DSP algorithms

[brok landers]

The best of all : Brass is equal to gold, because it looks nearly the same...

like i said, you don`t read my posts.
i never said that.
i`m off ...

[Borogove]

Hey pustekuchen -

You say "unisono" is "a stack or layer of tone generation with the same synthesis thechnique and probably very equal (but not identical) sound."

But you don't accept merely detuning a couple of oscillators to achieve "unisono".

If the generators are, just for sake of argument, running a static wavetable with a given spectrum, then it should be obvious that generator 1 playing wavetable A and generator 2 playing wavetable B could be replaced with just generator 1, playing a wavetable consisting of the sum of A+B, but using half the processing of the two generators.

So I'm curious as to what you accept as a valid difference between the generators, which would achieve a effect which you're willing to call "unisono".

[jmh]

well, maybe because it is, what it is:
a stacking of freerunning voices, which are detuned.
so, if you want to achieve this, it is simply done by that. that`s maybe why there is no paper about it.
there should be indeed a paper which allowes to achive similar sounding while not consuming the cpu power of the summ of stacking voices, as this would , as pustekuchen said, use not only the osc`s, but the whole cloning of the signal path (i.e. each "cloned" voice would also have it`s own filter).
but as long as one would implement it as usual, the result would be monophone (using all available voices), therefore consuming cpu as if one would press all keys playing all voices.
this is not neccessarily so, but most synths use it that way, as it would cost way too much cpu, if it would be polyphone. however, there are some syths where the user can define how much voices should be used for unisono, and how much voices should be allowed to play that unisono polyphone (i.e. korgs polysix vst).
but it`s, like i said, horrible on the cpu.
that`s why some developers use the compromise to _only_ stack the osc`s, feeding the regulat path of _one_ voice, which is cool, as long as you don`t use filterdrive, which then would sound different, because _one_ drive would be applied to _multiple_, detuned osc`s.

(I'm addressing this this generally, not you directly, quoting you is just a practical way of referring to this as I'm referring to some things you've said)

So... I'm not trying to fight this issue, hence I'll try to think what I'm saying...

We have a definition for unison. Something most parties agree on. It's either done that way, or it isn't unison.

Doing it as described, stacking detuned oscillators... that's what the term 'super' attached to a waveform name does. While not being a practical or descriptive name for it, it is somewhat a common term for this feature.

What seems to be the issue here is that some of 'us' would prefer that the term unison wouldn't be applied if it's not the real deal. Some manufacturers are doing it properly as you've pointed out, and it is very demanding on the CPU.

I do see the need for seeking a similar sounding process that could be applied to a synth afterwards, and could surely use such a plugin if someone comes up with one. I'd just prefer it not be called as unison, as per the definition we seem to have an agreement that it isn't

Attaching another term to it which already has a definition - such as 'ensemble' bears the same burden - if you're not doing things the way the ensemble effect on old synths work, it shouldn't be called that.

Funny how this at the same time looks like hairsplitting and discussing an important issue, how certain effect names are already associated with a certain way of processing signals and how such terms are misused - I personally don't always see it as a crime that should result in capital punishment, but it could be that such terms are used misleadingly for marketing reasons, etc. Even funnier that such an issue leads to threads going like this... it's the KVR effect - and we need a name for that

(I'm of course as quilty of contributing to this effect as can be witnessed in several threads )

[DISCLAIMER - What I'm about to say is purely my own opinions and thus not reflect anyone else's, despite using quotes freely from other parties =)]

So I'm curious as to what you accept as a valid difference between the generators, which would achieve a effect which you're willing to call "unisono".

I apologize for replying to this as it isn't directed at me... but "unisono", read above. What it should be is "a stack or layer of tone generation with the same synthesis thechnique and probably very equal (but not identical) sound." - as said. What the difference is compared to just using detuned oscillators is that the minute differences between the stacked voices provide the sometimes subtle, sometimes not, effect - as each voice might be a tad detuned, oscillator phases differ, filter cutoff might be a little different, different envelope stages vary and so on (for about each possible parameter). This is of course - as you know - coming from the land of the mighty analogue, where the differences were provided courtesy of the technology.

