[sdiy] In praise of the ATM STM32F303

[Eric Brombaugh]

I'm fairly certain that it's mostly down to the crosstalk from digital 
circuits on the same die. The F4 parts errata sheets specifically 
mention that the flash cache system interferes with the ADC and they 
have a number of workarounds listed which unfortunately don't really 
seem to help much. The best approach with the F4 ADCs is to do a lot of 
filtering and hysteresis / guard-banding of the digitized values. This 
is fine for slowly varying CVs and pots, but does tend to limit their 
application for audio-rate modulation.

While the F3 CPU which runs slower requires a less aggressive cache 
system and hence has less on-chip interference, there may be other 
circuit-level differences in the analog section - on the F303 there is a 
high-speed SAR ADC which performs pretty well and as Olivier mentioned, 
the F373 SD ADCs exhibit very good ENOB. I've got an F373 prototype 
running that does stereo audio input/output without an external codec 
and sounds quite good.

Note that I've done some research on the F302 parts and the ADC in that 
family seems to have issues that I've not seen on other F3xx parts. I 
suspect that there may be some compromises on those lower-cost members 
of the F3 series.

Eric

On 02/13/2016 04:05 AM, Simon Brouwer wrote:
> Hi Eric,
>
> Any idea whether that noise is due to worse (differential) linearity or to worse
> crosstalk from digital circuits?
>
> I am starting out with the STM32F446 for a one-off semimodular monophonic synth,
> and was thinking of using one of its built-in DACs for VCO control voltage. My
> plan is to deal with its nonlinearities by calibrating its output voltage at
> each semitone, then using a lookup table with correction values.
> But if the DAC output voltage contains LSB's of uncorrelated noise, I may be
> better off using an external SPI DAC.
>
> Best regards
> Simon
>
>> Op 13 februari 2016 om 4:37 schreef Eric Brombaugh <ebrombaugh1 at cox.net>:
>>
>>
>> A couple guesses why:
>>
>> 1) for high-volume production such as Waldorf usually does, every cent makes a
>> difference. In large quantity F3 parts like the 303 are a few bucks cheaper
>> than F4 parts, so that adds up.
>>
>> 2) I've used both F3 and F4 parts in products so I've got some experience with
>> them. The F3 parts analog sections are considerably higher quality than the
>> F4. Noise on F3 ADCs and DACs are a few lsbs lower amplitude than F4 parts.
>>
>> Eric