Choosing your FPGA development board used to be simple and straightforward. You selected a device for your project, called up your distributor, and a few days and dollars later you were the proud owner of “the kit,” which included the standard development board, a version of the design software, and maybe a reference design or two. Your new board had an FPGA in the middle, some configuration circuitry and perhaps a couple of peripherals, and a few standard connectors along the edges.
As time passed, your demands increased. With the newer generations of FPGAs, you needed more interfaces, more memory options, and more on-board connectivity. If you were in a specialized field, you probably felt you’d reach breadboard nirvana if they’d just perhaps add the “obscurobus” driver and connector (known only by you and two buddies at your competitor’s company), so that your development system could connect seamlessly to all of your proprietary prototyping platforms.
At the same time, the FPGA companies were getting lots of skepticism from wary designers. “Can your device really interface to DDR2 memory at speed – on a board, and not just in a databook?” Intent on proving their products, FPGA companies responded with bigger, more complex development boards.
The problem with the incredible expanding board was the incredible expanding price tag. The more options, interfaces, connectors and peripherals that went on the board, the more expensive, expansive and complicated the systems became. The bloated board became jack-of-all-trades, master of none, and suppliers started to look for a better way to meet the conflicting needs of demonstration, evaluation, development, prototyping, and even production use.
Today, there are a daunting number of options available in the “useful boards with FPGAs on them” category. Before you slap down the corporate Visa, it pays to know what the choices are and to match your board carefully to your needs so you and your boss aren’t disappointed with the results.
Before you’ve even bought into the idea of using programmable logic in your design, the vendors need to offer specialized boards to win you over. Designers are a skeptical lot, and they’ve been burned by the datasheet many times in the past. There’s only so much they’re willing to take on good faith, and the health and welfare of their career isn’t part of that. “Electrical engineers give only a certain amount of credibility to waveforms in a software simulator,” says Richard Terrill, senior manager of Spartan solutions for Xilinx. “but if you show them waveforms on a scope attached to a working board, they feel comfortable.”
In order to achieve that comfort factor for some of the more challenging interfaces and applications, FPGA companies have created elaborate demonstration hardware to prove that their silicon is capable of doing what the datasheet says. Don’t believe you could get your favorite low-end FPGA to work at speed with DIMM-based DDR2 DRAM? Just walk up to the table at a trade show and watch’er go. Burn your finger on a couple of heat sinks. Twiddle the knobs on the scope to be sure the whole thing is real. Now you’re convinced. Skeptical about claims of umpteen Gigahertz serial communication over a backplane? Step right up and see it work. Feast your eyeballs on the eye diagram and become a true believer. Now go home and break out the corporate wallet.
Beyond proof of performance, however, these boards have little utility. Once you know that a particular interface works on their board, you figure you can probably get it to work on yours. The demo board itself isn’t required for your project. Its life is best spent on the road with applications engineers, starring in show booths, then slipping away in a circuit board limo for its next celebrity engagement.
Once you’re pretty sure that an FPGA is in your future, you’ll want to make sure that you’re also up to the task. This is where the evaluation board comes in handy. Once they’ve proven to you that their chips can do the job, FPGA companies have to prove to you that you can do it too. If you’re an old hand at programmable logic design, you probably don’t give this a second thought, but if you’re a newcomer, the idea of starting with a blank chip, an empty vi window, and a board with lots of protruding wires and connectors can be a little intimidating.
The idea of the evaluation board, then, is to make FPGA design as easy and accessible as possible. A typical evaluation kit includes a bare-bones board, stripped down to the lowest possible cost and sold with little (or even negative) profit margin, plus a starter set of software tools and some canned reference designs that will have your LEDs blinking in nothing flat.
If you can pay a paltry sum for a system that will let you ease your way into successful FPGA development, all the while watching the results of your initial efforts in glorious hardware hello-world equivalents like marching ants, time and temperature displays, and square-wave speaker tricks, you’re much more likely to jump with confidence into a production deployment of programmable logic. While these boards may not be the ideal platform for real-world engineering development, they get you up and running at minimal cost with maximum confidence. For the FPGA companies, that means a new customer and a great return on their development investment.
Every FPGA vendor offers some variation on the eval-board theme and reports similar success in product proliferation as a result. “If we just sent out sample chips, they’d end up in a drawer somewhere,” continues Xilinx’s Terrill. “By putting out very low-cost boards, we get people to actually try out the devices. We even once inserted a small Spartan sample board into a Japanese technology magazine. The issue sold out.”
“We felt Actel needed an introductory board that allowed people to become familiar with our technology, including silicon, tools, IP, and design flow,” says Saloni Howard-Sarin, director of antifuse products and tools marketing at Actel. “In 2003, we introduced our starter board for our Axcelerator family, and since then we’ve blown past every forecast in demand. These kits are intended as one-stop boxes with a programmer, our Libero design software, and a board. We’ve now followed that success with similar starter kits for ProASIC Plus and ProASIC-3.”
If you’re past the proof and evaluation stage, you’ll be most interested in a board that lets you do your day-to-day development work with maximum effectiveness. Once we reach this stage, our requirements change somewhat. With development boards, we are much less sensitive to cost, because a day or two of engineering time getting around a problem with a cheap board costs far more than the typical upgrade to industrial-strength development hardware.
The keys to success on a development board are rich connectivity so you can hook up to the other parts of your system and the peripherals you need, and enough quality so that the problems you encounter will be with your design rather than your development board. At this stage also, boards start to become more specialized and application-specific. The reason for this is simple. Building a board with every possible proprietary peripheral, interface, and connector would be prohibitively expensive, and no single design team would use more than about 10% of the capabilities. It pays to build more focused hardware that better addresses specific markets.
