18

Sensor and Actuator

Characteristics

18.1 Range
18.2 Resolution
18.3 Sensitivity
18.4 Error
18.5 Repeatability
18.6 Linearity and Accuracy
18.7 Impedance
18.8 Nonlinearities
18.9 Static and Coulomb Friction
18.10 Eccentricity
18.11 Backlash
18.12 Saturation
18.13 Deadband
18.14 System Response
18.15 First-Order System Response
Joey Parker 18.16 Underdamped Second-Order System Response
University of Alabama 18.17 Frequency Response

Mechatronic systems use a variety of sensors and actuators to measure and manipulate mechanical,
electrical, and thermal systems. Sensors have many characteristics that affect their measurement capa-
bilities and their suitability for each application. Analog sensors have an output that is continuous over
a finite region of inputs. Examples of analog sensors include potentiometers, LVDTs (linear variable
differential transformers), load cells, and thermistors. Digital sensors have a fixed or countable number
of different output values. A common digital sensor often found in mechatronic systems is the incremental
encoder. An analog sensor output conditioned by an analog-to-digital converter (ADC) has the same
digital output characteristics, as seen in Fig. 18.1.

18.1 Range
The range (or span) of a sensor is the difference between the minimum (or most negative) and maximum
inputs that will give a valid output. Range is typically specified by the manufacturer of the sensor. For
example, a common type K thermocouple has a range of 800∞C (from -50∞C to 750∞C). A ten-turn
potentiometer would have a range of 3600 degrees.

©2002 CRC Press LLC

FIGURE 18.1 Analog and digital sensor outputs.

FIGURE 18.2 Sensor sensitivity.

18.2 Resolution
The resolution of a sensor is the smallest increment of input that can be reliably detected. Resolution is
also frequently known as the least count of the sensor. Resolution of digital sensors is easily determined.
A 1024 ppr (pulse per revolution) incremental encoder would have a resolution of
1 revolution 360 degrees degrees
------------------------------ × ------------------------------ = 0.3516 -----------------
1024 pulses 1 revolution pulse
The resolution of analog sensors is usually limited only by low-level electrical noise and is often much
better than equivalent digital sensors.

18.3 Sensitivity
Sensor sensitivity is defined as the change in output per change in input. The sensitivity of digital sensors
is closely related to the resolution. The sensitivity of an analog sensor is the slope of the output versus
input line. A sensor exhibiting truly linear behavior has a constant sensitivity over the entire input range.
Other sensors exhibit nonlinear behavior where the sensitivity either increases or decreases as the input
is changed, as shown in Fig. 18.2.

18.4 Error
Error is the difference between a measured value and the true input value. Two classifications of errors
are bias (or systematic) errors and precision (or random) errors. Bias errors are present in all measure-
ments made with a given sensor, and cannot be detected or removed by statistical means. These bias
errors can be further subdivided into
• calibration errors (a zero or null point error is a common type of bias error created by a nonzero
output value when the input is zero),
• loading errors (adding the sensor to the measured system changes the system), and
• errors due to sensor sensitivity to variables other than the desired one (e.g., temperature effects
on strain gages).

©2002 CRC Press LLC

18.5 Repeatability
Repeatability (or reproducibility) refers to a sensor’s ability to give identical outputs for the same input.
Precision (or random) errors cause a lack of repeatability. Fortunately, precision errors can be accounted
for by averaging several measurements or other operations such as low-pass filtering. Electrical noise and
hysteresis (described later) both contribute to a loss of repeatability.

18.6 Linearity and Accuracy

The accuracy of a sensor is inversely proportional to error, i.e., a highly accurate sensor produces low
errors. Many manufacturers specify accuracy in terms of the sensor’s linearity. A least-squares straight-
line fit between all output measurements and their corresponding inputs determines the nominal output
of the sensor. Linearity (or accuracy) is specified as a percentage of full scale (maximum valid input), as
shown in Fig. 18.3, or as a percentage of the sensor reading, as shown in Fig. 18.4. Figures 18.3 and 18.4
show both of these specifications for 10% linearity, which is much larger than most actual sensors.
Accuracy and precision are two terms that are frequently confused. Figure 18.5 shows four sets of
histograms for ten measurements of angular velocity of an actuator turning at a constant 100 rad/s. The
first set of data shows a high degree of precision (low standard deviation) and repeatability, but the

FIGURE 18.3 Linearity specified at full scale.

