensing and algorithm adjustments. These same types of shown in Figure 2 (next page).

An Engineer’s Guide to Industrial Robot Designs 31 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

DC link Motor

RSHUNT L1

RSHUNT L3
RSHUNT
L2

AMC1306Mx
3.3 V AVDD DVDD 3.3 V
AINP CLKIN
AINN DOUT
AGND DGND TMS320F28x7x
AMC1306Mx
3.3 V AVDD DVDD 3.3 V SD-D1
SD-C1
AINP CLKIN
AINN DOUT SD-D2
SD-C2
AGND DGND
AMC1306Mx
SD-D3
3.3 V AVDD DVDD 3.3 V
SD-C3
AINP CLKIN
SD-D4
AINN DOUT
SD-C4
AGND DGND
AMC1306Mx
CDCLVC1104 PWMx
3.3 V AVDD DVDD 3.3 V
AINP CLKIN
AINN DOUT
AGND DGND

Power board Control board

Figure 2. Reinforced isolated phase current-sense reference design with small delta-sigma modulators.

The result is a highly robust signal that significantly decreases C2000™ MCUs designed for high-performance computation
problems with setup and hold times, which can occur with and programming ease, along with a peripheral set for use in
changes in operating temperature, thus simplifying the design control systems. The C2000 DesignDRIVE software platform,
and routing of three-phase motor-control systems. with support for a wide variety of motor types, helps
speed algorithm development and system implementation.
Manchester coding with the clock embedded in InstaSPIN™ motor-control solutions provide algorithms,
the data stream tools and reference designs for evaluation, fast learning and
To help implement the AMC1306 modulator, TI created the quick development. TI’s DSP and analog expertise combine
Reinforced Isolated Phase Current Sense Reference in comprehensive solutions that save developers time when
designing advanced motor drive control.
Design with Small Delta Sigma Modulators reference
design. Figure 2 shows the functions of the reference design, Precise motors for integrated manufacturing
including the AMC1306 used for current-, temperature- and
Industries are moving ahead with more precise control,
voltage-sensing signals. (The device is used for current
greater communication among machines, a wider range of
sensing on all three power signals to the motor shunts, but
sensing inputs, and new capabilities in robotics and artificial
to reduce detail, Figure 2 shows only one signal.) The dotted
intelligence. These advances are leading to a new level of
red line within the reference design’s circuits, indicates that it
integrated automation and data exchange known as the
is effectively isolated for safety. Breaks within the AMC1306
triangles indicate specific points of reinforced isolation, as fourth industrial revolution, or Industry 4.0.
do the dotted red lines within the UCC5320 and UCC23513 Precisely controlled motors make an important contribution
insulated gate bipolar transistor (IGBT) switch drivers, which to Industry 4.0, since they drive almost every motion made
also feature TI’s integrated capacitive isolation technology. by industrial machines. TI’s advanced technology has played
a significant role in bringing about high-resolution motor
Figure 2 shows control processing using a TMS320F28379D
control, and it continues to help manufacturers push motor
Delfino™ 32-bit floating-point MCU, part of TI’s family of
and motion control to higher levels.

An Engineer’s Guide to Industrial Robot Designs 32 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

2.2.9 Getting the most out of your power stage How does the temperature error degradation of
at the full temperature range the SOA affect system performance?
When designing a power stage for motor control, you Figure 1 illustrates the effect that temperature error has in
can drive down the total system cost if you make special regards to RMS current at ambient temperature, assuming
considerations regarding efficiency. This includes optimizing that the NTC has an error of 3.9°C and that the TMP235A2
the field-effect transistor, switch node and control algorithms. has an error of 2.0°C. Using the SOA curve on the maximum
During design, you will need to protect the system against phase current with regards to ambient temperature, it is
an over-temperature condition. If the system gets to a certain possible to define the maximum phase current possible
temperature level, the components on the printed circuit before cooling is required. You can use this maximum phase
board (PCB)are outside their specification ranges, which current to calculate the power degradation of the power
can damage the components and cause the drive system to stage given a specific temperature sensor error.
malfunction. Figure 2 is based on calculations found in the three-phase
Temperature sensors monitor and protect power-stage inverter reference design.
components to keep the drive system within the safe

Maximum apparent load at 85°C ambient (W)

550
operating area (SOA). An SOA is a defined operating
540
temperature range for a system at a specific apparent load Power reduction:
530 4%
or root-mean-square (RMS) current in a phase that the drive
520
can support without extra cooling abilities. The temperature 8%
510
range for industrial equipment is typically an ambient
500
temperature of -40°C to 85°C. Safety margin
490
Maximum power
Figure 1 shows an SOA curve generated with 48-V/500-W 480
Three-Phase Inverter with Smart Gate Driver Reference 470

Design for Servo Drives. The curve is defined based 460

on results from thermal camera testing and an efficiency 450

0 2.0 3.9
measurement at 10 ARMS. With an assumed zero temperature
Ideal TMP235 NTC
error, it is possible to use this curve as a reference against Measurement error (°C)
a negative temperature coefficient (NTC) thermistor and TI’s
TMP235A2 sensors. The difference in the SOA is a result of Figure 2. The usable apparent load at 85°C using different temperature
the temperature error of the sensors and indicates the need sensors.
for a safety margin in order to ensure operation within the
You can see that the power stage can support 539 W – if
SOA of the drive.
it was possible to measure the temperature without error.
90 Now, because of the temperature error of the sensor, you’ll
88
need to add a safety margin. This safety margin means that
Ambient temperature (°C)

86
the power stage needs to be degraded by 4% or 8% of the
84
82
potential power usage of a servo drive power-stage module
80 system. If the power stage needs to support 500 W (which
78 NTC is clearly possible, as shown in Figure 1) but you choose to
76 TMP235A2
use an NTC, you would need to add additional cooling to the
74
Ideal
system to support the system’s full temperature range, or
72 redesign the system for higher efficiency.
70
8 8.5 9 9.5 10
Phase current (ARMS)

Figure 1. SOA differences caused by sensor temperature error

compensation.

An Engineer’s Guide to Industrial Robot Designs 33 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

To ensure that the power stage is only used under To avoid working outside the recommended temperature
recommended temperatures, a temperature sensor monitors range, you need to ensure that your system does not
the temperature and shuts down the power stage in the exceed temperatures higher than 125°C. In the smart gate
event of an over-temperature condition. The accuracy of driver reference design, a thermal camera measures the
the temperature sensor affects the power stage’s maximum temperature. In a real system, either a thermistor or IC sensor
temperature limit due to the temperature safety margin. would measure this temperature. In the comparison shown
Below we discuss how to generate the right configurations in Figure 1, a small error is introduced into the calculations
to get the most out of your power stage, and some of the by incorrectly assuming that the thermal camera has no error
considerations you should keep in mind for your design. and is ideal. A resistance temperature detector probe would
Above, we identified that temperature sensor error introduces offer a more accurate baseline method of sensing.
a safety margin that affects the safe operating area of the The various methods of the various sensor types add a
power stage. Aside from sensor error, another error not measurement error, which needs to be compensated from
accounted for in the comparison shown in Figure 1 is the the ideal measurement. The smaller the error, the more
analog-to-digital converter (ADC) error, since both the power the three-phase inverter can supply before the system
TMP235A2 and negative temperature coefficient thermistor must turn off due to an over-temperature error. In other
(NTC) are delivering an analog signal. The ADC error will words, the temperature sensor will limit your overall system’s
affect both signals equally, however at hotter temperatures apparent load capabilities without even looking at the gate
NTCs become very nonlinear which adds to the challenge of driver, FETs or any of the other factors affecting power-stage
obtaining an accurate measurement. A silicon temperature efficiency. This is why temperature sensor accuracy is so
sensor like the TMP235 maintains its linearity across important.
temperature, making it easier to compensate the error. What does this mean for your design? Let’s use the smart
A digital temperature sensor like the TMP117 can eliminate gate driver reference design as an example and compare the
the ADC error and achieve an apparent load that’s much temperature error of an NTC thermistor with the TMP235A2
closer to the “ideal” curve achieved using a thermal camera. IC temperature sensor. First, how do you quantify the
When your system is operating at an ambient temperature temperature error?
of 85°C, the PCB and ICs are in fact working at a higher Looking at the TMP235A2’s data sheet, it has a temperature
temperature due to the self-heating of the ICs. A typical error of ±2°C for the full temperature range. There is a small
industrial-grade IC has an operating temperature of 125°C footnote in the data sheet that refers to the lookup table used
or higher. to correct the device for temperature drift.
When the ambient temperature is 85°C, the self-heating of For the NTC, it takes a little more time to calculate the error.
the ICs can only raise the IC up to 40°C. If the IC gets to a Mathematical equations can generate a model that you can
higher temperature the IC is working outside of temperature use to calculate the error in Celsius degrees. The equations
range. For more details about self-heating, see the 48-V, can typically be found on the NTC data sheet or NTC
500-W Three-Phase Inverter with Smart Gate Driver application notes. With these equations, you can generate
Reference Design for Servo Drives design guide. the curves representing the NTC model, as seen in the
TMP6131 parameter simulator.

An Engineer’s Guide to Industrial Robot Designs 34 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

TMP6-System Parameter Simulator

Divider circuit parameter input Temperature calculations
Parameters
Notes
PTC1 PartNum: TMP6131
Select TMP613x device,10/Ω 1% Temperature error
Divider topoloty: Bottom
Select thermistor position in voltage divider Show NTC
Linearization Order: 4 6
Select output polynomial fit order
Show TMP6
4

Temperature error (°C)

Parameters Value Unit Notes Calibrate at 25C
Min temperature -40 degC Minimum operating temperature 2
Max temperature 125 degC Maximum operating temperature 0 NTC temp err
VCC: 3 V Divider bias coltage, typical is 3V
-2 -50 0 50 100 150 min
VCC tolerance: 0 % Use 0% for ratiometric mode
Rbias: 10000 Ω Nominal bias resistor value, default is 10 k Ω NTC temp err
-4
Rbias tolerance: Typical tolerance is 1% max
1 %
Rbias TCR: Bias resistor tempco, typical is 100ppm/C -6
100 pmm/C TMP6 temp err
-8 min
Temperature (°C) TMP6 temp err
max

Figure 3. Temperature error calculation of an NTC using the TMP6131 system parameter simulator.

The TMP6131 system parameter simulator can simulate the total apparent power of the power stage and provided
the NTC model and estimate the accuracy when applying understanding on the effect the temperature sensor error has
polynomial fitting, as opposed to a lookup table. Figure 3 on the power stage system performance.
shows the setting used for the simulation and the result.
The blue curve in Figure 2 defines the temperature error. 2.2.10 Anything but discrete: how to simplify a
The area used is from -40°C to +125°C. You can see that
48-V to 60-VDC-fed three-phase inverter design
the maximum negative and positive error is -3.6°C and
+4.3°C, respectively. For ease of calculation let’s normalize Imagine that you’re designing the next power stage of
this temperature error range to ±3.9°C of the NTC. Due to a servo, computer numerical control (CNC) or robotics
the NTC’s nonlinearity over temperature, the compensation application. In this instance, the power stage is a low-voltage
should occur in the temperature range specified for the DC-fed three-phase inverter for voltages ranging from 12 VDC
system. to 60 VDC, with power ratings less than 1 kW. This voltage
rating covers the range typically used for battery voltages
There are two basic ways to linearize thermistors in software:
in battery-operated motor systems or low-voltage DC-fed
polynomial fitting and a lookup table. These methods will
motor systems. On top of this, your boss says, “By the way,
get the NTC, PTC and or TMP235A2 measurements closer
you need to design this to work without extra cooling of the
to the ideal measurement and actual temperature. In the
power stage. It has to be as small as possible to fit the target
simulation tool, polynomial fitting will reduce the linearization
application needs and of course it needs to be low cost.”
error. Using a higher polynomial fitting order increases the
NTC’s accuracy, but also increases the time it takes for the No problem, right?
processor to calculate the temperature. This extra calculation
Well, in this case, there is a approachable solution for
time results in additional power consumption, which also
designing an inverter that meets the demands of this
affects system efficiency. The lookup tables require processor
hypothetical – albeit demanding – boss.
memory.
But before you start defining the specified power stage,
Calibration errors are not considered in the simulation, but
current sensing and protection circuit, it is important to
will affect both the sensors similarly.
consider a very real and accessible reference design for
This chapter section defined the NTC accuracy error a 48-V/500-W three-phase inverter with smart gate driver
as ±3.9°C and the TMP235A2 error as ±2°C, and servo drives.
discussed how these errors translate into a derating of

An Engineer’s Guide to Industrial Robot Designs 35 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

The 48V/500W Three-Phase Inverter with Smart Gate The covered modules of the low-voltage DC-fed servo drive
Driver Reference Design for Servo Drives achieves a small in Figure 1 have a huge impact on system performance and
form factor using highly integrated circuits that include three affect the design considerations.
half-bridge gate drivers with 100% duty-cycle operation It is possible to build a robust system by adding fault
and selectable source/sink currents from 50 mA to 2 A. detection to the half-bridge gate drivers for VDS sensing
VDS sensing enables overcurrent protection, which prevents and soft shutdown. These integrated features enable the
damage to the power stage and motor. The VGS handshake gate-driver system to detect typical overcurrent or when
feature protects the power stage against shoot-through short-circuit events take place, without adding extra current
caused by erroneous pulse-width modulation configurations. sensing or hardware circuits to enable dead-time insertion.
A typical low-voltage DC-fed servo drive power stage can be Thus, the MCU can’t provide a wrong drive signal and
partitioned like Figure 1, which is based on the DC-fed servo damage either the power stage or the motor with a shoot-
drive power-stage module. The boxes outlined in red are the through short circuit.
modules.
Brake
control Non-isolated
Safe gate driver
torque DC link Regen brake power stage Brake
off input High-side resistor
Isolation smart switch DC/DC

Diagnostics/monitoring Non-
isolated AC
integrated motor
Logic
DC link (24-V to 100-V DC) gates drivers/
half bridge FET’s Holding
Gate driver drivers brake
supply Temp
FET REF sensor
3-phase power stage
Amp

PWM
controller
PWM & Gate driver supply
Isolated DC/DC main FLT
REF Comp Temp
24 V & sense Comp
DC link current & Brake Amp
other rails
voltage sense control Temp monitoring

LDO
EFuse
DC/DC PMIC ADC Amp ISO delta
24 V sigma
AUX Supervisor
Oring sequencer LDO REF
controller Multiple Amp
Non-isolated DC/DC rails Comp
power supply Fluxgate/hall
FET
Input power Signal Motor current &
protection MCU conditioning voltage sense
ASIC
CLK distribution FPGA
Clocking Isolation High-/low-side DC/DC
switch
Holding brake power stage
CAN Ethernet
Connection to
control module RS485
CMOS Isolation
LVDS Isolation Switch/LDO
Wired interface Digital processing FAN power stage Fan

Figure 1. DC-fed servo power stage.

An Engineer’s Guide to Industrial Robot Designs 36 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

One consideration in building a robust system is optimizing impedance, an amplifier with an additional integrated fixed
efficiency to reduce cost for the heat sink and radiated gain and zero offset further helps reduce system cost while
emissions (electromagnetic interference) versus the switching ensuring highly accurate current measurements.
speed. Implementing these features with 100-V single- or An 100-VDC buck regulator creates an intermediate rail
half-bridge FET gate drivers requires additional active and from the DC input to supply the gate driver and point of
passive components, which increases bill-of-materials load. The power stage needs to work at a high efficiency to
(BOM) cost and PCB size, while reducing the flexibility to reduce self-heating in order to meet the industry’s ambient
modify parameters like the strength of the gate drive. When temperature of operation, which is typically 85°C. Given this
analyzing the system efficiency the current sensing circuit, requirement the ICs used in the system need to support
FETs with low RDS(on) and a low gate charge to enable fast even higher temperatures, as the electronics will always have
switching affect the system efficiency performance. System some temperature increase (self-heating).
designers typically want to achieve 99% efficiency of the
power stage. The reference design for servo drives was tested from
0 W to 500 W of output power with a permanent magnet
To enable continuous phase-current sensing with minimum synchronous motor. Figure 2 shows that the motor load
losses, the reference design uses 1-mΩ in-line shunts. was controlled with a dynamometer.
The resistor value is a compromise between accuracy
and efficiency. Conclusion

A major challenge for nonisolated in-line amplifiers is the wide The three-phase inverter reference design shows how to
common-mode voltage (0 V to 80 V) used for the system, design a compact hardware-protected power stage with
considering that the shunt full-scale voltage is ±30 mV in low BOM count, in-phase current sensing, fault diagnostic
this reference design (designed for ±30 ARMS). This is a small capabilities and high efficiency. The reference design uses the
signal compared to the common-mode voltage of 48 V. TI’s DRV8530 100-V three-phase smart gate driver with buck
Therefore, you will need a current-sense amplifier with a large regulator and the INA240 80-V, low-/high-side, bidirectional
common-mode voltage range and very high DC and AC zero-drift current-sense amplifier with enhanced pulse-width
common-mode rejection. Because of the low shunt modulation rejection, which enables optimization of a low-
voltage DC-fed power stage.

