Timothy Stockman wrote,
| L+R/2 is almost universally used. FM broadcast is trasmitted this way for
| compatibility with monaural receivers. The "mono" switch on preamplifiers
| does this. However, this simple, almost universally used transformation
| does not provide a perfect mono version of a stereo source.
If that is not "perfect," then there must be some definition of perfection
here.
| The reason is that stereo channels add by *power* acoustically, ...
Oh, that's the perfection we're seeking: averaging the power, ...
| ... whereas the electrical sum L+R/2 adds by *voltage*.
... rather than the voltage.
The sample amplitudes are voltages, right? And power is proportional to the
square of the voltage, and volume proportional to the log of the power, right?
| Therefore, when recordings mixed with prominent
| material in the center of the sound image, such as vocals, bass, etc. are
| electrically mixed to mono (L+R/2), the sounds in the center of the stereo
| will be 3dB higher in the resultant mono mix compared with those at the left
| or right edge (known in the recording industry as the "3dB center buildup").
3 dB more volume means twice the power ... OK.
| How to solve this? sqrt(L^2+R^2) is one method. This method would be
| particluarly good for a DSP implementation.
Where L=R, that will result in a 41% increase (as perhaps it should: it
should seem louder to hear the same volume from both channels than to hear
it from only one); yet I've been finding a decrease, which (L+R)/2 doesn't
explain either.
Also, sqrt(L^2+R^2) loses sign information [and so would my earlier thought
about the geometric mean]. Again, I'm not sure which numbers we're dealing
with there: amplitude? power? volume? Power would never be negative, so
there is no sign information to worry about. That brings up another problem
with using the geometric mean: silence in either channel would mute the other
channel. The harmonic mean is also impossible, as it couldn't be calculated
when one channel is silent.
Maybe ... add the squares, take the square root of the sum, scale down to
70% (2^-.5) of the root, translate back to amplitude and figure out how to
assign the signs ...
but still, that's an ideal, and first I'd like to know what Sony decks cur-
rently do so that we can cope with it.
| Another method is to rotate the phase of the channels with an "all-pass"
| filter so that they are 90 degrees apart and add in quadrature. This
| method is better for an analog implementation.
Sorry, the engineering concept of the rotation is over my head, and I've no
idea at all of what adding "in quadrature" might mean.
| I'd love to see a deck that implemented the first solution!
The first solution you mentioned [(L+R)/2] or the first one you suggested
[sqrt(L^2+R^2)]?
Ah well, I'm just plain confused.
-----------------------------------------------------------------
To stop getting this list send a message containing just the word
"unsubscribe" to [EMAIL PROTECTED]
