Topic: MS Mix Question

[china_rider]

Hey all... I record Al Howard a few weeks ago.  I set up MS on stage and had both channels set so they peaked at about -1.  So when I do the MS decode the mix sounds best to me when mixing both mid side at a positive DB setting.  The problem is that doing so with the Voxengo plugin raises the levels over +0.

Logically I would think that as long as the ratio of mid to side is equal, a positive adjustment should produce the same stereo mix as a negative one.  Going down those lines I figured I could use the negative mix and then normalize for the final result.  It may just be in my head, but when I do this the positive mix sounds fuller than the negative normalized version.

Is the difference all in my mind or am I thinking about this whole thing wrong?

Thanks,
Dana

[easy jim]

I do not know the Voxengo plug-in at all, but it seems normal/logical that the summing during the M/S matrix process is producing levels above +0.  The 'fuller' part of the sound may actually be distortion from overdriving the plug-in.

Theoretically, you should be able to lower the dB gain of both channels evenly by 1-2 dB prior to plug-in and achieve the same mix (sans distortion) as long as the m:s ratio remains the same.

I assume this is a pre-mix of the stage source before mixing with the SBD tracks for a stage+SBD matrix? 

[rowjimmytour]

I agree and w/ my TL's>R4 running MS I peak at -6dB and then for the mix most of the time run mid close to +1>0 and side +1>3 w/ the voxengo plug on SF 8.

[china_rider]

Agreed that I should have run my levels lower.  I guess my question is as long as the ratio of mid to side is the same the actual stereo mix should be the same just at a lower level, correct? 

T+ for the help.

[easy jim]

^ It should be.

[china_rider]

Ok... Now that I am thinking (Oh no).... Maybe the lower levels are not better. 

So given that you are able to achieve the same stereo mix as long as the decode ratio is the same maybe it is beneficial to run the channels hot just as you would in a normal recording?  Each pre-decode channel would have more detail and you can just mix down during the decode?  Probably more of a mixing pain but does it give you added benefit?

[easy jim]

Ok... Now that I am thinking (Oh no).... Maybe the lower levels are not better. 

So given that you are able to achieve the same stereo mix as long as the decode ratio is the same maybe it is beneficial to run the channels hot just as you would in a normal recording?  Each pre-decode channel would have more detail and you can just mix down during the decode?  Probably more of a mixing pain but does it give you added benefit?

Well, if you are adjusting gain (up or down) in addition to just allowing the m/s signals to sum, you are theoretically introducing a greater possibility of quantization errors.  So, there is certainly something to the concept of getting your levels 'right' during the recording/capture stage.

Assuming you're then going to mix this m/s source with another two tracks/stereo track (SBD?), you'll also want to keep in mind that there will be additional summing when mixing the decoded m/s stereo track with the second stereo pair.  If your decoded m/s stereo track peaks above ~ -2 dBFS or so, you'll probably want to back it off some (reduce gain) before mixing it with another stereo pair (SBD tracks) so that the subsequent summing does not also push the levels of the final mix above 0 dBFS.

I think Jim was right on above in suggesting that ~ -6 dBFS is a good level to shoot for as your highest peak for both the raw mid and raw side tracks - pre decoding - especially if mixing with a second stereo pair.  The extra headroom is handy, and may prevent the need for any more than a slight gain adjustment on the final mix to peak normalize.

[SmokinJoe]

I don't use your software so my workflow is different... bottom line is that you probably need to drop both mid and side by 3db to avoid clipping in mix.

In a related question/observation... which may be relevant here:
When I record a show Mid/Side with my LSD2... I notice that generally I am getting a lot more sound coming in from the Mid than the side... but I'm cranking the knobs like I was recording XY... in other words to get target levels I have to run the gain a lot less on the Mid than I do on the side... i.e. my left (mid) gain is at 9:00 and my right (side) gain is at 12:00 to both peak at -3db let's say.  That's just the way it is... if I'm set up balcony rail, a fair distance from the stacks, more sound is hitting the front than the side.  That's reasonable.

The issue is that by trying to get -3 on both gains I have artificially added about 12db gain to the side with respect to the mid. Then I get home, try to do a 50/50 mix, and of course I have way too much side because I boosted it during recording.  In this particular recording http://www.archive.org/details/gptn2008-02-23.lsd2.flac16f I did the mix with the mid about 9db more than the side (to compensate for the overly boosted side during recording), and I think it's still a little bit "phasey" from too much side.

I think next time I'm going to try setting the the gain knobs the same... both at 9:00 for instance.  The side channel will be peaking at about -15 and the mid at -3, but then when I come home it will make more sense.

Thoughts?

[DSatz]

This type of question is a little frustrating because each piece of software is different, and the limitations of one program's feature or plug-in may not apply to another's.

Generally, of course, you want to record at more or less the highest levels you can without risk of overload. The question is really whether your M/S matrixing software is written in a way that avoids overload during the sum-and-difference arithmetic that it has to perform.

Fortunately the task of an M/S matrix is so simple that you can suss it out by simple means. In most editing software there is some ability to generate test tones. So you can generate a stereo pair of sine waves (say, at some midrange frequency such as 500 Hz) at some exactly equal level between -3 dB and 0--what would ordinarily be a mono signal, since it's the same in both channels. But if you treat these as M and S inputs and run the M/S matrix on them, you will see whether or not one channel is driven into clipping as the two (identical) waveforms are added together.