Figure 2. Test setup of the motor drive power stage.

An Engineer’s Guide to Industrial Robot Designs 37 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

2.2.11 Selecting amplifiers for shunt-based 110 to 690 VAC or from 12 to 60 VDC. To get the motor
current sensing in three-phase motor drives phase currents, shunts are typically placed either at the
DC-link return to ground, between the bottom switch and
Accurate phase-current sensing has a significant impact on
ground, or in-line with the three-phase power to the motor
the performance of vector-controlled three-phase inverters
(see Figure 1).
for industrial motor drives. The motor phase currents can be
measured through Hall-effect, fluxgate or transformer-based Each shunt placement has advantages and challenges,
magnetic sensors, or through shunt resistors. Magnetic and specific requirements so that the amplifier can convert
sensors offer inherent isolation and a wide current range, the small shunt voltage into an analog or digital signal for
while shunt solutions offer cost-effective, highly linear and processing by a microcontroller (MCU). Figure 2 shows the
high-bandwidth sensing options. Phase currents can be as ideal shunt current versus the phase current over one pulse-
high as 100 A with three-phase inverters operating from width modulation (PWM) cycle for each shunt placement.

VDC

u
v
Up to 1200 V

VW , i W
w

PWM W-LS
3 2 or 3 In-line motor shunts

i DC i LS-W 2 2 or 3 Low-side shunts

Power
GND

1 DC-link single shunt

Figure 1. Current-shunt options in three-phase inverters.

1 Ideal DC-link shunt current over one PWM period 2 Ideal low-side shunt current over one PWM period 3 Ideal in-line shunt voltage and current over
one PWM period for phase W to motor
Sample DC-link current Example:
twice to reconstruct the Almost zero common-
i U = -4 A mode shunt voltage
3-phase currents
shunt current, i L S-W
Ideal shunt current, i DC

VW
voltage and current
Ideal low-side

Example:
Ideal in-line shunt

i V = -1 A
iW = 5 A Ideal current-sampling point i U = -4 A
(PWM center aligned) for
i V = -1 A iW
inverted phase current
i DC = +i W

iW = 5 A
i DC = –i U

0
0
0
Very high common-mode shunt voltage and
Bottom PWM

PWM U-LS common-mode transients, 10 kV/µs

i LS-W = i W
PWM V-LS
Bottom

Bottom
PWM

PWM

PWM W-LS PWM W-LS PWM W-LS

PWM cycle, e.g. 8 kHz PWM cycle, e.g. 8 kHz PWM cycle, e.g. 8 kHz

Figure 2. Shunt current vs. phase current and common mode depending on shunt placement.

An Engineer’s Guide to Industrial Robot Designs 38 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

From a system view, the motor in-line shunts offer major Placement No. 1: a single DC-link to ground shunt
performance advantages, while from an amplifier view, the
A single DC-link shunt is more common in low-cost, low-
low-side shunts enable lower-cost solutions, as shown in
power, vector-controlled fans and pumps than in industrial
Table 1.
AC and servo drives. The DC-link current must be measured
In shunt-based systems, the shunt resistance and package twice per PWM cycle at two different PWM switching
are a compromise between accuracy, thermal performance, states to reconstruct the three-phase currents. The short
PCB size and cost. In motor drives, the shunt resistance is measurement cycle at small voltages requires amplifiers like
such that the voltage drop at the maximum phase current is the OPA835 from TI. This amplifier offers a high large-signal
typically from ±25 mV to ±250 mV. The subsequent amplifier unity-gain bandwidth of at least 20 MHz and high slew rates
converts the small bipolar shunt voltage into a typically (>10 V/µs) to settle in <1 µs. The method won’t work for
unipolar output voltage, with bias offset matched to the zero phase voltage since all three PWM duty cycles are 50%,
ADC’s 3-V to 5-V input range. Gain settings are typically from unless extended with sophisticated PWM compensation
10 to 100 V. algorithms.
For each of the three shunt placements, the shunt resistance
Placement No. 2: low-side current shunts
tolerance and drift, as well as the amplifier’s gain, input offset
and related drift over temperature, have similar impacts on Using low-side current shunts are attractive for compact
accuracy. AC-line-fed inverters up to approximately 5 kW and for
12- to 60-VDC-fed motor drives, where the control MCU is
Consider a shunt with a ±50-mV maximum voltage (100-mV
nonisolated and connected to power ground. The shunts
full-scale input range), and assume that each parameter
might be placed in two or three legs of the three-phase
will not contribute more than ±0.1% to the absolute error
inverter.
over the industrial temperature range. The amplifier input-
offset voltage cannot be ≤100 µV and the offset drift must The amplifier should operate from a single supply, as should
be ≤1 µV/°C. The amplifier’s gain-set resistors, as well as the subsequent ADC. Since the shunt voltage drop is
the shunt, are required to have 0.1% tolerance with a drift of referenced to ground, an input common-mode voltage near
≤10 ppm/°C. Of course, not all drives require this accuracy the ground negative rail is crucial. To decouple from ground
and parameter scale. Unlike the gain error, the offset error is bounces during switching, an amplifier in a differential to
often more critical, as it contributes to an absolute error that single-ended configuration will convert the small bipolar
is independent of the current magnitude, and thus especially shunt voltage into a unipolar voltage, typically 0 V to 3.3 V
impacts inverter performance at lower currents.

Advantages Challenges Accuracy

In-line • Direct, continuous phase current sense. • Amplifier requires high common-mode input voltage with High
motor • Allows averaging phase current over on PWM a high common-mode refection ratio (CMRR).
shunts cycle for higher accuracy and/or a current • AC line-fed inverters typically use an isoated approach.
controller can run twice per PWM.
• Detection of phase-to-phase and phase-to-GND shorts.
Low-side • Can detect shoot-through. • Indirect, discontinuous inverted phase-current sense. Medium
shunts • Lower system cost: due to near zero common mode Can be measured only when low-side switch is on.
input voltage a non-isolated current sense amplifier/op • Cannot average over one PWM period and run the
amp can be used. current controller twice per PWM
• Cannot detect phase-to-GND shorts.
DC-link • Works with both vector control and trapezoidal control. • Requires two measurements synchronized to the PWM Low
single • Lower system cost: due to near zero common pattern in each PWM cycle.
shunt mode input voltage a non-isolated current sense • Requires a minimum duration of each active PWM pattern
amplifier/op amp can be used. and PWM compensation algorithms to ensure a minimum
duration.
• Requires an amplifier with a high bandwidth and slew rate.

Table 1. Comparison of shunt placement for motor phase-current measurement.

An Engineer’s Guide to Industrial Robot Designs 39 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

with 1.65 V mid-bias, to drive the ADC. Key parameters for fixed gain settings like the INA181, INA2181 (dual channel),
the amplifier are: and INA4181 (quad channel) current sense amplifiers.
• A rail-to-rail input with a near-zero input common-mode Placement No. 3: in-line motor shunts
voltage. For 12-V to 60-VDC-fed inverters, non-isolated current-sense
• A rail-to-rail output. amplifiers referenced to GND of DC– are attractive due tof
• A single supply voltage. their system cost. The major challenge is the huge common-
mode voltage, which is 100 to 1,000 times higher than even
• Offset and offset drift, which might not be critical because it the full-scale shunt voltage. This requires amplifiers with:
is possible to measure offset at each PWM cycle when the
low-side switch is off. • A very high DC and AC common-mode rejection ratio
(CMRR) for accurate current measurements without long
• Bandwidth and slew rate, which impact the minimum recovery ripple after transients. The DC CMRR should be
settling time, and which should be smaller than the specific at least -100 dB and the output should settle within a few
minimum on-time for the low-side switch. microseconds. Table 3 outlines the impact of CMRR.
When using three shunts, a workaround for a low-side • A wide common-mode voltage range from at least -1 V
on-time that is too small is to only consider the two phases to 70 V for margin during switching and DC-link voltage
with the highest on-time and calculate the third phase. This increase during motor braking.
approach won’t work for the two-shunt solution, however;
the amplifier has to settle at least within the minimum on-time The amplifier should operate from a single 3.3-V supply, as
specified, typically even at half the minimum on-time since should the subsequent ADC or MCU-embedded ADC. This
the current is often sampled symmetrically to the PWM. eliminates the need for clamping diodes to protect the ADC
input. An amplifier bandwidth at a configured gain of 400 kHz
Table 2 provides example settling times for an amplifier allows overcurrent detection as fast as ≤1 µs (10% to 90%).
with a unity gain bandwidth of 10 MHz, like the TLV9062. It’s not possible to compensate for offset and gain error easily
The TLV9062 meets these specifications and offers dual in this configuration, especially over the operating temperature
amplifiers in a single 8-pin package hence minimizing the range. As outlined before, the offset and offset drift are critical
BOM for systems that use the two low-side leg shunt for the inverter’s low-current performance and the acceptable
approach. offset error depends on the desired current-measurement
To further reduce the BOM, designers can eliminate the need accuracy.
for external gain setting resistors and amplifiers with internal

Minimum slew rate Settling time Settling time

Operational amplifier Gain Bandwidth at 3.3 V to 1% to 3%
10 MHz 20 500 kHz ≥ 3 Vµs 1.5 µs 1.1 µs
10 MHz 50 200 kHz ≥ 1.2 Vµs 3.7 µs 2.7 µs

Table 2. Settling time versus amplifier gain bandwidth.

Accuracy versus
Parameter Value ±50 mW Comment
CMRR (DC) -120 dB 0.1% At 48-V common mode
CMRR (AC) -90 dB 3.3% At 0-V to 48-V common-mode transient, fast
settling required “PWM rejection”

Table 3. Impact of CMRR on accuracy.

An Engineer’s Guide to Industrial Robot Designs 40 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

Figure 3 shows a transient response of a current-sense Electrotechnical Commission (IEC) 61800-5-1 electrical-
amplifier with enhanced PWM rejection (the INA240) with a safety requirements, basic or reinforced insulation is required.
48-V three-phase GaN inverter. Thanks to high DC and AC Basic or reinforced isolated amplifiers and delta-sigma
common-mode rejection, the phase current settles within modulators serve this purpose.
around 2.5 µs. Assuming a center-aligned sampling, the
minimum PWM on or off time required to accurately measure Isolated delta-sigma modulators
the corresponding phase current is 5 µs. For lower on/off Figure 4 (next page) shows the isolated phase-current
times, the three-shunt approach enables the calculation measurement with an in-line shunt and isolated delta-sigma
of the third-phase current from the other two phases with modulator. The measurement can be applied to three phases
higher on/off times. or two phases, with the third phase current calculated
accordingly. The floating shunt voltage is low-pass filtered,
12 to 60 VDC
amplified and fed into a second-order delta-sigma modulator,
which is isolated from the output. The isolated output
is a bitstream of ones and zeros at the modulator clock
3.3 V frequency, typically from 5 MHz to 20 MHz. A decimation
filter in the MCU must process the bitstream to get an
VOUT accurate high-resolution result.
V CM PWM
rejection From a system perspective, isolated delta-sigma modulators
VREF/BIAS
should offer:
INA240
• A gain amplifier with an anti-aliasing filter.
– A ±50-mV input range reduces shunt losses by 80%
Setup: INA240, Gain = 20, Shunt = 5 m Ω, IPhase = 10 A compared to the traditional ±250-mV range.
– Very low gain, offset and related drift are crucial for
(50 V/div)
VCM

0-V to 48-V common-mode voltage

with transients, 20 V/ns, at shunt accuracy, since it is difficult to compensate for them. A
V OUT settles after~25 µs very low 50-µV offset with 1-µV/°C drift contributes to
(500 mV/div)

less than 0.11% error over a temperature range from

VOUT

Large common-mode AC 25°C to 85°C.

transient from 0 V to 48 V at
20 V/ns, attenuated to less
than 150-mV peak
– An integrated anti-aliasing filter attenuates noise above
half the modulator clock frequency to avoid it folding
Impulse noise coupled to scope-probe GND
back and impacting accuracy in the band of interest.
Time (2 µs/div)
– The common-mode input voltage should be at least half
Figure 3. In-line current-sense amplifier and transient response over one the negative full-scale input range.
PWM cycle at 48 V.
• A delta-sigma modulator running a 20-MHz clock to enable
Isolated in-line phase-current sensing highly accurate, highly linear current sensing at low latency.
Modulators with the Manchester-coded bitstream option
For AC line-fed inverters with DC-link voltages from 300 to
enable easier clock routing from the processor to each of
1,200 VDC, an isolated amplifier or delta-sigma modulator
the three modulators.
provides accurate phase-current sensing with in-line shunts.
The isolation function enables rejection of the large common- • A wide-range high-side supply voltage and low current
mode voltage and transients shown in Table 1. Since consumption, preferably with an integrated low-dropout
industrial motor drives are required to meet International regulator such as the AMC1304, to enable use of the
floating-gate drive supply.

An Engineer’s Guide to Industrial Robot Designs 41 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

DC+ Floating gate-driver

300 to supply: +15 V Note: Only one phase/half-bridge shown
1200 VDC
LDO 3 Low-side supply
3.3 V to 5 V Isolated delta-Sigma 3.3 V to 5 V
or
Zener modulator MCU
diode with SINCx
4 Diagnostics decimation
IPHASE 5 filters
6

Reinforced or basic
Driver

isolation barrier
1 1-Bitstream, uncoded or
Floating GND SINCx
∆Σ Manchester coded data
Gain &
AA filter modulator
2 Driver/ Clock
receiver

LVDS or CMOS
interface
DC – Isolated
Floating GND GND

Figure 4. In-line shunt-based phase-current sensing with an isolated delta-sigma modulator.

• A diagnostic function that detects a loss of high-side power The processor’s decimation low-pass filter, such as a
to avoid unpredictable measurements. sinc filter, sets the bandwidth and resolution of the output
• Basic or reinforced isolation, with high immunity against signal by cutting off high-frequency noise. The effective
electromagnetic fields and high common-mode transient number of bits (ENOB) and settling time increase with the
immunity (CMTI) of at least 10 kV/µs to reject switch-node sinc filter order and oversampling ratio; see Figure 5. The
transients. advantage of digital filters is that resolution versus bandwidth
and settling time can be configured in software, and it is
• A complementary metal-oxide semiconductor or low- possible to apply two or more filters to the same bitstream.
voltage differential signaling (LVDS) digital interface option. This advantage enables a high-resolution phase current
In noisy environments or for longer traces, LVDS offers for accurate control (for example, 12 ENOB with a sinc3
higher common-mode noise immunity. filter and 64 times oversampling) and very-fast overcurrent
detection (for example, 1.2 µs with a sinc3 filter and 8 times
oversampling).

Legend:
High-resolution current filter Fast overcurrent detection filter
Sinc 3, OSR = 64, ENOB = 12.5, Settling time = 9.6 µs Sinc 3, OSR = 8, ENOB = 5, Settling time = 1.2 µs

16 16

14 14

12 12 Sinc 3
Sinc 3 Sinc 2
10 10 Sinc 2
ENOB (Bits)
ENOB (Bits)

8 8

6 6
Sinc 1
4 4
Sinc 1
2 2

0 0
1 10 100 1000 0 2 4 6 8 10 12 14 16 18 20
OSR Settling Time (µs)

Figure 5. ENOB vs. oversampling ratio and settling time for a 20-MHz modulator clock (AMC1306).

