DSP Robotics Support

Tepeix wrote:
... I wanted a delay that keep the same maximum time in any sample rate, but reinitialize the memory size to not use too much. (even if it's not a big amount of memory in most situation)
But the memories is not reset using selector in this way..

You may know this already, but just in case…

All the DSP and ASM delays assign a maximum number of samples, not time. So a memory size of 44100 samples would give a maximum of 1 second at a sample rate of 44.1 kHz. If you go to 96 kHz sample rate the maximum time would be about 0.46 seconds. So I would be inclined to set a maximum number of samples based on the maximum expected sample rate. Of course the actual time set should make used of the sample rate prim so the delay time adjusted will not be affected by the sample rate. That means if the sample rate increases, the number of samples requested in the delay input would increase in proportion.

Something to be aware of is using stream delays in the poly white path. The voice count (number of notes sounding) will have to create new memory allocation and for longer delays that can cause clicks. In mono blue the memory allocation is done on loading or recompilation and remains fixed.

The mono delay prim might have no limit or be sensitive to sample rate, but I never tested it because, apparently, it’s not very CPU efficient.

Hum maybe i could do a delay in white, then use it with mono to poly.
But it seams a little complex task..
There's also some mem solution.

But i was thinking of plugin with very large memory and finally i'm not sure i will do with so much for now..
Seams better to do like you said even if it waste a little memory in normal sample rate, it might be better for cpu that to try to realocate memory.

Or maybe using the zipped selector, i could use a normal delay with less memory (enough for 48000) then if a greater SR is detected move to a bigger one. (But will be the mem of the second code be declared even if it's never selected ??)

I also think about a Sample Rate reductor before the delay.
Sometime if we have a low pass in the delay it could be at lower SR.

(When i come back to Flowstone i was also using delay with hop(2) whithout hearing that without a low pass they could make some aliasing)

Tepeix wrote:... Or maybe using the zipped selector, i could use a normal delay with less memory (enough for 48000) then if a greater SR is detected move to a bigger one. (But will be the mem of the second code be declared even if it's never selected ??)
...
(When i come back to Flowstone i was also using delay with hop(2) whithout hearing that without a low pass they could make some aliasing)

If the delay is in mono blue I believe the memory will be assigned even if not selected. I base this on my experience of switching to a delay FX using a selector and hearing a previously playing sound. That leads me to my conclusion because otherwise I wouldn’t hear the “old” sound. You can check this yourself.

Using hop in an audio stream will introduce distortion. I tried it once and it was horrible!

The hop do very bad thing.
But here's an example of using it with delay.
(Like an oversample or downsample, using filter to avoid aliasing)

...Hum, it works when not using modulation. (very little inaudible aliasing)

But i don't understand clearly why, it's don't works with the modulation...

Maybe in this case the filter i use is insufficient.. (it's a naive one)
When the depht is low it's ok, but when frequency go to much up it breaks..

It's possible, but need a better filter. Hum, maybe it need also something else ?

(I was not using those filter and not hearing so much the problem with my old pioneer amp .)

I’m not sure why you would want to hop a delay, unless you’re looking for that type of distortion of course!

I think the issue of wide modulation is down to the integer delay. I’ve attached an interpolated fractional version with my favourite delay which was originally supplied with MyCo’s analogue kit years ago. I also de-zipped the Amount knob in the LFO. It sounds smooth to me now.

Finally i get it to works.
A better placing of the output filter in the feedback loop.

Very low tone; a more fast transition band is needed to preserve more high at 22100 SR..

Didn't see your response when posted my last.
It works well and preserve more high !

>> oups now i understand my error !) I use the non interpolated delay to get some modulation !!

That's even crazy that the filter could make this almost audible in the version 2 !)
But a very bad and destructive way..

My idea was to use a hop but only when the user use more than 48000 SR.
The delay will be a little darker but not so much. But also sometime we only need bass resonance.

Well, again this is maybe a premature optimization idea... The size of memory is not so big.
Maybe i like to break my head on some stuff..



Also i find that the zipped selector idea could sometime be bad..
With a preset manager switching every selector in a short time there could make double cpu spike..
When a selector is acceptable in most selecting situation...

But i plan a preset version that shut the volume before activating the new code..

Finally, an hop(4) version of the analog tool kit one !)

I suspect it do some noise..
But they are not so much audible, with the vibrato and feedback it's hard to know if it come from the signal or is some noise added.

The filter are not adapted, but the transition band is bad enough to attenuate strongly one more octave.