M/S oops with P8 - undo???

[springer]

Hi all,
This definitely goes down as biggest ooops so far for me.
Last week I was doing a simple jazz trio (2)UMT70s->True P8(channels 1 & 2)->MR1000

I got there late due to traffic and setup in a hurry. Monitored from edge of stage and wasn't able to hear much except levels were set ok. I wrapped up after the short 60 min. set and listened to it the next day. Sounds weird!?! No matter how I massaged the tracks I couldn't get it right. Can't put my finger on it - just "off".
Yesterday I dug out the P8 to play some more and noticed that the MS matrix was ENGAGED!!!! This explains it all...
Now - I am quite sure there is no way to run the 2 tracks thru something to "un-encode" it, but thought it might be worth a shot to ask and definitely worth it to remind P8 users to use channels 3 & 4 as a defacto standard when going to 2 track.
Feeling Homer-ish lately...
Patrick

[DrDeltaM]

Originally Posted by springer

Now - I am quite sure there is no way to run the 2 tracks thru something to "un-encode" it, but thought it might be worth a shot to ask

You can decode MS to Stereo again, no problem, you just won't be able to do it in DSD format tho. Most easy will be in PCM digital with some plugin. But you could do it in analog by running your audio thru the MS matrix once again. It's possible you'll have to half the levels after the MS matrix depending on the design of the matrix.

[just.sounds]

You could go to an analog mixer or ms matrix.

Analog mixer:

Split the S signal in two channels
pan them to the middle
flip the phase of one channel
adjust till it cancels
then pan the two channels L and R(the flipped channel R, stays flipped)
Mix the M in to taste panned to mid.

[locosoundman]

Split the S signal in two channels
pan them to the middle
flip the phase of one channel
adjust till it cancels
then pan the two channels L and R(the flipped channel R, stays flipped)
Mix the M in to taste panned to mid.

Maybe I am misunderstanding, but I think the problem is that the signal already IS encoded MS -> LR. Don't you need a way to separate the out-of-phase material (S) from the in-phase material (M)? Would just running it back through an MS matrix fix the problem?

Otherwise, the only way I can think to do it would be to record two passes: first sum the two tracks to mono, eliminating the S signal and leaving only the "M" signal. Record this to one track.

Then reverse polarity on the right channel and sum it with the left and record this to another track. I think this would get rid of the in-phase material (M) and leave you with the "S" only.

The "M" signal should correspond to whatever was in channel 1 on the P8 and the "S" should correspond to channel 2.

I should leave you with the disclaimer that I have never tried this - I would be very interested to hear the results.

[DrDeltaM]

Originally Posted by locosoundman

Maybe I am misunderstanding, but I think the problem is that the signal already IS encoded MS -> LR. Don't you need a way to separate the out-of-phase material (S) from the in-phase material (M)? Would just running it back through an MS matrix fix the problem?

He recorded in MS, meaning that on channel 1 he recorded M signal (L + R) and on channel 2 he recorded S (L - R), so these 2 signals are seperated already.

Originally Posted by locosoundman

Otherwise, the only way I can think to do it would be to record two passes: first sum the two tracks to mono, eliminating the S signal and leaving only the "M" signal. Record this to one track.

Then reverse polarity on the right channel and sum it with the left and record this to another track. I think this would get rid of the in-phase material (M) and leave you with the "S" only.

That's exactly what would happen if you run an MS signal thru an MS matrix again.

Originally Posted by locosoundman

The "M" signal should correspond to whatever was in channel 1 on the P8 and the "S" should correspond to channel 2.

Since this is the case, it will work fine.

[springer]

Easy Lads... Yes I recorded with the MS matrix engaged but I should add that it was NOT a MS recording in that I setup with 2 cards. I tried doing a couple of gyrations of splits and phase inversions, but I think it is a lost cause. I wouldn't get your knickers in a rough over this as I have conceded my loss.

[DrDeltaM]

Originally Posted by springer

Easy Lads... Yes I recorded with the MS matrix engaged but I should add that it was NOT a MS recording in that I setup with 2 cards. I tried doing a couple of gyrations of splits and phase inversions, but I think it is a lost cause. I wouldn't get your knickers in a rough over this as I have conceded my loss.

Did you try just running it back thru the MS-matrix?

The actual mic setup is irrelevant to the MS matrix, it doesn't know what mic you place where and how. All it does is these simple maths:

OUT1 = IN1 + IN2
OUT2 = IN1 - IN2

So if you'd run your 'encoded' signal thru it again, it will be decoded perfectly, as you'd get something like this:

NEWOUT1 = NEWIN1 + NEWIN2
NEWOUT2 = NEWIN1 - NEWIN2

Replacing those names with what you recorded (being OUT1 and OUT2 signal):

NEWOUT1 = OUT1 + OUT2
NEWOUT2 = OUT1 - OUT2

Replacing OUT1 & OUT2 with what they consist of (being combinations of IN1 and IN2):

NEWOUT1 = (IN1 + IN2) + (IN1 - IN2)
NEWOUT2 = (IN1 + IN2) - (IN1 - IN2)

Working this out gives:

NEWOUT1 = 2 x IN1
NEWOUT2 = 2 x IN2

Hence, you're back at your original signal (at double volume). Some MS-matrixes compensate for this volume half during 'encoding' and half during 'decoding' so that the result is same level.

Hope this helps!

