Mixing 3 mics/lavs in post

[James Huenergardt]

Hi,

So here's what I ended up doing. I filmed three guys having a round-table discussion using 4 cameras and 3 lav mics.

2 of the lav mics went into a Tascam DR-100 recorder which then went into a 5D.
1 of the lav mics went into a Zoom recorder which just recorded onto a card.

I'm having some slight issues trying to mix the audio.

Once I sync up my clap hits on the audio and video files, and then play it from the timeline, the audio has a slight echo to it. Sometimes, if I turn off one of the tracks, it goes away, sometimes it doesn't.

I found I could sync this up in Adobe Soundbooth, but not in my NLE. Is this normally the case?

I guess using a field mixer to record ALL tracks onto ONE file would have caused this to not happen.

Any ideas, work-arounds that keep me from syncing all this stuff up in Soundbooth?

[Richard Crowley]

Originally Posted by James Huenergardt

Once I sync up my clap hits on the audio and video files, and then play it from the timeline, the audio has a slight echo to it. Sometimes, if I turn off one of the tracks, it goes away, sometimes it doesn't.

The default is to never have more than one track enabled at any give moment unless two people are talking over each other.

I found I could sync this up in Adobe Soundbooth, but not in my NLE. Is this normally the case?

Not sure what "sync this up" means here? It doesn't sound like you have a sync problem. You have a mixing problem (which is easily solved.)

I guess using a field mixer to record ALL tracks onto ONE file would have caused this to not happen.

Eh, maybe. But then you would have had to have a dedicated sound mixer performing the mix live in real-time. And they would have to be really GOOD to get as good a mix as you can do in post. I would rather record each mic iso as you have done, and do the mix in post. Of course most of the factors that go into this kind of trade-off decision were not revealed here, so we are talking generically.

Any ideas, work-arounds that keep me from syncing all this stuff up in Soundbooth?

I think you are approaching it all wrong if you think it is a "sync" problem.

I would have at least run each mic to its corresponding camera so that the sound and picture for each individual would always be in sync. But beyond that it is a matter of turning up ONLY the active channel. That essentially eliminates the "echo" problem altogether.

[Steve House]

Adding a footnote to Richard's post, the reason the "echo" occurs when you have two tracks potted up is due to the fact that with one mic on each person in the conversation, a given sound arrives at each mic at slightly different times. The mic on B is several times farther away from A's mouth than is the mic on A himself. Due to that distance, a sound uttered by A arrives at B's mic significantly later than it arrives at A's own mic. When you mix them, B's mic will sound like an echo of A's because in a sense, it is.

[Greg Miller]

I would bet that the two mics on the 5D are pretty close in sync to each other, since they're recorded as a stereo pair in sync with the pic.

But the separate mic on the Zoom might be measurably further out of sync with the 5D tracks. That's because the Zoom might not be perfectly in sync even at the beginning ... just synced to the nearest frame (depending on your procedure).

Further, the Zoom can "walk" out of sync during the course of the take because it has a different time base than the 5D has. (But of course at random times, the Zoom might "walk" back INTO sync with the 5D, too... all based on the randomness of the two different timebases.)

Having said all that, that should address the question of why you hear the echo. But still, the correct procedure is to have only one live mic at a time, depending on who's speaking. That eliminates all potential echo and phase issues (such as comb filtering), and also eliminates 2/3 of the background noise (assuming the noise in all three tracks is fairly random, and roughly equal in level).

Out of curiosity, what are the physical dimensions of the round table? What's the distance between the three participants?

[James Huenergardt]

There was technically no round table. The guys were sitting on bar stools about 2'-3' apart.

If I go into Soundbooth and take the audio from the 5D and the audio from the Zoom and perfectly sync it up, the echo goes away.

I was pretty sure the 'echo' was from the same audio being played milliseconds apart causing it to sound like an echo.

My work-around will be to sync it all up in Soundbooth, and then have the proper track on for whoever is talking.

I guess I also could have recorded everything into the Zoom using the 4-track recording option. At lease they would ALL be in the same file and I wouldn't have the echo issue.

[Brian P. Reynolds]

I thought that the H4n in 4 track mode actually records it as 2 seperate files which need to be synced together also. (Or thats what I have found using mine in 4 track mode)

[Steve House]

Originally Posted by James Huenergardt

There was technically no round table. The guys were sitting on bar stools about 2'-3' apart.

If I go into Soundbooth and take the audio from the 5D and the audio from the Zoom and perfectly sync it up, the echo goes away.

I was pretty sure the 'echo' was from the same audio being played milliseconds apart causing it to sound like an echo.

My work-around will be to sync it all up in Soundbooth, and then have the proper track on for whoever is talking.

I guess I also could have recorded everything into the Zoom using the 4-track recording option. At lease they would ALL be in the same file and I wouldn't have the echo issue.

It's not that they are different files, it that it is two different mics. When A is speaking, his voice is picked up on mic A 8 inches from his mouth and ALSO a few milliseconds later on mic B 3 feet away. Playing both tracks at once, whether they're in two files or the same file, and mixing the two tracks will produce the echo.

