DSP Projects

We were fairly late converts to DSP devices. The first was as an external supervisor to a M.Tech student for his thesis
B.W.L.R.L. Paul (1997 Jan), Signal Analysis and Synthesis in Real-Time Data Processing. Basil developed hardware using an Analog Devices ADSP-21xx processor (cannot recall the exact part number). For testing, several filters were written with comparisons done between input and output on a dual trace oscilloscope. Our actual work on the ADSP-21xx was very limited — we did configure the tools and compile some demo code for an evaluation board, but had little interest as we were running RISC devices at ten times the frequency. We also had no work prospects in the Port Elizabeth area for DSP development.

Analog Devices

In 2001, the second Analog Devices DSP was the SHARC, which was used as a communications controller by a company that used DSPs for everything, even if unsuitable. The SHARC was a 32-bit only device, and to work with ARCnet or UART chips, plenty of shifting, masking and OR’ing was required. The tools were upgraded between milestones, which was a serious error.

Texas Instruments

The Texas Instruments range included the TMS320C6201 integer VLIW DSP, and the TMS320C6701, which was the floating point equivalent, however, most of the imaging evaluation was on the PCI board with the 6201. The above processors ran at 133 or 200 MHz, with a jumper to select the different crystals. A short while later, Texas Instruments would bring out a very low-cost 6211 evaluation board. The work other than the 6211 was as a contractor in 2000 and 2001, but the 6211 was for performance evaluation for private projects.

Dual-core ARM-DSP

In 2009, becalmed in the middle of the eye of the global financial storm, we bought a Spectrum Digital/ Texas Instruments OMAP-L137 board to evaluate the ARM, however, the DSP came up first and we ended up writing for the 674xx compatible VLIW DSP. We did not test any DSP algorithms, but programmed the on-board and on-chip peripherals in C. From the software view of the board (without interrupts enabled), the code was no different to that of any RISC. The processor can access bytes on any boundary. We are not sure if the promised Linux port ever shipped for the L137 board, but two years on, it is “shelfware” as newer devices are just so much more capable — the technology treadwheel never stops!