Fusion Adds ARM
It slices, it dices, and it has flash-memory. It gets your whites whiter, removes those collar rings, and flaunts flexible FPGA fabric. It has more flavor, is less filling, and boasts programmable analog to boot. It softens your hands while you do the dishes and offers a wide selection of pre-configured IP. Now how much would you pay? Well don’t answer that because now, Actel’s already incredible versatile Fusion programmable system chips also pack an available ARM7 processor.
Actel’s claims about Fusion sound like something from one of those impossibly overblown daytime soap opera commercials. Fusion is an incredibly versatile, in-system reprogrammable system-on-a-chip platform that incorporates programmable analog, custom digital logic fabric, embedded flash and SRAM, and versatile IO options. Now, just when the embedded systems equivalent of the Swiss army knife was almost too big to fit in the pocket of your Gore-Tex pants, Actel has added the option of an embedded configurable 32-bit ARM7 core and peripherals.
Just about the only thing you’ll need to fuse to your fusion device to build whatever it is you’re building is some commodity-grade DRAM. After that, you can probably just start shipping boards while your marketing department decides whether your product will be a motor controller, an MP3 player, or a digital multi-meter. OK, that might be a stretch. But the point is, Fusion will let you design just about anything on a single chip with the flexibility to reprogram it in the field.
For the risk-averse system designer, the first-to-market innovator, the we’re-not-quite-sure-what-we’re-making marketer, or the (oh, I’m so jealous, why didn’t we have these when I was in college?) precocious university engineering student, an ARM-enabled Fusion device on a development board will let you light up some serious LEDs with your idea faster than just about any other approach.
Sure, Fusion is a serious semiconductor product serving embedded systems designers in the consumer, automotive, industrial, and communications markets, but just reading the datasheet makes any red-blooded engineer want to smuggle one home for that secret personal robot project hidden in the back of the garage. Whether your thing is software, hardware, or (hopefully these days) both, you’ll probably be able to teach Fusion to do what you want. Actel’s approach to Fusion’s development environment is inspired by the knowledge that probably not everybody is an expert analog, digital, and software engineer.
Before we delve into the design environment, we should look at the particulars of the new product. Fusion’s heritage is Actel’s highly successful ProASIC flash-based FPGA family. ProASIC (now in it’s third incarnation with ProASIC3) differs from the conventional SRAM-based FPGAs offered by other companies in that its basic logic cell is based on non-volatile flash memory rather than SRAM. Flash-based FPGAs have several unique properties. Since they’re non-volatile, they’re live at power-up and don’t require the intricate configuration dance and external configuration circuitry of other FPGAs. They also tend to sip less power and boast better security than their better- known cousins. They are also more resistant to radiation-induced single-event upsets that can plague normal FPGAs even at ground level.
The flip-side of those benefits is typically lower density and performance when compared to an SRAM FPGA of the same process geometry. Compounding that disadvantage, flash is harder to implement on the process side, and so flash-based FPGAs usually lag at least one process generation behind SRAM. If not for these shortcomings, flash would have probably taken over the FPGA world long ago. Despite the disadvantages, flash’s special properties have made it the go-to choice for many applications where raw performance and density are not the primary considerations.
Last year, Actel announced it was cleverly capitalizing on some of the special characteristics of ProASIC3 (like its wide power supply voltage rails) by adding programmable analog, thus creating the Fusion family. While Fusion brought unparalleled flexibility to the FPGA-based system design game, it wasn’t difficult to predict the company’s next move. At about the same time Actel introduced Fusion, they also announced a cooperative agreement with industry-leader ARM to license the super-popular ARM7 core for a special version of ProASIC3. It was obviously only a matter of time before the ARM7 core found its way into Fusion.
Fusion posed a problem for Actel’s design support team, however. When you reach the level of integration and convergence where you’ve equipped a low-cost semiconductor device with processors, peripherals, embedded software, analog, memory, and IO, how do you find the renaissance engineer with the breadth to design something with it? Since the industry long ago burst through the terminology bounds of simple “design tools,” Actel reasoned that a product as ambitious as Fusion clearly required an “Ecosystem.” That’s right, a design flow isn’t enough. We’ve gotta have digital and analog predators and prey, an IP nitrogen cycle, and system-design symbiosis to boot the bios on today’s programmable embedded system chip.
OK, maybe we’re making fun of their marketing a bit, but it is true that Actel has (by necessity) gone far, far beyond simple simulation, synthesis, and place and route in their Fusion design environment. Actel claims that, in addition to traditional FPGA development tools, Fusion’s terrarium already includes almost 100 IP cores, app notes, generators for specialized Fusion blocks, industry-standard compilers for the ARM core including RealView, Keil, IAR and GNU, available RTOS including uCLinux, Nucleus, and Integrity, APIs and drivers including Jungo and GAO Research, debuggers including RealView, GDB, and IAR, and a selection of software development environments. You get the idea. The tools and support required to design effectively expand exponentially with the versatility of the platform.
The ARM7-enabled version of Fusion is designated M7. The licensing deal with ARM is such that you need only to purchase an ARM-enabled device from Actel, and you get the rights to use the ARM core on that device – no complicated multi-party IP licensing issues involved. M7 Fusion boasts from 600K to 1.5M system gates, ~13K-34K bytes of on-chip SRAM, 512K-1M bytes of Flash, analog inputs at +/- 12V, and analog-to-digital conversion at up to 600K samples per second. The ARM7 is a 32-bit soft processor (implemented in the FPGA fabric of the Fusion device) that is user-configurable and optimized in size/speed for the Fusion family.
Actel envisions that the M7 Fusion devices will be adopted in applications as diverse as home networking, multimedia, automotive, motor control, robotics (Of course! Maybe now my bionic ferret project can finally get off the ground.) telecom and networking cards, and handheld radios. The price point of Fusion makes it an attractive option for a wide variety of applications, and the small footprint, security, and meager power budget keep it viable even for projects with fairly specialized requirements.
Fusion samples will be available in April 2006, with production slated for Q3. Software tools will be available in April, so you’ll be able to get started on your masterpiece right away. Development boards won’t be around until Q3, though, so you’ll be forced to experiment with Fusion in the virtual world for awhile.
In reality, Actel probably has no idea what kind of applications will find their way into Fusion. When you create a technology with so many options, it is typically impossible to predict what people will think of to do with it. Of course, that’s part of the fun.