KVR Audio

aciddose · Post by **aciddose** » Tue Mar 11, 2008 4:01 pm

"if you give me a signal plot I can totally draw in the perfect envelope"

you'd better learn some theory because what you've said is impossible. any amplitude modulation at any rate will add "distortion". amplitude modulation itself is a form of distortion. please don't be silly.

the only "perfect" envelope would be a flat line.

here is a simple way to observe that what i have said is true:

consider a sine wave at any frequency. apply by multiplication any shape, including another sine wave or a linear slope to that wave. observe that at certain points the delta of the wave has increased, at others it has decreased. this results in sideband harmonics. QED.

http://en.wikipedia.org/wiki/Amplitude_modulation

LOSER · Post by **LOSER** » Tue Mar 11, 2008 6:08 pm

aciddose wrote:"if you give me a signal plot I can totally draw in the perfect envelope"

you'd better learn some theory because what you've said is impossible. any amplitude modulation at any rate will add "distortion". amplitude modulation itself is a form of distortion. please don't be silly.

the only "perfect" envelope would be a flat line.

here is a simple way to observe that what i have said is true:

consider a sine wave at any frequency. apply by multiplication any shape, including another sine wave or a linear slope to that wave. observe that at certain points the delta of the wave has increased, at others it has decreased. this results in sideband harmonics. QED.

http://en.wikipedia.org/wiki/Amplitude_modulation

Sorry I said I'd shut up, but this is just too much.

So I'll reference the wikipedia article you so nicely linked me to:
The function of a modulated signal is y(t) = (A+m(t))*c(t), which makes the function of the envelope E(t) = A+m(t) (since the envelope contains the message to be transmitted, the +A to shift it up so it enwraps the modulated function and not just sits around the 0 line). Note this is the prefect envelope for the signal y(t) (because it is mathematically calculated, thus it has to be correct), and now by all means, this it NOT a flat line. (Or should I show you the matlab plot of the function as a prove?)

Though I'm getting your point, that EVERY dynamic alteration of a signal is de facto an AM (I know this, because I wrote my final thesis for my Abitur about AM and FM). BUT I wrote (to quote myself) that "[also] regarding your modulation theory thing ... IMO should there be no modulation going on on a steady signal since its envelope is constant[...]." Which I assume is basically what you meant with the perfect envelope being a flat line.

So there is nothing wrong there.

Once attack and release get involved it is a totally different thing, I won't, will never and never did argue that. But for signals with a constant level the envelope will (in a perfect case) be a flat line, which means there should not be any modulation.

Also I'm pretty sure you're not trying to mock me (you seem like a really nice guy), nor do I want to mock you, because I'm sure you know your stuff, but so do I. Plus I really don't want this to become anything nasty or something, that maybe just started because I may have misread, mistyped something, or you misread, misinterpreted something I wrote, etc...

So I said what's to say, and in case you want to further discuss this matter I'd suggest you PM me and include a FULL quote of the part of the post you refer to.

Regards,
Michael.

aciddose · Post by **aciddose** » Tue Mar 11, 2008 6:21 pm

my original point was that applying a test of distortion for a fixed signal will not represent the over-all distortion behaviour of the processor.

sure, a flat line can be generated for a flat signal, but there will always be a state in which the envelope makes a "mistake". in other words there will always be a case in which the "perfect for cases A, B, C" doesnt work, case D. so while measuring distortion on several flat tones of different compositions may be useful to judge how the processor will preform in those cases, it will be absolutely useless to judge the processor for it's performance on signals with dynamic content.

the point is: while it is certainly possible to generate a flat line envelope for a flat signal using various methods, and this is indeed the "perfect" envelope for that particular signal; the envelopes generated in other cases will not be perfect and may even be worse.

the guy posting the thread i'm sure didn't want to compress steady sine tones. a better test may be one which processes the same mixtures of sines with common dynamics applied.

ultimately: there is no single test which can be performed to give a "quality" index for compressors if your ideal is zero distortion because all amplitude modulation introduces sideband harmonics, meaning that a compressor yielding this ideal isnt simply ideal requiring infinities or so on, it's impossible.

Shy · Post by **Shy** » Tue Mar 11, 2008 6:54 pm

If it was possible, don't you think someone would have made it already? Then, we wouldn't have to argue about transparency. Heck, we'd have absolutely perfect, pumping free compression/limiting with absolutely no audible distortion without any effort..

Wait, I think that's what companies with over 50 years of experience keep trying to achieve.. they're so stupid.

William Sharkey · Post by **William Sharkey** » Tue Mar 11, 2008 8:26 pm

HRM.... Think of: clipped peak distortion-- that is a special case of a compressor. Like how a square is a kind of rectangle... (attck=0,ratio=inf,knee=0,release=0...). So in that case, the compressor with the least amount of distortion would be the compressor that's not doing what you asked!

And I agree that any form of envelope following multiplied (that's really what it is, perhaps scaled some way, but essentially multiplied) by an original signal creates distortion.

So if we are hunting for a product that controls dynamics and has low distortions, we picked the wrong category. I think that we want Limiters, no?

But if my guess is right, that does not silence the arguments, because:

The topic of least distortion still makes sense in a way for compressors--- In the case that we are using them for functionality equal to a limiter.