An Engineer’s Guide to Industrial Robot Designs 42 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

DC+ Floating gate-driver

300 to supply: +15 V Note: Only one phase/half-bridge shown
1200 VDC
LDO Low-side supply
or 3.3 V to 5 V
3.3 to 5 V
Zener MCU
diode Isolated amplifier with 12-bit
ADCs
IPHASE
7
Analog SAR

Reinforced or basic
8
low-pass

isolation barrier
Floating GND ADCs
∆Σ filter
Gain & Differential
AA filter modulator to single-ended
Window
Oscillator comparator
Differential
analog interface

DC – Isolated
Floating GND GND

Figure 6. In-line shunt-based phase-current sensing with an isolated amplifier.

Isolated amplifier a single-ended 12-bit successive approximation register

ADC. Short-circuit detection requires an additional window
Figure 6 shows the phase-current sensing with an isolated
comparator per phase.
amplifier. The nonisolated subsystem of the isolated amplifier
(in red) is the same as the isolated delta-sigma modulator A major system advantage is the simple analog interface to
in Figure 4. The main difference is the inclusion of an output a wide range of MCUs with an embedded ADC. For single-
filter (the subsystem in blue). An active low-pass filter with ended ADCs, you will have an additional operation amplifier
a fixed cut-off frequency removes the high-frequency (op amp) that will not degrade performance. For better noise
quantization noise in the bitstream and provides a high linear immunity, place the op amp close to the MCU to keep the
differential analog output. The features that a delta-sigma analog traces differential as long as possible. From a system
modulator should have listed earlier also apply to isolated performance perspective, the isolated delta-sigma modulator
amplifiers. However, the analog bandwidth and settling system is superior. Table 4 provides a comparison.
time are hardware-fixed and depend on the device-specific
oscillator clock and low-pass filter of the isolated amplifier. Conclusion

The isolated amplifier-based phase-current sensing system A TI reference design is available for each system
has three conversion stages: the isolated amplifier, an configuration described in this article, with detailed hardware
additional differential-to-single-ended amplifier and (typically) design guidelines and system test results.

Parameter Isolated amplifier Isolated delta-sigma modulator

Single analog-to-digital conversion: 16-bit resolution with 14-
Three conversion stages: system resolution also impacted by
Resolution/accuracy bit accuracy (ENOB) possible, pending digital filter configura-
external ADC, typically ≤ 12 bit.
tion. See Figure 5.
Fixed. High-performance amplifiers offer 300-kHz bandwidth
Bandwidth/settling time Flexible. Pending digital filter on processor. See Figure 5.
and less than 3-µs settling time.
Short-circuit detect Requires additional analog hardware (window comparator). No additional hardware required; calculated on the processor.
CMOS or LVDS interface. Requires a higher-performance
Analog differential interface: easy to interface to any MCU with
Interface to processor MCU/microprocessor unit (MPU) with an integrated delta-
an embedded SAR ADC, but requires an additional amplifier.
sigma interface or field-programmable gate array (FPGA).
High to very high, due to digital signals and LVDS interface
EMC immunity Medium, due to analog differential output signals.
option.

Table 4. Comparing an isolated amplifier and an isolated delta-sigma modulator.

An Engineer’s Guide to Industrial Robot Designs 43 2Q 2020 I Texas Instruments

Chapter 2: Robot system controller

2.3.1 Servo drive-related reference designs for

robotic systems

48V/500W Three-Phase Inverter with Smart Gate Driver Efficiency, protection and integration are important design
Reference Design for Servo Drives factors for compact DC-fed drives up to 60 VDC. This
reference design shows a three-phase inverter with a
nominal 48-VDC input and 10-ARMS output current.
Basic Isolated Three-Phase Compact Power Stage This three-phase, compact power stage reference design
Reference Design for Industrial Drives for industrial drives, which uses a UCC5350 gate driver to
support basic capacitive isolation requirements, provides an
increased lifespan and better propagation delay match over
optocouplers to minimize inverter deadband distortions and
losses.
Reinforced Isolated Phase Current Sense Reference This reference design realizes a reinforced isolated three-
Design with Small Delta Sigma Modulators phase inverter subsystem using isolated gate bipolar
transistor (IGBT) gate drivers and isolated current/voltage
sensors. AMC1306E25 delta-sigma modulators perform
precise in-line shunt-based motor phase current sensing.
Three-Phase Inverter Reference Design for 200-480 VAC This reference design realizes a reinforced isolated three-
Drives with Opto-Emulated Input Gate Drivers phase inverter subsystem using isolated IGBT gate drivers
and isolated current/voltage sensors.

Find more reference designs for your robot system

controller.

Contributing authors: Eddie Esparza, Jason Reeder, Martin

Staebler, Nagarajan Sridhar, Mateo Begue, Mamadou Diallo,
Scott Hill, Manny Soltero, Harald Parzuber, Kristen Mogensen
and Martin Staebler.

An Engineer’s Guide to Industrial Robot Designs 44 2Q 2020 I Texas Instruments

Chapter 3: Robot arm and driving system (manipulator)

3.1.1 How to protect battery power VREF

management systems from thermal damage
Nowadays, consumers expect their personal electronics
RBIAS R1
to have a longer battery life, a shorter charge time and a VTRIP
smaller form factor. The increased charge and discharge VTEMP
currents, as well as the smaller form factor, make battery +
packs vulnerable to thermal damage. In addition, different
RTMP61 -
battery technologies have different charging and discharging
requirements that are sensitive to temperature, as shown in R2
Table 1.
Charge Discharge
Battery type temperature temperature
Figure 1. Thermistor + comparator for threshold detection.
Lead acid -20°C to 50°C -20°C to 50°C
Since battery applications require protection at both hot
NiCd, NiMH 0°C to 45°C -20°C to 65°C
and cold temperatures, a temperature window comparator
Li-ion 0°C to 45°C -20°C to 60°C
is a better solution. Figure 2 illustrates an example of this
Table 1. Common charge and discharge temperature limits for various output. In this example, the trip points are set to 60°C and
batteries. 0°C with a 10°C hysteresis.
Batteries are typically discharged over a wider temperature
range, but the charge temperature is limited. It is possible
to perform fast charging safely if the cell temperature is TH = 60°C
50°C
between 10°C and 40°C. These temperature limits are tied
to battery-cell chemistries and temperature-dependent
chemical reactions. If charged too quickly, the cell pressure 10°C
5°C
can build up and lead to venting and reduced battery life. TL = 0°C

If the operating temperature is too high, cell degradation can

occur and result in thermal runaway and explosion. If the
operating temperature is too low, irreversible cell chemical Changes
HYSTSET1
reactions can occur and shorten battery life. Thus, battery hysteresis
from 10°C to 5°C
temperature monitoring is critical for battery management
HYSTSET0
systems.
Forces OUT pin
Thermal protection solutions SOH high regardless
of temperature
Both discrete or integrated temperature-sensing solutions
OUT
can protect battery management systems from thermal
TimeTime
damage. A discrete solution comprises a thermistor,
comparator and voltage reference, as shown in Figure 1.
Figure 2. Example temperature window comparator output behavior.
This approach provides real-time thermal protection without
interrupting the control processing system.

An Engineer’s Guide to Industrial Robot Designs 45 2Q 2020 I Texas Instruments

Chapter 3: Robot arm and driving system (manipulator)

Note that the set output high (SOH) in Figure 2 is a system 3.1.2 Protecting your battery isn’t as hard as
diagnostic test feature that enables you to force the output you think
high, independent of the temperature.
When it comes to any type of protection, the solution should
TI provides a broad portfolio of temperature switches be simple. Protection should be something that you design
and thermistors like the TMP303, TMP390 and TMP61. and set up once and don’t worry about again; at least that’s
The TMP303 uses a window comparator and offers how it should be. But when it comes to more and better
design flexibility through an extra-small footprint (small battery protection, you might worry about what it might cost
outline transistor-563), low power (5 μA maximum) and them going forward.
supply-voltage capability as low as 1.4 V. No additional
Given that battery protection circuitry usually sits inside of
components are required for operation and the window
battery packs out of sight, it isn’t typically considered a cool,
comparator can function independently of microprocessors
sleek new application feature, and you may not give it much
or microcontrollers. Seven trip points are available through
thought. But battery protection can cause major headlines if
different device options, which are programmable at the
not done properly.
factory to any temperature.
For any protection device, you want the setup to be simple:
The TMP390, shown in Figure 3, is a resistor programmable
an integrated circuit that protects your system but does not
dual-output temperature switch with two internal
come with a high current consumption. The TI BQ77905
comparators and two outputs. The TMP390 is offered in the
family of battery protectors for three- to five-series cells and
same small package, has ultra-low power (1 μA maximum)
beyond can help by providing the protection your system
and low supply-voltage capability (1.62 V).
needs with low power drain.
VDD VDDIO In battery applications, you always need to have a primary
Comparators protector that will serve as the first line of defense; any
Outputs
SETA protection after will have the role of secondary protection.
A OUTA Secondary protection is a last-resort type of battery
RA Threshold protection that is usually simple overvoltage protection, like
and
hysteresis BQ7718 family.
SETB B OUTB The BQ77905 shown in Figure 1 is a primary protector for
RB applications such as power tools; garden tools; vacuums;
and robotic applications like drones, robotic vacuums and
Temperature
sensor robotic lawnmowers. These types of industrial consumer
applications take a toll on their battery packs, since
GND consumers want their power tool or robotic vacuum to
perform as if it were AC-powered. Typical continuous current
Figure 3. TMP390 block diagram.
consumptions for these applications can go as high as
Both the hot and cold trip points are configurable to any 50 A (in power tools) and as low (but still high) as 15 A (in
desired temperature window, with hysteresis options vacuums).
between 5°C and 30°C using just two resistors. The separate
In addition to supporting high current draws, the internal
hot and cold trip outputs generate independent warning
battery pack circuitry needs to consume ultra-low power for
signals to be interpreted by the microprocessor.
longer battery lifetimes and overall runtime. That is where the
  BQ77905’s 6 µA of average current consumption comes in
handy.

An Engineer’s Guide to Industrial Robot Designs 46 2Q 2020 I Texas Instruments

Chapter 3: Robot arm and driving system (manipulator)

and even 20S. To accommodate these various sizes, creating

a common platform for battery-pack design eliminates the
engineering costs associated with redesign and unfamiliarity
between different IC architectures. The BQ77905 also offers
stacking capability to provide cell-count flexibility to designs.
If additional features like cell balancing and hibernate mode
are needed, upgrading to the BQ77915 may be a good
choice. Cell balancing helps in high cell count applications
where the application needs to prolong the pack life by
keeping each individual cell appropriately balanced.
In general, protection should be straightforward, easy to
Figure 1. BQ77905 3S to 5S advanced stackable low-power battery use and should not cost much (power consumption, price,
protector EVM. safety). In addition, battery protection devices should provide
flexibility and enable scalable approaches to various cell
Industrial consumer applications typically include battery-
counts, which help manage overall design costs. Protection
pack sizes of 3S (small power tools or drones), 4S (drones),
should never limit what the application can do.
5S (professional power tools), 6S (industrial drones), 7S
(vacuums), 10S (garden tools or larger power tools like saws)

3.1.3 Position feedback-related reference

designs for robotic systems

1 to 4 Series Cell Li-Ion Battery Pack Manager This reference design features the BQ40z50-R1 battery
Reference Design pack manager, with integrated gas gauging and
protection for completely autonomous operation of 1S
to 4S cell lithium-ion and lithium-polymer battery packs.
The architecture enables communication between the
fuel gauging processor and an internal battery charger
controller through SMBus broadcast commands.
High EMC Immunity RS-485 Interface Reference This high electromagnetic compatibility (EMC) immunity
Design for Absolute Encoders reference design demonstrates a RS-485 transceiver
for use with both the drive and with encoders such
as EnDat 2.2, BiSS and Tamagawa. EMC immunity –
particularly immunity against inverter switching noise – is
important for position encoder feedback systems within
industrial drives.

Find more reference designs for your robot system manipulator.

Contributing authors: Miguel Rios and Bryan Tristan

An Engineer’s Guide to Industrial Robot Designs 47 2Q 2020 I Texas Instruments

Chapter 4: Sensing and vision technologies

4.1 TI mmWave radar sensors in robotics sensors, cameras, inertial measurement units (IMUs), GPS
applications and others. One relatively new technology in robotic sensing
is complementary metal-oxide semiconductor (CMOS)
When you conjure up an image of robots, you might envision
millimeter-wave (mmWave) radar sensors. CMOS
massive machine arms with visible coils and wire harnesses
mmWave radar sensors enable the accurate measurement
along a factory floor, with welding sparks flying. These
of not only the distance of objects in their field of view but
machines are very different than robots portrayed in popular
also the relative velocities of any obstacles. These sensing
culture and science fiction, which present a future where
technologies all have advantages and drawbacks, as shown
robots are ubiquitous assistants in everyday living.
in Table 1.
Today, breakthroughs in artificial intelligence technology
One important advantage that mmWave sensors have
are driving the advancement of robotics for service robots,
over vision- and LIDAR-based sensors is their immunity to
unmanned aerial vehicles and autonomous vehicles.
environmental conditions such as rain, dust, smoke, fog or
As robotic technologies advance, so do complementary frost. Additionally, mmWave sensors can work in complete
sensor technologies. Much like the five senses of a human darkness or in the glare of direct sunlight. Mounted directly
being, combining different sensing technologies offers the behind enclosure plastics without external lenses, apertures
best results when deploying robotic systems into changing or sensor surfaces, the sensors are extremely rugged
and uncontrolled environments. and can meet Ingress Protection (IP) 69K standards. TI’s
mmWave sensors are also small, lightweight and produce
Sensor technologies in robotics
designs that are three times smaller and half the weight of
Sensor technologies in robots include force and torque miniature LIDAR range finders.
sensors, touch sensors, 1D/2D infrared (IR) range finders,
3-D time-of-flight light detection and ranging (LIDAR)

Detection Detection Range Detectable Night Detection

Sensors range angle resolution information Bad weather operation performance
mmWave
Narrow and Velocity, range, Robust and
Long Good Good Yes
wide angle stable

Camera Complexity
Target to calculate
Medium Medium Medium Poor No
classification object
coordinates
LIDAR
Narrow and Velocity, range, Poor in bad
Long Good Poor No
wide angle weather

Ultrasonic

Short-range
Short Wide Good Range Good No
applications

Table 1. Sensor technology comparison.

An Engineer’s Guide to Industrial Robot Designs 48 2Q 2020 I Texas Instruments

Chapter 4: Sensing and vision technologies

Detecting glass walls Using the demo software and visualization tools included with
the EVM in the mmWave demo visualizer, the results shown
Figure 1 illustrates the use of glass walls and partitions in
in Figure 3 clearly demonstrate the TI mmWave sensor
modern architecture and service robots that vacuum or mop
detecting the glass wall surface, as well as the wall behind
floors, for example, and need to sense these surfaces to
the glass.
prevent collisions. These elements have proved difficult to
detect using camera- or IR-based sensors. But mmWave
sensors can detect the presence of glass walls as well as
materials behind them.

Figure 3. Test results showing the detection of glass and wall panel.

Using TI mmWave sensors to measure ground

speed
Figure 1. Modern architecture makes extensive use of glass surfaces. Accurate odometry information is essential for the
To demonstrate this capability, we set up a simple experiment autonomous movement of a robot platform. It’s possible
using the IWR1443 single-chip 76-GHz to 81-GHz TI to derive odometry information simply by measuring the
mmWave sensor evaluation module (EVM) with a pane of glass rotation of wheels or belts on a robot platform. This low-cost
positioned 80 cm away. We then placed a wall panel behind approach is easily defeated, however, if the wheels slip on
the glass at a distance of 140 cm, as shown in Figure 2. surfaces such as loose gravel, dirt or wet areas.
More advanced systems can assure very accurate odometry
through the addition of an IMU that’s sometimes augmented
IWR1443 EVM Glass panel Wall panel with GPS. TI mmWave sensors can supply additional
odometer information for robots that traverse over uneven
terrain or have a lot of chassis pitch and yaw by sending
chirp signals toward the ground and measuring the Doppler
shift of the return signal.
80 cm
Figure 4 (next page) shows the potential configuration of a
140 cm ground-speed mmWave radar sensor on a robotics platform.
Whether to point the radar in front of platform (as shown) or
behind the platform (as is standard practice in agriculture
Figure 2. Test setup for detecting glass walls.
vehicles) is an example trade-off. If pointed in front, then you

An Engineer’s Guide to Industrial Robot Designs 49 2Q 2020 I Texas Instruments

Chapter 4: Sensing and vision technologies

can use the same TI mmWave sensor to also sense surface Historically, a safety curtain or keep-out zone has been
edges and avoid an unrecoverable platform loss, such as used around the robot’s field of operation to ensure physical
going off the shipping dock in a warehouse. If pointed behind separation, as shown in Figure 6.
the platform, you can mount the sensor at the platform’s
center of gravity in order to minimize the pitch and yaw effect
on the measurement, which is a large concern in agriculture
applications.

