IDT 574 June 2001 Design

IDT574 to Altera Ep1K100 interface

Routing trial for pin assignments for the IDT574 and Altera’s EP1K100 FPGA.

The June design never made it out of OrCAD.

IDT574 CPU with Altera FPGA placement

A Stellenbosch company specialising in radar for measuring projectile speed (muzzle velocity) were interested in measuring the speed of golf balls, tennis serving and cricket bowling velocities. We started in April, 2001, commuting between Port Elizabeth and Cape Town. The first task was to propose an architecture for future products. The architecture selection, based on prior work, was presented in May. The choices were PowerPC, MIPS and Texas Instruments 62xx DSPs. ARM was nowhere for throughput in 2001. MIPS was chosen for multi-core devices, 64-bit architecture, clock rates well over 500 MHz, plus we had free compilers.

One of the design proposals was to handle 500 frames per second of video with a 1024×1024 resolution. The IDT574 cannot handle 500 MBytes a second but we were starting a design for replacing several smaller devices used in an underground coal mine radar application. The paper design was completed in June (FPGA schematic entry for interrupt controller, boot control, Flash programming, Flash data width conversion, proposals for the free real-time kernel, memory configuration, etc.). The rats nest shows the connections and the initial placement. Note that there was no Ethernet, no video input or graphics output.

Requirements for image processing of nine LED probes for neurosurgery were added to the target applications; changing the design. The neurosurgery application was lower data rate, but would require tight integration into external systems. Ethernet was essential. For local debugging or verification, graphics would be necessary as well. Typical imaging debug problems are lack of visual output of the raw image (what does the camera see if anything, is it focused?), partial results, delay through the processing stages measured in frames or milliseconds (which affects actuators), and a long list of others.

Other projects run in parallel were to decypher a CRC algorithm for messaging between some Joy mining equipment (one of the other engineers cracked the code), graphic output of radar waterfall diagrams for FFTs of tennis ball data to debug intermediate results (done on Linux with Motif/ X11 graphics).