Kjaerhus GAC-1: Mid-side routing?

[pekadan]

The routing called "mid-side setup" in GAC-1, how does it work? The signal is encoded into 2 bands, mid and side, the manual says. The two bands are processed separately then decoded back into L/R. What does the encoding do and what are these two bands exactly?

Side-note (mid-note being my question I guess... ): What an amazing plugin this is! I'm trying parallell compression on everything these days!

[kylen]

Have a look at this article:
http://paia.com/msmicwrk.htm

m/s is a nice little rabbit hole to stumble into - I'm still down it myself. Now you have a cool way to directly manipulate Left/Right/MID (mono)/SIDES which is really nice for repairs to existing mixes too if that's the only way out!

m/s is one of the earier ways to get in and more or less directly access the psychoacoustic "centre" of the stereo field by manipulating the MID. Even though MID is actually Left+Right it's a heck of a lot faster for me to touch that than trying to tune a center channel extractor or enhancer and deal with the phasey artifacts & such (ex: Adobe Audition Center Chan Extractor).

Very powerful for adjusting ambience too - independently compress and/or adjust the gain of the sides.

[pekadan]

Wow, really cool! This can certainly be useful I imagine. Can't wait to play around with it to get a better idea of the possibilites. Heaps of thanks for helping, kylen!

[Kingston]

Even though MID is actually Left+Right it's a heck of a lot faster for me to touch that than trying to tune a center channel extractor or enhancer and deal with the phasey artifacts & such (ex: Adobe Audition Center Chan Extractor).

Very powerful for adjusting ambience too - independently compress and/or adjust the gain of the sides.

You know, left+right=mid=mono. Quite simple. It's not the psychoacoustic "center" that our brain hears (the center extraction you mentioned), but the whole mono sum. In fact when you think about it, you're just treating the whole mix or bus (or whatever) as mono when you're compressing or EQing the mid "channel". Side-channel is a difference signal of left and right, that's literally everything that that is panned even slightly to left of right, usually a big part of the ambience/reverbs as well. Excellent bags of tricks for mastering, or stereo mic recording, for example.

[kylen]

You know, left+right=mid=mono. Quite simple. It's not the psychoacoustic "center" that our brain hears (the center extraction you mentioned), but the whole mono sum. In fact when you think about it, you're just treating the whole mix or bus (or whatever) as mono when you're compressing or EQing the mid "channel".

Yes - good point Kingston. It took me a while to figure out (and remember) that just cause I was doing something to the MID I wasn't actually manipulating a discrete center. So that's why I mentioned it and it's a good thing you mention it to hammer it home. The MID is the whole mix in mono - not the center with the sides removed.

I just noticed this recently when I was trying to remaster a cassette tape that had minor tracking issues - thinking that if I went to the MID there would be no side information. Well the m/s encoders don't do that - they fold the sides into a huge mono like you mentioned. The realm of the center channel extractors is to identify & extract the elusive "center". But that comes at too high of a price for me to fool with - sonically.

[Kingston]

But that comes at too high of a price for me to fool with - sonically.

(slightly off topic.. not with my implementation, it sounds clean. but it doesn't even run real-time, and only exists as a matlab simulation.. One of these days I have to crunch this algo to workable real-time solution.)

[kylen]

Yummy Kingston! I can bet it would eat cpus alive. You'd have to be able to run it real-time long enough to do a render I guess. Good Luck!

[Kingston]

I've been putting off doing a plugin out of it because it would mean I have to write my own low-level wavelet routine.. surprisingly enough I've always found "something better to do" everytime I even think about it.

[kylen]

Get back to work then!

[pekadan]

Thinking about this... I guess encoding L/R to mid/side, you could get pretty funny results if you're processing a stereo signal that's not mono compatible, no? I mean, certain sounds would get cancelled out in the L+R band. Parts of it would be brought back from the side band during decoding, but then probably sound weak and weird in stereo. Or?

[kylen]

...Parts of it would be brought back from the side band during decoding, but then probably sound weak and weird in stereo. Or?

You can punch the sides up directly in GAC-1 or with MSED:
http://www.voxengo.com/product/msed/

But if you think you're losing something in the stereo field you can always try and rotate it (another rabbit hole to go down) either pre or post m/s processing. The easiest one I use for that (when I have to) is C_SuperStereo:
http://userpage.fu-berlin.de/~thman/VST/vst-alleff.htm

BetaBugs has something I've tried briefly also to change the phase of the stereo field:
http://www.kvraudio.com/get/1767.html

My knees are shaking just thinking about messing around with that stuff!

[pekadan]

Thanks! It's not a problem I'm having actually. I was merely speculating a bit about what sort of surprises, if any, L/R -> M/S -> L/R processing could bring.

[kylen]

Got ya pekadan...one more thing. Just to keep my mind straight while in m/s I keep a test wav file around sometimes for when things start going crazy. It has a kick panned hard left, snare panned hard right, crash cymbol panned center and each one hits after the preceding ones tail has finished. I run that thru whatever m/s processing chain I've "invented". Sometimes I just need to hear what I've actually mangled! It's not always pretty...

[pekadan]

Good tip kylen! Thanks again!