KVR Audio

BertKoor · Post by **BertKoor** » Wed Oct 10, 2012 2:56 pm

So then what IS the correct formula to derive the true mid channel then? Must be possible to do without FFT...

hvz · Post by **hvz** » Wed Oct 10, 2012 3:01 pm

I don't think there is one.

Consider 2 channels in perfect anti-phase. Then what should the mid channel be? I don't think there's a single perfect answer to this. If you choose either channel the sound will be good. So there are actually 2 completely different "best" solutions - unless you can think of one that's better than randomly picking one of the channels...

hvz · Post by **hvz** » Wed Oct 10, 2012 3:39 pm

I'll make it a bit easier (or harder, depening on how you look at it):

Take a one sample pulse, in anti-phase. So there's silence with 1 non-zero sample. And the question to you is: What should the center channel be?

loopdon · Post by **loopdon** » Wed Oct 10, 2012 5:17 pm

hvz wrote:I'll make it a bit easier (or harder, depening on how you look at it):

Take a one sample pulse, in anti-phase. So there's silence with 1 non-zero sample. And the question to you is: What should the center channel be?

Trying to follow the discussion. Apart from that - Stereo Tool - i did not know til today and i think it is absolutely great. Just needed to get that out, even if i have only touched the surface!

Ah_Dziz · Post by **Ah_Dziz** » Wed Oct 10, 2012 5:18 pm

hvz wrote:I don't think there is one.

Consider 2 channels in perfect anti-phase. Then what should the mid channel be? I don't think there's a single perfect answer to this. If you choose either channel the sound will be good. So there are actually 2 completely different "best" solutions - unless you can think of one that's better than randomly picking one of the channels...

in this scenario there is no mid channel. Test it and see.

gavriloP · Post by **gavriloP** » Wed Oct 10, 2012 5:29 pm

BertKoor wrote:
docdued wrote:Mid = L+R
Definitely incorrect, that's just the sum of both. A hard-panned left or right signal does not belong in the Mid channel, it's in the Side only.

This is totally wrong. MID is L+R and that is the standard way to collapse stereo signal to mono. If what you say would be right, most of sixties stereo recordings would be more or less silent in mono

Only when L is 180 degrees out of phase with R, the mono result becomes silent. If you have track panned 100% to left it can't be heard from right channel, but when it is converted to mono (with DAW or even amplifiers stereo/mono switch) you hear that hard panned stuff, but only little quieter.

dreamkeeper · Post by **dreamkeeper** » Wed Oct 10, 2012 5:54 pm

hvz · Post by **hvz** » Wed Oct 10, 2012 7:04 pm

Ah_Dziz wrote:in this scenario there is no mid channel. Test it and see.

I know, the discussion however was about trying to find a formula to generate a mid channel without loosing audio. I was using this example to show that there cannot be a single perfect solution.

sauli · Post by **sauli** » Wed Oct 10, 2012 7:10 pm

s_t wrote:
docdued wrote:...it's impossible to precisely isolate the "center signal" in the stereo field. Which is kind of weird, because "in our head" we can do that (e.g. we hear that there's only the drums and the bass sitting exactly in the center of a stereo field recording).
Long ago there was a plugin that achieved this. If I remember correctly, it used FFT processing and suffered from some artifacts because of that, but it worked.
Unfortunately I can't remember what was the name of the plugin and who made it.
cheers

PS: Maybe our brains use FFT

Anyone remember what this is?

I asked for something like this in the forum long time ago and only got m/s plugin suggestions from people misunderstanding me

hvz · Post by **hvz** » Wed Oct 10, 2012 7:19 pm

sauli wrote:Anyone remember what this is?