Hence... to provide proper "unisono", what's really needed is duplicating the whole signal path for each layer and then providing all the little variations to a bunch of parameters to reach the effect known as "unisono".

Sorry, can't put it more eloquently right now I'm going around repeating the same thing, so apologies for that too

Regards,

JMH

Regards,

JMH

[brok landers]

We have a definition for unison. Something most parties agree on. It's either done that way, or it isn't unison.

correct.

Doing it as described, stacking detuned oscillators... that's what the term 'super' attached to a waveform name does. While not being a practical or descriptive name for it, it is somewhat a common term for this feature.

well, i didn`t say this.
i said stacked _voices_, not stacked osc´s (which i stood allready corrected for), and i discribed it by telling that the whole signalpath of one voice is involved. in another post i only statded that it`s a common technique to only stack the osc`s, which of course is no real unison.
but it brings most users a very similar effect.

What seems to be the issue here is that some of 'us' would prefer that the term unison wouldn't be applied if it's not the real deal. Some manufacturers are doing it properly as you've pointed out, and it is very demanding on the CPU.

thats correct, too, but the initial post of galloneer was directed to "something similar" as unison, so there was a wider range of possible achievements.
however, i do agree, only the real thing should be called unison.
the only thing that i was going against was a statement of pustekuchen, where he said that "by definition" in unison subtle changes of a variety of parameters were achieved by _implementing_, which isn`t correct.
in analog synths this sinply wasn`t necessary, as theese subtle changes are always there, because of the tolerances of the used units in the synths.
the osc`s were free running, they were not stable in tuning, also the filters were not, nor the pulsewidth, if a pulsewave was selected.
in digital synths of the old days this huge implementation simply was not possible because of the poor dsp`s which had not enough power, or were way to expensive for a commercial synth.
even the "real" unison, which i mean by "complete signalpath-stacking" was very heavy on the dsp`s, and that`s why a lot of dev`s in the old days were only stacking the osc`s, and giving the possibility to detune them.
to call this unison of course is missleading, but most of the users didn`t even recognise the difference to real unison, as theese digital synths mostly didn`t have a filtersaturation (polyphon), which is what could make this "wrong" or "cheap" implementation obvious.

Attaching another term to it which already has a definition - such as 'ensemble' bears the same burden - if you're not doing things the way the ensemble effect on old synths work, it shouldn't be called that.

which is exactly what i stated in one of the earlier posts.

Funny how this at the same time looks like hairsplitting and discussing an important issue, how certain effect names are already associated with a certain way of processing signals and how such terms are misused - I personally don't always see it as a crime that should result in capital punishment, but it could be that such terms are used misleadingly for marketing reasons, etc. Even funnier that such an issue leads to threads going like this... it's the KVR effect - and we need a name for that

you couldn`t have said this better.

Hence... to provide proper "unisono", what's really needed is duplicating the whole signal path for each layer and then providing all the little variations to a bunch of parameters to reach the effect known as "unisono".

that would be the ideal way to achieve that on the digital domain.correct.
but, let`s put it that way:
we have one way which it should be done.
and then there`s hundreds of dev`s using hundreds of different ways to implement a compromise.

i said it before, and i`ll say it again:
there`s gold, and there`s brass.
if most people couldn`t tell the difference, why not brass, if it`s cheaper?
most people can`t tell the differece of a cheap and a good vino, sad, but that`s how it is.
ironically, when you think of it:
analog vs. digital.
even very good knowledged users couldn`t tell the difference easily.
the same applies to fake and real unison, though not generally imo.
so one is light on cpu, and one is heavy on cpu.
the differences are too subtle to the general user, so they all decide (exeptions tighten the rule of course) to implement a compromise.
like i said, even in a synth, that got it`s character mainly from it`s unison (beside it`s crossmodulation) and got very famous because of it`s fat unison sound in the 80`s it was _only_ stacked voices which were then detunable:
the mks 80.
back to topic, there never was a question of real unison, but something like that.
so again, there are different possibilities for his request.
but as you say .. funny, how a thread can turn ...

