SQ Shadow Vector Soundfield Mapping

[Lynn Olson]

I was hiding under the couch for the first several weeks, given that I'm in a "high-risk" group (over 65, male, and a bit overweight). We're still staying away from the stores by having groceries delivered to our doorstep, and then soap-and-washing the items that go into the refrigerator, and garage-isolating the ones in cans and cardboard boxes for 24~72 hours. As recommended by the Colorado governor, we're wearing masks when going outside walking the doggo. At least we get a view of the glorious Rocky Mountains as we walk the dog. The mountains themselves, though, are closed, since it was the ski resort towns that brought in the first infections from Italy, and they remain hardest hit. It does look like our state is rounding the corner, though, and we will have a limited re-open at the end of the month. The governor says the most vulnerable groups (us) need to stay indoors for several more weeks. It's beginning to look like mask-wearing will become the new social norm, at least until a vaccine arrives.

I'm curious how your Shadow Vector handles anomalies like digital clipping and/or cartridge mistracking. These gremlins can throw off a fast-acting logic section, which I partly averted by the aforementioned 500 Hz ~ 5 kHz bandpass and inverted Fletcher-Munson response shaping. The "clicks" usually happen in only one channel, which gives a false hard LF or RF localization. I suspect this is what is creating glitches in the DTS Neo:6 system, or maybe some other fast-acting transient.

The other weird problem that I never solved with the original prototype were LP's with wacko encoding. Specifically, there was one LP, Loggins & Messina "Full Sail" in SQ which had a harmonica at LB and a violin at RB (I know, I know, but this was CBS in the early Seventies). That drove the decoder bonkers, because the musicians, without knowing it, were able to phase-lock each other, so the sound spun around the room. The CBS professional decoder was so slow that it didn't try to track what was going on, so the mix engineers missed it. These kind of weird recordings were rare, one-off events, but they ONLY happened with SQ recordings, where the engineers just couldn't help themselves with some gimmicky set of locations. If the engineers had the good taste to use the Front Encoding scheme, instead of the more conventional 4 -> 2 encoder, those problems never arose. I think, but am not 100% sure, that EMI used their in-house Front Encoder to make sure all their recordings sounded good. (The other difference is that EMI used only 6 poles and an all-discrete-transistor circuit with the phase-shifter section in parallel, while CBS used a long cascade of 301 op-amps and 10 poles of phase-shifting. So the EMI and CBS encoders sounded quite different sonically.)

 

[proufo]

If the engineers had the good taste to use the Front Encoding scheme, instead of the more conventional 4 -> 2 encoder, those problems never arose.

Where can I learn more about Front Encoding?

Thanks in advance.

 

[Sonik Wiz]

Where can I learn more about Front Encoding?

Thanks in advance.

SQ was problematic so they had several different variations such as forward, side & backward oriented encoders:
https://issuu.com/disclord/docs/sq_recording_techniques__position_encoder_Edit: check out Disclord's archive on the QQ front page. Info thats hard to find anywhere else.

 

[proufo]

Many thanks Sonik Wiz.

 

[par4ken]

SQ was problematic so they had several different variations such as forward, side & backward oriented encoders:
https://issuu.com/disclord/docs/sq_recording_techniques__position_encoder_Edit: check out Disclord's archive on the QQ front page. Info thats hard to find anywhere else.

I've seen this before I think that it was published in "Quadraphony". Too bad you can't download it, I like to have copies of articles like this on my hard-drive.

 

[Sonik Wiz]

I've seen this before I think that it was published in "Quadraphony". Too bad you can't download it, I like to have copies of articles like this on my hard-drive.

Toaly agree. I purchased that AES Quadraphony so long ago it is beat up, water stained, but so much fantastic & accurate information! I call it The Blue Bible of Quad. When my wife sees me come up from the basement with it she moans, "Oh no. Here we go again. Another project."

 

[sjcorne]

These kind of weird recordings were rare, one-off events, but they ONLY happened with SQ recordings, where the engineers just couldn't help themselves with some gimmicky set of locations.

Almost every CBS pop/rock quad mix I've heard has different mono elements (rhythm guitars, strings, backing vocals, drums & bass, etc) isolated in the back corners - I always presumed this was done because the rear center position was off-limits in SQ, not because the engineers sought to mix in a 'gimmicky' fashion.

 

[proufo]

"Rear center position", easy to bleed but impossible to mix.

