Polarity and Phase in Multitrack Recording
From douzzer Thu May 8 03:41:24 EDT 1997
From: Daniel Pouzzner <douzzer@mit.edu-antispam>
Newsgroups: rec.audio.pro
Subject: Re: Phase of mix question
References: <19970507225100.SAA12804@ladder01.news.aol.com>
Sender: <douzzer@mit.edu-antispam>
Organization: (private)
>When I am recording or mixing, If I want to see if my mix is in phase do I
>have to have an osciloscope to do that?
That's something of an oddly put question. Do you mean polarity or
phase? They are _really_ not the same, and they are abusively
interchanged with alarming regularity. If you mean polarity, do you
mean between pre-mixdown signals, or post-mixdown between L and R? If
the latter, you need nothing but your ears. You can't miss it. If
the former, boy are you in for a wild goose chase. If the source is
anything other than a microphone, you're dealing with approximately a
50/50 chance that the polarity is non-inverting. Note that most
sources can be considered to have symmetric waveforms, and that truly
symmetric waveforms (a theoretical impossibility because of the finite
nature of all realworld signals, but in practical terms a reasonable
approximation) are not altered by a polarity reversal (so polarity can
be safely neglected). Percussive waveforms are _not_ symmetric, and
this asymmetry can be seen not only on scope traces, but heard and
succesfully ABX'd by someone of modest talent. Thus it is quite
important to preserve the polarity of percussive sources, and not too
hard either (though a storage scope or DAW is necessary if you want to
do this reliably and easily without a priori knowledge of the signal
chain).
If you really do mean _phase_, then things are much hairier and
there's no easy answer. Let's say you're doing a studio session with
5 mic's on the drum kit (say, a binaural pair, a mic on each of two
kick drums, and a highlight mic on an orientalesque inverted cymbal),
a couple mics for miscellaneous percussive effects (bells, why not a
didgeridoo?), 3 vocalists with a mic each, a straight-feed guitar, a
mic'd-cabinet guitar, a mic'd-cabinet bass, a straight-feed keyboard,
a mic'd accordion, an absorbent divider to attenuate spill from the
kit to the rest of the mic's, headphones for monitoring by the
players, and walls and ceiling covered in deep absorbent foam with a
bass trap at one end of the space. This is a moderate-complexity 24
track job (though realistically at least 32 for multiple takes of
isolated parts), happens every day. On the face of it, things seem to
be carefully controlled. So, what sort of complexity is involved in
terms of phase?
It is well to invoke Einstein's principle of relativity at this
juncture. This principle says that the reality perceived by an
observer is meaningful only within the context of the observer's
location and velocity (and derivatives thereof). Similarly, each
microphone in the above scenario receives sound from each of the sound
radiators in the recording space with different delay (hence phase
difference) and different spectral contour and relative amplitude.
The ideal in a multitrack framework is to have each track record only
a single, isolated instrument, but in practice this happens only when
the sounds are recorded at different times, in acoustically isolated
booths, or not by microphones at all (i.e. by guitar pickups or from
synthesizers, drum machines, et al). Even if you use a purist
recording technique, e.g. a single coincident-cardioid or dummy-head
binaural rig, the phase relationships between the various sources will
vary dramatically depending on where you put the microphones. As most
any symphonic recording engineer can attest, the placement of these
purist rigs not only alters the final product substantially, but is in
fact critical to the capture of a listenable, balanced, well-defined
recording (though phase per se is only a small piece of the
explanation for why this placement is critical).
In a multitrack environment, you could veritably lose your mind if
your console gave you the ability to dial in delays with
sub-millisecond accuracy on each channel to permit tweaking of
coherency. Compare to the procedure an optometrist uses to measure
your eyes - but instead of 3 axes and two eyes, there are now 3 axes
(delay, eq, and soundstage placement (comprising relative amplitude,
panning, etc)) and over a dozen eyes which must be carefully
configured with respect to eachother. Some day, probably not far off,
the process of demuxing the signals captured by mic's in controlled
live-ensemble studio session will be automated (n sources and n
microphones is a solvable system algebraically - the sources can, in
theory, be perfectly isolated, provided the precise (within less than
a centimeter X/Y/Z) locations of all sources, walls, and microphones
are known, and their characteristics of spectral reflection and
transmission are also known).
But wait, there's more. The various processing boxes - distortion
boxes, eq's, guitar cabinets, harmonizers, pitch shifters, etc. - all
wreak havoc with phase relationships within a single channel. This is
the territory of group delay and such - and furthermore, the territory
of potential euphony and potential earbleed. Phase *accuracy* is not
even relevant to synthesizers as such, except in the case of
manipulated samples with an acoustic origin.
High-end digital reverberators - stuff from the likes of Lexicon and
Yamaha - are successful at rendering a soundspace because they
carefully preserve the phase, placement, and spectral contouring of
each discrete synthesized reflection. The fact that such anal
retentivity is prerequisite for a convincing illusion of soundspace
hints at the largely neglected importance of phase relationships in
mixdown.
In the next ten years, some very exciting and truly revolutionary
changes will be seen in the way high-end multitrack production is
done.
-Daniel Pouzzner
System Architect