Figure 6. Robotic arm with a physical safety cage.

Sensors make it possible for a virtual safety curtain or
Figure 4. Ground-speed radar configuration on a robotics platform. bubble to separate robotic operation from unplanned human
interaction, and also to avoid robot-to-robot collision as
Equation 1 calculates velocity under uniform ideal density and operation programmability increase. Vision-
conditions. based safety systems require controlled lighting, which
fd = ( 2V / λ ) i cosθ increases energy consumption, generates heat and requires
(1) maintenance. In dusty manufacturing environments such
as textile or carpeting, lenses need frequent cleaning and
Expanding Equation 1 enables you to compensate for attention.
velocity-measurement errors for variables such as uneven
terrain that result in sensor pitch, yaw and roll and introduce Since TI mmWave sensors are robust at detecting objects
a rotational velocity component. regardless of lighting, humidity, smoke and dust on the
factory floor, they are a good fit to replace vision systems,
Safety guards around robotic arms and can provide this detection with very low processing
latency – typically under 2 ms. With a wide field of view and
As robots interact more with humans – either in service
long detection range, mounting these sensors above the
capacities or in flexible, low-quantity batch-processing
area of operation simplifies installation. The ability to detect
automation tasks – it is critical that they do not cause harm
multiple objects or humans with only one mmWave sensor
to the people with whom they interact, as shown in Figure 5.
reduces the number of sensors required and reduces cost.

Point-cloud information generated by mmWave

sensors
TI mmWave radar sensors convert radio-frequency (RF)
front-end analog data to a digital representation through
an analog-to-digital converter. This digitally converted
data requires high-speed external data buses to bring the
data stream to the processing chain, where a series of
mathematical operations generate the range, velocity and
angular information for points detected in the sensor’s field of
Figure 5. Robots of the future will be more interactive with humans. view. Because these systems are traditionally large and

An Engineer’s Guide to Industrial Robot Designs 50 2Q 2020 I Texas Instruments

Chapter 4: Sensing and vision technologies

ADC data
RF front-end ADC

IWR1443
Point cloud
(range, velocity, angle)
Pre-processing 1st dim FFT 2nd dim FFT 3rd dim FFT Detection
(interface mitigation) (range) (velocity) (angle arrival)

IWR1642
Object
Clustering Tracking Object classification

Figure 7. TI mmWave sensors processing chain.

expensive, TI sought to integrate all of this functionality onto Mapping and navigation using TI mmWave
a single monolithic piece of CMOS silicon, thus reducing sensors
size, cost and power consumption. The additional digital
Using the point information for objects detected by the
processing resources now handle data post-processing IWR1443 EVM, it is then possible to demonstrate the
for such tasks as clustering, tracking and classification, as use of TI mmWave radar to accurately map obstacles in a
shown in Figure 7. room, and to use the free space identified for autonomous
A person walking in front of a TI mmWave sensor generates operation. In order to quickly demonstrate the use of a single
multiple reflection points. Each of these detected points mmWave radar in mapping and navigation applications,
can be mapped in a 3D field relative to the sensor (as we chose the Robot OS robotic open-source Turtlebot 2
shown in Figure 8) within the robot operating system (ROS) development platform and mounted the IWR1443 EVM on it,
visualization (RVIZ) tool. This mapping collects all points over as shown in Figure 9.
a quarter-of-a-second time period. The density of the point
information collected provides a good amount of fidelity with
leg and arm movement visible, enabling object classification
as a moving person. The clarity of the open spaces in the 3D
field is also very important data for mobile robots so that they
can operate autonomously.

Figure 9. IWR1443 EVM mounted on a Turtlebot 2.

Implementing a basic driver for the EVM (ti_mmwave_
rospkg), we integrated the point-cloud information into
the navigation stack using the OctoMap and move_base
libraries, as shown in Figure 10 (next page).

Figure 8. Point cloud of a person shown in RVIS, captured with the

IWR1443 EVM.
An Engineer’s Guide to Industrial Robot Designs 51 2Q 2020 I Texas Instruments
Chapter 4: Sensing and vision technologies

“move_base_simple/goal”
geometry_msgs/PoseStamped
Navigation stack setup
move_base “/map” Octomap_server
nav_msgs/GetMap (pre-existing map)
Amci global_planner global_costmap

Sensor “/tf” internal sensor topics

recovery_behaviors ti_mmwave_rospkg
transforms tf/tfMessage nav_msgs/Path sensor_msgs/LaserScan
sensor_msgs/PointCloud
Odometry “odom”
local_planner local_costmap
source nav_msgs/Odometry

“cmd_vel” geometry_msgs/Twist

Provided node
Base controller
Optional provided node
Turtlebot_bringup Plaform specific node

Figure 10. Robot OS library navigation stack used with the IWR1443 EVM-equipped Turtlebot 2.

We placed obstacles in an interior office environment

and drove the Turtlebot 2 through the area to build a 3D
occupancy grid map using the OctoMap library. Figure 11 is
a screen shot of the occupancy grid using RVIZ.

Figure 12. Using the IWR1443 EVM occupancy map for autonomous
navigation of a Turtlebot 2 with the Robot OS move_base library.

Conclusion
TI mmWave sensors were initially expensive and large, and
Figure 11. Generating an occupancy map using the OctoMap library in
required multiple discrete components. Now, however, driven
Robot OS.
by TI’s integration of RF, processing and memory resources
We used the map generated from OctoMap with move_base, onto a single monolithic CMOS die, it is now realistic to
inputting a final destination and pose position as shown by say that mmWave sensors will complement or displace
the green arrow in the screen capture of Figure 12. The established sensing technologies in robotics.
Turtlebot 2 successfully and efficiently navigated to the
Here are the advantages of TI mmWave sensors vs. other
selected spot and then rotated to the appropriate pose,
technologies:
avoiding both static and dynamic objects in its path. This
demonstrated the efficacy of using a single forward-facing • TI mmWave sensors are not sensitive to environmental
mmWave sensor for basic autonomous robotic navigation conditions such as direct sunlight, shadows or light
quickly in the Robot OS environment. reflections off of water.
• TI mmWave technology can detect glass walls, partitions and
furnishings where light-based sensing solutions could fail.

An Engineer’s Guide to Industrial Robot Designs 52 2Q 2020 I Texas Instruments

Chapter 4: Sensing and vision technologies

• TI mmWave sensors provide Doppler velocity information

on objects, and thus can help augment robot odometry
where wheel slippage on wet surfaces is possible.
• TI mmWave sensors are less mechanically complex,
and thus reduce manufacturing alignment and error-
calibration processes. Without apertures or lenses,
they are mountable directly behind enclosure plastics.
Integrated calibration means less in-line manufacturing
complexity. Wide field-of-view possibilities eliminate the
need for mechanical rotating sensor mechanisms.
• Integrated single monolithic CMOS TI mmWave sensors
enable all processing to occur within the sensor. This Figure 1. mmWave sensing helps monitor the area around machinery
lowers BOM costs, makes for a small sensor and reduces for real-time incident management.
the million instructions per second (MIPS) needed from Additionally, having the ability to detect multiple objects
the central controller processor vs. vision-based systems. and process data with only a single sensor reduces overall
TI mmWave sensor technology enhances the intelligent system costs.
operation of robotics while increasing robustness in real- Immunity to environmental conditions such as dust, smoke
world environments. The application of this technology will and variable lighting is another important consideration in
further accelerate the rapid adoption of robotic systems. a factory setting. TI mmWave sensors can operate in any
of these conditions – and be mounted behind enclosure
plastics – without the need for external lenses, apertures
4.2 Intelligence at the edge powers
or sensor surfaces. All of these attributes enable mmWave
autonomous factories
sensors to perform well in industrial sensing applications.
A variety of robots, from traditional industrial robotic
TI mmWave technology enables more than just distance
systems to today’s latest collaborative robots, rely on
measurement
sensors that generate and process large volumes of
highly varied data. This data facilitates real-time decision- Intelligent edge processing allows factory machines and
making in autonomous robots, resulting in smarter incident robots to interact with humans and reduce incidents.
management while maintaining productivity in dynamic real- For instance, TI mmWave sensors can be configured to
world environments, as shown in Figure 1. monitor a specific area of interest around machinery, define
keep-out zones and trigger a warning to people in the area.
How TI mmWave sensors enable intelligence at These zones can be partitioned so that the sensor can
the edge in factories react accordingly based on zone occupancy or proximity of
Texas Instruments mmWave sensors have an integrated a person.
processor that can process data on-chip for real-time
decision-making. This integration enables smaller designs
compared to some light- or vision-based sensors.

An Engineer’s Guide to Industrial Robot Designs 53 2Q 2020 I Texas Instruments

Chapter 4: Sensing and vision technologies

Figure 2. A person walks slowly past a machine and the danger signal flashes at 1 m.

Figure 2 illustrates this feature, with zones marked safe take more predictive action, such as stopping the machine,
(green), warning (yellow) and danger (red) to indicate proximity based on how fast objects are approaching the sensor.
to the machine. Figure 3 shows how quickly the machine triggers a danger-
TI mmWave sensors enable the accurate measurement of not zone warning based on the speed of the person approaching
only the distance of objects in their field of view, but also of the machine.
the relative velocities of any obstacles. This enables robots to

(a) (b)
Figure 3. The danger sign triggers at 1 m when a person is walking slowly (a); and at 2 m when a person is walking fast (b).

To improve productivity, you want the machine to avoid stops

due to false triggers. The example in Figure 4 shows how
integrated tracking algorithms enable the sensor to accurately
determine a person’s direction. When a person is walking
away from the machine, it does not turn on a warning signal
and takes no other action.

Simplify design and speed development

To simplify the design of robotic systems and reduce
development time, the Area scanner using mmWave sensor
with integrated antenna-on-package reference design uses the
IWR6843 EVM, which works on the 60-GHz frequency band
and integrates a complete radar processing chain onto the
Figure 4. The sensor does not indicate a danger signal, since the person
device.
is walking away from the machine.

An Engineer’s Guide to Industrial Robot Designs 54 2Q 2020 I Texas Instruments

Chapter 4: Sensing and vision technologies

4.3 Use ultrasonic sensing for graceful robots would be to use ultrasonic sensors embedded on the sides of
the vacuum to give it full 360-degree coverage. The spacing
We’re not far from the day when robots perform many of
and number of sensors would depend on the shape of the
the tasks done by humans today. We already have robotic
vacuum and the field of view (FOV) of the ultrasonic sensor.
vacuums cleaning our homes and robotic mowers cutting
the grass in our yards. On factory floors, robots are building As the robot vacuum moves around, the ultrasonic sensor
many of the products we use, from toothbrushes to cars. network maps obstructions, calculates the distance from
Robots are serving food in China and Japan, while drones are them and feeds this information to the central processing
fertilizing farms and delivering goods. unit (CPU) to navigate around them. A similar approach with
integrated ultrasonic sensors would work for robot lawn
So it won’t be long until robots will build our homes, lay our
mowers, robot interactive toys, or restaurant or retail service
roads and drive us around. But one of the key requirements
robots, as shown in Figure 1.
for the reality of such a future is for robots to have senses
similar to humans.
One of the biggest challenges with robotic devices is how
they find their way around without crashing into walls,
furniture, equipment, humans or other robots. To avoid
obstructions and do their job effectively, a robot should be
able to detect obstacles from a few feet to a few centimeters
away so that they have time to navigate elsewhere.
Figure 1. Examples of service robots.
Common technologies for detecting obstacles include:
As a second example, consider assembly-line robots and
• Ultrasonic sensing, which transmits ultrasonic waves and
robots that move raw materials or finished goods within and
listens for the echoes that reflect back from any obstacles.
between a factory floor and a warehouse.
• Optical time-of-flight (ToF) sensors, which use a photodiode
In today’s factories, robot arms assemble products by
to capture reflected light waves from obstacles.
moving around to pick up and place parts and install nuts
• Radar sensors, which use radio-frequency waves and the and bolts, as shown in Figure 2. A main concern among
returning echoes from objects to determine the direction factory owners and robotic system manufacturers is installing
and distance of a moving object. sensors on robotic arms to prevent collisions between
The first technology, ultrasonic sensing, is a low-cost, slower- multiple robots on the floor. Ultrasonic sensors installed
speed alternative to radar for robots that don’t need to reach in appropriate locations on robotic arms or mobile robotic
high speeds in homes and factories. Ultrasonic sensing is vehicles can provide intelligence about objects nearby, along
more reliable than optical ToF sensing for obstacle avoidance, with distance information that the CPUs of these robotic
as ultrasonic sensing is not affected by the amount of systems can use to avoid collisions.
available light reflected off of obstacles. Another benefit of
ultrasonic sensing is its ability to sense glass or any other
transparent surface, since it uses sound waves instead of
light to detect objects.

Many applications for robotics

Consider a robot vacuum that either on command or
following a set schedule leaves its base and moves around a
home to clean the floors. A good way to design this system Figure 2. Assembly line robotic system components.

An Engineer’s Guide to Industrial Robot Designs 55 2Q 2020 I Texas Instruments

Chapter 4: Sensing and vision technologies

The components of an ultrasonic-based obstruction • A CPU. This section of a robotic system uses the ToF
avoidance system used in robots would include: information from multiple ultrasonic sensors around the
• Ultrasonic transducers. These are piezoelectric crystals robot to map obstructions and either stop or help navigate
that oscillate to generate ultrasonic sound waves when away from them, depending on its programming.
an AC voltage is applied, and vice versa when the echo Figure 3 is an example of an ultrasonic transceiver module
returns. There are two types of transducers: closed top, that combines an ultrasonic transducer and the TI’s PGA460
where the piezoelectric crystal is hermetically sealed ultrasonic signal processor and driver IC. The design files for
(protecting it from the environment); and open top, where the module are available as a reference.
the crystal is exposed or covered by something similar
to a speaker mesh. Closed-top transducers require a Get started with ultrasonic sensing
higher drive voltage that necessitates an additional system Ultrasonic sensing is a cost-effective, reliable and practical
component: a transformer. solution for home and factory robotic systems. TI offers a few
• A transformer. A single-ended or center-tap transformer different devices and a wide variety of collateral to help you
will generate the large voltage needed to drive closed-top quickly develop a design based on ultrasonic sensing.
transducers.
• An ultrasonic signal processor and transducer driver.
As an example, TI’s PGA460 drives the transformer,
processes the returned electrical signals from the echoes
and calculates the ToF data for each of the relevant echoes
in real time.

Transformer &
transducer pair Closed top
transducer
Central
processing
unit PGA460

(CPU)

Figure 3. Ultrasonic transceiver module example.