 

[Lynn Olson]

CBS had four different encoders, not including the specialized Ghent microphone. The “standard” SQ encoder was simply a black box with 4 inputs (LF, RF, LB, RB) and 2 outputs (Lt, Rt). The input signals assumed the use of pairwise panning ... no 4-channel sums (which SQ cannot encode) nor 3-channel sums. Using either 4 or 3 channel sums will give unpredictable results, and is not recommended under any circumstances.

SQ also has the bizarre capability of encoding diagonal splits, with smooth pans between LF and RB possible, and similarly, pans from RF to LB (but not at the same time). Any SQ decoder can decode these, and a competent dynamic decoder will provide full separation during the pan. Less competent decoders might struggle with detenting, a sort of stickiness in the corners. But the capability is an inherent part of the SQ spec.

It gets weirder. OK, so any decoder can handle diagonal spilts (and I’ve heard it on some recordings). But ... the standard speaker phasing of LB and RB has one split in-phase, and the other out-of-phase! No joke! This was done so the decoder can render Center Back (which is a legitimate and permissible localization) with both LB and RB speakers in phase with other.

But I could hear the diagonal-split asymmetry and found it a little annoying. Even though it’s a rare localization, a little always leaks through, so I reversed the phase of one of the rear channels to get rid of the asymmetry. Yes, in-phase splits sound different than out-of-phase spilts. The in-phase splits tighten the soundfield, while the out-of-phase splits open it up and make it more diffuse.

There is a similar weirdness on the encoding end. You see, the standard SQ encoder doesn’t really follow the ideal SQ pattern as the signal is panned from LF to LB, or RF to RB. I forget which way the deviations go, but for one side, the pan-locus drifts towards the front, while the other side drifts towards the rear. All of the “standard” 4 to 2 encoders do this, and it’s too much to ask even a sophisticated dynamic decoder to anticipate this.

This is where the Front and Back encoders come in. They have much better side-location encoding (it’s symmetrical for one thing), but there’s one thing they can’t do: a Front encoder can’t encode Center Back, and a Back encoder can’t encode Center Front. For obvious reasons, a Back encoder is pretty useless, but a Front encoder is often a better and more natural-sounding encoder than the standard model. (Think of both Front and Back encoders as more-accurate 270-degree encoders, while the “standard” encoder is a less-accurate 360-degree encoder.)

What do I mean when I say a Front encoder can’t encode Center Back? Well, you can go ahead and input matched LB and RB signals, but the encoders then emits an encoded Center Front signal! It literally “won’t go there” if you try to pan across the rear of the room. If you try, the signal sweeps from one rear speaker, to the center front, then sweeps to the other rear speaker. So a rear pan is forbidden.

But ... how many recordings use a 360-degree circular pan? Some, true, but it’s mostly a novelty that might get used for one track, just to show off that it’s a quadraphonic recording. If image integrity is important, then the Front encoder does a better job. In subjective terms, Center Back is an unpleasant and unstable localization, and not where you want to place an instrument. Worse, it disappears in mono, which is a serious no-no for AM radio or other low-fi environments. This is also a subtle disadvantage of QS encoding, where LB and RB are nearly 10 dB down relative to Center Front when played in mono.

CBS has always been a commercially-focused company. The biggest reason for the SQ system, for all of its weird properties, was 100% compatibility with existing stereo *and mono* broadcasts and recordings. My personal preference these days is EV4 with dynamic decoding, but in all honesty, mono compatibility isn’t great, with the rear channels well down in level compared to the fronts.

Oh yeah, I nearly forgot, there’s the “Position Encoder” as well. This is not a 4 to 2 encoder, but a box with an array of clickstop switches that cover a full 360 degrees. Each input channel is assigned it’s own switch, which then selects the location for that track. Obviously, no pans are possible (without making click noises), but the device does create an ideal SQ matrix, with no deviations from intended locations.

I think you can see my perspective a little more clearly now. I view all matrix systems as a more efficient use of a 2-channel storage and transmission medium, and recognizing that 2-channel has been the dominant standard since 1958, when the 45/45 Westrex disc was chosen as the market standard. The dominance of 2-channel was further cemented when the Sony/Philips alliance forbade the production of 4 or 5 channel Compact Discs. (This was technically possible with existing early-eighties technology, but Sony and Philips wanted costs as low as possible, and for all discs to play on all players.)

Given the dominance of 2-channel media for music, and the enormous back catalog, a highly compatible (up and down) matrix offers a lot of advantages for the music lover. All that’s required is a modern decoder, which is where our Australian and British friends come in.

