Vember Audio Shortcircuit is now open source!

[baconpaul]

Yeah honestly I would rather do things that sound “worse” (like a 12 bit pipeline at the generator or a fixed low sample rate which resamples at the end of group) than a higher latency tiny improvement in the current reconstruction

And I guess that’s sort of part of my philosophy in things that excite me personally with the project. Stuff I can’t already do with the instruments I have easily combined with access to the collection of dsp code we have in surge land and a modulation first instrument feel is kinda why I get excited to type in code!

But as we always say it’s an open project. If someone here really wants to add a much wider fir reconstruction implementation you just need to show up and write the code! Heck I could even give you the pointers you need. Just I don’t think it will make the 1.0 cut. Since one day we have to ship something stable otherwise what’s the point!

Love the enthusiasm for the project here by the way. Exciting!

[wikter]

Zero Order Hold = no math at all. Just take the value of the previous sample. Lots of Aliasing.
Linear interpolation = trivial math. Draw lines between samples and place the new samples on the line. Less aliasing, but still a lot.
Cubic interpolation = better math. Instead of lines between samples, draw curves. Even less aliasing, but still some.
Sinc = even better math. Still draw curves, but even better curves. Can get arbitrarily close to alias-free.

ZOH is. Indeed no modern DAC does it. But tracker software apparently does.

MilkyTracker & OpenModplug use the term "No Interpolation" for that ZOH you wrote there. That's the idea. And about "resampling" it's about the same. It's not just about old Hardware smplers, but about oldschool trackers: Impulse Tracker, Fast Tracker, Octamed... Most of them avoided interpolation to gain voices on old 386/486 machines, and the aliasing & lofi was part of that sound.
The fact is that Trackers are back in hardware format for some reasons. The sound is one.
https://www.youtube.com/watch?v=roBkg-iPrbw

[wikter]

A pair of samples created with MPT & a pattern by SagaMusix.

No interpolation.
https://www.hispasonic.com/index.php?co ... _id=219991
Cubic Spline Interpolation.
https://www.hispasonic.com/index.php?co ... _id=219992

The difference between interpolations is easier to notice on high frequencies, that's where the reflected frequencies start to sound louder. Anyway, I think it's something easy to understand by using a tracker or just this plugin, Amigo Sampler:
https://www.potenzadsp.com/amigo/
https://www.youtube.com/watch?v=WZk6l4tKzm4&t=2s

[baconpaul]

neat - thanks for sharing!

[baconpaul]

also i'm just adding all the surge filters, so there's now a comb in place if you want to do the phase delay from that amigo video

[rafa1981]

No interpolation generally doesn't work on single sample rate output machines like our computers (certain vintage samplers had a DAC per voice, and DACs would be separately clocked to change the tuning of the sample, or something to that extent) without quite some DSP being involved. It is possible to use sinc interpolation as a per voice DAC emulator, which is exactly what the Wavetable oscillator in Surge is already doing (and this is also SC's default interpolation method IIRC).

I think I have seen a paper about BBD emulations doing some kind of clever approximation to do this with no latency. Only in the downsampling direction of course. I have probably seen on the sources for the Chow BBD.

The tracker thing is diferent. If my memory serves me right OpenMPT does it too, so it must be just taking a look at the sources. I guess it's just plain ZOH.

[audiojunkie]

Oh, yes!! Please do put it in the to do list! Get the thing out the door first, of course. But I really, really want a full blown, high quality open source sampler that does time stretch, formant shifting, great quality interpolation, and direct from disk support! To have that and have it be open source is the Holy Grail!

[audiojunkie]

Yeah honestly I would rather do things that sound “worse” (like a 12 bit pipeline at the generator or a fixed low sample rate which resamples at the end of group) than a higher latency tiny improvement in the current reconstruction

And I guess that’s sort of part of my philosophy in things that excite me personally with the project. Stuff I can’t already do with the instruments I have easily combined with access to the collection of dsp code we have in surge land and a modulation first instrument feel is kinda why I get excited to type in code!

But as we always say it’s an open project. If someone here really wants to add a much wider fir reconstruction implementation you just need to show up and write the code! Heck I could even give you the pointers you need. Just I don’t think it will make the 1.0 cut. Since one day we have to ship something stable otherwise what’s the point!

