DSP algorithms

[Music Engineer]

hey, what happened to the text-formatting?

[Urs]

Hey braindoc,

I think to remember that Detunator was written using a phase vocoder (see www.dspdimension.com ). But I may be wrong, havn't followed the thread intensively (too much garbage).

Using two delay lines for pitch shifting does work like you describe, but don't expect any good results due to a lot of phase cancellations...

Cheers,

Urs

[jmh]

The info on how Detunator works is buried here somewhere, might be hard to find though due to all the noise

And it sounds friggin' sweet on some things

Regards,

JMH

[jmh]

Here we go, the first post on the plugin... too tired to dig up the continuation, something for the others to do

Well I started an "UniClono" effect project now. That is actually my part of work to do the prove of concept. But also the prove of being able to do more tha only "fighting" with words.

But the result should be quite usable as an effect finally for everybody.

As far the conception is, I'll do a stereophonic effect with up to 8 voices (= max. 16 clones pitch shifted from the original),

- an UniclonoDetune parameter and
- a StereoSpread Parameter.

I think, -+ 100 cent detuning as a maximum should do the job for normal Uniclono detune range. Althought there are amounts up tp +- 24 semitones possible ...

The effect uses delay line based pitch shifting, which has a lower quality - but good for performance reasons and to be able to perforn polyphonic material *and* monophonic material this way.

If there are some whishes to implement, please say it now. Because I have planned to be ready tomorrow with that.

.

Regards,

JMH

[;-)]

Hey braindoc,

I think to remember that Detunator was written using a phase vocoder (see www.dspdimension.com ). But I may be wrong, havn't followed the thread intensively (too much garbage).

Using two delay lines for pitch shifting does work like you describe, but don't expect any good results due to a lot of phase cancellations...

Cheers,

Urs

Right. Too much garbage here.
Detunator don't use the phase vocoder. And that source code at DSP dimension would not even allow to use 4 pitch shifters at ones due to too heavy performance usage.

Detunator uses a maximum of 16!!! pitch shifters (8 times stereo) with very resonable performance (25 % on a good old DURON processor).

And by the way, phase cancellations is a remarkeble feature with unisono. You cannot prevent it, even not with the phase vocoder. Because it lies in nature of pitch detuning with many voices.

.

[;-)]

after 2 hours or so sitting in front of the monitor, heaving read the whole thread, i think

here something interesting going on. i have listened to the examples of detunator - very

impressing. but the download link for the plugIn seems to be down. i need that plugIn! i

remember "blümchen" writing, that the host cannot turn off the pluIn, so you implement your

own bypass, if the input is zero. you seem to know a lot, although your style is sometimes

quite offending. so maybe i have overlooked something, but maybe you should have a closer

look at the getTailSize function in the VST-SDK.

i'm interested in, how delay-line based pitch shifting works. the first thing which comes to

my mind is a delay-line with two tap-outs where the first one starts at the end of the line

(the maximum delay), fading in, while the second one starts in the middle, fading out (when

shifting upwards so that the tapOut is faster than the tapIn). when the second one passed

the tapIn it's volume should be zero and then it wraps around to the end of the line while

the first one has reached the middle and its maximum volume. this would be

granular-synthesis with a triangular window and an overlap of 50%. is that the way - or in

any way similar to - how detunator works?

Yes, what's with your text formatting?

Did you try to contact the creators of Detunator? Maybe there is a way to get a copy?

[Urs]

Ah. ok.

And by the way, phase cancellations is a remarkeble feature with unisono. You cannot prevent it, even not with the phase vocoder. Because it lies in nature of pitch detuning with many voices.

Well yeah, in that case it might be okay. But you can't get any high quality pitch shifting with that

Regarding the phase vocoder... it depends on the actual implementation. The code at dspdimension is optimized for readability, not for speed. An optimized version can be *very* fast. - You can for instance create several detuned "images" at once, you don't need to do the whole process (FFT->scale->iFFT) for each voice, that's only two OLAPs, not 16 or 32.

Cheers,

Urs

[;-)]

And why don't you do it then, if it is so easy and that effect is so hot needed by the people?

By the way: what you said about the FFT->IFFT, that was already all (very detailled) said here inside the thread.

But there are several other problems with the phase vocoder, you obviously don't see.

I assume you never really tried it with the phase vocoder. Because then you would actually see, that it just sounds more worse than with the delay based pitch shifter in an "unisono effect".

