[sdiy] Some Audio DSP prototypes

Brian Willoughby brianw at audiobanshee.com
Sat Apr 16 03:21:15 CEST 2022


If supply chain issues are making it difficult to obtain the most popular, general purpose chips - ARM - then it seems like a good time to consider designing with chips that have special purpose instructions sets intended for signal processing. DSP families like the TMS320 and SHARC will process more audio per Watt than an ARM - even an ARM that has one or two token "DSP" instructions.

I've designed with the TMS320VC5506, and it has an internal bus that's 80 bits wide, plus DMA and SRAM allowing parallel access of multiple data sets. There are floating point models in the family. They're not as easy to program as a general purpose CPU, but the effort of learning their special tricks is well worth it.

A few hundred MHz on a DSP architecture is more than enough to compete with general purpose GHz CPUs, and then you don't have to source expensive ARM-A chips.

Brian Willoughby


On Apr 14, 2022, at 16:05, Eric Brombaugh wrote:
> Hey folks,
> 
> A while back there were some lengthy threads that veered off into discussions of what to do about audio DSP in these times of MCU shortages. I mentioned some projects with RP2040 and ESP32 that I've been puttering around with for the last few months. I've gotten them to a presentable stage now and have published them on my github here:
> 
> https://github.com/emeb/RP2040_Audio
> 
> https://github.com/emeb/ESP32S2_Audio
> 
> You'll notice a lot of similarities - that's intentional as I wanted to be as much apples:apples as possible and it's allowed me to evaluate them with very similar use-cases.
> 
> Overall I prefer the ESP32 for the faster speed and PSRAM, but both systems are usable and not difficult to work with. Definitely worth considering if you're casting around for alternatives for the high-powered ARMs that we can't get these days.
> 
> Eric




More information about the Synth-diy mailing list