FIGURE 18.4 Linearity specified at reading.

©2002 CRC Press LLC

FIGURE 18.5 Examples of accuracy and precision.

average accuracy is poor. The second set of data shows a low degree of precision (high standard deviation),
but the average accuracy is good. The third set of data shows both low precision and low accuracy, while
the fourth set of data shows both high precision, high repeatability, and high accuracy.

18.7 Impedance
Impedance is the ratio of voltage and current flow for a sensor. For a simple resistive sensor (such as a
strain gage or a thermistor), the impedance Z is the same as the resistance R, which has units of ohms (W),

V
Z R = ---I = R

For more complicated sensors, impedance includes the effects of capacitance, C, and inductance, L.
Inclusion of these terms makes the impedance frequency sensitive, but the units remain ohms:

V 1 V
Z C = ---I = ----------
jCω
- and Z L = ---I = jL ω

where j = – 1 is the imaginary number and ω is the driving frequency. The impedance form is
particularly nice for analyzing simple circuits, as parallel and series inductances can be treated just like
resistances. Two types of impedance are important in sensor applications: input impedance and output
impedance. Input impedance is a measure of how much current must be drawn to power a sensor (or
signal conditioning circuit). Input impedance is frequently modeled as a resistor in parallel with the

©2002 CRC Press LLC

input terminals. High input impedance is desirable, since the device will then draw less current from the
source. Oscilloscopes and data acquisition equipment frequently have input impedances of 1 MΩ or
more to minimize this current draw. Output impedance is a measure of a sensor’s (or signal conditioning
circuit’s) ability to provide current for the next stage of the system. Output impedance is frequently
modeled as a resistor in series with the sensor output. Low output impedance is desirable, but is often
not available directly from a sensor. Piezoelectric sensors in particular have high output impedances and
cannot source much current (typically micro-amps or less). Op-amp circuits are frequently used to buffer
sensor outputs for this reason. Op-amp circuits (especially voltage followers) provide nearly ideal cir-
cumstances for many sensors, since they have high input impedance but can substantially lower output
impedance.

18.8 Nonlinearities
Linear systems have the property of superposition. If the response of the system to input A is output A,
and the response to input B is output B, then the response to input C (= input A + input B) will be
output C ( = output A + output B). Many real systems will exhibit linear or nearly linear behavior over
some range of operation. Therefore, linear system analysis is correct, at least over these portions of a
system’s operating envelope. Unfortunately, most real systems have nonlinearities that cause them to
operate outside of this linear region, and many common assumptions about system behavior, such as
superposition, no longer apply. Several nonlinearities commonly found in mechatronic systems include
static and coulomb friction, eccentricity, backlash (or hysteresis), saturation, and deadband.

18.9 Static and Coulomb Friction

In classic linear system analysis, friction forces are assumed to be proportional to velocity, i.e., viscous
friction. With an actuator velocity of zero, there should be no friction. In reality, a small amount of static
(no velocity) or Coulomb friction is almost always present, even in roller or ball type anti-friction
bearings. A typical plot of friction force vs. velocity is given in Fig. 18.6. Note that the static friction force
can assume any value between some upper and lower limit at zero velocity. Static friction has two primary
effects on mechatronic systems:
1. Some of the actuator torque or force is wasted overcoming friction forces, which leads to ineffi-
ciency from an energy viewpoint.
2. As the actuator moves the system to its final location, the velocity approaches zero and the actuator
force/torque will approach a value that exactly balances frictional and gravity loads. Since static
friction can assume any value at zero velocity, the actuator will come to slightly different final
resting positions each time—depending on the final value of static friction. This effect contributes
to some loss of repeatability in mechatronic systems.

FIGURE 18.6 Static and Coulomb friction.

