Lynn Olson
Active Member
- Joined
- Aug 29, 2010
- Messages
- 98
Now for a deeper dive into allpass phase shifters. What are these gizmos, what do they do to the sound, and why are they essential for SQ and QS encoding and decoding, but not used in EV4 or DPL-II?
An allpass phase shifter, or allpass filter, is an odd little device that shifts the phase around 180 degrees, 360 degrees, or more while retaining a completely flat response. Square waves and impulses look really weird after passing through these things, but oddly enough, it's barely audible at all, despite the horrific appearance on a scope screen. I've met a handful of audiophiles who claim to hear absolute phase, but it has to be kept in mind there is no standard for absolute phase, and there are plenty of records where the absolute phase is different from track to track, since they were recorded in different studios. In any event, the kind of steady phase rotation produced by an allpass phase shifter is remarkably difficult to hear, even on direct and immediate A/B switching. The physical device is a simple op-amp circuit, or done with discrete transistors. The allpass function can also be realized in DSP software with more precision than physical capacitors and resistors allow.
When the previous post referred to 0 and -90 degree phase shifters, that does not mean the 0 degree function is directly connected to the input. No, not at all. Both 0 and -90 degree outputs came from phase shifters, just with slightly different tuning. The pair are set up so there is a constant 90-degree difference between the two outputs, but compared to the original input, the phase spins all over the place. In fact, the original, non-phase-shifted input and the output of the phase shifters can never be combined anywhere in the decoder, or anywhere else in the audio chain, or there will be very deep nulls and peaks (called a "comb filter" in the world of audio). Compared to the input, the phase from the two shifters spins around and around, with more spins with more poles, but the phase relation between the two shifters stays at a constant 90 degrees though the audio band.
Why do decoders use these things? For QS decoders, it allows a smooth 360-degree pan without a phase jump of 180 degrees at Center Back. This phase jump happens in EV4 and DPL-II, so Center Back is usually avoided, or special techniques are used to encode the circular pan.
How does QS do this? At first glance, LB is made of +0.924L and -0.383R, and RB is made of -0.383L and +0.924R. In reality, each rear channel is shifted by 90 degrees, so LB is really 0.924L at +90 degrees and 0.383R at -90 degrees, and RB is made of 0.383L at +90 degrees and 0.924R at -90 degrees, so they are actually in phase with each other. The two front channels are at 0 degrees, but this is NOT the direct input, but the output from the 0-degree phase shifter. If the fronts don't pass through the 0-degree phase shifters, the phase relation between the front and rear channels will be completely random, and no image will be possible on the side walls.
SQ relies even more heavily on phase shifters, since the LB and RB channels are entirely made of the outputs of the 90 degree phase shifters, and as with QS, the front channels are derived from the 0 degree phase shifters, NOT the direct input. In both QS and SQ, when a signal is panned in a 360 degree circle, the pair of front speakers are in-phase with each other, the pair of rear speakers are in-phase with each other, and there is a 90 degree phase shift between the pairs of speakers on the side walls.
In both QS and SQ decoders, all of the audio that passes through the matrix, whether static or dynamic, passes through the sets of 0 and 90 degree phase shifters. If the matrix is fed a combination of direct, non-phase-shifted audio and phase-shifted audio, the result is massive cancellations in the audio band as well as random phase relations between the front and rear speakers. I've read that early versions of DPL-II actually used phase shifters on the rear channels only, which results in random phase in relation to the fronts, and prevents any sidewall images from forming.
The previous post describing the Shadow Vector mentioned 0,-90, 180, and +90 degree phase shifters. This is really simple. The actual phase shifters generate 0 and -90 degrees, and there are a pair of inverters that create the 180 and +90 degrees. Repeat for the other channel and you have a set of 8 signals the VCAs can use in the dynamic matrix.
P.S. Every SQ and QS recording has passed all of the audio through a set of phase shifters, with anywhere from 360 to 720 degrees of phase rotation across the audio band. So absolute phase may not too meaningful for these recordings ... which includes quite a few beautiful EMI classical recordings made in the Seventies.