reg,

[jmh]

We seem to be saying the same things, in a long way =) I guess in my case it's long-post-envy towards Squids

The only thing I can further comment on is the gold vs brass analogy... we should aim for the gold - if for no other reason then just to educate the masses

(One thing related to this, that has also been discussed on KVR recently, is complaining about the CPU usage in synths that provide unison done the "right" way... even after explaining it, it still seems to be too hard a concept to grasp to a lot of folks. Which in a way is sad as the rule of thumb is that quality always costs. But reasonable compromises do have their uses, especially if someone out there makes a chorus/ensemble/whatever plugin that sounds good enough I won't hesitate to use it =))

Regards,

JMH

[nick at artsacoustic]

wrong thread, sorry.

[brok landers]

We seem to be saying the same things, in a long way =) I guess in my case it's long-post-envy towards Squids

The only thing I can further comment on is the gold vs brass analogy... we should aim for the gold - if for no other reason then just to educate the masses

(One thing related to this, that has also been discussed on KVR recently, is complaining about the CPU usage in synths that provide unison done the "right" way... even after explaining it, it still seems to be too hard a concept to grasp to a lot of folks. Which in a way is sad as the rule of thumb is that quality always costs. But reasonable compromises do have their uses, especially if someone out there makes a chorus/ensemble/whatever plugin that sounds good enough I won't hesitate to use it =))

Regards,

JMH

i see we`re on the same ship ...

reg,

[blümchen]

[edit]

.

[blümchen]

Hey pustekuchen -

You say "unisono" is "a stack or layer of tone generation with the same synthesis thechnique and probably very equal (but not identical) sound."

But you don't accept merely detuning a couple of oscillators to achieve "unisono".

If the generators are, just for sake of argument, running a static wavetable with a given spectrum, then it should be obvious that generator 1 playing wavetable A and generator 2 playing wavetable B could be replaced with just generator 1, playing a wavetable consisting of the sum of A+B, but using half the processing of the two generators.

So I'm curious as to what you accept as a valid difference between the generators, which would achieve a effect which you're willing to call "unisono".

Hi!

I accept it, if an Unisono is produced merely by detuning the voices! Did not claim something different.

What I don't accept, ist the fact, that you could ever produce an acceptable or even justificable Unisono with merely one tone source (as an effect (thus: after the process of tone generation)).


Because (my personal) definition is, that there must be multiple independant *processes* for generating the original partial tones or voices, which then mix up to that Unisono finally. Because simple pitch modification of clones is quite different and somehow completely different.

Uni sono == all sounding together and not all cheap cloned from one source and then multiplied with some modifications. Althought, that may sound very similar under certain circumstances... but merely *may*.

So call it "UNI-CLONO" then !!!


The sonic variousity of a real Unisono is for some obvious reason much more complex than with merely applying pitch scaling multiple times to one single original source. You should know that.

For instance, if there are Random or polyphonic (free running) LFO generated modulations inside each single voice which differ clearly from each other. But such are not *required* to justify being Unisono finally.


But well.
Maybe I am wrong and you all are right (except JMH, which is then with me on the wrong side - again )

Cheers

.

[blümchen]

By the way: We can make that test!

Build that ominous Unisono effect!
Then we will listen to the result, whether *at least* the final sound is identical or even the same quality as a real Unisono processing ...

But also in that case I am not willing to accept the "new" Unisono definition, because that finally would be merely a kinda "Virtual Unisono Effect" and not Unisono processing.

.

[blümchen]

I did some googling about Unisono.

