[sdiy] Cheapest good sounding digital reverb?

Mikko Helin maohelin at gmail.com
Mon Mar 22 14:31:46 CET 2021 

The latest Teensy 4.x version of the CS42448 board adds an extra A/D for
total 8 channels of audio input:

https://hackaday.io/project/2984/logs


On Mon, Mar 22, 2021 at 3:16 PM Mikko Helin <maohelin at gmail.com> wrote:

> Teensy 4.1 with a multi-channel I/O (6 in / 8 out) using Cirrus Logic
> CS42448 codec:
> https://blog.oshpark.com/2017/05/01/tdm-support-for-many-channel-audio-io/
>
> For basic Schroeder type reverberator you will need quite a lot RAM for
> the delay lines though...
>
> https://ccrma.stanford.edu/~jos/pasp/Freeverb.html
>
>
>
> On Mon, Mar 22, 2021 at 9:42 AM Vladimir Pantelic <vladoman at gmail.com>
> wrote:
>
>> check the reverb that's part of Mutable Instruments Clouds. IIRC it uses
>> very
>> little CPU and memory and can be easily adapted to run standalone.
>>
>> On 3/22/21 12:31 AM, cheater cheater wrote:
>> > Yeah, I'm mostly looking for that sort of information about
>> > algorithms. Something that'll run on a small mcu while also having a
>> > pretty good sound.
>> >
>> > On Mon, Mar 22, 2021 at 12:09 AM Tom Wiltshire <tom at electricdruid.net>
>> wrote:
>> >>
>> >> The cost isn’t the memory, so the decay time isn’t a big deal.
>> >>
>> >> The cost/quality trade-off mostly comes from the ADC/DAC, I'd say.
>> Cheap convertors gives a cheap result, quality convertors gives a more
>> expensive result. Mike B proposes one way of doing the conversions at very
>> low cost.
>> >>
>> >> However, I have to say that I think a simple delay line won’t satisfy
>> as a reverb. Keith Barr’s impressively simple reverbs all used other tricks
>> beyond delay to make the reverb tail smoother and denser - combinations of
>> allpass and delay, mostly.
>> >>
>> >>> On 21 Mar 2021, at 20:45, cheater cheater <cheater00social at gmail.com>
>> wrote:
>> >>>
>> >>> Right, but what can be done regarding reverb in a small inexpensive
>> >>> microcontroller? Remember, we're not talking about decays of several
>> >>> seconds here.
>> >>>
>> >>> On Sun, Mar 21, 2021 at 6:13 PM Tom Wiltshire <tom at electricdruid.net>
>> wrote:
>> >>>>
>> >>>> There isn’t much at the low-end of the DSP market, so your options
>> are limited.
>> >>>>
>> >>>> There’s the FXCore, which can deal with 4 ins/4 outs, but needs
>> external codecs (but that lets you choose for best quality or lowest price)
>> >>>>
>> >>>> http://www.experimentalnoize.com/product_FXCore.php
>> >>>>
>> >>>> Otherwise, you might be best moving to a bigger, faster chip and
>> then expecting the one device to process 16 channels of audio for you.
>> You’d still need a multi-channel codec, and the hardware is going to be all
>> throughly modern and SMD-tiny!
>> >>>>
>> >>>>
>> >>>> On 21 Mar 2021, at 14:34, cheater cheater <cheater00social at gmail.com>
>> wrote:
>> >>>>
>> >>>> It's 2-4x too expensive, but I'll check the youtube demos anyways -
>> >>>> thanks a lot.
>> >>>>
>> >>>> On Sun, Mar 21, 2021 at 3:23 PM Tom Wiltshire <tom at electricdruid.net>
>> wrote:
>> >>>>
>> >>>>
>> >>>> I don’t know that it fits your definition of cheap in small volumes,
>> but the Spin FV-1 chip is about your best option, I’d say. It’s a
>> simple-to-use almost all-in-one option and there are loads of good reverb
>> algorithms for it freely available.
>> >>>>
>> >>>> The standard application uses a cheap watch crystal (so 32KHz
>> sampling) but you can run the chip faster if you need a little bit more
>> hi-fi. Honestly, I doubt this is necessary for reverb. The high end is
>> absorbed most quickly and hardly appears in any reverb signal. But it’s