 

[malcolmlear]

Hi Lynn, I'll have a look out for Loggins & Messina "Full Sail" (if I can locate it), sounds like a good test )))
Just got auto selection of s/pdif and ADAT input (48kHz) working and it works and sounds great. Its just mesmerizing, selecting different decode modes on the front panel )). I'm very happy with the specs and performance now, so its time for more documentation and finalize the PCB and parts lists for pre-production.
Hopefully I'll have time after this to look into an advanced variable matrix SQ encoder just to see how good the system might have been.

 

[JonUrban]

CBS had four different encoders, not including the specialized Ghent microphone. .........

Hi Lynn,

Richard (OD) has asked me to post the following information from him that you may see it. Here's what he sent me:

"I would like to add that reversing the Rear Right channel (which is what I suspect you did) just causes more disturbance and reduces the possibility of creating a clean stable detailed image, as you made mention of. This does not need to be though. There is a cure to the known issue you came across. I too found the issue related to the Rear Right channel in my work, and finally found the solution that eradicates it with zero side effects with either the Center Rear, Right Rear or Right side areas. In fact it enables my process to produce the most stable image ever heard from an SQ source, and I'm sure that if you continue work on the actual cause of the issue you'll find similar improvements. I must add that this will only work when used on decoders based on 'Ben Bauer's published decoder specifications. Good luck!"

 

[MidiMagic]

I have that article "Advances in SQ Recording Techniques, sent to me by Ben Bauer himself.

If you read in the article how to connect a mixer to these encoders, you can do what you want.

There were actually 5 different encoders, but they could overlap on the mixer. This shows how correct the encoding is for each pair.

SQ ENCODER	LF-RF	LB-RB	LF-LB	RF-RB	LF-RB	LB-RF
4-Corners Matrix	BEST	BEST	Odd	Odd	BEST	No
Acroperiphonic	BEST	BEST	Odd	Odd	Odd	Odd
Diagonal Split	BEST	No	Odd	Odd	BEST	BEST
Forward-Oriented	BEST	No	BEST	BEST	Odd	Odd
Backward-Oriented	No	BEST	BEST	BEST	Odd	Odd

For instance, all of the encoders except the backward-oriented encoder and the Acroperiphonic encoder used the same front encoder.

SQ ENCODER	FRONT ENCODER	BACK ENCODER
4-Corners Matrix	Front oriented	Back oriented
Acroperiphonic	Straight mixing	Back oriented
Diagonal Split	Front oriented	Diagonal Split
Forward-Oriented	Front oriented	Front oriented
Backward-Oriented	Back oriented	Back oriented

So a 6-bus mixer can be used with 3 front buses and three back buses to get all 5 encoders.

Use of the mixing bus selectors lets you choose a different encoder for each channel strip. And by pushing bus selector buttons during panning, you can even do a full circle pan.

The main purpose of the 4-corners encoder was to take an existing discrete tape and encode it.

 

[Lynn Olson]

Midimagic, that’s definitely new information to me. Then again,
I was never part of the recording community; Audionics was just another SQ licensee, and they rather selectively disclosed information to us. The four-person engineering team (as I recall it) seemed to be rather surprised to discover that our team was just me, and that Shadow Vector was a fully operational dynamic decoder designed and built in the wilds of Oregon. The dual-beam Tektronix scope connected to the 3 logic-sense lines subtly reminded them we had technology out there too.

What the heck is a “Acroperiphonic” encoder, and what type of encoding does it favor? No clue on that one. I also didn’t know there was a dedicated diagonal-split encoder ... news to me. I imagine if multiple encoders were used on a SQ mix, the pole structure would, of necessity, be identical. From what I saw of the CBS 4 to 2 encoder, it was a cascade of 301 op-amps (ugh, very low slew rate) added up to 10 poles of phase shifting. The other CBS encoders must have had an identical phase shifter array if the system allowed them to be used in parallel.

The EMI encoder looked very different. It was all discrete transistor, no op-amps anywhere, and only 6 poles, using a parallel instead of series cascade. The parallel approach is electrically cleaner, but requires careful hand-selection of precision capacitors for good performance. This was also the approach we used for Shadow Vector.

Sonically, I preferred the EMI encoder, and I liked their approach of releasing their entire classical catalog in SQ format ... no stereo release at all. No price differentiation, no hideous gold frame around the album, just everything in SQ. This made a ton of sense, considering the entire reason for the SQ matrix was superb compatibility with stereo recordings. That was kind of the point. Seamlessly transition to well-recorded quadraphonic, and let the buyers transition at their own pace, as decoders gradually improved, and recording technique advanced.