An allpass phase shifter, or allpass filter, is an odd little device that shifts the phase around 180 degrees, 360 degrees, or more while retaining a completely flat response. Square waves and impulses look really weird after passing through these things, but oddly enough, it's barely audible at all, despite the horrific appearance on a scope screen. I've met a handful of audiophiles who claim to hear absolute phase, but it has to be kept in mind there is no standard for absolute phase, and there are plenty of records where the absolute phase is different from track to track, since they were recorded in different studios. In any event, the kind of steady phase rotation produced by an allpass phase shifter is remarkably difficult to hear, even on direct and immediate A/B switching. The physical device is a simple op-amp circuit, or done with discrete transistors. The allpass function can also be realized in DSP software with more precision than physical capacitors and resistors allow.
When the previous post referred to 0 and -90 degree phase shifters, that does not mean the 0 degree function is directly connected to the input. No, not at all. Both 0 and -90 degree outputs came from phase shifters, just with slightly different tuning. The pair are set up so there is a constant 90-degree difference between the two outputs, but compared to the original input, the phase spins all over the place. In fact, the original, non-phase-shifted input and the output of the phase shifters can never be combined anywhere in the decoder, or anywhere else in the audio chain, or there will be very deep nulls and peaks (called a "comb filter" in the world of audio). Compared to the input, the phase from the two shifters spins around and around, with more spins with more poles, but the phase relation between the two shifters stays at a constant 90 degrees though the audio band.
Why do decoders use these things? For QS decoders, it allows a smooth 360-degree pan without a phase jump of 180 degrees at Center Back. This phase jump happens in EV4 and DPL-II, so Center Back is usually avoided, or special techniques are used to encode the circular pan.
How does QS do this? At first glance, LB is made of +0.924L and -0.383R, and RB is made of -0.383L and +0.924R. In reality, each rear channel is shifted by 90 degrees, so LB is really 0.924L at +90 degrees and 0.383R at -90 degrees, and RB is made of 0.383L at +90 degrees and 0.924R at -90 degrees, so they are actually in phase with each other. The two front channels are at 0 degrees, but this is NOT the direct input, but the output from the 0-degree phase shifter. If the fronts don't pass through the 0-degree phase shifters, the phase relation between the front and rear channels will be completely random, and no image will be possible on the side walls.
SQ relies even more heavily on phase shifters, since the LB and RB channels are entirely made of the outputs of the 90 degree phase shifters, and as with QS, the front channels are derived from the 0 degree phase shifters, NOT the direct input. In both QS and SQ, when a signal is panned in a 360 degree circle, the pair of front speakers are in-phase with each other, the pair of rear speakers are in-phase with each other, and there is a 90 degree phase shift between the pairs of speakers on the side walls.
In both QS and SQ decoders, all of the audio that passes through the matrix, whether static or dynamic, passes through the sets of 0 and 90 degree phase shifters. If the matrix is fed a combination of direct, non-phase-shifted audio and phase-shifted audio, the result is massive cancellations in the audio band as well as random phase relations between the front and rear speakers. I've read that early versions of DPL-II actually used phase shifters on the rear channels only, which results in random phase in relation to the fronts, and prevents any sidewall images from forming.
The previous post describing the Shadow Vector mentioned 0,-90, 180, and +90 degree phase shifters. This is really simple. The actual phase shifters generate 0 and -90 degrees, and there are a pair of inverters that create the 180 and +90 degrees. Repeat for the other channel and you have a set of 8 signals the VCAs can use in the dynamic matrix.
P.S. Every SQ and QS recording has passed all of the audio through a set of phase shifters, with anywhere from 360 to 720 degrees of phase rotation across the audio band. So absolute phase may not too meaningful for these recordings ... which includes quite a few beautiful EMI classical recordings made in the Seventies.
Last edited: