Topic: Matrix problem - phase cancellation?? - please help

[yates7592]

I made 2 stereo recordings of a show (4 channels) and when played as 2 shows (i.e. 2 channels in separate files) they sound absolutely fine. However, when I merge them in a 4 channel matrix I am getting intermittent areas where there seems to be phase cancellation or inversion for want of a better term. The  sound changes to a tinny sound, I lose a lot of the bass and punch, and it just sounds crap. As the 2 independent sources are fine I assume it is the mixing effect and some kind of cancellation. Both files are stereo, but the mics were all on my lapels so there's not a lot of separation there. I am using Wavelab 6, not that i think this is the source of the problem. Any ideas anybody? Any advice gratefully received!

[kcmule]

Have you tried reversing the phase on one channel of one source?

[page]

First off, did you use two separate recorders or a recorder with a single clock.

Second, to some degree you will always get phase cancelation when mixing sources, the degree of such is dictated by how much common information is in both (the more common the information, the worse the cancelation). An onstage recording and a sbd feed don't have as much in common as two mic sources where the two sets of mics are in the same general area. Regarding the later, unless there is something specifically bad about both, I'd pick a single set and just play with that one, but ymmv.

So assuming that you still want to do this, and stuff is clocked correctly, and as kcmule points out, make sure the polarity or phase is correct amongst the channels, then you can line them up as best as possible (I recommend synching on a snare hit to start with personally), and start delaying one source over the other by 2 or 3 samples and listen. I've found a VST plugin works best as you can let it play and change the delay and listen to it. If you're plugin works in terms of milliseconds, then adjust by 1ms at the most and listen. I've found thats enough to hear shifts in balance and cancelation when mixing. good luck.

[Gutbucket]

Since you indicate that the mic pairs were in close proximity to each other there should not be large time arrival differences which could cause sigificant time-based incremental phase differences.  Assuming the two files are correctly synced, it's most certainly swapped polarity on one stereo pair leading to cancellation.   If you recorded all four channels on the same recorder then the files will be synced without needing to do further work to sync them.  In that case, invert polarity on one pair only then try mixing them again.

If recorded to two different machines that were clock synced to each other, you'll need to align to two sources in time as Page mentions.  If recorded to two different machines which were not clock synced, you may also need to stretch one source in addition to just lining them up, else they may slowly shift out of sync again as the recording plays.

[yates7592]

Thanks for the speedy advice guys!

I used 2 separate recorders. I used one show in one big file, and then chopped up the other and aligned each track incrementally to get a good time match between the 2 sources throughout. I think the were very well matched in time.

I haven't reversed the phase on one channel yet but I wil try. As this phasing sound comes and goes, it might solve it in some parts, but not others? Should I do this on both channels of one source, one from each source, or just one from one source?

I will also try a delay plug in. Should I apply this as part of the matrix-forming process (i.e. before the matrix file is finalised or fully rendered) or as a separate exercise in post on the newly formed matrix file?

[Gutbucket]

If it was simply reversed polarity of one stereo pair, you'd swap polarity on both channels of one pair or the other and you should get solid bass when both sources have the same polarity.

However, you just stated that the phasing sound comes and goes, and that indicates your chop & align method of clock syncing isn't accurately syncing the two sources sufficiently.  You are probablyl hearing phase interactions between the two sources as they drift slightly in time relative to each other (assuming the 'phasing' sound is not present in each source by itself).  That is a drift on the order of milliseconds.

I think you need to use another method to re-sync your sources to eliminate the phasing.   Try:

1) Time stretching one entire source to match the other instead of matching chopped chunks, zoom into the waveform check that transients are closely aligned at both the begining and end of the files.  There are different ways to do that depending on the software used.  Some are simple and some complex.  This is your best best.
-or-
2) The manual cheat verison is to play both files back in realtime off the original recorders using their original clocks.  Each clock will compensate for it's own error for the most part.  I used to do this for playing back two simultaneous seperate sources for multichannel listening.  That required aligning the two manually with quick jabs of play/pause and a careful listening.  Pretty crazy.   In your case you could do something less crazy which achieves more accurate results than jabbing by playing back one source from it's original recording machine and re-record its analog output to the other machine (doing this makes an D/A/D conversion or requires resampling for a digital transfer).  Then take that resulting recording and time align it with the ohter source (which was originally recoded to the same machine) as Page describes.  The two should then drift only minimally.  Obviously doing 1) is better, but 2) is simple.. just takes time for real time transfer and introduces an analog stage or a resample.

A time delay plugin just helps to optimally align the two sources once they are clock-synced.  It's an easy way of nudging things slightly one way or another after aligning transients by zooming in.

[yates7592]

Yes, I think you're dead right, the in and out phasing must be something to do with time, not polarity. I will look into the stretch method and get this project sorted! I think the chop method may be ok on short tracks but most of my tracks were 10-15 minutes each. Thanks again for all the help.

[Gutbucket]

No problem, hope it helps.  This is a big reason why using a single machine whcih can record 4 channels become very attractive to tapers doing lots of matrixing, and makes the process much faster and easier.  Stretching is a PITA.  Time-aligning is less so, but still.  A 4 channel recorder takes care of both syncing and aligning the channels, but sometimes you can still get polarity inversions on one pair if some other gear in the chain inverts.  I had this happen on Sunday.  Both (co-located) pairs sound fine alone, but the bass drops out when mixed together.

One visual indication of polarity which you can look for when zoomed way in to align transients is checking if the waveform moves in the same direction for that transient on each channel.  For most things centered in the stereo fileld, the waveform should spike up (positive) and go down (negative) more or less symetrically at the same point across all tracks.   If you use mic configs with spacing between capsules, the Left/Right waveforms won't line up perfectly for sounds arriving from off center, but if it sounds aligned, yet the waveform of one pair looks upside down compared to the other when you zoom way in, that pair may be polarity reversed.

[runonce]

Have to wonder if the lapel mount isnt in play here - if you move, the phase shift might sound more exaggerated in a matrix...

[page]

Have to wonder if the lapel mount isnt in play here - if you move, the phase shift might sound more exaggerated in a matrix...

possible, but having stuff that isn't clocked together will contribute to it as well.

[Gutbucket]

Have to wonder if the lapel mount isnt in play here - if you move, the phase shift might sound more exaggerated in a matrix...

Could be. It depends on how close together the same-side mics of the two pairs are to each other in combination with that movement.  If the phase shifting isn't audible when listening to each pair alone and the two Left mics were co-located close to each other (same for the Right side mics of course) then we can pretty much rule out human micstand movement.  If the pairs weren't very close together then phasing like that is possible but I sort of doubt that's the case unless the movement was significant and the mics were spaced apart enough on the lapel to cause those kinds of phase interactions between the coresponding mics of each pair.

(Note that the other way to avoid that kind of movement phasing would be to mount the pairs much farther apart form each other, but the distance neccesary to do that isn't available on a lapel.  See 3:1 rule, which applies to phasing from combining the corresponding mics of the two individual pairs)