On 5/31/2013 5:08 PM, William Sommerwerck wrote:
The correct term is "isolate" or "extract", not synthesize. The rear
channels are always present. Advanced decoders (such as Tate SQ and
VarioMatrix QS) selectively cancel the interfering crosstalk, based
on which channel is momentarily dominant.

I would perhaps resort to semantic quibbling in this case, since the
'isolation' or 'extraction' of a left or right rear channel would
presume that they had been encoded into the mix in some explicit way to
begin with, and could thus be extracted using some reciprocal process or
decoding scheme. The original front channels did not possess the
bandwidth nor the dynamic range to permit separate channels to be
encoded, and any scheme which claims to fold 4 channels into two and
then magically permits the original 4 to be regenerated would need to
use alternate modulation schemes, thereby rendering downward
compatibility with existing stereo to be none existent.


This is absolutely true mathematically -- but it is not true
psycoacoustically. The ear can be tricked.

Indeed it can. I was talking in the parlance of an electrical engineer,
ultimately predicated upon the underlying mathematics of communications
theory and its vocabulary.

It is possible to have significant material on all four channels at
the same time, with the resulting effect seeming fully "discrete".

Yes, this is true, but this ear/brain trickery comes at a price.
Engineers would not call this a discrete system since the effect is
artificially created, aka 'synthetic'.

Actually, the "alternate modulation schemes" you refer to, do allow
full backward compatibility, just as stereo FM broadcasts can be heard
in mono without losing anything.

Only one alternate modulation schemes I mentioned does offer backwards
compatibility, which is why JVC chose it for their CD4 vinyl LP system,
at the expense of rapid wear-out and very noisy rear channels, mitigated
somewhat by companding and severe filtering of highs in the rear. The
other modulation schemes I described do not offer backwards
compatibility unless the original front left and right channel
performance is degraded.



Fundamentally,
you cannot take two channels of 20 KHz bandwidth and (let's say) 70 dB
of dynamic range such as may be found in a standard LP record and
somehow encode anything additional without either spoiling the original
stereo L and R pair, eliminating conventional stereo playback, or
creating a new and different encoding scheme from scratch.

Again, yes and no. SQ encodes the front left and front right channels
as if they were conventional stereo, so they sound pretty much the
same as they would on a stereo record -- or when an SQ disk is played
in stereo.

It is worth noting that Ambisonic UHJ encoding allows
psychoacoustically correct playback without logic circuits.

Of course, the availability of "discrete" delivery systems largely
eliminates the issues of compatibility.

The kernel of your semantic distinction in that we are dealing with
extraction and isolation of rear channel information which has been
encoded and added into 2 standard front audio channels, ostensibly
without compromise to the original front channel pair.

I entirely agree that psychoacoustic techniques permit the illusion of 4
(or more) channels to be constructed in the listener's mind. The brain
has a lot of adaptive power, and mp3 recordings with less than 15% of
the originally encoded music are generally accepted as reasonable
approximations to the original recording as well. Perhaps we hear what
we want to hear or what we choose to hear.

To the engineer however, the distinction between extracting an isolated
signal which is independently signaled versus the synthesis of a derived
signal which is not explicitly and discretely separable are two entirely
different methods. The fact that the human brain can be fooled to think
that the more complex discrete version can be adequately imitated by the
less complex derived version really doesn't change the technical
distinction between real versus synthetic.