I asked for something like this in the forum long time ago and only got m/s plugin suggestions from people misunderstanding me

It could be mine - it does use FFT's and merges stereo to mono without loosing sounds: http://www.kvraudio.com/forum/viewtopic ... c&start=26

Harry_HH · Post by **Harry_HH** » Wed Oct 10, 2012 7:23 pm

hvz wrote:
Ah_Dziz wrote:in this scenario there is no mid channel. Test it and see.
I know, the discussion however was about trying to find a formula to generate a mid channel without loosing audio. I was using this example to show that there cannot be a single perfect solution.

This all is very interesting, although much of this is purely intellectual.

There might be several target with people joining this discurse.
My point was not that much to find a forumula to generate a mid channel without loosing audio - more to find a formula (or rather a plugin) to generate a mono file from the stereo file without loosing audio.
Only if mid = mono the above is correct, but then it's not the mono I mean (or call "supermono" if you want, i.e. mono file which includes all the stereo information (or if not "information", at least includes all the sound which can be heard in the mid+side of the stereo file.). H.

hvz · Post by **hvz** » Wed Oct 10, 2012 7:26 pm

Lol, I like the name "supermono". I called it "Flat Stereo" because it IS mono but if you're not between the speakers it sounds as full as the original stereo recording.

gavriloP · Post by **gavriloP** » Wed Oct 10, 2012 7:53 pm

Harry_HH wrote: This all is very interesting, although much of this is purely intellectual.

There might be several target with people joining this discurse.
My point was not that much to find a forumula to generate a mid channel without loosing audio - more to find a formula (or rather a plugin) to generate a mono file from the stereo file without loosing audio.
Only if mid = mono the above is correct, but then it's not the mono I mean (or call "supermono" if you want, i.e. mono file which includes all the stereo information (or if not "information", at least includes all the sound which can be heard in the mid+side of the stereo file.). H.

Problem is, that L+R is the most optimal mono. When you add Side to it, you are always decreasing it. Sound changes and maybe sometimes for the better but it is bound to be weaker. The only case where one should need side channel information is the case where the original stereo mix is so messed up with widening that the difference is greater than sum. But that is one phasey horror scenario, and I cannot think a situation where it would occur in normal mixing.

To recap: if your samples or tracks sound much weaker in mono than in stereo, there is bound to be some phase issues. Then you can try MS EQ or indeed that MSED trick that I wrote about to heal the sound.

Like BertKoor mentioned before, it is important with mono subwoofer. If your lowend is out of phase (the more it has that "side" channel data, the worse it is) then when your sub filters and combines your stereo signal to L+R, you lose bottom, and in worst case (low end totally out of phase) there is no bass at all from the sub. Oh, and if you have sub output from amp, then it is the amp that does the L+R summing.

In real life, I've never encountered this kind of problem but it is possible.

Ah_Dziz · Post by **Ah_Dziz** » Wed Oct 10, 2012 8:37 pm

Well if the goal is to have all information collapsed to mono without losing any then you will need a tool to adjust the phase of the side signals so that they are no longer out of phase. This could be done with a phase vocoder but it will sound different than the original. Another method is to use tiny delays on one half of the side signal until you get minimal cancellation. However you will not be able to combine the out of phase (side) portion of the stereo signal to mono without loosing it unless you are willing to modify the signal first.

hvz · Post by **hvz** » Wed Oct 10, 2012 9:22 pm

gavriloP wrote:Problem is, that L+R is the most optimal mono.

I don't agree. In many songs there are stereo effects that you don't want to loose when going back to mono. Think of reverb on voices for example.

For normal recordings of - say - rock bands, there's very little difference between a normal (L+R) change to mono and one that keeps all sounds. But things are very different if you play for example (electronic) trance music. Changing that to mono by just averaging the channels causes a huge loss in audio - the result sounds very 'thin' - basically you keep the bass and highs but loose most of the mids.

By the way, IF you use my algorithm on music with little phase problems, there is no noticeable difference with the L+R method. It only does different things when there are problems in the first place, so that's why I would say it's 'more optimal' than simply using L+R.

Stereo to mono plugin wanted