Achronix Semiconductor Corporation

The unprecedented performance standard set by the Speedster^® family – based on its patented picoPIPE™ acceleration technology – is a huge advantage in signal processing applications. Such applications work with analog signals:

• Digitizing input signals
• Performing complex calculations on them
• Converting back to analog for transmission

Key examples of applications using digital signal processing are:

• Wireless data networking (WiFi)
• Cell-based voice/data communication (GSM, CDMA, WiMax, LTE)
• Radar systems and military electronics
• Signals intelligence and Measurement and Signature Intelligence (MASINT)
• Image processing (high-end projection systems, real time mapping, targeting systems, etc.)

Specific types of processing used include:

• Signal correlation (target ranging and identification, radar pulse identification, RF communications identification, etc.)
• Fourier transformation (channelization in WiMax/LTE, etc.)
• Cosine transformation (e.g., Video Compression, etc.)
• Voice coding (voice compression & decompression)
• Image manipulation (conversion, scaling, keystoning, rotation, etc.)
• Signal synthesis (Direct Digital Synthesis (DDS), etc.)
• Filtering (FIR, IIR, etc.)

These are all computationally intensive tasks, and historically have often been addressed by the use of DSPs – processors executing sequences of instructions from a stored program. Such sequential processing cannot match the performance of hardware-based processing, especially that of the 1.5 GHz Speedster FPGA family. The Speedster advantage is even more pronounced in high-end, high-throughput systems in which multiple signals are being processed simultaneously. But fast logic is only part of the picture. Speedster devices offer a fully balanced approach, with:

• Fast reprogrammable logic (1.5 GHz)
• Fast embedded memory (850 MHz)
• Fast multipliers (1.5 GHz)
• Fast I/Os (up to 1.066 Gbps)
• Fast SerDes (up to 10.375 Gbps)

A common feature of the computational processing outlined above is the need for high-speed multiplication of streaming data. The multiply-accumulate operation is used in almost all of these cases. Matrix operations, used in image processing and elsewhere, place huge demands on multiply resources. The 1.5 GHz throughput of the 18x18 Speedster multipliers – matching the other fabric components – results in an unbeatable performance. For applications needing lower resolution, each multiplier can be used in split mode – as two independent 9x9 multipliers, doubling multiply capacity.

The message is clear. For high-throughput, high-performance signal processing applications, Speedster provides the best reprogrammable platform for system development.