©2002 CRC Press LLC

18.10 Eccentricity
The ideal relationships for gears, pulleys, and chain drives assume that the point of gear contact remains
at a fixed distance from the center of rotation for each gear. In reality, the true center of the gears pitch
circle and the center of rotation will be separated by a small amount, known as the eccentricity. Small
tooth-to-tooth errors can also cause local variations in the pitch circle radius. The combination of these two
effects can lead to a nonlinear geometrical relationship between two gears like that of Fig. 18.7, where
the nonlinear behavior is greatly exaggerated for clarity. Eccentricity impacts the accuracy of position
measurements made on the input side of the gear pair, as the output gear is not exactly where the sensor
measurement indicates.

18.11 Backlash
If two otherwise perfect gears are not mounted on a center-to-center distance that exactly matches the
sum of the pitch radii, there will be a small clearance, or backlash, between the teeth. When the input
gear reverses direction, a small rotation is required before this clearance is removed and the output gear
begins to move. Gear backlash is just one of many phenomena that can be characterized as hysteresis,
as shown in Fig. 18.8. Clearance between shafts and bearings can cause hysteretic effects also. Backlash
exhibits effects similar to those for eccentricity, i.e., a loss of repeatability, particularly when approaching
a measured point from different directions. The gear backlash problem is so prevalent and potentially
harmful that many manufacturers go to great lengths to minimize or reduce the effect:
• gears mounted closer together than the theoretically ideal spacing,
• split “anti-backlash” gears that are spring loaded to force teeth to maintain engagement at all times,
• external spring-loaded mounts for one of the gears to force engagement, or
• specially designed gears with anti-backlash features.

FIGURE 18.7 Gear eccentricity.

FIGURE 18.8 Gear backlash.

©2002 CRC Press LLC

18.12 Saturation
All real actuators have some maximum output capability, regardless of the input. This violates the linearity
assumption, since at some point the input command can be increased without significantly changing the
output; see Fig. 18.9. This type of nonlinearity must be considered in mechatronic control system design,
since maximum velocity and force or torque limitations affect system performance. Control systems mod-
eled with linear system theory must be carefully tested or analyzed to determine the impact of saturation
on system performance.

18.13 Deadband
Another nonlinear characteristic of some actuators and sensors is known as deadband. The deadband is
typically a region of input close to zero at which the output remains zero. Once the input travels outside
the deadband, then the output varies with input, as shown in Fig. 18.10. Analog joystick inputs frequently
use a small amount of deadband to reduce the effect of noise from human inputs. A very small movement
of the joystick produces no output, but the joystick acts normally with larger inputs.
Deadband is also commonly found in household thermostats and other process type controllers, as
shown in Fig. 18.11. When a room warms and the temperature reaches the setpoint (or desired value)

FIGURE 18.9 Saturation.

FIGURE 18.10 Deadband.

FIGURE 18.11 Thermostat deadband.

©2002 CRC Press LLC

 o

FIGURE 18.12 System stability.

on the thermostat, the output remains off. Once room temperature has increased to the setpoint plus
half the deadband, then the cooling system output goes to fully on. As the room cools, the output stays
fully on until the temperature reaches the setpoint minus half the deadband. At this point the cooling
system output goes fully off.

18.14 System Response

Sensors and actuators respond to inputs that change with time. Any system that changes with time is
considered a dynamic system. Understanding the response of dynamic systems to different types of inputs
is important in mechatronic system design. The most important concept in system response is stability.
The term stability has many different definitions and uses, but the most common definition is related to
equilibrium. A system in equilibrium will remain in the same state in the absence of external disturbances.
A stable system will return to an equilibrium state if a “small” disturbance moves the system away from
the initial state. An unstable system will not return to an equilibrium position, and frequently will move
“far” from the initial state.
Figure 18.12 illustrates three stability conditions with a simple ball and hill system. In each case an
equilibrium position is easily identified—either the top of the hill or the bottom of the valley. In the
unstable case, a small motion of the ball away from the equilibrium position will cause the ball to move
“far” away, as it rolls down the hill. In the stable case, a small movement of the ball away from the equi-
librium position will eventually result in the ball returning, perhaps after a few oscillations. In the third
case, the absence of friction causes the ball to oscillate continuously about the equilibrium position once
a small movement has occurred. This special case is often known as marginal stability, since the system
never quite returns to the equilibrium position.
Most sensors and actuators are inherently stable. However, the addition of active control systems can
cause a system of stable devices to exhibit overall unstable behavior. Careful analysis and testing is required
to ensure that a mechatronic system acts in a stable manner. The complex response of stable dynamic
systems is frequently approximated by much simpler systems. Understanding both first-order and second-
order system responses to either instantaneous (or step) changes in inputs or sinusoidal inputs will suffice
for most situations.

