Topic: This is a new one.. weird channel imbalance with 4-channel mix (Audacity)

[acidjack]

Trying to mix two mic sources from a show.  Both from an R-44 so perfectly synced. I am using Audacity.

When I split stereo track and check the peak amplification, it is off by no more than .5dB in either direction between channels in each of the 2-channel pairs.  The waveforms look nearly identical. When played back individually, each channel peaks out pretty close to the same.

But when played back together, one channel (left) peaks dramatically higher - like one is peaking 6dB or more differently.  Given that the waveforms are both approximately channel balanced, can anyone give me any idea of why this would be?

[hoserama]

Might one of the channels gotten inverted somewhere along the way? Play with the phase inversion and that may fix it.

[page]

That would be my first check as well. Second would be it sounds like you are experiencing some phase cancelation (or phase amplification).

[acidjack]

How would "phase inversion" have happened?   I apologize for not understanding - I assume you mean it's as if there are 2L, 2R (a 4ch mix) but one channel now has 3 channels and the other only has one? 

How would you check for this? And, given that all I did was take the raw recordings and import into Audacity, how could it have occurred? I tried re-importing the second set of channels, but it did not help...

[page]

How would "phase inversion" have happened?   I apologize for not understanding - I assume you mean it's as if there are 2L, 2R (a 4ch mix) but one channel now has 3 channels and the other only has one? 

How would you check for this? And, given that all I did was take the raw recordings and import into Audacity, how could it have occurred? I tried re-importing the second set of channels, but it did not help...

polarity inversion. In audacity, it's the Invert plugin. Some hardware have settings that will, or bugs that cause it (the V3 had a bug on certain serial numbers which would invert the polarity on one channel).

[runonce]

You have your individual source's levels too high...

When you combine - some frequencies reinforce each other and create higher peaks than in either source alone.

This is a common problem when combining sources...and often overlooked.

One of the beauties of 24 bit audio is being able to record at a lower level so you still have some headroom when you combine sources.

Try reducing both sources 3db - before you combine -  until you dont get any overs...

[acidjack]

You have your individual source's levels too high...

When you combine - some frequencies reinforce each other and create higher peaks than in either source alone.

This is a common problem when combining sources...and often overlooked.

One of the beauties of 24 bit audio is being able to record at a lower level so you still have some headroom when you combine sources.

Try reducing both sources 3db - before you combine -  until you dont get any overs...

...but would this explain overloading only, or would it also explain one channel suddenly being dramatically higher?   The sources do not overload at all - they still have about -3dB of headroom, even on the higher-peaking one.

I'm going to try "invert" and this and see what I come up with.

[runonce]

Have you exported the mix and inspected the wav form or is this just on playback via audacity?

Id export it at least once - rule out a playback bug...

If the invert works - then I guess you have to start taking a look at the gear involved?

[runonce]

Another thought...

Is it possible that one channel is somehow - slightly out of sync?

Perhaps L+L is correct and has the higher peaks...but R+R is out of sync - and not getting the higher peaks that result when combining...

[page]

Another thought...

Is it possible that one channel is somehow - slightly out of sync?

Perhaps L+L is correct and has the higher peaks...but R+R is out of sync - and not getting the higher peaks that result when combining...

Also a possibility, especially if you are using 2 mic sources compared to a board and a set of room mics.

Can we get a 30 second sample of both sets?

[acidjack]

^^^ I will try and do that.  Just to summarize:

Original sources were 2 mic pairs evenly spaced, running into an R-44.  Thus, they are not out of sync time-wise.

The individual stereo WAVs are within .5dB of each other consistently (visually and if you try the "Amplify" function to tell you how far the peaks are from 0)

I tried reversing the phasing, but all that did was result in the same effect in reverse.

I went ahead and exported the mixed WAV.  I then rebalanced the channels on the mixed WAV, and it seemed to be fine. 

I will try and post samples of each source individually plus a raw mixed later today or tomorrow.

[kcmule]

I'm voting for phase inverting as well.  I often have to do this
when taking 4 channel to my R4 (two ambient, two sbd).  I
generally only have to apply the phase inversion to one of the
channels on one of the sources, not to both channels of one
source.

[runonce]

For one track to be out of time - you would almost have to do this on purpose...(or by mistake)

And to be clear - I understand that the tracks are clock-synced...but Im suggesting one track has lost its time-sync with the other tracks...(although still clock synced)

Not too familiar with the R44 as far as importing goes...does it import all 4 tracks at once..?

I assume you are still moving one or another track set forward or backward in time using the Time Shift tool?

[page]

I'll go with general phase cancelation due to placement, but I'll wait till I see a 4ch sample to pass judgement. I've had it happen before.

[live2496]

But when played back together, one channel (left) peaks dramatically higher - like one is peaking 6dB or more differently.  Given that the waveforms are both approximately channel balanced, can anyone give me any idea of why this would be?

Quite possibly there is some DC offset in that one channel.

I don't have experience removing it with Audacity but it does seem to have the function in the Normalize dialog. Perhaps it can be convinced to remove DC offset without normalizing the file.

Alternatively, you could try eq'ing out anything below 20 Hz or lower.

Gordon