Love the enthusiasm for the project here by the way. Exciting!

I understand where you are coming from, but please understand that there are several lower quality samplers available for all three OSes, but there are no high quality ones with GUIs in the open source world—for any OS.

[baconpaul]

Yeah that’s why our default works to be high quality. Have you played it and found cases where it distorts a sample surprisingly today? It’s a pretty accurate reconstruction

[baconpaul]

Oh, yes!! Please do put it in the to do list! Get the thing out the door first, of course. But I really, really want a full blown, high quality open source sampler that does time stretch, formant shifting, great quality interpolation, and direct from disk support! To have that and have it be open source is the Holy Grail!

I had a very good idea why thst doesn’t exist in the open source world. It is a very very hard feature set. But there’s some existing libraries which do some of thst we can integrate over time if the project succeeds.

[audiojunkie]

Yeah that’s why our default works to be high quality. Have you played it and found cases where it distorts a sample surprisingly today? It’s a pretty accurate reconstruction

I haven’t tried it yet. I’ve been waiting for things to get a little closer to being a beta, rather than a pre-alpha.

[audiojunkie]

Oh, yes!! Please do put it in the to do list! Get the thing out the door first, of course. But I really, really want a full blown, high quality open source sampler that does time stretch, formant shifting, great quality interpolation, and direct from disk support! To have that and have it be open source is the Holy Grail!

I had a very good idea why thst doesn’t exist in the open source world. It is a very very hard feature set. But there’s some existing libraries which do some of thst we can integrate over time if the project succeeds.

I’m glad to hear that the libraries exist!

[Andreya_Autumn]

I understand where you are coming from, but please understand that there are several lower quality samplers available for all three OSes, but there are no high quality ones with GUIs in the open source world—for any OS.

I sympathize with the point! Though my perception is that the output reconstruction quality (say, your 64 point sinc suggestion) is a red herring in this case. The current strategy is really quite clean. Once you've sped/slowed a sample for enough to make re-sampling artifacts remotely audible, the chipmunk/blue whale effect is 100 times more problematic than said artifacts. As Paul said, if someone can show a case where that's false I'll certainly stand corrected, but I doubt that'll happen.

Formant/Pitch shifting would be far more impactful for the goals you seem to seek (although they come with artifacts of their own), and it is is on the list! And the prospect seems a lot more hopeful for that Bungee library existing! One thing at a time though.

[Andreya_Autumn]

Yeah honestly I would rather do things that sound “worse” (like a 12 bit pipeline at the generator or a fixed low sample rate which resamples at the end of group) than a higher latency tiny improvement in the current reconstruction

And I guess that’s sort of part of my philosophy in things that excite me personally with the project. Stuff I can’t already do with the instruments I have easily combined with access to the collection of dsp code we have in surge land and a modulation first instrument feel is kinda why I get excited to type in code!

But as we always say it’s an open project. If someone here really wants to add a much wider fir reconstruction implementation you just need to show up and write the code! Heck I could even give you the pointers you need. Just I don’t think it will make the 1.0 cut. Since one day we have to ship something stable otherwise what’s the point!

Love the enthusiasm for the project here by the way. Exciting!

I wanna say I love the enthusiasm too! And this comment gave me an idea, which I'll share that in the dev channel.

[audiojunkie]

I understand where you are coming from, but please understand that there are several lower quality samplers available for all three OSes, but there are no high quality ones with GUIs in the open source world—for any OS.

I sympathize with the point! Though my perception is that the output reconstruction quality (say, your 64 point sinc suggestion) is a red herring in this case. The current strategy is really quite clean. Once you've sped/slowed a sample for enough to make re-sampling artifacts remotely audible, the chipmunk/blue whale effect is 100 times more problematic than said artifacts. As Paul said, if someone can show a case where that's false I'll certainly stand corrected, but I doubt that'll happen.

Formant/Pitch shifting would be far more impactful for the goals you seem to seek (although they come with artifacts of their own ), and it is is on the list! And the prospect seems a lot more hopeful for that Bungee library existing! One thing at a time though.

Fair enough. I’ll withhold judgement until after I’ve had a chance to actually use it. You may indeed be correct—the chipmunkization is a real negative as well, and fixing that may indeed need to take precedence. I’m really glad that Bungee library was found!