Recording - November 2014 - 45

This means that I'm likely to choose microphones for controlled-leakage
sessions with an ear toward those whose response off-axis is similar to onaxis. This usually means a pronounced slant toward small-diaphragm mics
like the KM 140s I used on the clips; most large-diaphragm mics have offaxis responses that sound, well, peculiar.
I remember a session where I recorded local musician Bob Abrams
singing and playing guitar, using Neumann U 87 large-diaphragm mics on
both his voice and guitar. When Bob started wailing on a loud blues number, his voice leaking into the guitar's U 87 sounded terrible; his mouth was
about 90° off axis, and the microphone's frequency response at that angle
was a lumpy "smile" curve. The session was saved by a pair of smalldiaphragm Neumann KM 84s on the guitar, so I could throw away the guitar U 87 track. I'd learned a lesson.
But there's a bigger fly in the ointment when recording multiple musicians
in a studio with the frim-fram sauce of controlled leakage: the Phase Monster.

Figure 1

Figure 2

Enter the Monster
The Phase Monster creeps out of its cave when a sound wave hits two
microphones at different times, or takes two different-length paths to a single
microphone. (We'll discuss the second scenario later.) Here's how it happens.
Picture a pair of microphones; one is 2' away from the sound source, the
other 3'. The absolute distances aren't crucial here; what matters is the difference between the two distances. This is known in audio engineering circles as the path length difference, and it's usually abbreviated D.
Now, you know that sound waves in air have a wavelength-the distance
between successive points of maximum compression. And, for any distance,
there is a sound wave which has that wavelength, and that sound wave has
a frequency. There's a simple formula for figuring out the frequency of the
sound whose wavelength corresponds to the path length difference D:

Frequency (F) = Speed of sound / Path length difference (D)

Figure 3

The speed of sound in air is typically 1130 feet per second. If, in our
example, D = 1', then the frequency whose wavelength corresponds with
that distance will be:

F = 1130 feet per second / 1' = 1130 Hz
Well and good. What will happen at half that frequency F (which works
out to 565 Hz)? Take a look at Figure 1.
In that picture, you can see that the bottom waveform (from the mic 3' from the
source) arrives slightly later than the top one (from the mic 2' from the source).
You can also see that, after the first half-cycle, the second waveform goes up when
the first one goes down, and vice versa. That means that, when you add the two
waveforms together in a mic mixer or a DAW, they'll cancel one another out, and
(except for that initial half-cycle) you'll be left with no signal.
Now let's look at what happens at the frequency F which has a wavelength of 1 foot, which we figured is 1130 Hz. In Figure 2, you can see
that the later waveform is delayed by a full cycle, and that when the first
wave goes up, the second also goes up. When you add them together, they
reinforce one another; after the first cycle, the combined waveform is twice
the amplitude of the original waves-that's a level jump of 6 dB.
Finally, Figure 3 shows what happens at the frequency which is delayed
11/2 cycles when D is 1 foot. That frequency is 1695 Hz (11/2 times 1130
Hz), and once again the signals at that frequency cancel one another out
after the initial transient.
You can see the pattern: signals that are delayed 1/2 cycle, 11/2 cycles,
and for that matter 21/2 and 31/2 cycles, etc. will cancel; signals that are
delayed 1, 2, or 3 cycles, etc. will reinforce. You can express this in terms
of frequencies; if F is the frequency whose wavelength corresponds to the
path length difference D, then frequencies which are 1/2, 11/2, 21/2 (etc.)
times F will cancel, while frequencies that are 1, 2, 3 (etc.) times F will reinforce. Taking a cue from the childhood game of jacks, I like to say that the
"halvesies" cancel, while the "onesies" reinforce.
(Meanwhile, I hear someone wailing, "But these are pictures of sine
waves. I don't record sine waves-I record music!" Actually, you do record
sine waves; all musical waveforms are made up of many sine waves, each
of which will suffer these effects. The full story's a bit too involved for this
article, but perhaps another time....)

If you plot the frequency response of the combined
microphones, you get a picture that looks like Figure
4. The picture's resemblance to a coarse comb (the
sort you'd use to comb fleas from a cat) has led to this
pattern of frequency response being labeled comb filtering. The exact frequencies that cancel and reinforce
will vary, depending on the size of the path length difference, but the overall pattern remains the same.
What about the popular remedy of flipping the
polarity of one of the signals? If you look at Figs. 1-3
and imagine one of the signals with flipped polarity,
Figure 4

RECORDING November 2014

Recording - November 2014

Table of Contents for the Digital Edition of Recording - November 2014