Brainworx bx_digital and Neutrons Xenon Codec

[symbiote]

On 2006-02-26 09:39, Shroomz wrote:
btw . Alfonso's Example M/S modular patch has potential as a processing tool (route the mid signal from before the insert also to output 3 etc )

Totally! For processing you don't really care about phase as much/at all, and for basic tasks like balancing M vs S (boosting/cutting one or the other), the simple M/S devices are perfectly fine at the mastering stage (might want to test it out with some test/reference signals first if you are going to print 100,000 CDs , but should be fine.)

You just have to be careful if you process the whole mix, since you'll be affecting low frequencies along with everything else. But, for example, you could add some very slight reverb to just the S signal to stereoify things a bit, without any problems, without the need for a 600 euros device.

That algo you posted looks interesting, but he says himself it's not mathematically correct, so probably not a good idea for mastering . FFT itself doesn't have perfect reconstruction even, so definitely not a good idea.

His phase-correction routine is related to the method he uses, i.e. FFT, but he doesn't really explain the logic behind it. It might have just been cooked up by trial-and-error while he was developing it.

The first part is figuring out the phase difference between left and right signal, divided by pi to have it over the 0-2 interval instead of 0-2pi (since the atan2 stuff returns 0-2pi stuff, you know, rad vs degrees etc.) and then making sure the value is between 0 and 1 (in other words, taking the angle on the positive "y" side of the circle (eh?))

But pow(1 - u*u*u, 24)??? I dunno. I like how it looks (pow! u*u*u 24) , but I don't get it. He uses it to scale his previously-found center values though. Since "u" will have a value between 0 and 1, cubing (is that a verb?) will also be between 0 and 1, and so pow(24) will also be between 0 and 1.

That's awesome! But I don't get it. Sorry! Maybe it's the mantissa part of a float (32bit with 24bit mantissa,) or he assumed 24bit signals, or he tweaked the value of the pow! until it sounded nice, or maybe the aliens forced him with mind-control rays?

Who knows. I'm way too lazy to try it in Scilab. And besides, I don't have my tinfoil hat on so the results wouldn't be reliable!

[wolf]

If you load a pair of normal equalizers as inserts in that device, and EQ the M and S signals differently, you will definitely get some phase issues when you reconstruct the L and R signals from M and S.

nope, phase differences in the m/s-domain translate into different left/right placements and perhaps (slight) frequency manipulations in the stereo domain. You don't affect the phase in the resulting stereo output. However that hihat, which was previously on the left side, might be on the middle-right side now and the whole mix sounds a bit more mono

Not to mention having to set the sample delay by hand every time might be a bit of a hassle, and might change with the EQ's parameter, e.g. when adding/removing peak/etc EQ in a parametric EQ, the total sample delay imposed by the plugin might change, forcing you to re-adjust the delays. It's a bit distracting in a mastering situation.

The funny thing in scope is, it really doesn't matter, how the eq's bandwidth frequency/bandwidth/gain is set, but it is crucial, where the device is inserted and how much bands are activated (at least when used with my eq ).
However, the panning difference is easily checked by ears and, believe me, needs only one attempt.
Another task is to do this automatically ..

The easy way to fix this is to just use a minimum/linear phase EQ

well, this makes sense in the stereo domain, but what is used in m/s-domain, are two mono eq's, not one stereo eq. It doesn't make sense to link each filter in regard to phase, as the mid path might have other settings as the side path. There are other problems to solve

cheers
Wolfgang

[garyb]

thanks all.

[symbiote]

nope, phase differences in the m/s-domain translate into different left/right placements and perhaps (slight) frequency manipulations in the stereo domain. You don't affect the phase in the resulting stereo output.

That's pretty much what I meant. By "phase issues", I did not mean that the phase of the stereo signal was going to be affected, I meant that the L and R signals' reconstruction were going to be affected by the phase difference in the M and S signals (i.e. slight freq manipulations.) "Slight" is usually enough to make a difference at mastering . Sorry if I wasn't clear.

I shouldn't have used the term "phase-correction", more like "correction of the effects of phase differences in M and S signals at the L/R reconstruction", but that's a bit tedious, and anyway I am notoriously lazy!

The easy way to fix this is to just use a minimum/linear phase EQ

well, this makes sense in the stereo domain, but what is used in m/s-domain, are two mono eq's, not one stereo eq. It doesn't make sense to link each filter in regard to phase, as the mid path might have other settings as the side path.

Yep, I was pretty much assuming different EQ/other settings on both paths. I was just saying linear/minimum-phase EQ would take the effect of phase difference due to EQ out of the equation. With good enough ears/monitoring, it doesn't matter too much, but if you do mastering on a regular basis, that's a hassle you might not want to deal with too often. With lin/min-phase, you can just boost M and cut S and that's what'll happen, instead of having to also compensate for panning/balance issues.