I agree, the language "envelope followers" limits the discussion. That IS a limitation in thinking, because it assumes we are using a real time product that is unable to read the future. Some products use look ahead. And it's conceivable to write a program that looks at entire tracks first.

This is my definition of the ideal transparent compressor/limiter: A product that makes the same decisions and outputs the same audio as I would have through manually tweaking the volume of the track.

(notice how the definition does not include the concepts: attack, ratio, knee, release etc. That's because I'm trying to de-couple my ideal definition from current real-world implementation. Easy to get trapped in old thinking.)

William Sharkey · Post by **William Sharkey** » Tue Mar 11, 2008 8:29 pm

delete

William Sharkey · Post by **William Sharkey** » Tue Mar 11, 2008 8:53 pm

aciddose wrote:"if you give me a signal plot I can totally draw in the perfect envelope"

you'd better learn some theory because what you've said is impossible. any amplitude modulation at any rate will add "distortion". amplitude modulation itself is a form of distortion. please don't be silly.

By perfect envelope he meant something different. Of course he did not mean something that does not produce distortion. He meant a perfect envelope straight up.

Specific example: You have a sine wave generator. The sine wave generator has an amplitude control knob. So you play the sine wave, and hit record on your daw. Now you wobble that amplitude knob while it is recording.

Then you play that recorded audio through an envelope generator. A perfect envelope generator would output exactly the same values as you wiggled on your volume knob.

That is a very difficult thing to do, because it requires guess work (similar to uncertainty in position vs speed of particles). The exact same audio that was recorded could, infact have been produced by different envelope generation means. What if, instead, there was a constant dc offset, but you happened to be able to modulate it's amplitude at audio rate, to archive the same audio output as before.... that's two wildly different envelopes that could describe the same audio right there!

So an amplitude envelope should attempt to reconstruct the forces that we commonly associate with amplitude. (It should reconstruct how hard we hit that piano key for example) So a "perfect" amplitude envelope generator would be some sort of creature, capable of traveling back in time and observing the situation that created the audio in the first place.

It needs to know context, so that it knows how tightly to envelope the waveforms.

aciddose · Post by **aciddose** » Tue Mar 11, 2008 10:11 pm

what you're saying is exactly the point i was making though - there are multiple ways that a particular apparent envelope could have been created and this means there is no "perfect" envelope outside of a very defined case, like your volume knob case. if we want to define things this way that is fine, but testing to see how well a compressor meets your definitions of "ideal" doesnt give you a general quality score. you only get a score for exactly how much the compressor meets your specific criteria. different compressors are going to sound more "transparent" in different situations, so there is no "most transparent" compressor.

take for example two sines very close together, beating. what you get is an apparent sine tone amplitude modulated at the difference between the two tones. if our compressor were to "perfectly" crush this tone, eliminating sidebands, we will be left missing one of our tones - obviously not "perfect". also, which tone would be kept and which will we take away? which side-bands should be removed?

there is no definition of an "ideal" envelope or "ideal" compressor. such a thing is impossible.

we know based upon the tests in this thread that a certain compressor does the least to a steady waveform with up to apparently three zero crossings. what about four? five? six? when does the envelope generation method this compressor uses break down and make a "mistake" ? all we have seen here is that for steady sine tones and steady sine tones only, a certain compressor does the least compression.

bduffy · Post by **bduffy** » Wed Mar 12, 2008 4:15 am

Math war!!!

dickiefunk · Post by **dickiefunk** » Sun Mar 16, 2008 5:58 pm

Ok there's been over 3000 views and only 78 votes

ganym3d · Post by **ganym3d** » Sun Mar 16, 2008 7:11 pm

deleted

ganym3d · Post by **ganym3d** » Sun Mar 16, 2008 7:13 pm

fred2112 wrote:http://waves.com/content.aspx?id=237&l=4

Hmmmm... Not really FREEWARE, isn't it?

Hink · Post by **Hink** » Sun Mar 16, 2008 7:27 pm

hmmm....not going to get into the war of specs, but I want to point out that there is a whole category of free comps that seem not to be mentioned (I skimmed the thread so I could be wrong).

Without a host a comp is kinda useless, so I'm just going to say it's a safe bet everyone here uses a host

. With that in mind the comps that come with different hosts are in one sense of the word "free". I can't speak for every host, but I can for the ones I have and I think that there are some good comps. Samp has some great comps and a very nice multi-band dynamic processor. The same with Sonar, FL and AA (AA2.0 has a nice multi-band made by izotope).

Debutante · Post by **Debutante** » Sun Mar 16, 2008 8:48 pm

Might the DVC have made the list?

rifftrax · Post by **rifftrax** » Mon Mar 17, 2008 8:39 pm

jens wrote:
Rottweiler wrote:
jens wrote:it seems you're out of luck...

I looked for a license-file but didn't find any - I'm not sure whether I accidently deleted it or if it came without one - the manual (which I have) doesn't touch this point - sorry.
license file?

i thought it was FREEWARE

Freeware usually also has a license.

um...no.

KVR Audio

Cleanest, least colored freeware compressor? POLL ADDED!!!!

Most uncolored freeware compressor