Exploring Embedded ARM Cores in FPGAs Beyond ZYNQ

Soft Cores vs Hard Cores

Embedded processor hard-core FPGAs, also known as SoC FPGAs, integrate hard-core processors into the hardware design from the outset, allowing for pure hardware implementation without consuming FPGA logic resources. The hard-core processor and FPGA logic are largely independent of each other. In simple terms, SoC FPGAs combine an ARM processor and an FPGA chip into a single chip.

For more information, see: Differences Between FPGA Hard-Core and Soft-Core Processors

1. Xilinx’s ZYNQ-7000 Series

The FPGA invented by Xilinx has revolutionized the semiconductor world, establishing the Fabless semiconductor model. The Zynq®-7000 series integrates an ARM Cortex-A9 processor, offering programmability of ARM software and hardware programmability of FPGA. This enables important analytics and hardware acceleration while highly integrating CPU, DSP, ASSP, and mixed-signal functions on a single device.

ZYNQ chip internal block diagram

Technically, ZYNQ cannot be classified as an embedded ARM hard-core FPGA; the official term used is scalable processing platform. The Zynq-7000 scalable processing platform is the latest product series utilizing the same 28nm programmable technology found in Xilinx’s next-generation FPGAs (Artix-7 and Kintex-7). The programmable logic can be configured by the user and interconnected through the “interconnect” module, providing custom logic functions that enhance the performance and functionality of the processing system.

ZYNQ chip resources are mainly divided into two parts: PL and PS, where PL refers to the programmable logic part (FPGA) and PS refers to the programmable system part (ARM processor). These can communicate via a bus. For an embedded software engineer unfamiliar with FPGAs, it can be used like an ARM MPU, utilizing the hardware environment set up in the SDK for development. If during software debugging, certain algorithms are too slow, the FPGA logic can be used to optimize this part, typically achieving speed improvements of ten to twenty times. The switch from FPGA logic development to ARM software development can be done entirely within Xilinx’s own development environment.

The PYNQ series is an upgraded version of ZYNQ, which can be simply understood as: Python + ZYNQ. As for how Python controls hardware,

With the support of Python, FPGA development becomes very convenient, fully utilizing Python’s flexibility and FPGA’s hardware resources, simplifying algorithm design for image processing and artificial intelligence.

2. Altera’s Cyclone V Series

In December 2015, Intel acquired Altera for $16.7 billion. As the second player in the FPGA market, Altera needed to catch up with Xilinx after the launch of ZYNQ! Compared to Xilinx’s SoC FPGA series, Altera has a wider range of embedded ARM hard-core FPGAs, and I have compiled a table here:

Altera® Arria® V GT FPGA development kit, official price $3,995

Overall, it seems that ZYNQ has a better ecosystem with more resources available, whereas Altera’s SoC FPGA documentation is not as plentiful.

3. Microsemi’s SmartFusion Series

In the FPGA domain, the most recognized companies are the two mentioned above. However, both of these mainstream FPGAs are based on SRAM technology, meaning an external SPI Flash chip is required for program storage. This poses a problem: if only program downloading is done without program solidification, it can lead to data loss during power outages, and due to the time required to read and write external devices, power-on cannot start immediately. In contrast, Microsemi’s FPGAs are based on Flash architecture, meaning they have internal Flash for program storage, eliminating the need to distinguish between program downloading and solidification. Data will not be lost during power outages, and they start immediately upon power-on. Microsemi primarily serves markets in medical institutions, military aviation, automotive, and industrial control.

You might not be very familiar with Microsemi, but we must mention its history. Microsemi initially produced power electronic devices, while Actel focused on Flash-based FPGA chips. In 2010, Microsemi acquired Actel, and in 2018, Microchip acquired Microsemi, making Microchip the third largest FPGA vendor today.

2019 FPGA market share | Data source: MRFR

Compared to Xilinx and Altera’s FPGAs with embedded ARM Cortex-A series MPU processors, Microsemi’s FPGAs seem relatively basic, based on a 130nm process. They mainly include two series: SmartFusion and SmartFusion 2, which are first and second generation, both featuring embedded ARM Cortex-M3 hard cores, sharing the same core as STM32. The peripherals are also quite common, such as UART, SPI, IIC, EPROM, RTC, etc. Therefore, this chip is relatively easy to use for those with a microcontroller background. The FPGA and ARM parts are independent and can exchange data via the APB bus. ARM development tools can include common MCU environments like Keil and IAR.

SmartFusion first generation FPGA internal block diagram

Conclusion

It seems that Xilinx’s ZYNQ ecosystem is well-developed, with ample tools, documentation, community support, and related literature available online, while Altera’s materials can be harder to find.

Recommended Reading

• Summary | Selected Articles | Search

• Differences Between FPGA Hard-Core and Soft-Core Processors

• I Plan to Build My Own Satellite

• [Open Source] I Used STM32MP1 to Create an Epidemic Monitoring Platform

• Why Choose Aluminum-Based PCB?