[sdiy] The Owl: ARM fx pedal

[rsdio at sounds.wa.com]

On Apr 24, 2013, at 12:22, Eric Brombaugh wrote:
> On 04/24/2013 11:57 AM, Martin Klang wrote:
>> I've come across this presentation paper just now:
>> http://www.arm.com/files/pdf/dspconceptsm4presentation.pdf
>>
>> It uses an example of a 7-band EQ (designed in Audio Weaver!).
>> As compared to the M3 requiring 1291 cycles, the M4 performs this  
>> in 299 cycles.
>>
>> Perhaps as an illustration of chip capabilities, this would be  
>> more useful.
>> We really just want to communicate an idea of what the chip is  
>> capable of, in terms of DSP, rather than give any exact benchmarks.
>>
>> In the paper they also show optimised FIR code which computes in  
>> 1.6 cycles/sample, as opposed to 12 unoptimised, but I think it  
>> would be misleading to present that as an achievable result in the  
>> general case.
>
> I think that paper speaks for itself and not in the way that the  
> authors intended - the complexity of hand-optimizing the ARM DSP  
> code vs the simplicity of the traditional DSP code to which they  
> are comparing themselves.
>
> That said, I'm by no means denigrating the capabilities of the  
> STM32F405 MCU - I've used it in a number of audio projects so far  
> and it is by far the best trade-off of cost, ease of use and  
> features that I've come across to date. It is quite capable of  
> handling a wide variety of audio processing tasks without requiring  
> laborious optimization, and with optimization you could likely take  
> it somewhat further. Just be aware that you're not likely to get  
> the same sort of performance from the Cortex M4F that you would  
> from something like a Blackfin or TMS320-C64x.

Good points.

I recently designed with the TMS320VC5506. It costs the same as the  
STM32F405 and supports USB on-chip as well as parallel, I2C, SPI and  
I2S. The C5000 series is Texas Instruments' Ultra Low Power DSP and  
the efficiency of optimization certainly means it should use less  
power to get the same tasks done. Including 48 channels of audio  
(which require a lot of analog chips), my last C5506 product used  
only 250 mA from the USB power.

In my opinion, if the primary raison d'etre for your processor is to  
handle audio, then you should start by selecting a full DSP, even if  
that means the high-level code is less efficient. The bulk of the  
cycles will be spent on audio, with very few cycles reading knob  
settings, handling MIDI, or flashing an LED. ARM is great for general  
purpose computing, but don't let it's general popularity overshadow  
the fact that other processor designs are way more suited for audio,  
especially in a battery-powered setting.

SHARC is certainly a contender, but I have less experience with ultra  
low power potential with that chip family.

Brian Willoughby
Sound Consulting