An Engineer’s Guide to Industrial Robot Designs 56 2Q 2020 I Texas Instruments

Chapter 4: Sensing and vision technologies

4.4 How sensor data is powering AI in robotics What are some of today’s key trends in robotics and
automation? What technologies will tie AI to the data that
From traditional industrial robotic systems to today’s latest
it needs to be intelligent? And finally, how are TI’s sensors
collaborative robots (cobots), robots rely on sensors that
being used (and fused) into AI systems?
generate increasingly massive volumes of highly varied data.
This data can help build better machine learning (ML) and Pushing AI processing for robotics to the edge
artificial intelligence (AI) models that robots rely on to become
There are two main parts of ML: training and inference,
autonomous, making real-time decisions and navigating in
which can be executed on completely different processing
dynamic real-world environments.
platforms. The training side usually occurs offline on
Industrial robots are typically placed in caged environments, desktops or in the cloud, and entails feeding large data sets
and will stop when a human enters its environment. But into a neural network. Real-time performance or power is
limiting human/robot collaboration prevents the realization of not an issue during this phase. The result of the training
many benefits. Autonomous capabilities would enable their phase is a trained AI system that when deployed can
safe and productive co-existence of humans and robots. perform a specific task, such as inspecting a bottle on an
Sensing and intelligent perception in robotic applications assembly line, counting and tracking people within a room, or
are important, because the effective performance of robotic determining whether a bill is counterfeit.
systems – particularly ML/AI systems – greatly depends on But in order for AI to fulfill its promise in many industries,
the performance of sensors that provide critical data to these the fusion of sensor data that happens during inference (the
systems. Today’s wide range of increasingly sophisticated part that executes the trained ML algorithm) must happen in
and accurate sensors, combined with systems that can fuse (near) real time. So ML and deep-learning models must be on
all of this sensor data together, are enabling robots to have the edge, deploying the inference into an embedded system.
increasingly good perception and awareness. Examples of
Say that a cobot is built to work in close collaboration with
sensing abilities in robotics can be seen in Figure 1.
humans. It relies on data from proximity sensors as well as
vision sensors to ensure that it successfully protects humans
The growth of AI
from harm while supporting them in activities that would be
Robotic automation has been a revolutionary technology in challenging for them. All of this data needs to be processed
the manufacturing sector for some time, yet the integration in real time, but the cloud is not fast enough for the real-time,
of AI into robotics is poised to transform the industry over the low-latency response that the cobot needs. To address this
next few years. bottleneck, today’s advanced AI systems are pushed to the
edge, which in the case of robots means onboard.
Vision
The decentralized AI model
A decentralized AI model relies on highly integrated
processors that have:
Hearing • A rich peripheral set for interfacing to various sensors.
• High-performance processing capability to run machine-
vision algorithms.
Vibration
• A way to accelerate deep learning inference.
Temperature/ All of these capabilities also have to work efficiently and with
humidity
a relatively low-power and small-size footprint in order to
Proximity/
touch exist at the edge.

Figure 1. The different senses of a robot.

An Engineer’s Guide to Industrial Robot Designs 57 2Q 2020 I Texas Instruments

Chapter 4: Sensing and vision technologies

Power- and size-optimized “inference engines” are As cobots become more capable in demanding industrial
increasingly available as ML grows in popularity. These environments, manufacturers will increasingly add them to
engines are specialized hardware offerings aimed specifically the factory floor, particularly those manufacturers with strict
at performing ML inference. return-on-investment objectives and a desire to improve
An integrated SoC is often a good choice in the embedded product cycle times.
space, because in addition to housing various processing
Logistics robots
elements capable of running deep learning inference, an
SoC also integrates many components necessary to cover Logistics robots are mobile units that operate in
the entire embedded application. Some integrated SoCs environments where people may or may not be present, such
include display, graphics, video acceleration and industrial as warehouses, distribution centers, ports or campuses.
networking capabilities, enabling a single-chip solution that Logistics robots fetch goods and bring them to a packing
does more than just run ML/AI. station, or transport goods from one building of a company
site to another; some are capable of picking and packing
Sitara™ AM57x processors are good examples of goods as well. These robots typically move within a particular
processors running AI at the edge. These processors have environment and need sensors for localization, mapping and
multiple high-speed peripherals for interfacing to multiple to prevent collisions (especially with humans).
sensors – like video, ToF, LIDAR and mmWave sensors – as
well as dedicated hardware in the form of C66x digital signal Until recently, most logistics robots used pre-defined routes;
processor cores and embedded vision engine subsystems to they are now capable of adjusting their navigation based
accelerate AI algorithms and deep learning inference. on the location of other robots, humans and packages.
Ultrasonic, infrared and LIDAR sensing are all enabling
Let’s look at some of the top robotic trends today. technologies. Because of the robot’s mobility, the control
unit is located inside, often with wireless communication to
Cobots
a central remote control. Logistics robots are now adopting
Humans can’t generally get near traditional industrial robots advanced technologies such as ML logic, human-machine
while they are operating without placing themselves in peril. collaboration and environmental analysis technologies.
Cobots are, in contrast, designed to operate safely alongside
humans as shown in Figure 2, moving slowly and gracefully.
As defined by International Organization for Standardization
TS 15066, a cobot is a robot capable of being used in a
collaborative operations where robots and humans work
concurrently within a defined workspace for production
operation (this excludes robot-plus-robot systems or co-
located humans and robots, which are operating at different
times). Defining and deploying cobots to foresee potential
collisions between physical portions of the robot (or virtual
extensions like lasers) and the operators makes the use of
sensors to determine the exact position and velocity of the
operator more important. Figure 2. Cobots working alongside humans in a factory environment.
Cobot makers must implement a high level of environmental Rising labor costs and stringent government regulations are
sensing and redundancy into robot systems to quickly contributing to the higher adoption of logistics robots. Their
detect and prevent possible collisions. Integrated sensors popularity is also rising because of a decrease in the cost of
connected to a control unit will sense an impending collision equipment, components like sensors, and the cost of (and
between a robot arm and a human or other object, and the time required) for integration.
control unit will turn the robot off immediately. If any sensor or
its electronic circuit fails, the robot also turns off.

An Engineer’s Guide to Industrial Robot Designs 58 2Q 2020 I Texas Instruments

Chapter 4: Sensing and vision technologies

Last-mile delivery robots Temperature and humidity sensors

In a product’s journey from warehouse shelf to customer Many robots need to measure the temperature and
doorstep, the last mile of delivery is the final step of the sometimes the humidity of both their environment and their
process: the point at which the package finally arrives components – including motors and main AI motherboards
at the buyer’s door. In addition to being key to customer – to ensure that they are operating in safe ranges. This is
satisfaction, last-mile delivery is expensive and time- especially important for robots, because when a motor is
consuming. under a heavy load, it can draw a lot of power and heat up.
Last-mile delivery costs represent a substantial percentage of Accurate temperature monitoring protects motors, while
the total shipping cost: 53% overall. As such, making the last better temperature accuracy enables motors to be driven
mile of delivery more efficient has become a focus for where harder before hitting safety margin limits. In addition, just
to develop and implement new robotics technologies that about every other sensor is sensitive to temperature and
can drive process improvements and increased efficiency. benefits from thermal compensation. By knowing the
temperature, you can correct for the temperature drift of
ToF optical sensors other sensors to get more accurate measurements.
These sensors rely on the principle of ToF and use a In equatorial factories and in tropical climate zones,
photodiode (a single sensor element or an array) along with temperature and humidity sensors can predict dew points for
active illumination to measure distance. The reflected light electronic system protections and predictive maintenance.
waves from obstacles are compared with the transmitted
wave to measure the delay, which in turn is a representation Ultrasonic sensors
of distance. This data then helps create a 3D map of the Vision sensors may not work if the robot is blinded by a
object. bright light or finds itself in a very dark environment. By
TI’s ToF chipsets enable ToF-based sensing that goes transmitting ultrasonic waves and listening for echoes that
beyond proximity detection to enable next-generation reflect back from objects (similar to how bats maneuver),
machine vision. The chipsets allow for maximum flexibility to ultrasonic sensors perform excellently in dark or bright
customize designs for robot vision and other applications, conditions, overcoming the limitations of optical sensors.
with tools that include an evaluation module and a highly Ultrasonic sensing is a low-cost, slower-speed alternative
configurable camera development kit, which provides a to radar for robots that don’t need to reach high speeds.
3D location of each pixel for accurate depth maps that aid Ultrasonic sensing is more reliable than optical ToF for
customization. Discrete solutions leverage topologies and obstacle avoidance, as ultrasonic sensing is not affected by
semiconductor technologies such as time-to-digital converter the amount of available light reflected off of obstacles. For
and GaN, as demonstrated in the LIDAR Pulsed Time
example, ultrasonic sensing provides the ability to sense
of Flight Reference Design and the Nanosecond Laser
glass or other transparent surfaces because it uses sound
Driver Reference Design for LIDAR.
waves rather than light to detect objects.
A 3D ToF sensor like TI’s OPT8320 enables robots to
determine the exact angle of a screw and then fine-tune the Vibration sensors
screwdriver so that screws consistently align without human Industrial vibration sensing is a crucial part of the condition
intervention. A ToF-based analog front end like the OPT3101 monitoring necessary for predictive maintenance. Integrated
can help identify the distance of a robotic arm to a target and electronic piezoelectric sensors are the most common
help in accurate positioning. vibration sensor used in industrial environments.
For higher-resolution 3D sensing, flexible structured lighting Vibration sensors enable robots to be aware if some of its
– as enabled with DLP® technology and demonstrated in the mechanics are damaged or aged, facilitating preventive
3D Machine Vision Reference Design Based on AM572x maintenance before operations become endangered. Using
Processor with DLP Structured Light – can help bring AI/ML can take the accuracy of these predictions to the next
resolutions to micrometers or below. level.

An Engineer’s Guide to Industrial Robot Designs 59 2Q 2020 I Texas Instruments

Chapter 4: Sensing and vision technologies

mmWave sensors
mmWave sensors use radio waves (Figure 3) and their
echoes to determine the direction and distance of a moving
object by measuring three components: velocity, angle and
range. This enables robots to take more predictive actions
based on how fast objects are approaching the sensor.
Radar sensors deliver excellent performance in the dark and
can sense through materials like drywall, plastic and glass.
As stated in the white paper “TI mmWave radar sensors in
robotic applications,” CMOS mmWave radar sensors enable
the highly accurate measurement of not only the distance of
objects in their field of view, but also the relative velocities of Figure 3. Robotic arm using mmWave sensors.
any obstacles.
Complementary metal-oxide semiconductor (CMOS) TI solutions for the entire AI robotics signal chain
mmWave radar sensors enable the highly accurate The signal chain for adaptive, self-learning AI robotic systems
measurement of not only the distance of objects in their field requires a fusion of diverse sensor data in real time. The
of view, but also the relative velocities of any obstacles. sensors of a cobot are in some ways like the five human
TI’s highly integrated single-chip mmWave radar sensors senses, and all of our senses are critical for fully autonomous
are small, lightweight and enable real-time processing to operation. Each of our senses use different parts of the brain
occur within the sensor edge, often removing the need for and different amounts of the brain’s processing. For example,
additional processors. vision requires more brain power than hearing or smelling.

mmWave technology enables designs that are three times Analogously, robots will have more and more sensors
smaller and half the weight of miniature LIDAR range finders. connected to AI and ML systems that run inside the robot,
This lowers BOM costs, makes for a small sensor, and with the key challenge for AI robotic system manufacturers
reduces the million instructions per second needed from the being the need to address multiple AI systems running and
central controller processor versus vision-based systems. communicating together for a hybrid ML system driven by
Mounted directly behind enclosure plastics without external data from hybrid sensors.
lenses, apertures or sensor surfaces, the sensors are Robot developers depend on advanced IC solutions to
extremely rugged and can meet the Ingress Protection 69K minimize the headaches of circuit design and certification,
standard. speeding the development of products that they can deliver
mmWave sensors were initially expensive and large, and to industrial customers quickly. The ICs that enable advances
required multiple discrete components. Now, however, in industrial robots must provide precise sensing, high-speed
driven by TI’s integration of radio-frequency, processing and sensor signal conversion, fast computation/signal processing
memory resources onto a single monolithic CMOS die, it is for real-time response and high-speed communications.
reasonable to say that mmWave sensors will complement or ICs also enable high-efficiency and small-form-factor power
displace established sensing technologies in robotics over supplies in conjunction with advanced semiconductors like
the coming years. GaN FETs. New ICs also bring new standards to the industry,
such as single twisted-pair Ethernet and power over single-
More advanced radar sensing systems can ensure very
twisted pair, which reduces cabling complexity and enhances
accurate odometry through the addition of an inertial
reliability.
measurement unit that’s sometimes augmented with
GPS. mmWave sensors can supply additional odometer
information for robots that traverse over uneven terrain or
have a lot of chassis pitch and yaw by sending chirp signals
toward the ground and measuring the Doppler shift of the
return signal.

An Engineer’s Guide to Industrial Robot Designs 60 2Q 2020 I Texas Instruments

Chapter 4: Sensing and vision technologies

TI can provide everything from the sensors to the processors Training and inference
needed for next-generation robotics, with a broad product
The two main pieces of deep learning, training and inference,
and solutions portfolio that spans the entire AI robotics
can be executed on completely different processing
signal chain. From sensor input to actuator or motor output,
platforms, as shown in Figure 1. The training side of deep
from individual equipment units to factory-level control and
learning usually occurs offline on desktops or in the cloud
beyond, TI solutions handle the signal chain as well as the
and entails feeding large labeled data sets into a deep neural
processing and power required for robotic applications.
network (DNN). Real-time performance or power is not an
Products include features like reinforced isolation and are
issue during this phase.
tested and qualified for use in harsh industrial environments.
As stated in the white paper, How sensor data is powering
AI in robotics, the result of the training phase is a trained
4.5 Bringing machine learning to embedded neural network that when deployed can perform a specific
systems task, such as inspecting a bottle on an assembly line,
It is hard to understate the promise of machine learning, the counting and tracking people within a room, or determining
latest evolution of which, deep learning, has been called a whether a bill is counterfeit.
foundational technology that will impact the world to the The deployment of the trained neural network on a device
same degree as the internet, or the transistor before that. that executes the algorithm is known as the inference.
Brought on by great advancements in computing power Given the constraints imposed by an embedded system,
and the availability of enormous labeled data sets, deep the neural network will often be trained on a different
learning has already brought major improvements to image processing platform than the one running the inference. For
classification, virtual assistants and game playing, and will the purposes of this discussion, the terms deep learning and
likely do the same for countless industries. Compared to machine learning refer to the inference.
traditional machine learning, deep learning can provide
improved accuracy, greater versatility and better utilization of Machine learning at the edge
big data – all with less required domain expertise. The concept of pushing computing closer to where sensors
In order for machine learning to fulfill its promise in many gather data is a central point of modern embedded systems
industries, it is necessary to deploy the inference (the part – that is, the edge of the network. With deep learning, this
that executes the trained machine learning algorithm) into an concept becomes even more important to enable intelligence
embedded system. This deployment has its own unique set and autonomy at the edge.
of challenges and requirements.

Large data sets Trained network End product

Training Format conversion Inference – deployment

(PC/GPU) (if necessary) (embedded processor)

Figure 1. Traditional deep learning development flow.

