Lynn Olson
Active Member
- Joined
- Aug 29, 2010
- Messages
- 98
We have to do a deep dive into what's called "pairwise mixing", which is what was common back in the quad days. Basically, to do a circular pan, you mix a one pair of channels at a time ... if the pan is clockwise, you pan from LF to RF, then RF to RB, then RB to LB, then LB to RF, and keep going if desired. If a joystick is used, then more than 2 channels may be present for some portions of the pan.
The gotcha with "discrete" recordings is that pairwise mixing doesn't necessarily deliver the most natural impression in the rearward 270-degree arc. It's fine for the frontal arc, from LF to RF, but kind of falls apart on the sides or rear. You can test this for yourself by routing a normal stereo recording between LF and LB, or RF and RB. If you turn 90 degrees and face the speaker pair, the image will only be mediocre if the speakers are dissimilar (crossover slopes and frequencies don't match), or if the heights are different. Then face forward and notice the image quality gets even worse (because the ear doesn't have very good localization for sounds coming from the side, which is why it's instinctive to face towards a new sound). Typically, the phantom-images between the pair will be diffuse and uncertain, and lower quality than the frontal phantom images. It's very common in consumer-grade surround systems for no phantoms to form at all, leaving a localizations pinned to either front or rear speakers, with nothing except a vague wash of sound in-between.
The same applies with greater force to the rear arc. In fact, a real, physical sound from directly behind you can sound very much like a sound directly in front, and if this is a phantom image created by LB and RB speakers, most likely it will have no apparent localization at all, unless you turn around 180 degrees and face the speaker pair.
The implication is that even "discrete" recordings will have vague localization in the rear 270 quadrant unless special mixing techniques are used, for example, Ambisonic's deliberate introduction of carefully phased crosstalk. This is another way of saying that pairwise mixing is satisfactory for conventional stereo (although not optimum) but only fair for quadraphonic or 5-channel.
So a conventional pairwise-mixed discrete recording should not be the gold standard for a matrix system, because it has a somewhat unnatural sound as the localization is swept through the full circle. The reference standard for any system, discrete or matrix, should be reality itself, not another recording system.
If ... and this is a big if ... a matrix system is carefully designed and artifact-free (this is pretty rare), it can sound better than a conventional pairwise-mixed discrete recording. For my own system back in the day, I always preferred the Shadow Vector decoding to a not-so-great CD-4 disc recording, but that was partly because of the low fidelity of the JVC CD-4 system, and show-offy "discrete" mixes that tended to localize sounds right at the corners ... the quad equivalent of "ping-pong" recordings. For some reason, CD-4 recordings with a natural sense of ambience were fairly rare, while nearly every EMI SQ recording was very natural (this includes the legendary Dark Side of the Moon in EMI SQ).
My personal gold standard is a uniform, evenly distributed 360-degree soundfield, without sound piling up towards the speakers. Ideally, the speakers and the listening room should completely disappear, with no sense of a frontal "proscenium" or stage, and no awareness of speaker location. If the speakers are low diffraction (rounded corners) and have a rapid-decay time signature (less than 1.5 mSec), that will release the sound from the speaker location, and if the decoder is not corner-focused, all of the machinery of sound reproduction will disappear. If Malcolm has decent speakers, this is what he will hear ... and best of all, the decoder can do that with stereo recordings, provided the producers didn't go too far with multimiked-mono techniques and the reverb is good quality.
Even some fairly poor SQ and QS decoders can sound quite "discrete" if they are fast and aggressive in their action, but if the underlying dynamic matrix is not well thought through, artifacts will be noticeable ... clicks, momentary harshness that comes and goes, an uneasy "swimming" sensation for the ambience, these are all unwanted artifacts that detract from quad or 5-channel reproduction. Another unpleasant artifact is the leading transient ending up with a different location than the rest of the sound ... and this is the primary challenge for any dynamic decoder.
The ideal is open space, freedom from localization artifacts, and a natural, real-world tonal quality. This is a challenge for the amps and speakers as well, and an area where most home-theater receivers fall down pretty badly. If they can't do plain old 2-channel stereo without sounding kind of nasty, then quad or 5-channel will entertaining, but not natural, and over time, fatiguing. This is the bad rap that quad and 5-channel have been fighting against since the 1970's ... unnatural sound and listening fatigue. It takes hard work to get around those problems.