[Greg Miller]

Originally Posted by James Huenergardt

There was technically no round table. The guys were sitting on bar stools about 2'-3' apart.

If they were that close, the physical distance between mics would not cause any "echo." That's only 2 or 3 millisecond delay. That's basically like standing 2 feet from a wall and talking. You certainly will hear the reflection, but it's too soon to be perceived as a distinct echo. It's just perceived as "liveness."

In fact, this amount of delay does cause some noticeable comb filtering, which is definitely undesirable because it's detrimental to the voice quality.

You can contrive an experiment to demonstrate this (as I've just done myself).

Record a clean voice track. Copy it to your clipboard. Mix/Paste the copy on top of the original, but move the start point 3 msec after the original start point. (Or 6 msec if you want to simulate 6' distance.) Paste it in at a slightly lower level (I used -6dB) since the "distant" mic will pick up at a lower level than the "speaker's" mic. When you play back this track, you won't hear an echo, but you will definitely hear the comb filtering. That's why it's absolutely necessary to keep only one mic at a time live.

According to my venerable old Audio Cyclopedia (the pages aren't too yellow yet), "To be an echo, the reflected sound must be 1/20 of a second or longer behind the original sound."

[Rick Reineke]

It could also be early reflections (echo) from the room as well. For instance the person wearing mic-1 is a low-talker, so that mics overall level is hotter, and would make the delay significantly noticeable when mic-2 or 3 are used by persons speaking. (normally or loud) In any case, if there on separate tracks, it should be fixable in post.

[Greg Miller]

That's a good point, Rick, I hadn't considered additional reflections within the room.

If you go by the 1/20 second definition above, the total path would need to be roughly 20 feet. In that case, I think that the overall attenuation would be significantly more than the 6dB of my test. But you're right that a discrepancy in mic gain could somewhat counteract that.

Also, with a total path of 20 feet, I think you'd be getting a lot of diffuse reflections within the room, so the problem would include some reverberation as well as distinct echo.

All this is rather moot, though, because the OP says he can get rid of the echo by getting the tracks perfectly in sync. That tells me that the problem described as "echo" really was a sync problem between tracks, and not any sort of acoustic phenomenon.

And, at any rate (even if it's caused by a full moon or high tide), the solution is to have only one live mic at any given time. Just a little more work in post. Job security. ;-)

[Steve Oakley]

the room echo is from the 2 non-speaking mics picking up room slap. one mic isn't too loud, but have 3 or 4 mics open and they pick up a LOT more room tone.

as you have discovered, you have to shut off ( the term is duck actually ) the mics of the non-speakers to get rid of echo / room tone / noise in general to get a good mix.

is it a lot of work ? yes. thats just how it is to do it the right way.

you could try your NLE's gate filter on the track, but probably its not going to work as good as doing it the right way - manually ducking the mics.

[Daniel Epstein]

Actually this sounds like a combination of factors. The NLE is not allowing you to sync the files perfectly as it only allows movement of the audio 1 frame at a time. This could be from the recording sources being different. The camera and the zoom probably have different delays for the A to D conversion which is smaller than a frame but enough to create a phase issue when mixed together. I haven't been able to sync up 7D camera tracks perfectly with zoom tracks in FCP 6. The camera Long Gop recording may be the reason. Separate tracks will have to be manipulated independently to minimize the issue. Soundbooth sounds like it will allow you to deal with these issues better than the NLE.
To avoid issues between two mics on separate subjects recording the same voice it is best to know the old 3 to 1 rule. If the first mics distance is 6 inches from the 1st subjects mouth the same style mic on the second person should be greater than 18 inches away from the 1st subjects mouth to avoid issues. For instance try not to put a lavalier on one subjects left lapel and the next mic on the right lapel if they are sitting shoulder to shoulder. The distance may not give you the 3 to 1 ratio you want. If you are live mixing them then you would have to pot down and up one or the other of the mics in the mix than if they are spaced according to the 3 to 1 rule. If you record them on separate tracks then you might have to do this in your NLE after the fact.

[Paul R Johnson]

Daniel has it pinned here - if you must have all channels up, then simplest way is to hive off all three tracks to your audio editor, sync them together there, trim off the beginnings so they all line up, and export those audio files, bring them into the video editor, and sync one, then slide the others so their starts align, and that's as close as you can go. Most audio editors can edit down to the individual samples - so are much more accurate than the picture editor, which is stuck with the frame rate as the smallest movement that works.

[Greg Miller]

I think the OP has already discovered this, per his post of Jan.11 at 10:37 PM.

But of course you can't achieve "perfect" sync this way. It will depend on who's speaking at the point where you align the tracks. If person "A" is speaking, his voice on track B and C will be delayed (by the physical distance) so you'll need to slide B and C to the left to get them in sync with A. But then when person "C" is speaking, his voice on track C will come before his voice on track A, so you'll need to slide A to the left (or C to the right)... etc. Kind of like an audio version of whack-a-mole.

And of course this procedure doesn't compensate for the fact that the different recorders are likely to drift apart over the duration of a long take.

So yes, the OP knows he needs to get the tracks reasonably close. And then ducking is the final answer.