Some software has an output level control, but where that control fits into the process is not entirely clear until you experiment. It might "pre-scale" the inputs, or it might allow the inputs to matrix and over/underflow, then "scale" that (already distorted) result. If it's the latter, then you have to take care of your recording levels before matrixing. Ideally the documentation would explain this, but who am I kidding ...

Several things that you might see include: One channel clips (flat tops and bottoms of the waveform); one channel actually wraps around when the sum exceeds full scale (very bad for the sound quality!); and any gain control that you have might help avoid either type of clipping, or it might simply shrink the size of the clipped waveform after the damage has been done.

--Agreed, for certain recordings and tastes (and some playback equipment and listening environments) some relatively modest amount of low-order harmonic distortion can enrich the sound. I don't suppose anyone here has heard one of them newfangled electrical guitars, for example? These kids with their "rock and roll" music? Well, I heard--it's just a rumor, but--I heard that sometimes the players actually drive their electrical amplificators into distortion on purpose. They even step on petals (?) to do this, which just desecrates the flowers that they came from, I would imagine. Where will it end? Where's their sense of dignity?

--best regards

[carlbeck]

Well I am curious as well, when running MS is it best to just leave the levels on your pre or recorder while recording even like Joe suggested, i.e. both at 12:00 or should you get both channels to read even amounts of gain?

[DSatz]

carlbeck, if you're directly recording the M and S signals for later dematrixing, then you generally want both channels to peak at the highest recorded level you can get, as long as you don't risk overload.

My own experience with classical music recording has been that the S microphone (the sideways-facing figure-8) is a little less sensitive than the forward-facing microphone (cardioid, supercardioid, etc.) and also, that the peak sound levels at the sides are naturally not as high as the peak sound levels front and center. As a result I generally have the gain turned up several dB higher on the channel that's recording the S signal than on the channel that's recording M.

Choosing to let your S channel peak at a lower level is just throwing away signal-to-noise performance for no good reason.

This assumes, of course, that in playback you can adjust the gains of the two channels before their signals go into the matrix. But you need that ability anyway, so that you can choose the ratio of the two signals that sounds best to you over loudspeakers--the best combination of image width and reverberance.

--best regards

[anhisr]

yeah, what he ^ said.  That is how I run mine.  At the Pageant I run almost 5db higher on the side then the mid.

[china_rider]

Hey all... Thanks for the feed back.  More kudos to DSatz who always seems to enlighten me no matter what question I'm asking.  T+ to all.

From what I've got out of the responses here is it should be best to record each channel at the highest possible level without clipping (which I was doing) so you don't throw away any signal to noise benefits.  That being said the ease/quality of mixing those signals with higher levels really depends on the software package used.   I really did not realize there was that much difference between the way that different software packages handled the MS mix along with other post processing options.

For my particular Al Howard show I was able to get a great mix lowering the DBs with the Voxengo.  While I have the time I want to take the masters and redo the mix with a few other software packages just so I can see the differences for myself and determine what workflow I prefer.

My next task will have to be taking the SBD recorded on a R-9 and stretching it to make a matrix of the show.  I'll post the results in Kickdown for anyone interested. :-}

[china_rider]

DSatz a question for you...

... In one of your previous comments you mention software that has output level control and the uncertainty of where this is actually applied.  I was wondering when taking a MS recording that needs to be decoded, in general, what is your preferred choice of software is and your ideal work flow?

I'm not sure if what I am using can limit the output level so it does not clip in the MS mix process... But currently I have been using either WavLab5 or Audition3 with the Voxengo MS plugin.  With both packages when just using the plugin and adding DBs in the MS mix the resulting decoded wav comes out as clipped (flat wavforms).  What I have basically been doing is finding the ratio of mid to side that I think sounds best, then before doing the actual MS mix I adjust the gain on both mid and side so the resulting MS decode gets as close to 0db as possible without clipping.  It seems to be a matter of trial and error and I end up doing the gain adjustments and the MS mix a bunch of times (on the original wavs) before I get one I am happy with.  Once the decode is done I do any other processing that may need to be done.

I guess that really what I am asking is that what I am doing currently seems to be based on quite a bit of trial and error and I'm not sure if I am sacrificing quality in the process.  Do you have any tools/processes/tips that you use that helps with the process?

[SmokinJoe]

Well I am curious as well, when running MS is it best to just leave the levels on your pre or recorder while recording even like Joe suggested, i.e. both at 12:00 or should you get both channels to read even amounts of gain?

I think the answers given are correct... you might as well record at highest levels possible to get less noise, but when it comes to mixing, don't just blindly mix 50/50 of these tracks because that would be too much side.  I read somewhere use 3:1 ration of Mid:Side, and that makes sense if you assume the Side was boosted more than 3:1 when you started...

I think it's like "salt to taste".... "mix to a width you find pleasing".

By the way... for us poor folk who don't have some fancy program with M/S decoder... I do it in audacity... (I didn't invent this, I read it here somewhere).
- pull up file (mid is left, side is right)
- duplicate track (mid is on both lefts, side is on both rights)
- split both stereo tracks into mono tracks
- swap the 2nd and third tracks... move #3 up, or move #2 down.
- invert #3
- combine top 2 into stereo tracks, and combine bottom 2 into stereo tracks.
- At this point you have a stereo track at the top with MID in both channels, and on the bottom you have the inverted side and side.
- You can play this back and play games with the volumes... like -3 on the top and -12 on the bottom.  Listen and play until you find the mix you like.
- Save wav file as... your mix.wav