The gotcha with "discrete" recordings is that pairwise mixing doesn't necessarily deliver the most natural impression in the rearward 270-degree arc. It's fine for the frontal arc, from LF to RF, but kind of falls apart on the sides or rear. You can test this for yourself by routing a normal stereo recording between LF and LB, or RF and RB. If you turn 90 degrees and face the speaker pair, the image will only be mediocre if the speakers are dissimilar (crossover slopes and frequencies don't match), or if the heights are different. Then face forward and notice the image quality gets even worse (because the ear doesn't have very good localization for sounds coming from the side, which is why it's instinctive to face towards a new sound). Typically, the phantom-images between the pair will be diffuse and uncertain, and lower quality than the frontal phantom images. It's very common in consumer-grade surround systems for no phantoms to form at all, leaving a localizations pinned to either front or rear speakers, with nothing except a vague wash of sound in-between.
The same applies with greater force to the rear arc. In fact, a real, physical sound from directly behind you can sound very much like a sound directly in front, and if this is a phantom image created by LB and RB speakers, most likely it will have no apparent localization at all, unless you turn around 180 degrees and face the speaker pair.
The implication is that even "discrete" recordings will have vague localization in the rear 270 quadrant unless special mixing techniques are used, for example, Ambisonic's deliberate introduction of carefully phased crosstalk. This is another way of saying that pairwise mixing is satisfactory for conventional stereo (although not optimum) but only fair for quadraphonic or 5-channel.
So a conventional pairwise-mixed discrete recording should not be the gold standard for a matrix system, because it has a somewhat unnatural sound as the localization is swept through the full circle. The reference standard for any system, discrete or matrix, should be reality itself, not another recording system.
If ... and this is a big if ... a matrix system is carefully designed and artifact-free (this is pretty rare), it can sound better than a conventional pairwise-mixed discrete recording. For my own system back in the day, I always preferred the Shadow Vector decoding to a not-so-great CD-4 disc recording, but that was partly because of the low fidelity of the JVC CD-4 system, and show-offy "discrete" mixes that tended to localize sounds right at the corners ... the quad equivalent of "ping-pong" recordings. For some reason, CD-4 recordings with a natural sense of ambience were fairly rare, while nearly every EMI SQ recording was very natural (this includes the legendary Dark Side of the Moon in EMI SQ).
My personal gold standard is a uniform, evenly distributed 360-degree soundfield, without sound piling up towards the speakers. Ideally, the speakers and the listening room should completely disappear, with no sense of a frontal "proscenium" or stage, and no awareness of speaker location. If the speakers are low diffraction (rounded corners) and have a rapid-decay time signature (less than 1.5 mSec), that will release the sound from the speaker location, and if the decoder is not corner-focused, all of the machinery of sound reproduction will disappear. If Malcolm has decent speakers, this is what he will hear ... and best of all, the decoder can do that with stereo recordings, provided the producers didn't go too far with multimiked-mono techniques and the reverb is good quality.
Even some fairly poor SQ and QS decoders can sound quite "discrete" if they are fast and aggressive in their action, but if the underlying dynamic matrix is not well thought through, artifacts will be noticeable ... clicks, momentary harshness that comes and goes, an uneasy "swimming" sensation for the ambience, these are all unwanted artifacts that detract from quad or 5-channel reproduction. Another unpleasant artifact is the leading transient ending up with a different location than the rest of the sound ... and this is the primary challenge for any dynamic decoder.
The ideal is open space, freedom from localization artifacts, and a natural, real-world tonal quality. This is a challenge for the amps and speakers as well, and an area where most home-theater receivers fall down pretty badly. If they can't do plain old 2-channel stereo without sounding kind of nasty, then quad or 5-channel will entertaining, but not natural, and over time, fatiguing. This is the bad rap that quad and 5-channel have been fighting against since the 1970's ... unnatural sound and listening fatigue. It takes hard work to get around those problems.