[sdiy] Variable-rate drum sample playback

[rsdio at audiobanshee.com]

Yes, an external DAC wins for a couple of reasons. First, you get the analog signals out of the noisy CPU. Second, you’re able to select higher bit-depth converters that are more oriented towards audio. I’d say that a musical product pretty much requires an external DAC on the board.

By the time you get to 14-bit or better, you won’t need to dither.

Brian


On Jun 26, 2018, at 3:33 PM, Tom Wiltshire <tom at electricdruid.net> wrote:
> Thanks Brian
> 
> I was just trying to get it running, so it was a case of hack off the unwanted bits and see it works. It does, sort of!
> 
> TBH, if I was going to develop it further, I’d rather work out how to get a couple more bits of DAC accuracy or a better DAC. The 10-bit DAC doesn’t seem to have enough dynamic range to me. I suspect I’d get more impact with a few more DAC bits.
> 
> Tom
> 
> On 26 Jun 2018, at 08:01, rsdio at audiobanshee.com wrote:
>> Very cool. Nice to see something tangible come out of that thread.
>> 
>> 
>> Did you try dithering the 12-bit or 13-bit µ-Law decoded samples to 10-bit? instead of truncating?
>> 
>> You could try Uniform Distribution dither, or a Triangular Probability Distribution Function. Beyond that, you could even try spectrum shaping the dither to make it less audible. There are variable AES papers on the techniques that work best.
>> 
>> Just food for thought...
>> 
>> Brian
>> 
>> On Jun 25, 2018, at 2:06 AM, Tom Wiltshire <tom at electricdruid.net> wrote:
>>> There was a discussion recently about drum sample playback that got me inspired to have a play with it. I was asking about the u-Law DAC here as well, but in the end, I didn’t bother, at least not yet. I wrote up my experiment on my website:
>>> 
>>> https://electricdruid.net/experiments-with-variable-rate-drum-sample-playback/
>>> 
>>> What I found quite cool was that it’s possible to get three separate samples *each with their own variable sample rate* into a single PIC. That’s not half bad!
>>> That’s using 4K samples and a 16K chip (so three samples leaves 4K for code). That means the 8K samples for things like the Ride cymbal would probably need a chip of their own - but still, you can wrap up the sample memory, the DAC, and the VCO that clocks it all onto a single chip.
>>> 
>>> Hope you like it. I don’t know whether I’ll take the experiment any further - maybe one day.
>>> 
>>> Tom
>> 
>