Instant on and ready to go can be a selling point when choosing an FPGA. A part that powers up ready to roll has some advantages over statically configured FPGA's that need to be loaded and initialized upon powerup.
Instant on parts eliminate the extra PC board space required for a dedicated configurator IC. Similarly, instant on, won't stall a microprocessor/microcontrollers ability to start working while it waits for an FPGA to initialize.
But, there are advantages to static parts as well that must be taken into consideration. Design updates are easier to accommodate for one thing. Improved code can be put into the configurator memory chip with virtually no NRE or respin costs.
Key is the ability for an FPGA to change personalities on the fly more or less. True, Flash based FPGA's can be changed on the fly in some cases, but, it's a slower more tedious process compared to RAM based FPGA's.
It may not be intuitively obvious how a dynamically and on the fly configured part serves any real advantage, so let's look at a couple of examples that illustrate this potentially useful ability.
Take a protocol analyzer that sniffs a communications line that may carry different protocols in a packet based environment. A simple logic block in an FPGA can identify the specific protocol being sniffed at, then load it's internal logic array to decode and handle that packet in that protocol.
This is infeasible in a Flash world because the part will be constantly rewritten which is not a good thing for Flash. Also, the Flash write timing may make this too slow for any real and practical application.
But, a fast RAM based part can reconfigure quickly enough to make this feasible. Also, because not every protocol handler need be resident all the time, a lot of savings can be realized by using a smaller, cheaper part.
Another example is a piece of test equipment where the signal processing element is dynamically configured to perform the functions selected by the front panel. The signal stream from the D/A can pass through the dynamically configured FPGA to perform dynamic filters, spectral analysis, peak detection, average and RMS calculations, etc.
Bottom line, there is a need and a place for dynamically configured FPGA's, and, configurator memory chips are often key to this ability. Because of this, configurator IC's have gotten denser to match the high 'gate' counts of modern FPGA's. They have also gotten faster to pump data more quickly into the target parts.
Key is that they have also been getting smaller. There is already a PC board real estate penalty when choosing to use an FPGA configurator memory chip, so, manufactures want to ease your pain.
With all these factors in play, it prudent to be aware of advances that can help. One is the next generation FPGA configurator memory chip from Atmel.
Touted as the the world's only 16-Mbit, In-System Programmable (ISP) FPGA configuration IC in a small footprint (36 mm2) 8-pin Leadless Array Package (LAP), the new AT17F16 is the latest and largest FPGA configurator from Atmel's family.
It pumps data at up to 33 Mbit/s and has a page select pin options for storing different versions of the FPGA codes. I think they understand.
A high speed 2 wire protocol is used so it doesn't need to burn up FPGA I/O and resources too badly. It is also in system programmable which opens the door for other benefits like production line testing code vs run time customer code.
In system programming can be accomplished using Atmel's ATDH2200E programming kit or industry standard programmers. The 3.3V parts take direct aim at SRAM-based FPGAs, such as those from Xilinx and Altera.
Atmel's 16 Mbit AT17F16 configuration memories are available in the 36 mm2 8 pin Lap, a 20PLCC, a 44PLCC, and a 44TQFP package.