[d_fu]

If L=M+S and R=M-S, then there is a way to reverse this, but the problem might be determining just how much of S was added to/subtracted from M... But in this case, they might have just been using equal settings, if you were recording ORTF or something like that.
You will get M by just panning L and R mono (mind your pan law). S should then be L-M, i.e. phase reverse M and mix with (or subtract from) L. Hope I'm getting this right... If I am, then this should give you back your original channels, the reconstructed M representing the original left channel and S your right channel. Proceed to use as applicable.

Do you remember the gain settings, were they identical?

Daniel

[DrDeltaM]

Originally Posted by d_fu

Do you remember the gain settings, were they identical?

Good point, if they weren't identical, things can get a bit more complex

[locosoundman]

Replacing OUT1 & OUT2 with what they consist of (being combinations of IN1 and IN2):

NEWOUT1 = (IN1 + IN2) + (IN1 - IN2)
NEWOUT2 = (IN1 + IN2) - (IN1 - IN2)

Cool - thanks for showing the math. You are right (of course) - it does the same thing as what I was thinking (just less complicated).

Just curious: how does the gain really matter in this case? Regardless of the actual values of IN1 and IN2, the relative results should be more or less the same, shouldn't they? I mean, mathematically, you still end up with:

newOut1 = (IN1 + IN1)
newOut2 = (IN2 + IN2)

I think it is a lost cause. I wouldn't get your knickers in a rough over this as I have conceded my loss.

At least give it a shot and tell us what happens

[DrDeltaM]

Originally Posted by locosoundman

Just curious: how does the gain really matter in this case? Regardless of the actual values of IN1 and IN2, the relative results should be more or less the same, shouldn't they? I mean, mathematically, you still end up with:

newOut1 = (IN1 + IN1)
newOut2 = (IN2 + IN2)

You're right, it's not an issue. I was a bit fast on acknowledgeing gains could be a problem, they just are of influence on the final signal:

If you leave the input gains on both inputs flat (or identical) on the decoding pass, you'll actually end up with the original mic feed signals ('L' & 'R' or just '1' & '2'), with each their gain being (twice) the gain set during recording. This is the closest to what you wanted to record. (note that "twice" obviously is twice the volume, not twice the dB-value)

When you change the gain of one of the channels in decoding, you can end up with following effects, depending on the signal content:
- if your original mic signals were a true stereopair, with both signals same gain, changing gain on one of the channels during decoding will change stereo width
- if your original mic signals were a stereopair out of gainbalance, changing gain on one the channels during decoding will shift the stereobalance, but will never compensate it to fix it into a good stereopair. To achieve a centered mid in this case, you have to change volumebalance between the new L & R signals after the decoding.
- if your original mic signals are independent mics, not forming a stereopair, changing gain on one of the channels during decoding will influence their relative volumes and their crosstalk.

So, conclusion: you'll want to decode with flat input gain settings, and then maybe change the balance between the newly formed L & R (or 1 & 2) signals, AFTER decoding.

[Remoteness]

Patrick,

Let us know how everything went.
I'm very curious on how this dilemma turns out.

[springer]

OK - the experiment is on...
I am going to try this ITB (in the box). I have made 2 mono tracks from L and R signals. And I have LPR (left phase reverse) and RPR (right phase reverse) mono signals to work with. Some combination of these is what we are after but which?

I will include excerpt of original and correction when I am done in MP3. Thank you to all who are trying to explain this to me!!! GS rules.
Patrick

[locosoundman]

I have LPR (left phase reverse) and RPR (right phase reverse) mono signals to work with. Some combination of these is what we are after but which?

On my first post I was thinking of an "inside-the-box" solution. Here is/are the combination(s):

record two passes: first sum the two (L & R) tracks to mono, eliminating the S signal and leaving only the "M" signal. Record this to one track.

Then reverse polarity on the right channel and sum it with the left and record this to another track. I think this would get rid of the in-phase material (M) and leave you with the "S" only.

so first pass: record the original Left and Right channels panned dead center; record this to a mono audio track. Second pass: just invert polarity (i.e. "reverse phase") on the Right channel and record again to another mono track.

Good luck!

[springer]

Hey loco... thanks for your help but it didn't sound right. So I tried this as was posted earlier...
=====================
Split the S signal in two channels
pan them to the middle
flip the phase of one channel
adjust till it cancels
then pan the two channels L and R(the flipped channel R, stays flipped)
Mix the M in to taste panned to mid.
=====================
and it sounds correct. I will work on getting the whole set mixed this way - please be patient.
Patrick

[springer]

Well this was a real learning experience and thanks to all of you masterminds of the GS board I was able to recover fully. Here are the samples of before (mess) and after (fix).

[Matti]

As we know FM stereo broadcasts are sent MS matrixed and the reciever
decodes the signal back to L-R ( at least here in europe )

Matti

[DrDeltaM]

Originally Posted by springer

Well this was a real learning experience and thanks to all of you masterminds of the GS board I was able to recover fully. Here are the samples of before (mess) and after (fix).

Sound like you did more then just decoding on this sample, also EQ-ing and maybe something else? Good to hear it worked out for you

[d_fu]

Originally Posted by MATTI

As we know FM stereo broadcasts are sent MS matrixed and the reciever decodes the signal back to L-R (at least here in europe)

It's a little more complex than that..
http://en.wikipedia.org/wiki/Fm_radio#FM_stereo

[Matti]

Sure

Matti

[springer]

You are correct - EQ was applied to final and the levels were not matched. But you get the idea...