[sdiy] Chris Synths polysynth
Richie Burnett
rburnett at richieburnett.co.uk
Wed Oct 20 19:10:01 CEST 2021
I guess what you are describing is really an Elliptic lowpass filter with its notch in the stop-band Gordon. The transition-band slope should be nice and steep.
A deep notch in the filter response at the switching frequency works well when the PWM is being used to synthesise signals that are much lower than the switching frequency. This is because the "modulation sidebands" sit nice and close to the PWM carrier and also get heavily attenuated by the notch. It works less well when the frequency of the signal being generated is a significant fraction of the PWM frequency.
-Richie,
Sent from my Xperia SP on O2
---- Gordonjcp wrote ----
>On Wed, Oct 20, 2021 at 10:37:00AM +0000, Mike Bryant wrote:
>> If you really want cheaper DACs consider using PWM followed by RC smoothing. And not an FPGA but something like the new Pi 2040 MCU which has lots of I/O programmability. I suspect you could get about 20 to 24 8 bit resolution voltages for about a $1. And when if you need more resolution combine two PWM outputs with different summing Rs. 12 or 13 bits linearity is easily attainable this way, maybe a little more.
>
>I've never seen anyone else do it but I've always had super clean results from PWM by having a 12dB/oct lowpass filter at half the sample rate followed by a notch filter at the sample rate.
>
>It's super easy to calculate a Sallen-Key filter for this. For a Butterworth response with equal resistors the "ground" cap needs to be half the "feedback" cap, and if you pick 1nF and 470pF then 15kΩ resistors will give you 15kHz-ish and a Q of about 0.73, close enough. Add in a transistor and a couple of resistors for the emitter load and biasing and you've got a lot more components than a simple RC filter, but far far far better performance. An RC filter ought to give 6dB/oct rolloff with a Q of 0.5, which is such a slow transition and gentle curve that the Butterworth filter seems almost precipitous by comparison!
>
>A notch filter is harder to do but it'll completely scrape off the last of the PWM "carrier breakthrough" leaving a super clean output.
>
>--
>Gordonjcp
>
>_______________________________________________
>Synth-diy mailing list
>Synth-diy at synth-diy.org
>http://synth-diy.org/mailman/listinfo/synth-diy
>Selling or trading? Use marketplace at synth-diy.org
More information about the Synth-diy
mailing list