Here in the US, of course, things promptly fell into disarray, and dual inventory became the bane of retailers. CBS corporate just couldn’t resist squeezing another dollar out of the customer, which then undermined the whole SQ compatibility argument. Frankly, that kind of short-sighted profit-taking killed quad as much as anything. (Along with the JVC CD-4 system, which was very definitely not ready for prime-time ... for one thing, it wasn’t compatible with FM disc-jockey practice of back-cuing.)

Looking back, SQ and QS could have easily co-existed, but the completely incompatible JVC system marked the beginning of the end. It created a market-mania for “discrete”, never mind the substantially lower sound quality, extreme sensitivity to dust, a requirement for a pretty exotic cartridge, as well as special low-capacitance phono cables. Having THREE systems just to play records was a bridge too far for most customers. It was a very different consumer experience than the standardized Westrex 45/45 stereo system, which was fully compatible with standard mono LP’s (which were still pretty new at the time).

(I know JVC fans are gonna razz me for that “substantially lower quality” statement. Let’s be real here. A separate FM carrier on each groove wall is insane. Cartridge channel separation at 30 kHz is sketchy at best, FM phase-lock-loop demodulators in the early Seventies were primitive, and weren’t particularly low distortion in the presence of the interfering FM carrier from the other groove wall. And in the final blow, the JVC system is ALSO a matrix system, one that requires a test-disc adjustment of the baseband level since phono cartridges do NOT have standardized output levels. If the baseband level is not precisely set, the total F/B separation drops to 6 to 10 dB, which kind of makes the whole complex system pointless.)

 

[Lynn Olson]

Whoa, Midimagic, I finally just read what you said about the front and rear portions of each flavor of encoder. So the Acroperiphonic actually has a quasi-random phase relation between the front and rear channels! That’s wild! But that would allow a 4-channel sum, as well as a Center Front to Center Back pan. That’s usually QS and EV4 territory.

And the Front half of the standard encoder is trivially simple; LF goes through a 0-degree phase shifter (this is NOT the same as a direct connection) and then out to Lt. The same for RF; it goes through a 0-degree phase shifter, and then out to Rt. Unusually, the Acroperiphonic encoder bypasses the allpass filter and just feeds LF straight to Lt, and RF to Rt.

The back half of the encoder is where all the fancy matrix stuff happens, and I can see why this creates all the other variations. This is really quite clever, and a good match for multitrack studio technique.

And the virtue of the SQ encoding system is the treatment of the back channels, which are a 90-degree phase spread. This is a good mapping of encoded “rear” information in 2-speaker playback, so the rear encoding is diffuse and only moderately localized to the speaker with the leading phase.

The QS and EV4 systems present “rear” encoding as somewhat out-of-phase, so they appear as extra-width images (if the speakers are good enough).

Hmm ... the Acroperiphonic SQ encoder could be also used for purist classical recordings, with front-stage orchestral content bypassing the allpass filters (which would leave waveshapes intact). I wonder if EMI did that.

The Acroperiphonic encoder cannot localize any side images (thanks to random phase between front and rear), but that wouldn’t matter for a normal classical presentation. But it might confuse a dynamic decoder ... not sure about that. When I visited EMI Labs in 1975, the Shadow Vector handled EMI recordings just fine. I surmise that is exactly what Malcolm has found with his prototype.

P.S. For Malcolm: When I visited BBC Labs during that visit to the UK in 1975, they had a 4-track Studer with a discrete quadraphonic recording of “Last Night at the Pops” with a full-choral version of Beethoven’s 9th. Best sound I’ve heard before or since. I was able to directly compare discrete (recorded with a Soundfield microphone), BBC UHJ decoded with their decoder, and SQ decoded with the Shadow Vector. They were expecting SQ to be a distant third, but it wasn’t. I found it better than UHJ, and surprisingly close to the discrete master. Every bit as spacious as the discrete master, which really surprised the BBC engineers. They were expecting the usual dry SQ logic sound, but of course that’s not how Shadow Vector sounds. So I think you’ll be pretty pleased to hear how SQ encoding sounds with your modernized version.

I personally find sweetness and spaciousness to be the hallmark sound of Shadow Vector, and quite different than other SQ decoders. It’s more like QS, but without the blur and vagueness of the Vario-Matrix, or the dullness of DPL-II.

