Context:

BCA was contracted by a University’s Neural Research Center to shrink down a stationary rack of computers that perform digital signal processing for real time neural data. A driving requirement of this design was that it needed to be small enough to fit onto the back of a wheel chair.  The new, smaller form factored box would meet the high processing demand by leveraging a Xilinx Zynq chip set. When BCA stepped in the design was in the early stages of research, our engineers have been able to bring this project to a mature design, refining it to be closer to a markable product.

Solution:

The Xilinx Zynq chip set was selected due to its substantial processing power for its relative size, made possible by having a dual core processor and FPGA (Field Programmable Gate Array) on the same chip.  To reduce the form factor BCA designed custom PCBs and integrated them into a heatsinking enclosure that met the Universities’ requirements. The challenge of the project came in the creation of a framework where the processor is running a customized Linux image based on the Petalinux distribution which was integrated with a Vivado project that configured the FPGA’s hardware to perform the real time signal processing and decoding algorithms.  At a high level all of the hardware is described in a Vivado Project with Hardware Description Language and the software is configured and built in Petalinux and the bitbake build system to produce a final image where the parts of the system work together to efficiently accomplish the processing tasks.  The benefit of this technology is the flexibility in tasking the hardware (FPGA) side with high-speed processing and its ability to transfer that massive amount of data to the processor with its internal AXI bus.  The drawback is how complex all these moving parts are to coordinate, but BCA has developed a workflow and process that ties together the FPGA development with the processor coding that allowed us to be successful on this and other projects.