As field programmable gate arrays (FPGAs) diversify into new markets and become easier to develop with, the broader embedded community should understand the basics of the processing technology. Introduced to the market in 1985 by Xilinix, FPGAs first offered a flexible alternative to application specific integrated circuits (ASICs). The main distinguishing factor of FGPAs is that they are field-programmable which means they can be programmed by the end user to meet specific needs (during prototyping, post-deployment, etc.). This characteristic leads to a host of benefits such as reusability (sans antifuse types), high-performance parallel processing, and shorter time to market. FPGAs have changed significantly since their inception; the first FPGA had only around 8,000 gates, but FPGAs today have millions of gates which has significantly increased processing capacity and speed while decreasing product cost and energy consumption.
What actually is an FPGA? In short, an FPGA is a programmable silicon chip that is a type of programmable logic device (PLD). Generally, FPGAs have three main components: configurable logic blocks (CLBs), configurable input/output blocks (IOBs), and programmable interconnects. The CLBs perform logic functions, the IOBs connect internal components to external design, and the programmable interconnects link the CLBs and IOBs. FPGAs differ by the type of programming technology implemented: SRAM-based, Flash programming, and Antifuse technology. SRAM-based FPGAs cannot store data without a power source so they need external memory, but are the most widely-used type of FPGA because of their reprogrammability and the historical challenges with shrinking flash memory cells. Flash-based FPGAs can keep the configuration pattern on the chip (in non-volatile memory) without a power source which shortens power-up time. Antifuse-based FPGAs are one-time programmable due to burning antifuses to conduct currents which cannot be reversed.
A common way to enhance the capabilities of FPGAs is to embed a processor inside the FPGA structure. Embedding a processor in an FPGA offers advantages over a typical microprocessor solution such as fewer discrete components which can reduce costs, decreased microprocessor obsolescence, and customization. Embedded microprocessor life spans are decreasing due to market needs for greater performance combined with constant competitive pressures. With a soft processor core embedded in a FPGA, for example, the core is owned and can be reused with new FPGA hardware which helps prolong the utility and life of processor solutions.
Applications for FPGAs are very diverse and span across many industries. Where FPGAs see a lot of use is in hardware prototyping. Prototypes of ASICs and system-on-chips (SoCs) are often made using FPGAs. Prototyping with FPGAs provides a way to check functionality before producing the final product. A leading design application for FPGAs is for digital signal processing. FPGAs offer hardware reconfigurability and throughput advantages over most traditional discrete DSP processors. FPGAs are even found in embedded systems beyond our world. For example, radiation-tolerant SRAM-based FPGAs provide reprogrammable processing hardware for space applications. FPGAs were previously not as suitable for space because radiation would alter configurations, but newer radiation-tolerant FPGAs are less susceptible to the environment’s unique ruggedization requirements while enabling engineers to make hardware optimizations or improvements even after the system is deployed in space.
As FPGAs continue to improve in performance and ease-of-use, their applications and capabilities will continue to expand. With Intel’s acquisition of Altera, the market can expect to see additional SoC and FPGA integrations that will further FPGAs’ workload processing abilities and market addressability. Companies like Microsoft and Intel have been experimenting with using FPGAs for machine learning while others look to leverage the technology for earlier and faster testing/prototyping. Other applications such as genomic sequence analysis are also generating steam. Though FPGAs have been around for more than 30 years now, the technology is still evolving and carving itself more of the embedded market opportunity which will continue to attract more attention.
By Jamie Traverso, Research Associate
View the 2017 IoT & Embedded Technology Research Outline to learn more.