Outlook 2011: FPGAs find new meaning

4 min read

Let's go back 27 years to 1983, the year in which the first programmable logic device (PLD) company was founded. Shortly thereafter, on an average, at least one new PLD company was born every six months.

More than 60 companies dot the PLD genealogy chart today, but when you take a closer look, you'll see that some were absorbed by larger companies and some simply ended, not having enough resources to grow on their own. Just in the past few months, there has been a rise in merger and acquisition activities – as companies are trying to scale and offer much higher value to their customers. In our world, that of the FPGA, Actel was very recently acquired by Microsemi Corporation, long known for being an high performance analogue and high reliability discrete component supplier and now looking to move up the value chain. Clearly, this is a sign of things to come. Meanwhile, the market for PLDs has changed. PLDs are now more universally known as FPGAs – what's in a name, after all? FPGA means field programmable gate arrays and while these devices are field programmable, they are REALLY no longer gate arrays. The technology and usage of FPGAs have evolved. just as the purchasers of FPGAs have evolved. The same engineer who once used FPGAs as an afterthought is now starting their design with FPGAs in mind. As adoption of FPGAs grew rapidly in the 90s, the 'irrational exuberance' that defined that decade was also rampant in the semiconductor world. Business was heady, businessmen were giddy, cost and resources were not an issue and generating revenue was optional. The boom was headed for a bust. As history has proven, though, the meltdown of 2000 was nothing more than a self correction. Three things happened at the same time: common sense returned to business (which had evaporated in the internet bubble days); the world got flatter; and Moore's Law continued to hold true. The first two things made companies work smarter and more efficiently. With China, Taiwan and India's workforces available to help the Western workforce, companies could literally work round the clock and produce more products faster and cheaper. The third effect accelerated the cause of FPGAs FPGA or ASICs? FPGAs used to be a sea of gates and not much else. FPGAs started off as simple glue logic-a programmable, flexible, expensive, hot sea of gates. Primarily, they went into big boxes and were produced in small quantities. The very large (and expensive) ones went into space. As FPGAs fought to get into the heart of high volume designs, they invariably ran into ASICs, whose value proposition was 'exactly what you need, at the performance you want and at the lowest price'. Hiding behind those promises (and realities) were 'long lead times and high NRE (nonrecurring engineering) costs'. Moore's Law gave us smaller and smaller geometries, which became an Achilles' heel for ASICs. The rising tapeout (and retapeout in case of a bug) costs made NREs excessive and overall costs unappetising to management in the post telecom bust (year 2000). Today's mainstream FPGAs cover the whole spectrum of digital sockets and leading edge FPGAs are themselves complete systems on a chip. The smallest FPGAs are very small – they can be as small as 3x3 mm – a lot cooler (2mW), a lot less expensive (less than $1) and can be found in handheld products performing tasks from the mundane to the complicated. The largest FPGAs, meanwhile, have multiple millions of gates, high speed SERDES, processor cores and I/Os, supporting more than a dozen interface standards. Leading edge FPGAs have integrated analogue on the same chip with hard processor cores and hard microcontroller peripherals. It's all about flash From Actel's perspective, some strategic choices resulted in the company being where it is today. First and foremost was the investment in a nonvolatile flash process technology to create instant on, single chip FPGAs. This technology, in turn, provided security and reliability against single event upsets and made the products very attractive to the industrial, aerospace and defence markets. As technology nodes are shrinking, the same reliability is becoming attractive in control planes in the commercial and communications markets. A recent survey of FPGA users shows that the adoption of flash technology is increasing. The applications for FPGAs can be divided into two markets: datapath; and sense and control, where low power, security and integrated analogue (inherent to flash technology) are critical requirements. Datapath applications lie in the heart of telecom infrastructure, or 'the box that makes the internet hum'; like a utility box. Everything that connects to it requires sense and control. Cars, computers, medical devices, industrial controls, airplanes, games, smartphones, home entertainment centres - our entire lifestyle - all run through the internet and are connected, wired or wirelessly, to some communication infrastructure to process, send and receive data. Actel's flash technology is particularly suited to integrating analogue on a single chip for two reasons. First, flash technology is a triple well process that allows the coexistence of digital and analogue circuits on a single piece of silicon, without affecting signal quality and integrity. Second, flash technology is high voltage (30V), which enables erase and program functions for the flash cells; this feature is used to integrate high voltage bipolar voltage monitors needed for system management. Flash technology also enables the integration of a complete microcontroller and associated flash memory for running the program code. These inherent benefits make flash FPGAs ideal for single chip sense and control applications. What's an FPGA in the future? As devices get smaller, the need to integrate technologies is non negotiable. First out the gate were Actel's SmartFusion devices, aptly called intelligent mixed signal FPGAs, which are comprised of a full-featured FPGA, an ARM Cortex-M3 processor, embedded SRAM and flash memory, along with programmable analogue. This is integration at its best and a perfect match for today's sense and control markets. FPGA vendors have long been attempting to grow out of their $4billion box. Actel got on the MCU bandwagon three years earlier with the introduction of Fusion FPGAs, a mixed signal FPGA with an embedded soft ARM Cortex-M1 processor. Today, ARM is threatening Intel's turf and fast becoming the de facto processor in the embedded world. In first six months after the introduction of SmartFusion, more than 5000 designers snatched up evaluation kits, a sure sign that this kind of integration was the way to go. Sure enough, shortly thereafter, other FPGA vendors started announcing similar products, with different flavours of microcontrollers. Witness lately the partnerships announced by Xilinx and ARM, Intel and Altera. While they don't have the capability to embed analogue circuitry on their platforms, it's still apparent that integration is the strategy and moving up the value chain is the objective. The next move is to redefine what an FPGA really is. It's the sum of many valuable system parts – and it could very well be the entire system. Rich Kapusta is VIP of marketing and business development at Actel