 

[par4ken]

The EMI encoder looked very different. It was all discrete transistor, no op-amps anywhere, and only 6 poles, using a parallel instead of series cascade. The parallel approach is electrically cleaner, but requires careful hand-selection of precision capacitors for good performance. This was also the approach we used for Shadow Vector.

Sounds exactly like the Audionics 106 decoder.

 

[par4ken]

no hideous gold frame around the album

I love the gold frame, it makes the album look special. The A&M Audiophile series of LP's had a similar gold frame.

 

[MidiMagic]

All 5 of the encoders are in the article "Advances in SQ Recording Techniques" in the Sonik Wiz post. Click on it and read the article. The Interior switch activates acroperiphonic encoding. I have been using that article since 1972.

The fun part is that you can use a 12-bus STEREO mixer to do all of this. The encoders go into the bus inserts.

Use the main L-R bus for the interior (acroperiphonic) front
Use buses 1-2 for the Front-Oriented front encoding
Use buses 3-4 for the Back-Oriented back encoding
Use buses 5-6 for the Diagonal Split back encoding
Use buses 7-8 for the Front-Oriented back encoding (trade LB and RB)
Use buses 9-10 for the Back-Oriented front encoding (trade LF and RF)

Now you can use a STEREO pan pot to mix the whole thing. For each channel strip, select 2 mixing buses. The pan pot then makes the following pans:

Use L-R for front position encoding Interior (acroperiphonic)
Use 1-2 for front position encoding Front-Oriented, Diagonal Split, or 4-corners
Use 3-4 for back position encoding Interior, 4-corners, or Back-Oriented
Use 5-6 for back position encoding Diagonal Split
Use 7-8 for back position encoding Front-Oriented (LB and RB traded)
Use 9-10 for front position encoding Back-Oriented (LF and RF traded)

Use 1-6 for a diagonal split pan LF to RB
Use 5-2 for a diagonal split pan LB to RF
Use 1-8 for a side pan LF to LB in Front-Oriented
Use 7-2 for a side pan RB to RF in Front-Oriented
Use 3-10 for a side pan LF to LB in Back-Oriented
Use 9-4 for a side pan RB to RF in Back-Oriented

With a smaller mixer, use just the encoders needed.

 

[Lynn Olson]

Thanks again for the 1973 AES Ben Bauer article. I have my stash of AES Journals going back to 1973, but that's just a few months before I became a member. But I do remember reading that same article in the college library around the time I came up with Shadow Vector, back when I was living in Los Angeles, and also reading the Scheiber notation articles explaining in more understandable form the exact differences between EV4, QS, and SQ. The way to do a QS variable-matrix is fairly obvious from inspection, as well as the "half-logic" variable-blend SQ decoder (which improved Center Front and Center Back separation from 0 dB to 20 dB, but has no effect on the corners). The tricky part was imagining and building a variable-matrix SQ decoder from inspection of the SQ pan-locus, which is way more complicated than QS encoding.

As readers know, the Shadow Vector is a dynamic decoder that reverts to a static SQ decoder if the 3 control lines from the logic-sensing are all sitting at zero. Each of the 3 control lines can be modulated in the plus or minus direction (this is what we drove the Tek dual-beam scope with at the CES show in 1975). We also used the same Tek dual-beam scope to dynamically calibrate the SV with a spinning input signal, so it yielded equal separation under dynamic conditions.

From what I can tell, the CBS Paramatrix, the TATE DES, and I guess the Surround Master have an internal static SQ matrix, and additional audio signals are added to the LF, RF, LB, and RB outputs of the static matrix. In principle, they should come out exactly the same, but from listening, the CBS Paramatrix and the TATE DES seemed to be more corner-focussed than the Shadow Vector. Also dryer sounding, and I have no explanation for that, since I don't really understand how those gizmos work.

Shadow Vector relies on symmetrical control lines for all 3 axes of the Scheiber sphere, with a specialized shaping function to keep overall power levels constant and separation equal in all directions, including intermediate locations. I surmise the logic of the other three decoders work in a quite different fashion, but I've never clearly understood quite how they worked. The TATE DES patent is completely opaque, which I think violates the spirit of patent law, since patents are supposed to replicated by one "skilled in the art".

 

[proufo]

And by pushing bus selector buttons during panning, you can even do a full circle pan.

Wow. Quite complex stuff.

 

[Lynn Olson]

Thanks for the link to Disclord's wonderful postings ... a great thread here:
Disclord thread

I concur with another of his postings that the know-how of building surround decoders was being lost at the time, but Surround Master and Malcolm are bringing it back. He passed on too soon.