Altera offers domain-specific development boards in four key areas, including DSP-focused boards, embedded systems development boards, telecom boards, and PCI cards. In the embedded systems world, a development team is typically made up of both hardware and software developers, with the software developers outnumbering the hardware folks. The software developers require only a small subset of the capability that a hardware engineer needs, typically getting by with a “chip holder” board that has an FPGA and peripherals accessible for PC-based development and debugging of embedded applications.
Xilinx sees a similar trend in embedded developers approaching FPGAs. “The embedded software developer looking at our MicroBlaze processor cares about algorithms, operating systems, and drivers, not about soldering irons,” continues Xilinx’s Terrill. “For this engineer, Gerber is a type of baby food, not a plot, and we need to be sure we make it easy for them to get their job done with our products.” For the hardware developer, though, the workings of the board are of critical importance. Most FPGA companies also supply schematics with their development boards so design teams can replicate the board environment as closely as possible in their final product.
As boards become extremely application-focused, however, FPGA vendors embrace the involvement of third parties like distributors and OEMs to cover the territory. “A hardware engineer in the automotive industry might require a multi-media display, instrument panel, door and window controls, environmental monitors, and a host of specialized hardware,” says Ian Land, Senior Manager of IP Solutions Product Marketing at Actel. “If they’re trying to replace ten microcontrollers with an FPGA, it helps if they have a board like E2CAD produces, with stepper motors, LCD display, and a box that connects to all of the I/Os.”
While it might not make sense for an FPGA company to invest in development of specialized boards with small target audiences, OEMs and consulting companies can make a profitable business by focusing on specific high-value design problems. Distributors like Avnet and Memec have found value in the board development market as well. By supplying development boards that supplement those from the FPGA vendor, they can offer more capability for specific applications and keep customers aligned with the full range of components available on their line card.
The distributors have also found success with Lego-like plug-in expansion modules that allow generic boards to be customized for specific application areas. “The last 3 years, we’ve started to include capabilities like our P160 and P240 expansion slots to add application-specific features to general purpose boards,” says Jim Beneke, director of technical marketing at Memec/Insight. “Taking that a step farther, we are now doing more with overall applications, such as our starter kit for Linux, which includes tutorials, how-to guides, and reference designs. This is stuff that might take two or three months to do from scratch, but a customer can buy a kit from us that has everything they need including hardware.”
Not wanting to duplicate efforts or compete with themselves, FPGA vendors and distributors work to make sure that their offerings complement each other rather than overlap. “With our Spartan-3 and Spartan-3E boards, we worked to make sure that they fit in with our distributors’ portfolios,” says Xilinx’s Richard Terrill. “In each case we worked closely with the distributor to make sure their line card was well supported. This provides the customer with risk reduction, which is key. If they’ve seen a particular regulator, for example, that works well with a Spartan FPGA on the development board, they can use that same part in their design from their distributor and trust it.”
If you’re designing with high-end FPGAs, expect your development hardware to cost more as well. Creating a board that maintains signal integrity with gigabit serial I/O, or has high-quality D/A and A/D for DSP applications, or just works reliably with huge pin-count packages is a difficult and expensive proposition. It doesn’t pay to cheap-out on the development hardware and spend weeks debugging your design only to find out that the problem was in the board all along.
Prototyping to Production
Sometimes development teams want to do more than develop. If they have less time than money, they might want to find a pre-made FPGA board that can drop right into their product for prototyping, pre-production, or even final production for low-volume designs. This is particularly true in industries like industrial control where a reliable pre-fab module with an FPGA and a few other useful components can solve a wide variety of problems with identical hardware. For FPGA companies, this is both good and bad news. The good news is that customers that might never have gone directly to an FPGA supplier may be designing FPGA products into new systems. The bad news is that supporting a second-tier customer through a third party with little visibility into their design challenges is a difficult task.
Traquair Data Systems, Inc., a distributor of Orsys compact multi-chip modules, sees a key role for domain-specific products in transitioning customers to FPGA technology. The modules, which contain both a Xilinx FPGA and a TI DSP processor coupled to a firewire connection, allow signal-processing designers to tackle a wide range of challenges with the same hardware. “Some customers start out using just the DSP and the firewire connection, then gradually take advantage of the FPGA for acceleration and improved performance,” says Stephen Bradshaw, President of Traquair. “Others already understand the capability of the FPGA and use the DSP just to manage lower level tasks like the firewire interface. Part of our job is to help people make that transition.”
With such ready-made modules, there are considerations that go far beyond those associated with a development board. Once again, cost becomes an important factor because you aren’t just buying quantity one or two for your team. On the other hand, reliability, temperature range, vibration tolerance, and other quality considerations become important as well, as production use typically is much less friendly than the engineer’s desktop.
There’s also a cost-reduction angle to production modules. If the complicated board design surrounding the FPGA and closely connected peripherals, such as memory, configuration circuitry, and high-speed interfaces, can be localized to a pre-made but high-end (often 12-layer) PCB or module, the rest of the design can be implemented in a garden-variety simple 4-layer PCB. The burden of complex board design can be borne by the supplier, while the systems house is left with a much more manageable and less expensive integration task.
As FPGA applications become more diverse, we are bound to see a corresponding increase in the diversity of boards. From wooing new business and supporting development projects all the way to providing reliable, ready-for-production hardware based on FPGAs, suppliers are looking to leverage the flexibility benefits of programmable logic technology to get their customers’ designs out the door faster, easier, and at lower overall cost.
|For an excellent list of currently available development boards, go to: http://www.fpga-faq.com/FPGA_Boards.shtml|