[sdiy] Digital pots as the gain element in a filter

[Tom Wiltshire]

On 5 Jul 2011, at 18:37, David G Dixon wrote:

>> It's more likely to be a simple analog pot sent through an 
>> ADC on the Gaia's MCU/DSP - and the rest of the sound 
>> production (including the
>> filter) purely in software. The Gaia doesn't have analog 
>> filters (digitally controlled or not).
>> 
>> I would guess that the stepping is caused by the bad 
>> resolution of the ADC - forcing the designers to represent 
>> the pot position on 8 bits - rather than something in the 
>> filter code. Unless filter coefficients are stored in lookup 
>> tables with a coarse step size, but today there's plenty of 
>> computing power to not have to do that...
> 
> Yes, you are undoubtedly correct.  However, given that the parameter which
> the pot controls only has 8-bit resolution (for whatever reason), this is
> functionally equivalent to controlling cutoff with an 8-bit digital pot, is
> it not?  Hence, the observation that such control would be pretty crappy is
> still valid, I think.

Quite!

8-bits isn't enough by most measures, IMHO. The Rockit design uses the resistor-from-wiper-to-ground trick to make the pot response approximate a log curve, but that's only going to improve things, not fix the problem completely.

I noticed that he didn't use any hi-res slow filter sweeps in his demo. It would have showed the worst case scenario. The envelope didn't sound too bad though, but I guess it's fast enough you don't really notice. A slow decay would have been a killer too.

All in all, it's an odd design. Since it uses 3 digipots and about 8 op-amps, he could have used a couple of LM13700s and a MCP4822 dual channel DAC and had a lower chip count with less of the disadvantages. But that's been done, I suppose. And maybe running on +5V was important - I notice it's a low voltage audio path.

T.