>> easy to do if required. I think the chip is specced up to 50KHz or so, and
>> people have overclocked them faster than that - Spin left themselves a good
>> safety margin.
>> >>>>
>> >>>> The algorithms are stored on an external EEPROM, but there are also
>> 8 internal programs, including several reverbs, so if you use those you can
>> do without the external EEPROM, at which point it really is a one-chip
>> solution.
>> >>>>
>> >>>> Check out a few FV-1 effects pedals on Youtube and see what you
>> think.
>> >>>>
>> >>>> HTH,
>> >>>> Tom
>> >>>>
>> >>>> On 21 Mar 2021, at 13:19, cheater cheater <cheater00social at gmail.com>
>> wrote:
>> >>>>
>> >>>> I have been thinking recently about whether it would be feasible to
>> >>>> have a simple reverb of some sort per voice, and so I wonder if
>> anyone
>> >>>> had any suggestions on a cheap algorithm that could be executed on
>> >>>> inexpensive chips.
>> >>>>
>> >>>> what I need from the reverb: exponential decay of ~0.5 second, flat
>> >>>> frequency spectrum @ 22 Hz...22 kHz
>> >>>>
>> >>>> instrument: 16-voice
>> >>>>
>> >>>> architecture: vcos -> filters -> vca1 -> possibly vca2 (all stages
>> analog)
>> >>>>
>> >>>> I'd like to be able to insert reverb after the filter but before the
>> last vca:
>> >>>>
>> >>>> vcos -> filters -> vca1 -> rev -> vca2
>> >>>>
>> >>>> or possibly after the vco:
>> >>>>
>> >>>> vcos -> rev -> filters -> vca
>> >>>>
>> >>>> or after the filter:
>> >>>>
>> >>>> vcos -> filters -> rev -> vca
>> >>>>
>> >>>> or even:
>> >>>>
>> >>>> vcos -> rev1 -> filters -> rev2 -> vca -> rev3 -> vca2 -> rev4
>> >>>>
>> >>>> The reverb is meant to only "sweeten up" the sound by giving filter
>> >>>> sweeps, transients, and vco sweeps some more substance in the time
>> >>>> domain. I think this sort of thing could easily add a unique sound to
>> >>>> the synthesizer. I know some of you will mention the DSI Evolver, but
>> >>>> honestly I did not think that the digital part in that synth was of
>> >>>> high enough quality. So what I'm looking for is an inexpensive "hi
>> fi"
>> >>>> reverb.
>> >>>>
>> >>>> The considerations are either:
>> >>>> A) a single chip per voice/stage which only processes one stage in
>> one
>> >>>> voice. this chip would have to have high audio quality AD/DA, work
>> >>>> without a lot of additional circuitry, just enough processing power
>> to
>> >>>> perform the reverb, and be relatively inexpensive (up to ~5 per chip
>> >>>> at low volumes)
>> >>>> B) one global chip with a bunch of AD/DA. this chip would need to be
>> >>>> able to read from 64 AD and write to 64 DA, each at 16 bit.
>> >>>>
>> >>>> personally I prefer A because 1. it does not carry a bunch of digital
>> >>>> stuff around an otherwise analog board which can be a royal pain and
>> >>>> 2. drifting clocks (or ones shifted on purpose) will add variety to
>> >>>> the sound. So those two kind of kill B for me.
>> >>>>
>> >>>> What sort of chip would you all suggest for version A?
>> >>>>
>> >>>> What algorithm would you suggest to run on it?
>> >>>>
>> >>>> Thanks.
>> >>>> _______________________________________________
>> >>>> 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
>> >>>>
>> >>>>
>> >>>>
>> >>
>> >
>> > _______________________________________________
>> > 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
>> >
>>
>> _______________________________________________
>> 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
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://synth-diy.org/pipermail/synth-diy/attachments/20210322/3150aab2/attachment.htm>

More information about the Synth-diy mailing list