An Engineer’s Guide to Industrial Robot Designs 61 2Q 2020 I Texas Instruments

Chapter 4: Sensing and vision technologies

For many applications – from automated machinery and For example, an entire input image at high frames per
industrial robots on a factory floor, to self-guided vacuums second (fps) can run classical computer vision algorithms
in the home, to an agricultural tractor in the field – the to perform object tracking with deep learning used on
processing must happen locally. identified sub-regions of the image at a lower fps for object
The reasons for local processing can be quite varied classification. In this example, the classification of objects
depending on the application. Here are just a few of the across multiple subregions may require multiple instances
concerns driving the need for local processing: of inference, or possibly even different inferences running
on each sub-region. In the latter case, you must choose a
• Reliability. Relying on an internet connection is often not a processing solution that can run both traditional computer
viable option. vision and deep learning, as well as multiple instances
• Low latency. Many applications need an immediate of different deep learning inferences. Figure 2 shows an
response. An application may not be able to tolerate a time example usage of tracking multiple objects through sub-
delay in sending data somewhere else for processing. regions of an image and performing classification on each
• Privacy. The data may be private, and therefore should not object being tracked.
be transmitted or stored externally. • Choose the right performance point. Once you
• Bandwidth. Network bandwidth efficiency is often a key have a sense of the scope of the entire application, it
concern. Connecting to a server for every use case is not becomes important to understand how much processing
sustainable. performance is necessary to satisfy the application needs.
This can be difficult to understand when it comes to
• Power. Power is always a priority for embedded systems.
machine learning because so much of the performance
Moving data consumes power. The further the data needs
is application-specific. For example, the performance of
to travel, the more energy needed.
a convolutional neural net (CNN) that classifies objects
on a video stream depends on what layers are used
Choosing an embedded processor for machine
in the network, how deep the network is, the video’s
learning
resolution, the fps requirement and how many bits are
Many of the concerns requiring local processing overlap with used for the network weights – to name just a few. In an
those inherent in embedded systems, particularly power embedded system, however, it is important to try and get
and reliability. Embedded systems also have several other a measure of the performance needed because throwing
factors to consider that are related to or caused by the too powerful a processor at the problem generally comes
system’s physical limitations. There are frequently inflexible at a trade-off against increased power, size and/or cost.
requirements regarding size, memory, power, temperature, Although a processor may be capable of 30fps at 1080p of
longevity and, of course, cost. ResNet-10, a popular neural net model used in high power,
In the midst of balancing all of the requirements and centralized deep learning applications, it’s likely overkill
concerns for a given embedded application, there are a few for an application that will run a more embedded-friendly
important factors to consider when choosing a processor to network on a 244 x 244 region of interest.
execute machine learning inference for the edge. • Think embedded. Selecting the right network is just
• Consider the entire application. One of the first things as important as selecting the right processor. Not every
to understand before selecting a processing solution neural net architecture will fit on an embedded processor.
is the scope of the entire application. Will running the Limiting models to those with fewer operations will help
inference be the only processing required or will there be a achieve real-time performance. You should prioritize
combination of traditional machine vision with the addition benchmarks of an embedded-friendly network, one that
of deep learning inference? It can often be more efficient will tradeoff accuracy for significant computational savings,
for a system to run a traditional computer vision algorithm instead of more well-known networks like AlexNet and
at a high level and then run deep learning when needed. GoogleNet, which were not designed for the embedded

An Engineer’s Guide to Industrial Robot Designs 62 2Q 2020 I Texas Instruments

Chapter 4: Sensing and vision technologies

space. Similarly, look for processors capable of efficiently Additionally, support for open ecosystems like the Open
leveraging the tools that bring these networks into the Neural Network eXchange will support an even larger base of
embedded space. For example, neural networks can frameworks for development.
tolerate lots of errors; using quantization is a good way to There are many different types of processors to consider
reduce performance requirements with minimal decreases when choosing one for deep learning, and they all have
in accuracy. Processors that can support dynamic their strengths and weaknesses. Graphics processing units
quantization and efficiently leverage other tricks like sparsity (GPUs) are usually the first consideration because they are
(limiting the number of non-zero weights) are good choices widely used during network training. Although extremely
in the embedded space. capable, GPUs have had trouble gaining traction in the
• Ensure ease of use. Ease of use refers to both ease of embedded space given the power, size and cost constraints
development and ease of evaluation. As mentioned earlier, often found in embedded applications.
right-sizing the processor performance is an important Power- and size-optimized “inference engines” are
design consideration. The best way to do this correctly is increasingly available as deep learning grows in popularity.
to run the chosen network on an existing processor. Some These engines are specialized hardware offerings aimed
offerings provide tools that, given a network topology, will specifically at performing deep learning inference. Some
show achievable performance and accuracy on a given engines are optimized to the point of using 1-bit weights and
processor, thus enabling a performance evaluation without can perform simple functions like key phrase detection, but
the need for actual hardware or finalization of a network. optimizing this much to save power and compute processing
For development, being able to easily import a trained comes with a trade-off of limited system functionality and
network model from popular frameworks like Caffe or precision. The smaller inference engines may not be powerful
TensorFlow Lite is a must. enough if the application needs to classify objects or perform
fine-grain work.
When evaluating these engines, make sure that they are
right-sized for the application. A limitation on these inference
engines comes when the application needs additional
processing aside from the deep learning inference. More
often than not, the engine will need to exist alongside another
processor in the system, functioning as a deep learning co-
processor.
An integrated SoC is often a good choice in the embedded
space, because in addition to housing various processing
elements capable of running the deep learning inference, an
SoC also integrates many components necessary to cover
the entire embedded application. Some integrated SoCs
Figure 2. Example of object classification using embedded deep learning.
include display, graphics, video acceleration and industrial
networking capabilities, enabling a single-chip solution that
does more than just run deep learning.

An Engineer’s Guide to Industrial Robot Designs 63 2Q 2020 I Texas Instruments

Chapter 4: Sensing and vision technologies

An example of a highly integrated SoC for deep learning set; an industrial communications subsystem (ICSS) for
is the AM5729 device from TI, shown in Figure 3. The implementation of factory floor protocols such as EtherCAT;
AM5729 has two Arm® Cortex®-A15 cores for system and acceleration for video encoding/decoding and 3D and
processing, two C66x digital signal processor (DSP) cores 2D graphics, facilitating the use of this SoC in an embedded
for running traditional machine vision algorithms and four space that also performs deep learning.
embedded vision engines (EVEs) for running the inference. Choosing a processor for an embedded application is often
TI’s deep learning (TIDL) software offering includes the TIDL the most critical component selection for a product, and this
library, which runs on either C66x DSP cores or the EVEs, is true for many industry-changing products that will bring
enabling multiple inferences to run simultaneously on the machine learning to the edge.
device. Additionally, the AM5729 provides a rich peripheral

High speed interconnect 28 nm

Video acceleration Industrial communication

+ - * = IVA HD 1080p video,VPE substystem
Cortex®®=A15 Cortex®=A15 C66x DSP 4x Industrial Ethernet
+ - * =
ARM Graphics acceleration interfaces
Cortex®-A15 ARM L1 32K/32K L1
C66x DSP (2 PRU-ICSS Subsystems)
32K/32K L1 32K/32K 288KB L2 3D GPU BB2D
Cortex-M4 32-K/32-K L1 Supported industrial
32-K/32-K L1 2x SGX544 GC320 Ethernet protocols:
288-KB L2
TSN, Ethernet MAC, EtherCAT,
Display subsystem PROFINET, EtherNet/IP™,
2-MB L2 32-KB L1 PROFIBUS, POWERLINK,
64-KB RAM 3 LCD HDMI 1.4a
SERCOS3, HSR, PRP, and more...
1080p Blend/scale/convert
2.5-MB L3 Shared RAM w/ECC
Feedback interfaces
32-Bit DDR3/3L w/ECC 32-Bit DDR3/3L 2 x EVE (deep learning accelerator) Sinc3 sigma-delta filter,
EnDat 2.2, TMT
2 Video input ports 2x24b, 2x8b
Security acceleration

System services
EDMA 13 mailbox SDMA Secure WDT WDT Spinlock 16 timer KBD

Serial I/O Industrail and programmable I/O Storage I/O

1 8 2
DCAN + 1 CAN-FD McASP PCIe GPIO
GbE 2-port
switch 3 SD/SDIO
5 3 with 1588
QSPI SATA
IC
2
PWM/CAP/QEP G/MII, 1 eMMC/
RMII, SD/SDIO
10 4 RGMII*** NAND/
UART McSPI USB3/2 USB2
NOR

Figure 3. Block diagram of the Sitara™ AM5729 SoC.

An Engineer’s Guide to Industrial Robot Designs 64 2Q 2020 I Texas Instruments

Chapter 4: Sensing and vision technologies

4.6 Robots get wheels to address new • While the bigger industrial robots such as robotic arms
challenges and functions have been in used for many years, mostly in the automotive
industry, the smaller variants known as collaborative
With logistics centers multiplying to keep up with the growth
robots (cobots) are making a marked entrance as the
of online shopping, so are the numbers of wheeled robots,
opportunities of collaboration between human and robots
which handle many of the strenuous tasks in those logistics
increases.
centers. The next challenge for robots on wheels will be to
address the last mile of delivery to help reduce congestion in • Logistics robots are found in warehouses, distribution
urban areas. centers, ports or even campuses. These robots fetch
goods and bring them to a packing station, or transport
At the same time, human-friendly robots are starting to
goods from one building to another. These robots move
operate in brick and mortar shops to conduct real-time
within a particular environment and need a lot of sensors
inventory, which allows supermarkets to reduce shelf space
for localization and mapping, as well as sensors to prevent
per product and increase the variety of inventory. Wheeled
collisions.
robots are even making an entrance into hotels to provide
hospitality services ranging from check-in to room services. • Hospitality robots are found in supermarkets, airports and
hotels. These robots are virtual presence, welcoming and
It’s not all serious work for these wheeled robots; they will
guide customers/guests (Figure 2).
soon be bringing pizza (as shown in Figure 1) or coffee
to workplaces and campus dormitories. With the trend • Inventory robots are found in supermarkets or inventory
of restaurants moving to full-stack food delivery services, rooms. These robots scan the shelves regularly basis to
where one company runs all of the customer interaction ensure that the store never runs out of a product.
management, the cooking and the logistics, wheeled robots
will be in demand to deliver food sooner than later.

Figure 2. Example of a hospitality robot.

As mentioned in the sensor data powering AI in robotics

Figure 1. Example of a wheeled robot delivering pizza.
white paper, industrial robots are typically placed in caged
With the entrance of wheeled robots which some people environments, and will stop when a human enters its
consider to be robots in disguise, let’s look at the whole environment. But limiting human/robot collaboration prevents
portfolio of robots now in use in the industry: the realization of many benefits. Autonomous capabilities
• Industrial robots (or robot arms) handle tasks in factories would enable the safe and productive co-existence of
like welding, palletizing, sorting and lifting. They are typically humans and robots.
in a fixed position on the floor, ceiling or wall. They are
managed by a controller placed in the base or near the
robot arm.

An Engineer’s Guide to Industrial Robot Designs 65 2Q 2020 I Texas Instruments

Chapter 4: Sensing and vision technologies

Sensing and intelligent perception in robotic applications An application where sensing technology is especially
are important, because the effective performance of robotic important is in the operation of vacuum robots. A ToF
systems – particularly machine learning applications – sensor allows the robot to map an operational environment
greatly depends on the performance of sensors that provide accurately and ensures that the robot will accomplish its
critical data to these systems. Today’s wide range of task efficiently. Infrared “cliff” sensors can protect robots
increasingly sophisticated and accurate sensors, combined from falling down a flight of stairs or down steep drops.
with systems that can fuse all of this sensor data together, Just as humans rely on senses and intelligence to
are enabling robots to have increasingly good perception accomplish tasks, a plethora of technologies are required to
and awareness. Examples of these types of sensors include mimic what humans take for granted in robotics. TI provides
cameras, LIDAR, mmWave and ToF. solutions that address many of these technology needs,
including sensing, intelligence and power.

4.7 Vision and sensing-technology reference

designs for robotic systems

Area Scanner Using TI mmWave Sensor With This design leverages TI’s single-chip mmWave technology
Integrated Antenna-on-Package Reference Design to implement an area scanner capable of detection and
localization in 3D space.
Autonomous Robot Reference Design Using ROS on This design showcases autonomous robotics with the Linux
Sitara™ MPU and Antenna-on-Package mmWave processor software development kit (SDK) running on the
Sitara AM57x processor, and the mmWave SDK running on
Sensors
the IWR6843 evaluation module. The design demonstrates
the functionality of an embedded robotic system, where the
Sitara AM57x processor running the robot operating system
(ROS) processes point-cloud data from mmWave radar
sensing and is the main processor for the overall system
control.
Ultrasonic Distance Sensor with IO-link Reference This design features an ultrasonic distance sensor that fits in
Design a M12 housing due to its high integration and an optimized
layout. The design offers an IO-Link interface to communicate
with the system control, which makes it Industry 4.0 ready.
LIDAR Pulsed Time of Flight Reference Design LIDAR systems use the time taken by the light to fly back
and forth to an object in an effort to measure the distance
to this target. This design shows how to design the time-
measurement back end for LIDAR based on a time-to-digital
converter (TDC), as well as an associated front end.

Find more reference designs for your robot system.

Contributing authors: Dennis Barrett, Adrian Alvarez, Prajakta Desai, Ram Sathappan, Matthieu Chevrier, Mark Nadeski and
Lali Jayatilleke.

An Engineer’s Guide to Industrial Robot Designs 66 2Q 2020 I Texas Instruments

Chapter 5: Technologies for robotic tools (end effector)

5.1 Achieving accuracy in bin picking with TI DLP® technology provides a high-speed pattern projection
DLP® technology-powered structured light capability through a micromirror matrix assembled on top
systems of a semiconductor chip known as a digital micromirror
device (DMD), as depicted in Figure 2. Each pixel on a
An industrial environment handles parts of different shapes,
DMD represents one pixel in the projected image, enabling
sizes, materials, and optical properties (reflectance,
accurate pixel image projection. The micromirrors transition
absorption) every day. These parts have to be picked
at ~3 µs to reflect the light incident on the object through a
and placed in a specific orientation for processing, and
projection lens or onto a light dump. The former achieves a
the automation of these pick-and-place activities from an
bright pixel on a projected scene, while the later creates a
environment into containers is known as bin picking.
dark pixel.
This task poses a challenge for a robot end effectors –
devices attached to the end of a robotic arm – to know the
exact 3D location, dimensions and orientation of objects
it wants to grip. In order to navigate around the walls of a
box and as well as other objects that might also be inside,
the robot’s machine vision system needs to acquire depth
information in addition to 2D camera information.
A structured light technique can address challenges related
to the 3D image capture of objects for bin picking. A
structured light technology-based 3D scanner and camera
works by projecting a series of patterns onto the object being
scanned and captures the pattern distortion with a camera or Figure 2. DLP chips contain millions of micromirrors that are
sensor. A triangulation algorithm then calculates the data and individually controlled at high speeds and deliberately reflect light to
outputs a 3D point cloud. Image processing software like create projected patterns.
Halcon from MVTech calculates the object’s position and the DLP technology also offers unique advantages with its
robot arm’s optimal approach path (Figure 1). ability to project patterns over a wide wavelength range
(420 nm to 2500 nm) using various light sources like lamps,
LEDs and lasers.
For bin picking, DLP technology-powered structured light
offers these advantages:
• Robustness against environmental lighting. Lighting
conditions in factories (like low light exposure and high
contrast between differently lit areas) or flickering light
(which can interfere with machine vision systems) can be
a challenge in applications that need machine vision for
bin picking. DLP technology-powered structured light has
inherent active lighting, which makes it robust against those
conditions.
• No moving parts. Structured light systems capture the
Figure 1. Example of matching pipe couplings with their respective 3D whole scene at once, avoiding the need to either sweep a
model (Source: Halcon by MVTech). light beam over objects or move objects through a beam
(like in scanning solutions). A structured light system
doesn’t use moving parts at a macroscopic scale, making it
immune to wear and tear through mechanical deterioration.

An Engineer’s Guide to Industrial Robot Designs 67 2Q 2020 I Texas Instruments

Chapter 5: Technologies for robotic tools (end effector)

• Real-time 3D image acquisition. The micromirrors in DLP

chips are controlled at high speeds and offer custom pattern
projection at up to 32 kHz. DLP controllers provide trigger
output and inputs for use in synchronizing cameras and
other devices with the projected pattern sequence. These
features help enable real-time 3D image acquisition and thus
simultaneous scanning and picking.
• High contrast and resolution of projected patterns.
Because each micromirror either reflects light onto the
target or an absorption surface, it is possible to achieve Figure 3. Structured light setup including the DLP® Products
high contrast ratios and enable accurate point detection LightCrafter™ 4500 (left), a Point Grey forward-looking infrared Flea3
independent of the object’s surface properties. The camera (right) and calibration (back).
availability of high-resolution DLP chips with up to 2560-by- Recalibration is only necessary if the camera is moved
1600-resolution mirrors means that objects down to the relative to the DLP products board.
micron level are detectable.
After the setup is complete, it’s possible to create point
• Adaptable to object parameters. Programmable patterns clouds of real-world targets. These clouds are output by
and various point-coding schemes like phase shifting or the software in an arbitrary file format, which is then read
gray coding make structured light systems more adaptable and displayed by brief code in Halcon’s HDevelop platform.
to object parameters than systems using diffractive optical Figure 4 shows a point cloud with color coding for depth
elements. information taken of a box filled with coffee mugs.
• Accelerated development time. While robots offer
high repeatability, bin picking requires accuracy in an
unstructured environment, where objects keep shifting
position and orientation each time one is removed from
a storage bin. Working successfully with this challenge
requires a reliable process flow, from machine vision, to
computing software, to the robot’s dexterity and gripper.
Making everything work in conjunction can be a challenge
that consumes a lot of development time.
Figure 4. Captured mug (left) and acquired point cloud acquired with
TI’s evaluation modules for DLP technology enable fast
DLP® powered structured light of several mugs in a box displayed in
implantation of structured light into the machine vision
Halcon HDevelop (right).
workflow. To demonstrate this capability, TI mounted a DLP
LightCrafter™ 4500 evaluation board at a set distance and Halcon’s surface matching can determine the mug’s 3D
angle to a monochromatic camera. The DLP evaluation board pose by comparing the point cloud with a 3D computer-
is triggered by the camera through a trigger cable connecting aided design model of a mug. The robot arm now “sees” the
the two to each other; see Figure 3. object, making it possible to calculate the optimal approach
path of the robot arm so that it can pick objects from boxes
Both the board and camera are connected to a PC via
and avoiding obstacles in an unstructured and changing
USB, and the whole setup is directed at a calibration board.
environment.
Software from the Accurate Point Cloud Generation for
3D Machine Vision Applications Using DLP Technology Authored by: Franz Schellhase
reference design calibrates the camera and projector for
parameters like focal length, focal point, lens distortion,
translation and rotation of the camera relative to the
calibration board. The reference design user guide walks
through the process step by step.