18.15 First-Order System Response

First-order systems contain two primary elements: an energy storing element and an element which
dissipates (or removes) energy. Typical first-order systems include resistor–capacitor filters and resistor–
inductor networks (e.g., a coil of a stepper motor). Thermocouples and thermistors also form first-order
systems, due to thermal capacitance and resistance. The differential equation describing the time response
of a generic first-order system is

dy ( t ) 1
------------ + ---- y ( t ) = f ( t )
dt τ

©2002 CRC Press LLC

FIGURE 18.13 First-order system—step response.

where y(t) is the dependent output variable (velocity, acceleration, temperature, voltage, etc.), t is the
independent input variable (time), τ is the time constant (units of seconds), and f (t) is the forcing
function (or system input).
The solution to this equation for a step or constant input is given by
– t/t
y ( t ) = y ∞ + ( y 0 – y ∞)e
where y∞ is the limiting or final (steady-state) value, y0 is the initial value of the independent variable at
t = 0.
A set of typical first-order system step responses is shown in Fig. 18.13. The initial value is arbitrarily
selected as 20 with final values of 80. Time constants ranging from 0.25 to 2 s are shown. Each of these
curves directly indicates its time constant at a key point on the curve. Substituting t = τ into the first-
order response equation with y0 = 20 and y∞ = 80 gives
–1
y ( t ) = 80 + ( 20 – 80 )e = 57.9
Each curve crosses the y(τ) ≈ 57.9 line when its time constant τ equals the time t. This concept is
frequently used to experimentally determine time constants for first-order systems.

18.16 Underdamped Second-Order System Response

Second-order systems contain three primary elements: two energy storing elements and an element which
dissipates (or removes) energy. The two energy storing elements must store different types of energy. A
typical mechanical second-order system is the spring–mass–damper combination shown in Fig. 18.14.
1 2 1 2
The spring stores potential energy (PE = --2 kx ), while the mass stores kinetic energy (KE = --2 mv ), where k
is the spring stiffness (typical units of N/m), x is the spring deflection (typical units of m), m is the mass
(typical units of kg), and v is the absolute velocity of the mass (typical units of m/s).
A common electrical second-order system is the resistor–inductor–capacitor (RLC) network, where
the capacitor and inductor store electrical energy in two different forms. The generic form of the dynamic
equation for an underdamped second-order system is
2
d y(t) dy ( t )
2 + 2ζω n dt + ω n y ( t ) = f ( t )
------------- ------------ 2
dt

where y(t) is the dependent variable (velocity, acceleration, temperature, voltage, etc.), t is the independent
variable (time), ζ is the damping ratio (a dimensionless quantity), ωn is the natural frequency (typical
units of rad/s), and f (t) is the forcing function (or input).

©2002 CRC Press LLC

0066_Frame_C18 Page 10 Wednesday, January 9, 2002 5:14 PM

FIGURE 18.14 Spring–mass–damper system.

FIGURE 18.15 Second-order system—step response.

The response of an underdamped ( 0 ≤ ζ < 1 ) second-order system to a unit step input can be deter-
mined as:

ζ
y ( t ) = 1 – e –ζωn t  cos ω n 1 – ζ t + ------------------ sin ω n 1 – ζ t
2 2
 
1–ζ
2

This second-order system step response is often characterized by a set of time response parameters
illustrated in Fig. 18.15.
These time response parameters are functions of the damping ratio ζ and the natural frequency ωn:
• peak time, TP : the time required to reach the first (or maximum) peak

π
T P = -------------------------
ωn 1 – ζ
2

• percent overshoot, %OS: amount the response exceeds or overshoots the steady-state value
2
– ( ζπ / 1 – ζ )
%OS = 100 e
• settling time, TS: the time when the system response remains within ±2% of the steady-state value

4
T S = ---------
ζω n
• rise time, TR: time required for the response to go from 10% to 90% of the steady-state value.
Figure 18.16 shows the nondimensional rise time (ωnTR) as a function of damping ratio, z.
A frequently used approximation relating these two parameters is

ω n T R ≈ 2.16 ζ + 0.6 0.3 ≤ ζ 0.8

Figures 18.17 and 18.18 show the unit step response of a second-order system as a function of damping
ratio ζ.

