“WIDTH” cont’d...
So we’ve all heard some “big” sounds before that result from mid-side tweaks. But did you know
this? Let’s say you increase the width by just 1 dB (which is pretty subtle). What begins with maybe
100 dB dB of separation between left and right diminishes quickly. Now a sound that was panned hard
left will also be on the right side but out-of-phase and down about -20 dB. So, what seems only a hair
wider actually reduced left-right separation from 100 dB to 20 dB. Maybe this could be a concern if
you expect some people will listen on headphones. Maybe mixes with cool dramatic hard panned tracks
might actually LOSE some width along with losing center. Hmmm.
To do what we want, an analog circuit requires that 2 signals get subtracted (and added) accurately,
and this happens twice. It relates to a common spec called “common mode rejection ratio” or CMRR.
In general, good CMRR is spec’d for a low frequency because the numbers are best there and get worse
as the frequency gets higher. For example, a typical op-amp like a MC5534 can claim a best case sce-
nario CMRR of 90 dB at 60 Hz. Looking at the graph, however, that number drops to 40 dB at 20 kHz.
Not very impressive, if we hope for 80 dB across the entire spectrum as a base-line. A typical high
quality resistor may have a tolerance rating of 1% which means it is likely we only get 40 dB of separa-
tion in practice. Even a superb resistor might be 0.1% tolerance which translates to 60 dB. So to make
a long story short, it ain’t easy to get 80 dB of separation from 20Hz to 20KHz in a high performance
sum-diff and back circuit in analog. It is easy in digital when the operations are just a few additions
and subtractions, and 24 bit math virtually guarantees numbers in the range of 144 dB. But these are
just numbers and the previous paragraph suggests that once one actually adjusts or processes sum-diff
signals those numbers quickly become low anyways.
Perhaps some of the feeling that increasing the width is practically necessary in mastering might
be due to a few common modern studio techniques that may accidentally strip some subtle localization
cues. For example, the ear uses the initial onset of transient peaks as well as phase and level differences
as localization cues, and now we have more dynamic processors (compressor plug-ins on every track,
group and mix bus) in more studios than ever before in history. We also have more tracks and denser
arrangements. We time shift tracks and use more high-pass filters (messing with phase). Not that these
things are bad, but one should be aware that there are always trade-offs, and some say that “Engineer-
ing is the art of balancing trade-offs”.
14