An Engineer’s Guide to Industrial Robot Designs 68 2Q 2020 I Texas Instruments

Chapter 6: Industrial communications for robotics

6.1 Selecting the right industrial Industrial communication by use case

communications standard for sensors
Vision cameras
Greater factory connectivity and control are ushering in the 3D TOF
fourth industrial revolution, known as Industry 4.0, after the Gripper CNC

Performance
earlier revolutions of steam power, assembly lines and early Vibration monitoring
automation. Motor drives
Proximity Test and
This movement advances machine-to-machine
measurement
communication with an exponential growth in data,
Valves
bandwidth and networking, creating smart factories with Identification
Process data
more responsive automation at all levels.
Although large systems such as robots and coordinated
Complexity
assembly lines capture attention, the automation they enable
could not be possible without sensors and actuators, which IO-Link Industrial Ethernet
communicate with the programmable logic controllers (PLCs)
Figure 1. Industrial communication by use case.
that run production lines. Sensors and actuators, functioning
both locally and remotely, often greatly outnumber the IO-Link for low bandwidths
complex systems they support. Optimizing overall factory
communications is necessary to meet the wide variety of Sensors and actuators are the most basic units of
requirements from systems of all sizes as you can see in automation, feeding information into and acting on
Figure 1. instructions from networked systems. Traditionally, these
devices connect to control units through interfaces that
Protocols adapting Ethernet to industrial usage have proven provide little intelligence, and thus exchange little or no
popular as field buses on the factory floor. These industrial configuration and diagnostic information. Installing a new
Ethernet protocols, such as EtherCAT and Profinet, offer device requires configuration by hand at the point of use, and
high bandwidth, long physical connections, low latency and without diagnostics, it is impossible to perform just-in-time
deterministic data delivery, among other features required preventive maintenance.
in automated manufacturing. In addition, the field networks
based on these standards tie in easily to the larger plant data IO-Link (International Electrotechnical Commission [IEC]
networks and the internet. 61131-9) is an open standards protocol that addresses the
need for intelligent control of small devices such as sensors
For sensors and actuators, however, industrial Ethernet is and actuators. This standard provides low-speed point-to-
often excessively robust and powerful. These systems usually point serial communication between a device and a master
require point-to-point communications rather than a field bus, that normally serves as a gateway to a field bus and PLC.
and their bandwidth requirements are normally low. The intelligent link enables ease of communication for data
An innovative solution lies in IO-Link, a bidirectional exchange, configuration and diagnostics.
communication protocol based on standard cabling and An unshielded three-wire cable as long as 20 meters,
physical interconnection. IO-Link not only brings data from normally equipped with M12 connectors, establishes the
the factory floor to a PLC efficiently, but it also supports IO-Link connection. Data rates range up to 230 kbps
improved setup, diagnostics and maintenance, and is with a nonsynchronous minimum cycle time of 400 µs,
complementary to existing field-bus cabling. +10%. Four operating modes support bidirectional input/
Since IO-Link and industrial Ethernet are complementary, output (I/O), digital input, digital output and deactivation.
it can benefit designers of networked factory systems to Security mechanisms and deterministic data delivery are not
understand how the two standards work together. specified. A profile known as the IO device description (IODD)

An Engineer’s Guide to Industrial Robot Designs 69 2Q 2020 I Texas Instruments

Chapter 6: Industrial communications for robotics

contains communication properties; device parameters; industrial Ethernet uses a modified media access control
identification, process and diagnostic data; and information (MAC) layer to provide deterministic data delivery with low
specifically about the device and manufacturer. latency and support for time-triggered events. Support
The many advantages of an IO-Link system include for ring and star topologies, as well as traditional in-line
standardized wiring, increased data availability, remote connections, ensure safety and reliability in the event of a
monitoring and configuration, simple replacement of disconnected cable.
devices and advanced diagnostics. IO-Link permits factory Industrial Ethernet is not a unique single specification
managers to receive sensor updates and plan for upcoming but a large group of differing protocol implementations
maintenance or replacement. Swapping out a sensing or driven by various industrial equipment manufacturers for
actuation unit that needs replacement and configuring a new implementation in field-level applications. Popular protocols
one from the PLC through the IO-Link master eliminates include EtherCAT, Profinet, Ethernet/IP, Sercos III and CC-
manual setup and reduces downtime. Switching production Link IE Field, among others. The white paper “An inside
remotely from one configuration to another without visiting look at industrial Ethernet communication protocols”
the factory floor facilitates easier product customization compares these protocols and discusses older, non-Ethernet
(Figure 2). Factories can upgrade production lines readily serial field bus protocols such Controller Area Network (CAN),
to IO-Link, since it is backward-compatible with existing Modbus and Profibus.
standard I/O installations and cabling. Altogether, these Two of the most widely used protocols, Profinet and
capabilities result in reduced overall costs, more efficient EtherCAT, illustrate how the types of industrial Ethernet
processes and greater machine availability. differ with each other and with IO-Link. Both are specified at
100-Mbps transmission speeds and over distances up to
Field-level communication 100 m. Profinet requires a power supply that’s independent
Industrial Ethernet of the data cable power, while EtherCAT offers a version
(EtherCAT P) that includes power and data in the same
cable. Profinet supports full duplex traffic and is capable of
sending packets to each node on the network. The protocol
also offers three classes that allow users to match the level of
performance required to the network. By contrast, EtherCAT
sends a shared frame in one direction on the network in
which all slaves place their data – a scheme that supports
extremely fast forwarding times.
Both Profinet and EtherCAT have faster cycle times than IO-
Link, with much less tolerance. Both base timing on network
synchronization rather than from the start of communication,
Profinet IO-Link Discrete as with IO-Link. Additional protocols provide functional safety
for connections. Industrial Ethernet protocols in general offer
Figure 2. Field-level communication.
a number of services in order to simplify integration within an
automation environment.
Industrial Ethernet: the backbone of the smart
factory Although most sensors do not need the robust set of
features offered by an industrial Ethernet connection, an
In recent years, industrial Ethernet has demonstrated its
important exception is visual sensing. The massive data
value in highly automated factories, becoming the standard
created by a video camera is itself a sufficient reason for
of choice in large field networks that include complex
a higher data-rate connection than what IO-Link can offer.
systems, PLCs and gateways to support intercommunication
Visual and sometimes other types of sensing may provide
with external networks. Benefits such as high speeds,
essential inputs for real-time process control, thus requiring
common interfaces and long connection distances have
the deterministic delivery of industrial Ethernet.
made Ethernet ubiquitous for data networks. In addition,

An Engineer’s Guide to Industrial Robot Designs 70 2Q 2020 I Texas Instruments

Chapter 6: Industrial communications for robotics

For example, ToF applications track and anticipate the three- TI’s TIOL111x transceiver family provides complete IO-
dimensional movement of an object. A typical response Link functionality plus electrostatic discharge, electrical fast
would be a robot arm that moves to intercept the object. IO- transient and surge protection for sensors and actuators
Link may provide sufficient speed and resolution for limited in automated systems. An EVM allows you to review the
sensing of presence in these applications, but industrial devices in operation, and reference designs help speed
Ethernet offers sufficient bandwidth and low-enough latency development of transmitters, proximity switches, solenoid
to determine some characteristics of the object and its drivers, ultrasound and other applications.
surrounding space. Even higher levels of identification may Application designers who require greater bandwidth and
be possible using camera feeds via Gigabit Ethernet, but deterministic timing must decide how many industrial
the industrial Ethernet protocols discussed here have not Ethernet protocols to support in order to make their systems
yet been specified at these speeds. You can see a section compatible with multiple field bus environments. Traditionally,
example for ToF in Figure 3. adding protocols requires creating additional interfaces or
interchangeable modules that plug into the motherboard.
Selection example: Time of Flight (ToF)
Either decision involves additional hardware design, a larger
3D ToF bill of materials, and a longer cycle of testing and certification.
- High resolution
- High frame reate (>100 fps) Instead of adding hardware, the TI Sitara™ family of Arm®
- Sensor fusion processors offers an integrated programmable real-time unit
3D ToF and industrial communication subsystem (PRU-ICSS) that
Speed

- Object detection supports multiprotocol industrial Ethernet. Figure 4 (next

- Object dimension page) shows an example of a Sitara processor-based system
- Safety zone
communicating directly with IO-Link masters.
ToF The PRU loads industrial protocol firmware at device
- Presence run time, with options for EtherCAT, Profinet, Sercos III,
- Distance
Ethernet/IP and Ethernet PowerLink. The PRU-ICSS
handles real-time critical tasks that you would otherwise
Resolution
build into an application-specific integrated circuit (ASIC)
Figure 3. Selection example: ToF. or field-programmable gate array (FPGA), thus offering an
upgradable software-based solution if you need to add
TI technology for smart factory communications
new features or protocols. Based on the scalable Arm core
Many possibilities exist among the competing industrial (Cortex®-A8, A9 or A15, depending on the processor),
Ethernet protocols to serve cameras, motors, robots, Sitara processors enable a single-chip solution for factory
PLCs and other complex systems; even simple sensors automation using multiple industrial Ethernet protocols.
and actuators need to operate effectively within a larger
TI offers a wide line of interfaces (as show in Table 1 (next
industrial Ethernet environment. With so many options for
page) for industrial Ethernet and other standards such as
communication among automated equipment, industrial
CAN, both as stand-alone solutions and as technology
system designers need solutions that are flexible but easy
modules available in other integrated solutions. Many
to use. TI technology addresses these requirements with
TI network products feature reinforced isolation for the
integrated solutions for industrial communications based on
protection of circuitry and humans, while other devices
a range of standards, including IO-Link and the most popular
provide reinforced isolation to add to designs. In-depth
industrial Ethernet protocols.
development support includes software, tools, EVMs and
reference designs for a variety of applications in automated
industrial equipment.

An Engineer’s Guide to Industrial Robot Designs 71 2Q 2020 I Texas Instruments

Chapter 6: Industrial communications for robotics

IO-Link master with Sitara™ AM437x processor

TPS65218
Isolation
AM437x

PRU-ICSS 2 PRU-ICSS 1 Time trigger

DP838221 MII Time sync IO-Link 16-channel
100 Mbit IEP timer IEP timer
UART master IO-Link
Multi- ARM phys
IOLM master
protocol Cortex A9 OSSDe
MPU framer
DP838221 MII industrial Time capture
handler
100 Mbit ethernet
256-k RAM
Transformer
Security SPI
SPI
DP838671 RGMII Gigabit Peripherals
1000 Mbit Ethernet GPIO • Up to 16 channel IO-Link master
• Frame handler in firmware
Security • Channel independent or time sync
CC3xxx/ DDR3L Flash Functional safety
CC26xx

Figure 4. A Sitara™ processor-based system that communicates directly with IO-Link masters.

Feature IO-Link Profinet EtherCAT Comment

≤230 kbit, half duplex, 100 Mbit, shared packet, Only Profinet supports
100 Mbit, full duplex, 100
Physical layer 20 meter, power in same 100 meter, power in concurrent receive and
meter, separate power
cable same cable defined transmit of packets
Ethernet allows for large
Topology Point to point Line, ring, star Line, ring, star
scale networks
250 µs (31.25 ˜µs with IO-Link allows additional
Min. cycle time 400 µs + 10% 31.25 µs
DFP) tolerance of +10%
Based on communication PTCP <±1 µs, IRT test Distributed clocks IO-Link has no
Time syncronization
start <250 ns ±100 ns application time
All need additional
Limitations of no Profinet
Security Not available Not available security protocol for IT
traffic
connection
Only sign of live on Funcational Safety over Ethernet transmission is
Functional safety Profisafe
redundant channel EtherCAT seen as black channel
Smart sensor, fieldbus All support integration
Profidrive, CiR, system SoE, CoE, EoE, FoE,
Profiles and services integration, firmware into automation network,
redundancy, diagnostics AoE, EAP
update, OPC UA no drive profile on IO-Link

Table 1. IO-Link versus industrial Ethernet for manufacturing field communications.

Technology for the future of smart factories systems. IO-Link offers a straightforward alternative for point-
to-point connection between a field bus master and a sensor
The growth of smart factories depends on versatile
or actuator, aiding in configuration and maintenance. TI offers
networking that matches the requirements of individual
a broad portfolio of solutions and flexible technology that help
equipment units with the overall communication needs of the
designers make use of these complementary standards as
factory. Industrial Ethernet protocols provide high bandwidth
they create innovations in automation for the fourth industrial
and fast, guaranteed timing for field bus connections to
revolution.
PLCs, cameras, robots and other complex automated

An Engineer’s Guide to Industrial Robot Designs 72 2Q 2020 I Texas Instruments

Chapter 6: Industrial communications for robotics

6.2 Enabling robots to achieve new levels of place to handle tasks such as welding, painting, picking and
factory automation placing, assembling, and lifting objects to set them on pallets
or in containers. Control signals come from a robot controller,
For more than half a century, robots have played an ever-
which is a control unit in a cabinet usually located at the base
increasing role in manufacturing, successfully transforming
of or next to the robot.
industries ranging from automobiles to electronics to
consumer goods. Robots bring productivity, cost-efficiency Industrial robots are designed to perform tasks quickly,
and often greater safety to repetitive task performance. accurately and without direct interaction with humans. Thus,
they have no sensors to perceive the presence of people,
Robots continue to evolve, offering greater functionality,
and are not designed to accommodate people within their
flexibility, range of motion, speed and precision. Besides
operational space. When human interaction is necessary,
functioning in protected spaces on assembly lines, robots
the robot will usually be deactivated. For human safety and
increasingly operate side by side and interact with human
noninterference with operation, industrial robots are usually
beings, and in some cases move materials from place to
located inside fences, transparent walls, light-activated
place. For robots to operate in these ever-more-complex
barriers, arrays of floor mats that cut off power when stepped
ways, they must be able to process a great deal of sensing
on or other protective barriers.
data about the environment, communicate with each other
and with centralized control units, and perform control
functions that adapt to environmental changes and keep
them from harming humans.
By providing innovative electronic solutions for industrial
automation, TI offers a full range of IC products that enable
advanced robot system development. TI provides both
individual products and complete solutions for control,
communications, power and safety, from the highest control
layer in a factory down to actuators and sensors. TI systems
expertise is based on many years of engagement with
leading manufacturers in many industries, and the company’s
in-depth support helps simplify the design of robotic systems
and reduce development time.

