CND 111: Digital IC Design

Lecture Assignment #: 2

Section #: ----19-----

Submitted by:
Student Name ID
Malak Hamed Abdelhadi Hamed Mokhtar v23010696
Rawan Mohamed Edris v23010396
Mohamed Khaled Alahmady V23010489

Submitted to TA: Dr. Eslam & Eng. Sondos

Date: ----14-10-2023----
Write a report about 4 domains (ex: AI, DSP, Satellite communications, RF) and which FPGA
families would be suitable for them and why? Refer to two different FPGA production
companies.

▪ AI:
In recently released research, Intel compared two generations of their FPGAs to an NVIDIA
GPU. The major objective of the experiment was to determine whether FPGAs of the next
generation could compete with GPUs for AI application acceleration. In their research, Intel
draws the conclusion that the Intel Stratix 10 FPGA AI accelerator can outperform a conventional
GPU. When compact data types are used instead of conventional 32-bit floating-point data
(FP32), this is made possible. Calculations need to be made easier while yet being accurate
enough.
FPGAs might be the solution to both of these challenges:
• Finding a faster way to transfer and process data.
• Figuring out ways to make AI-based applications work better overall.
Inferencing needs both flexibility and low latency to be effective. FPGAs are useful in this
situation. These two issues can both be fixed with FPGAs. An FPGA's ability to be reprogrammed
provides the flexibility needed by artificial neural networks' dynamic structure. For real-time
inferencing of a model, FPGAs also offer the bespoke parallelism and high-bandwidth memory
needed. Also, FPGAs have the added benefit of being less power-hungry than traditional GPUs.
The use of cloud-based FPGAs for accelerating AI workloads and other types of applications
needing intensive computation is already being worked on by some FPGA manufacturers. For
instance, Intel powers the f1 instances Alibaba Cloud AaaS service. Users of the Alibaba Cloud
can also use the Acceleration Stack for Intel Xeon CPU with FPGAs, which provides two well-
liked software development flows, RTL, and OpenCL.
Microsoft is a significant firm that has entered the competition to create a powerful AI platform.
FPGA technology is available through their project Brainwave to speed up deep neural network
inference. They make use of Intel's Stratix 10 FPGA, just like Alibaba Cloud.

Although Intel undoubtedly dominates the FPGA market for AI application acceleration, Xilinx,
another significant FPGA vendor, plans to enter the competition. A new SDAccel integrated
development environment from Xilinx has been released to make it simpler for FPGA developers
to interact with various cloud platforms.

2|Page
▪ DSP
- Introduction
Digital Signal Processing (DSP) is a vital field in modern electronics, encompassing
the manipulation of signals using digital techniques. Field-Programmable Gate
Arrays (FPGAs) are widely utilized in DSP applications due to their flexibility and
high computational power. We will explore the importance of DSP in today's
technology landscape and recommend FPGA families suitable for DSP applications,
focusing on two leading FPGA manufacturers, Xilinx, and Intel (formerly Altera).
- Importance of DSP in Modern Technology
DSP is fundamental in a wide range of applications, including telecommunications,
audio and video processing, radar, medical imaging, and more. The key advantages
of DSP include:

Flexibility DSP allows for the implementation of complex signal processing

algorithms through software, making it highly adaptable to
changing requirements.
High Performance DSP algorithms can be optimized for parallel processing, enabling
real-time, high-performance signal manipulation.
Power Efficiency DSP hardware and algorithms can be optimized to achieve
excellent power efficiency, crucial in portable and battery-
operated devices
Noise Reduction DSP techniques enable the removal of unwanted noise from
signals, enhancing signal quality.

- FPGA Selection for DSP Applications

Selecting the right FPGA for DSP applications is essential for achieving optimal
performance. Two prominent FPGA manufacturers are Xilinx and Intel (formerly
Altera). Here's a comparison of FPGA families suitable for DSP applications from
both companies:
• Xilinx:
i. Zynq Ultra Scale+ MPSoC:
- Suitable for high-performance DSP application.
- Integrates ARM cores and programmable logic, providing a versatile processing
platform.
- Offers high-performance DSP slices, facilitating complex signal processing.
ii. Kintex UltraScale+:

3|Page
 - Offers high-performance DSP slices, facilitating complex signal processing.
- Features a considerable number of DSP slices for parallel processing.
- Offers high-speed transceivers for data streaming applications.
iii. Spartan-7:
- Designed for cost-sensitive DSP applications.
- Offers a balance between cost and performance.
- Suitable for applications with modest DSP requirements.
• Intel:
i. Arria 10:
- High-performance FPGA family suitable for demanding DSP tasks
- Offers high-speed transceivers and a significant number of DSP blocks.
- Ideal for applications like wireless communication and radar.
ii. Cyclone V:
- Offers a balance between cost and performance.
- Suitable for applications with moderate DSP requirements
- Can be used in image and audio processing applications.