There are two interesting facts:

- The most Unisono definitions on the web are in Geraman language.
- Many of those even claim, that Unisono defines "equal pitch" for all the voices . There is not even a note to any "detuning" or pitch shifting or whatever...


But all are conform in the definition, that Unisono means "Einklang" of many single and individual voice sources summing up to the whole result.


http://de.wikipedia.org/wiki/Unisono
http://66.102.9.104/search?q=cache:kFa3 ... o%3F&hl=de
http://magazin.klassik.com/lexikon/deta ... 4&letter=U

That is the essence so far.
Due the fact, that Unisono is also widely used in classical music - and in fact has its origin there - justifies my definition. Because orchestras of classical musicians are usually not "clones" but right tuned professional individuals ...


There are no "detuning knobs" in a classical symphony orchestra. You know...?
And you can be sure: they tuned their instruments perfectly before starting the concert!!! Otherwise the conductor would be very angry and those musicians carriere could find an end quickly that way ...


A synthesizer is a musical instrument, right?
So they once borrowed the name from classical music, right?

Then they probably have to "borrow" that exact definition of the term too.

Cheers.


.

[Borogove]

The most Unisono definitions on the web are in Geraman language.

That's because the word in English is "unison".

I accept it, if an Unisono is produced merely by detuning the voices! Did not claim something different.

What I don't accept, ist the fact, that you could ever produce an acceptable or even justificable Unisono with merely one tone source (as an effect (thus: after the process of tone generation)).

Okay, let me take the next logical step. If you're running two oscillators which are identical except for pitch, you'd call that unisono.

Now, if those are straight-up wavetable oscillators, you'll get the same result if you ran two oscillators at different pitches as you would if you ran one oscillator, recorded its output, stretched it, and overlaid it with the original -- disregarding for the moment issues of aliasing and the different length of the stretched and unstretched recording. Agree so far?

[blümchen]

Well, Unisono IS ALSO NOT exactly a German term...


But agreed. (For the result of that process.)

But that is then not Unisono.
Merely the sound is (approximately) like the result of Unisono.

Unisono is the process ->, but not the result.

And cloning from one sound source has even absolutely nothing common with the process of Unisono. Which is per definition an event, where many different individual tone sources *are playing* together. And that at the time point of sound generation (together), not afterwards nor cloned or whatsoever.

Also the individuality of each of those tone sources is, what the sound of Unisono finally makes.

Thus Unisono, which sounds like a single source due to missing interferrences, but merely louder, is special case of Unisono then (if the requirements of the process are filled) and dosn't really exist in the real (analog) world (merely in the digital domain).

Even the real world positioning of those sound sources in space would not allow that.


(Unisono has allot common with Polyphony (rather than with chorus-like effects). Nobody really ever would consider to build a "Polyphony Effect", right? So an "Unisono Effect" is and remains quite nonsense.)

Make Uniclono effects or Harmonizers or Multi Chorusses or Layered Pitch Shifters or whatever, if you want. But don't name it "Unisono" and lay the clueless users. Because that IS NOT AN innovation.


By the way: You should really use better oscillators, than those you described above ...

.

[Borogove]

But that is then not Unisono. Merely the sound is (approximately) like the result of Unisono.

Let go of process for a moment. How is the output different between what I've described and what you accept as unisono?

By the way: You should really use better oscillators, than those you described above ...

I'm trying to keep the argument clear, not implement a synthesizer.

[blümchen]

What a game is that?

Do that with your childs, if you have any...
... or your clients, if you are a doctor of psychology.

(Read all again, if you didn't understand it. There is no reason to ask again and again the same stupid (and by the way: leading wrong) questions.)

Unisono is a process or event, not a result of anything. So it would be impossible to make an (afterwards acting) "effect" out of that.

q.e.d


Try to find logical arguments against mine, not doing that "childs game" here. But now it will be too late, because I am outa here. There is all said already. Because the rest seems to be stupid repeation of not understood arguments and facts.


Cheers and finish.