Only two words here regarding those to expect heavy problems: bass frequencies and latency.

That summs finally up with the extreme usage of processor. You can optimize much, but probably never reach the speed of a delay line based pitch shifter with FFT.

And who actually does really claim, that a delay line based pitch schifter is always cheep sounding?

That is entirely wrong. Because the overlapping technique makes the quality here. I guess, you wouldn't even be able to hear the difference in quality with a good implementation. Probably you would rather hear the extreme problems the phase vocoder produces.

[Urs]

U talkin bullshit.

[Music Engineer]

in an - so called - unisono effect we need only a small amout of detuning by - say - one semitone maximum or so. in this case - i think - we do not need any frequency-based approach, because the artifacts generated by a granular approach will remain small. i didn' try, but maybe artifacts introduced by FFT/iFFT due to windowing and leakage could be more severe.

[Music Engineer]

hello earthlings,

this could be interesting for all.

I am looking for a well organized compendium kind of dsp effects algorithms book, resource, whatever ...

My focus is on coding a UNISON effect. But I never encountered a usefull guide.

thanks

// edited spelling

but, hey, besides the very intresting discussion of how to generate an unisono or uniclono or whatever-effect - still noone gave an answer to the question which started this thread. so i will do now. i strongly recommend the book "DAFX" (which is an abbreviation of digital audio effects) by udo zoelzer - it is really expensive, i paid 109 euros for it. but if you are really into audio-dsp and effect processing, this book is a must have. there is also a website on which every year very intrseting papers are published: www.dafx.de . this one is about effects. if you are intrested in synthesis, then the "Computer Music Tutorial" by Curtis Roads is a good resource. And a final recommendation: if you need a good and not over-mathematical explanation of digital signal processing in general, then the book "The Scientist and Engeneers Guide to Digital Signal Processing" is worth a look. it's downloadable for free at www.dspguide.com

[Galleoneer]

but, hey, besides the very intresting discussion of how to generate an unisono or uniclono or whatever-effect - still noone gave an answer to the question which started this thread. so i will do now. i strongly recommend the book "DAFX" (which is an abbreviation of digital audio effects) by udo zoelzer - it is really expensive, i paid 109 euros for it. but if you are really into audio-dsp and effect processing, this book is a must have. there is also a website on which every year very intrseting papers are published: www.dafx.de . this one is about effects. if you are intrested in synthesis, then the "Computer Music Tutorial" by Curtis Roads is a good resource. And a final recommendation: if you need a good and not over-mathematical explanation of digital signal processing in general, then the book "The Scientist and Engeneers Guide to Digital Signal Processing" is worth a look. it's downloadable for free at www.dspguide.com

Well, thanks for trying to put the thread to its original and noble direction. I actually could take a look inside the book you have recomended: it is more for the beginner (this means it has a wide but not deep spectrum). But it is I guess the only compendium so far a audio fx. thanks again. I actually know a little dsp so I try to dig a little deeper.

Somewhere in between this huge nonsense I tryed to explain that I am about to make a unisono effect by stacking multiple instances of the generators (By hosting them). This will be the Unisono in clasical way.

On creating a effect of a classical approach wich processes audio. Mrs. blümchen (current identity) has the best algorithm so far (detunator). But he has the optinion not to threat it as unisono.

And he wont tell how he did it even if urs tries hard to get it from him (only a joke).

I actually come from the video signal processing field. mathematics are not the problem therefore. but thaaaanks.

If you click on my sig you see some little dsp stuff I made so far -> more to come.

make code no war

[Music Engineer]

I actually could take a look inside the book you have recomended: it is more for the beginner (this means it has a wide but not deep spectrum).

assume you are referring to the "DAFX"-book. most of the content are basics, that's right. but have you also checked out the papers, they publish on their conferences. i think there is much of more advanced stuff in there - at least it looks to me quite advanced, but i am not a math-nerd

[Galleoneer]

I actually could take a look inside the book you have recomended: it is more for the beginner (this means it has a wide but not deep spectrum).

assume you are referring to the "DAFX"-book. most of the content are basics, that's right. but have you also checked out the papers, they publish on their conferences. i think there is much of more advanced stuff in there - at least it looks to me quite advanced, but i am not a math-nerd

no havent checked the papers yet. thanks for the hint

[;-)]

U talkin bullshit.

So?
Then prove, what you said.
I did it already. Your proove is missing so far.

.