©2002 CRC Press LLC

0066_Frame_C18 Page 11 Wednesday, January 9, 2002 5:14 PM

FIGURE 18.16 Rise time vs. damping ratio, ζ.

FIGURE 18.17 Second-order system step response vs. damping ratio, ζ.

FIGURE 18.18 Initial second-order system step response vs. damping ratio, ζ.

©2002 CRC Press LLC

0066_Frame_C18 Page 12 Wednesday, January 9, 2002 5:14 PM

18.17 Frequency Response

The response of any dynamic system to a sinusoidal input is called the frequency response. A generic
first-order system with a sinusoidal input of amplitude A would have the dynamic equation of

dy ( t ) 1
------------ + -- y ( t ) = f ( t ) = A sin ( ω t )
dt τ

where ω is the frequency of the sinusoidal input and τ is the first-order time constant. The steady-state
solution to this equation is

y ( t ) = AM sin ( ω t + Φ )
2 −1
where M = 1/ ( τω ) + 1 is the amplitude ratio (a dimensionless quantity), and Φ = −tan (τω) is the
phase angle.
Figure 18.19 is a plot of the magnitude ratio MdB and the phase angle Φ as a function of the non-
dimensional frequency, τω. Note that the magnitude is frequently plotted in terms of decibels, where
MdB = 20 log10(M).
The frequency at which the magnitude ratio equals 0.707 (or –3 dB) is called the bandwidth. For a
first-order system, the bandwidth is inversely proportional to the time constant. So, ω = 1/τ.
A generic second-order system with a sinusoidal input of amplitude A and frequency ω would have
the dynamic equation of

2
d y(t) dy ( t )
+ 2 ζω n ------------ + ω n y ( t ) = A sin ( ω t )
2
--------------
2
dt dt

The steady-state solution to this equation is

AM
- sin ( ω t + Φ )
y ( t ) = --------
ωn
2

where

1
M = -----------------------------------------------------------------------------
2 2
[ 1 – ( ω / ω n ) ] + [ 2 ζ ( ω /ω n ) ]
2 2

FIGURE 18.19 Frequency response for first-order system.

©2002 CRC Press LLC

0066_Frame_C18 Page 13 Wednesday, January 9, 2002 5:14 PM

FIGURE 18.20 Frequency response magnitude for second-order system.

FIGURE 18.21 Frequency response phase angle for second-order system.

is the amplitude ratio (a dimensionless quantity), and

−1 2 ζ ( ω /ω n )
Φ = – tan ----------------------------
1 – ( ω / ωn )
2 2

is the phase angle.

Figures 18.20 and 18.21 are plots of the magnitude response MdB = 20 log10(M) and the phase angle
Φ for the second-order system as a function of damping ratio, ζ. The peak value in the magnitude
response, MP , can be found by taking the derivative of M with respect to ω and setting the result to zero
to find (Nise, 1995)

1
M P = ------------------------
2ζ 1 – ζ
2

This peak value in M occurs at the frequency ωP given by

ωP = ωn 1 – 2 ζ
2

©2002 CRC Press LLC

©2002 CRC Press LLC

0066_Frame_C18 Page 14 Wednesday, January 9, 2002 5:14 PM

The peak value in an experimentally determined frequency response can be used to estimate both the
natural frequency and damping ratio for a second-order system. These parameters can then be used to
estimate time domain responses such as peak time and percent overshoot.

Reference
Nise, N. S., Control Systems Engineering, 2nd ed., Benjamin/Cummings, 1995.

©2002 CRC Press LLC