I went to grab a bite of lunch today with blog and podcast contributor Anthony Kosobucki. We were talking about tracking drums and he said he was having a problem fattening up a snare on a current project. I asked about mics, top and bottom. Then I asked about phase. He said he had flipped polarity every which way but couldn't get any meat in the sound. He said he put a parametric EQ on it and poured 200 Hz on and it just wouldn't come up. That was the tip off. It was definitely a phase issue.
We talked about some different ways to address it and then decided to just go to the box and take a look at it. Sure enough. Everything else in the project sounded pretty good but the snare was really anemic. We tried quite a few different things but the point of all of them was to take the sounds around 200 Hz that were out of phase between the two tracks and shift one or both of them until they stopped canceling and started adding together favorably.
We had the option of turning on a phase plugin but there's an easier way. The wavelength of sound from 200 to 250 Hz runs from about four and a half feet to five and a half feet. (I actually guessed six off the top of my head) So if they're lined right up and they're canceling, sliding one of them half a wavelength away should have some positive effect. We went with three seconds and when we started playing around with levels there started to be a little improvement. (Sound travels a tick faster than a foot per millisecond.)
I left at that point with Anthony starting to apply panning and gating to the rest of the drum tracks to make sure there weren't any additional phase issues floating around. A little later in the day I got a text that nine milliseconds did the trick. So that means that frequencies around 225 Hz have actually shifted all the way around to being in phase again and then about three quarters of the way around a second time.
So why did it take an adjustment like that to get the desired result. If the signal was a sine wave it would have made sense to only adjust it to being three quarters of the way around. But being a complex wave form with a decay, things get more complex. In this case it took a shift that long to get everything working right. To really figure it out you'd have to take things into account like the distance between the top and bottom mics and do a whole bunch of math. Just knowing how to approach the problem and listening while tweaking was good enough in this case.
The only remaining thought I have is that nine milliseconds is starting to be a pretty long time. I suggested that maybe he should shift one track back four and a half and the other forward four and a half to achieve the same relative delay while keeping the hit centered on the beat. But it's also a pretty jazzy track and not played to a click or quantized so maybe a tiny bit of slop in the snare is a tiny price to pay for a little meat in the sound.