▪ Satellite Communications:
- Reprogrammable (SRAM-based) FPGAs (RFPGAs), with their high levels of
flexibility, performance, and complexity, are becoming more and more crucial for
use in space applications. Reprogrammability in flight becomes a strict necessity as
satellite lifetimes extend well beyond 10 years, far beyond the usefulness of
telecom standards. The sole option in the future may be RFPGA if software solutions
are not feasible.
- With varying degrees of success, Field Programmable Gate Array (FPGA) devices
have been utilized in space for more than ten years. Due to their vulnerability to
involuntary reconfiguration brought on by Single Event Upsets (SEU) brought on by
radiation, few reprogrammable devices have up until this point been utilized
aboard European spacecraft. However, with the development of reprogrammable
devices with one million or more system gates, it is no longer practical to ignore
these technologies.
- Under the terms of an ESA (European Space Agency) contract, IASF Milano
developed the FLIPPER test system. A Xilinx FPGA's user flip-flops, configuration
memory, and reconfiguration control registers can all be corrupted with SEU-like
defects using FLIPPER. This makes it possible to examine the effects of configuration
SEUs on unprotected architectures and to assess the effectiveness of fault mitigation
techniques.

4|Page
 ▪ Radio frequency:
A) The Intel Direct RF FPGA family is suitable for RF applications as it offers a portfolio
of analog-enabled FPGAs with the following advantages:
1. Direct Analog RF Signal Conversion: These FPGAs can perform direct analog RF
signal conversion with multiple analog input and output channels at extremely fast
rates, supporting up to 16 channels with speeds as fast as 64 Gsps. This capability is
crucial for real-time, low-latency processing at the network's edge.
2. Innovative Packaging Technology: Intel employs embedded multi-die interconnect
bridge (EMIB) and Advanced Interconnect Bus (AIB) tile technology to combine RF-
capable analog-to-digital converters (ADCs) and digital-to-analog converters (DACs)
with Intel Stratix 10 and Intel Agilex FPGAs. This results in low latency and power
consumption when transitioning between analog and digital domains.
They are particularly well-suited for EW applications, where reducing size, weight, and
power (SWaP) while boosting performance is critical.

B) Microsemi SmartFusion2: is suitable for RF applications as it offers the following:

1. Security Features: The SmartFusion2 family offers security features, making them
suitable for RF applications where data integrity and protection are crucial.
2. Integration with ARM Cortex-M3 Processors: The integration of ARM processors in
the same chip allows for combined digital and RF processing.
3. Mixed-Signal Capabilities: SmartFusion2 FPGAs offer mixed-signal capabilities for a
range of RF system designs.

C) Xilinx RFSoC (Radio Frequency System-on-Chip): why it is suitable: In 2017, Xilinx

introduced the RFSoC, a monolithic device with eight 5 Gs/sec ADCs, eight 9.8 Gs/sec DACs,
Zynq UltraScale+ FPGA fabric, multicore Arm processors, and dual 100 GbE interfaces. This
device is fully qualified as a direct RF FPGA, supporting RF signal frequencies up to 6 GHz and
offering channel synchronous operation for phased arrays.
It is ideal for various applications, including defense applications like radar, countermeasures,
electronic warfare (EW), and communications.

5|Page
D) AMD Xilinx Versal ACAP FPGAs: why it is suitable: AMD's Versal ACAP devices feature a
combination of adaptable FPGA fabric, DSP engines, and AI engines.
These devices offer on-board direct RF ADCs and DACs, making them suitable for RF
applications such as beamforming, radar, FFTs, filters, and more. The integration of direct RF
data converters simplifies the architecture and enhances performance.

6|Page
▪ References:
- FPGAs for Artificial Intelligence: Possibilities, Pros, and Cons | Apriorit
- Can FPGAs Beat GPUs in Accelerating Next-Generation Deep Neural Networks? |
Proceedings of the 2017 ACM/SIGDA International Symposium on Field-Programmable Gate
Arrays
- https://www.esa.int/Enabling_Support/Space_Engineering_Technology/Microelectronics/Th
e_use_of_reprogrammable_FPGAs_in_space
- https://militaryembedded.com/radar-ew/rf-and-microwave/exploiting-direct-rf-
fpgas-for-electronic-warfare
- https://community.intel.com/t5/Blogs/Products-and-Solutions/FPGA/Intel-Delivers-a-
Game-Changing-Analog-Enabled-Direct-RF-FPGA/post/1417363
- Digital Signal Processing with FPGA – Digilent Blog
- How to Use a DSP FPGA in a Real-Time Application - RAYPCB

7|Page