Types of robotic applications in industrial Figure 1. Robot applications in industrial automation and service.
automation
Logistics robots are mobile units that operate in environments
Despite the widespread attention given to human- and
where there may be people present, such as warehouses.
animal-like robots, drones and even robot vacuums, robots
Logistics robots might fetch goods and bring them to a
used in industrial settings remain the mainstay of the robotics
packing station, or transport goods from one building of a
market. Many of the concerns that robot developers face
company site to another; one recent development has robots
when designing products for industrial use also apply to
delivering takeout, although they are currently accompanied
other areas of robotics, and the technology created to handle
by a human handler. These robots typically move within
factory requirements often enables new robot applications
a particular environment and need a number of sensors
outside of manufacturing.
for localization and mapping and to prevent collisions,
While all robots used in industrial automation are technically especially with humans. Ultrasonic, infrared and LIDAR
industrial robots, for the purposes of this discussion sensing are all possible technologies. Because of this robot’s
the focus will be on three groups of robot applications: mobility, the control unit is located inside, often with wireless
industrial robots as shown in Figure 1, logistics robots and communication to a central remote control.
collaborative robots. Industrial robots are units fixed in

An Engineer’s Guide to Industrial Robot Designs 73 2Q 2020 I Texas Instruments

Chapter 6: Industrial communications for robotics

Collaborative robots have the most complex interactions with All of these factors are especially important as the number
humans, often working directly with a person on the same of sensors and environmental stimuli increases. Robot
object at the same time. A collaborative robot might hold an developers depend on solutions that minimize the headaches
object while a human worker visually inspects it or performs of circuit design and certification, speeding the development
fine-tuning tasks. The robot might then set the object down of products that they can deliver to industrial customers
in an area where another robot can pick it up, possibly to quickly.
move it to a new location for collaboration with a different Advanced ICs must offer features that include:
worker.
• A high-efficiency, high-voltage power supply with circuit
Collaborative robot makers must implement a high level protection and low-noise emissions.
of environmental sensing and redundancy into robot
systems to quickly detect and prevent possible collisions. • Characterization for an extended temperature range.
Integrated sensors connected to a control unit will sense the • Support for industrial Ethernet and other widely used
collision between a robot arm and a human or other object; industrial communication standards.
the control unit will turn the robot off immediately. If any • Ease of programming for greater flexibility.
sensor or its electronic circuit fails, the robot also turns off.
• Fast, precise analog-to-digital and digital-to-analog signal
Collaborative robots are typically fixed in place with a control
conversion.
unit in a cabinet, but can also be mounted onto a vehicle.
• Reinforced isolation to meet industrial safety standards.
Technology requirements for industrial robotics
• Control redundancy for safety-critical applications when
Manufacturers investing in robots are seeking greater combined with other ICs.
productivity coupled with a good return on investment in
• A small footprint when placing circuitry in tight spaces such
a reasonable time frame. Achieving these goals depends
as mobile logistic robots, or for motor control in robot arms
on how precise a robot can be during difficult tasks, its
(along with other equipment with tight spaces, such as
performance speed for highly repetitive tasks, its ability
sensors and motor housings).
to maintain safety during dangerous tasks, or some
combination of these capabilities. • Low power consumption (critical for battery- or ambient-
powered equipment such as logistics robots and sensors).
Robots with flexible application capabilities such as the one
shown in Figure 2, often using cameras to see objects, can • Comprehensive support, including reference designs and
save manufacturers from having to invest in more specialized EVMs to minimize design time and let designers focus on
machines, while enabling more complete, shorter and value-added technology.
efficient production runs and new uses on the factory floor.
In addition, many factories today are adding more layers
of communication and control to production lines, bringing
together more data for better process control and equipment
maintenance while also making processes more responsive
to changing product demands. Robots and other equipment
that communicate among themselves and with higher-level
control are essential to fully integrated factories.
Robot developers rely on advanced IC solutions to meet
these requirements. IC products that enable advances in
industrial robots must provide precise sensing, high-speed
sensor signal conversion, fast computation/signal processing
for real-time response and high-speed communications.
ICs also enable high efficiency and small-form-factor power
supplies in conjunction with advanced semiconductors like Figure 2. An assembly line with robots interacting.
GaN FETs.

An Engineer’s Guide to Industrial Robot Designs 74 2Q 2020 I Texas Instruments

Chapter 6: Industrial communications for robotics

TI’s enabling technology for industrial robots • Proximity sensing. Collaborative robots require
sophisticated sensing of nearby things and people in order
TI offers the full range of advanced technology needed to
to make them safe. TI’s expertise in sensing technology
design flexible robots that operate within the environment of
includes solutions for proximity sensing in applications
today’s integrated manufacturing plants. From sensor input
that detect the presence of a target object and, if required,
to actuator or motor output, from individual equipment units
measure its distance. Techniques that TI supports for
to factory-level control and beyond, TI solutions handle the
proximity sensing include ultrasonic, magnetic, capacitive,
entire signal chain, as well as the processing and power
inductive and ToF.
required for robotic applications. Products include features
such as reinforced isolation, and are tested and qualified for • 3D ToF/optical sensing. TI products enable ToF-based
use in harsh industrial environments. To close the circle, TI sensing to go beyond proximity detection to next-
backs its IC products with in-depth support that simplifies generation machine vision. TI’s 3D ToF chipsets provide
design and speeds development. maximum flexibility for you to customize designs for robot
vision or other applications. Tools include an EVM and
Among the many solutions that TI offers for robots and other
a highly configurable camera development kit; the latter
industrial equipment, these are especially notable:
provides a 3D location of each pixel for accurate depth
• Sitara™ processors. The heart of any control unit is the maps that aid customization for a given application.
processor, and TI’s system-optimized Sitara processors
• GaN power. In addition to an extensive power-
are designed for flexible, fast design in robots and other
management portfolio of switching and linear regulators,
industrial equipment. Based on ARM® Cortex®-A cores,
switching controllers, power monitoring and other
Sitara processors provide flexible peripherals, connectivity
and unified software support to cover a wide set of supporting power-management devices, TI offers GaN
applications. A broad portfolio of single- and multicore modules, drivers and controllers that provide outstanding
devices provides a selection that offers the perfect balance power density for high-voltage power supplies in industrial
of integration, connectivity and performance for every systems.
application. A fully scalable software platform enables a
GaN technology greatly reduces switching losses and
unified software experience for simplified development
therefore enables faster switching speeds while reducing
and code migration across Sitara processors and TI DSP
or eliminating heat sinks. Easy-to-use modules provide
families. Pin-compatible options within processor families
a complete solution with optimized layout and efficiency,
make hardware upgrades seamless.
along with minimum electromagnetic emission and noise
for compliance with industrial standards. Complete support
Sitara processors are designed to meet industrial
includes EVMs, development boards and a quick-start
requirements for long-term applications, with product
tool set to speed design. For accurate control of precision
life cycles typically in excess of 10 years. The devices
drives, TI’s reference design library includes a 48-V
provide programmable flexibility when implementing
3-Phase Inverter with Shunt-Based In-Line Motor
specialized data-handling operations, custom peripheral
Phase Current Sensing Reference Design.
interfaces and fast real-time responses as short as 5 ns.
The programmable real-time unit-industrial communication • Industrial Ethernet. TI’s in-depth expertise in network
subsystem (PRU-ICSS), which exists as a hardware communications means support for a wide range of
block inside the Sitara processor family, replaces field- standards. The protocol includes essential setup tasks, and
programmable gate arrays (FPGAs) or application-specific open-source firmware enables product differentiation.
ICs (ASICs) with a single-chip solution. Easy access
to free software and design tools, plus a large base of
open-source community support, reduce barriers to
development.

An Engineer’s Guide to Industrial Robot Designs 75 2Q 2020 I Texas Instruments

Chapter 6: Industrial communications for robotics

Building better robots for the integrated factory This level of protection can simplify designs by eliminating
or greatly reducing the size of external transient voltage
As manufacturing continues to become more highly
suppression (TVS) diode components that originally provided
integrated at all levels, robots will play an ever-increasing role
protection, thus reducing the overall BOM and associated
in carrying out a wide variety of assembly tasks that increase
costs when compared to previous-generation or competing
production and make the workplace safer for humans.
solutions.
Traditional industrial robots, logistics robots and collaborative
robots have their jobs to do, and robot developers need The physical size of sensors continues to shrink. The smallest
solutions that enable accurate, safe, cost-effective operation of these sensors is likely the cylindrical sensor, as shown in
from all of them. Figure 1.
TI’s IC products for signaling, processing, communications and
power management help provide the solutions that robotics
manufacturers need. TI backs its semiconductor and IC
products with software tools, EVMs and reference designs, and
other forms of support that help make the work of designing
robots faster and more profitable. As robots are improving
manufacturing, TI is helping developers improve robots.

6.3 Is your factory smarter than a fifth grader?

Looking at the future of factory automation, IO-Link is an
enabling interface for Industry 4.0. You can use the IO-Link
bidirectional, manufacturer-independent communication
protocol to develop highly efficient and scalable “smart
factories.”
TI’s TIOL111 IO-Link transceiver and TIOS101 digital output
Figure 1. An IO-Link cylindrical sensor.
switch will help enable the next generation of sensors and
actuators in factories, while providing features that further The sensor at the top of Figure 1 is the finished product
optimize product offerings and simplify bill of materials. with a cylindrical enclosure. The middle wire is the sensor’s
internal PCB. This PCB measures 17.5 mm by 2.5 mm.
You can take advantage of the pin compatibility between
In order to fit in these small form factors, an equally small
the TIOL111 and TIOS101 devices to develop a complete
device is required to implement either IO-Link or SIO
portfolio of both IO-Link and standard input/output (SIO)-
output. Knowing these system requirements drove new
enabled sensors without using two separate PCBs for
package developments for the TIOL111 and TIOS101. The
each offering. Each device supports the interfaces intended
DMW package is one of the smallest, thermally enhanced
while also supporting a high level of integrated protection,
IO-Link packages available today. Measuring 3.0 mm by
including:
2.5 mm, the DMW package also supports a thermal pad
• 16-kV International Electrotechnical Commission (IEC) for heat conduction and a flow-through pinout. The flow-
61000-4-2 electrostatic discharge (ESD). through pinout (Figure 2) further aides in PCB layout and
• 4-kV IEC 61000-4-4 electrical fast transient (EFT) Criterion A. device placement by supporting the logic interface to the
microcontroller on one X-axis and the 24-V IO-Link interface
• 1.2-kV/500-Ω IEC 61000-4-5 (surge).
on the other.
• ±65-V transient tolerance.
• Reverse polarity up to ±55 V.
• Overcurrent/overvoltage/over temperature.

An Engineer’s Guide to Industrial Robot Designs 76 2Q 2020 I Texas Instruments

Chapter 6: Industrial communications for robotics

Driver current output Residual voltage

VCC_IN/
OUT
WAKE 100 mA 1.1 V
High side 200 mA 1.5 V
Logic interface

24-V interface
NFAULT L+

250 mA 1.75 V

2.5 mm
RX Thermal pad CQ
100 mA 1.1 V
TX L- Low side 200 mA 1.5 V

EN ILIM_ADJ
250 mA 1.75 V

Table 1. Power dissipation = residual voltage × 10.

3.0 mm A configurable current limit is an additional protection feature
that can protect the sensor and possibly prevent the network
Figure 2. TIOL111/TIOS101 flow-through package.
from shutting down. Through a 0-kΩ to 100-kΩ resistor,
The TIOL111 and TIOS101 have a low residual voltage of the TIOL111 and TIOS101 can support a 50-mA to 350-
just 1.75V – the size of sensors that use IO-Link and SIO are mA current limit. This limiter can notify the programmable
rugged, sealed and often very small. This small, enclosed logic controller (PLC) of an overcurrent condition and shut
form factor introduces numerous design challenges, with down the device’s output, with periodic monitoring of the
thermal performance one of the most challenging. By overcurrent condition.
supporting an ultra-low residual voltage of 1.75 V at 250 mA,
The configurable current-limit resistor is placed on the ILIM_
the TIOL111 and TIOS101 provide a superior base for power
ADJ pin (see Figure 3).
dissipation and related system thermal management. Table 1
shows the driver current output and residual voltage for high
side and low side.

TIOL111(x)
VCC_OUT Voltage L+
1 µF regulator 0.1 µF
10 V 10 kΩ 100 V
RX
10 kΩ
Sensor TX
CQ
Control

front-end Microcontroller EN IO-Link

logic

WAKE Diagnostics master

and
control
NFAULT

CUR_OK
STMP_OK L-
PWR_OK

ILIM_ADJ
RSET

Figure 3. TIOL111 application schematic.

An Engineer’s Guide to Industrial Robot Designs 77 2Q 2020 I Texas Instruments

Chapter 6: Industrial communications for robotics

As Figure 4 indicates, using a 0-Ω resistor on the ILIM_ADJ These and other features and benefits of the TIOL111 and
pin will default to the maximum current limit of 350 mA. TIOS101 can enable the smallest sensor form factors, while
providing the flexibility to support multiple platforms and
current configuration requirements.
400
Low Side
350 High Side
300
250
IO(LIM) (mA)

200
150
100
50
0
0 20 40 60 80 100
RSET (kΩ)

Figure 4. Current limit vs. RSET.

6.4 Industrial communications-related reference

designs for robotic systems

EtherCAT Slave and Multi-Protocol Industrial Ethernet This design implements a cost-optimized high
Reference Design electromagnetic compatibility immunity EtherCAT slave (dual
ports) with a Serial Peripheral Interface to the application
processor.
The hardware design is capable of supporting multiprotocol
industrial Ethernet and field buses using the AMIC110
industrial communications processor.
Redundant Dual-Channel Safe Torque Off (STO) This reference design implements safe torque off functionality
Reference Design for AC Inverters and Servo Drives in variable speed drives (per International Electrotechnical
Commission 61800-5-2) by using dual-channel isolated STO
signals to control the inverter that supplies power to the
motor.
Simple Open Real-Time Ethernet (SORTE) Master with The reference design implements a SORTE master with
PRU-ICSS Reference Design the programmable real-time unit (PRU) and industrial
communication subsystem. SORTE enables applications to
exchange process data between the master and devices in
a 4-µs cycle time. The design contains open-source PRU
firmware.
EtherCAT P One Cable for Power and EtherCAT This reference design shows a physical implementation of the
Reference Design power couple in circuitry of an EtherCAT P power-sourcing
device. The design is designed to meet all requirements from
the official EtherCAT P implementation guide.

Find more reference designs for your robot system.

Contributing authors: Thomas Leyrer, Miro Adzan, Anant
Kamath, Tobias Puetz and Russell Crane.

An Engineer’s Guide to Industrial Robot Designs 78 2Q 2020 I Texas Instruments

TI Worldwide Technical Support

TI Support
Thank you for your business. Find the answer to your support need or get in touch
with our support center at
www.ti.com/support
China: http://www.ti.com.cn/guidedsupport/cn/docs/supporthome.tsp
Japan: http://www.tij.co.jp/guidedsupport/jp/docs/supporthome.tsp

Technical support forums

Search through millions of technical questions and answers at TI’s E2E™ Community
(engineer-to-engineer) at
e2e.ti.com
China: http://www.deyisupport.com/
Japan: http://e2e.ti.com/group/jp/

TI Training
From technology fundamentals to advanced implementation, we offer on-demand and
live training to help bring your next-generation designs to life. Get started now at
training.ti.com
China: http://www.ti.com.cn/general/cn/docs/gencontent.tsp?contentId=71968
Japan: https://training.ti.com/jp

Important Notice: The products and services of Texas Instruments Incorporated and its subsidiaries described herein are sold subject to TI’s standard terms and conditions of sale.
Customers are advised to obtain the most current and complete information about TI products and services before placing orders. TI assumes no liability for applications assistance,
customer’s applications or product designs, software performance, or infringement of patents. The publication of information regarding any other company’s products or services
does not constitute TI’s approval, warranty or endorsement thereof.

The platform bar, C2000, LightCrafter, E2E, Delfino, InstaSPIN,

Hercules, Piccolo and Sitara are trademarks and DLP is a
registered trademark of Texas Instruments.
All other trademarks are the property of their respective owners.

© 2020 Texas Instruments Incorporated SSIY006

 IMPORTANT NOTICE AND DISCLAIMER

TI PROVIDES TECHNICAL AND RELIABILITY DATA (INCLUDING DATASHEETS), DESIGN RESOURCES (INCLUDING REFERENCE
DESIGNS), APPLICATION OR OTHER DESIGN ADVICE, WEB TOOLS, SAFETY INFORMATION, AND OTHER RESOURCES “AS IS”
AND WITH ALL FAULTS, AND DISCLAIMS ALL WARRANTIES, EXPRESS AND IMPLIED, INCLUDING WITHOUT LIMITATION ANY
IMPLIED WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT OF THIRD
PARTY INTELLECTUAL PROPERTY RIGHTS.
These resources are intended for skilled developers designing with TI products. You are solely responsible for (1) selecting the appropriate
TI products for your application, (2) designing, validating and testing your application, and (3) ensuring your application meets applicable
standards, and any other safety, security, or other requirements. These resources are subject to change without notice. TI grants you
permission to use these resources only for development of an application that uses the TI products described in the resource. Other
reproduction and display of these resources is prohibited. No license is granted to any other TI intellectual property right or to any third
party intellectual property right. TI disclaims responsibility for, and you will fully indemnify TI and its representatives against, any claims,
damages, costs, losses, and liabilities arising out of your use of these resources.
TI’s products are provided subject to TI’s Terms of Sale (www.ti.com/legal/termsofsale.html) or other applicable terms available either on
ti.com or provided in conjunction with such TI products. TI’s provision of these resources does not expand or otherwise alter TI’s applicable
warranties or warranty disclaimers for TI products.

Mailing Address: Texas Instruments, Post Office Box 655303, Dallas, Texas 75265
Copyright © 2020, Texas Instruments Incorporated