SQ Shadow Vector Soundfield Mapping

QuadraphonicQuad

Help Support QuadraphonicQuad:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
The patent isn't 100% clear on the nonlinear function between the control logic (3-axis direction sensing) and the presumably linear control voltage of the VCA's. Zero volts on the three control lines represent no dominant direction sensed, so the matrix coefficients are the same as a static SQ (or QS) matrix. Each control line has a plus or minus direction, depending on the magnitude and sign of the CF/CB, LF/RF, or LB/RB axis detected on the Scheiber sphere. (This is similar to the phase of the chroma subcarrier in NTSC or PAL color television representing hue, and the percentage magnitude of the subcarrier representing color saturation. Analog color TV is a 2-axis system, while SQ is a 3-axis system, which is why Scheiber notation is needed to understand it.)

But this leaves the exact relationship of the control magnitude and the percentage gain of the VCA's a nonlinear relation. Well, in practice, a string of resistor/diode pairs creates the quasi-exponential part of the curve ... but was the exact shape determined?

Fortunately, this was pretty easy, although not described in the patent (I guess it should have been, but it wasn't). All it takes is a scope with X/Y inputs and a pair of oscillators that are reasonably stable. Set both oscillators to 1 kHz at a moderate level (do not clip the sensing logic or the overall decoder), and offset them by about 5~10 Hz, which is well within the response speed of the direction-sensing logic. As you can imagine from the above drawings, this creates a rapid circular pan from CF -> RF -> CB -> LF -> CF and so on.

1) Connect the LF/RF control line to the horizontal axis of the scope, and connect the vertical amplifier to one of the decoder outputs, starting with LF output (all four will be examined in the same way). The 5~10 Hz offset will swing the display from left to right, and the decoded output will drive the vertical. What you will see is a smooth ramp as the output drops from LF to RF (no dynamic decoding in this region), and a suppressed region as the output swings towards CB (it should be zero at CB, and rise again as it approaches LF). Connect the vertical amplifier of the scope to the RF output; you should see the same thing, just on the other side of the display. Both LF and RF suppression of CB should be mirror-images of each other.

2) Connect the vertical amplifier to LB; the pattern should now be different, with the entire front half of the Scheiber sphere suppressed, and the magnitude only increasing as the incoming signal swings toward CB. This lets you examine the smoothness of the suppression region, and the accuracy of the nonlinear control function (the resistor/diode string). Our prototype, with a 6-pole passive phase shifter, had between 35 to 45 dB of suppression across the entire frontal arc. Connect the vertical amplifier to the RB output; you should see the same pattern ... in fact, this is a good opportunity to make sure that both LB and RB have exactly equivalent patterns. We used a Tektronix dual-beam scope to look at both LB and RB simultaneously, so we could exactly match the two nonlinear functions to each other, and make sure the entire system had symmetric suppression patterns with no bumps or nonlinearities present.

3) The rest is easy. Just use the same transfer function for the remaining 2 axis, CF/CB and LB/RB. Confirm by connecting the CF/CB line to the horizontal input of the scope, and looking at the four output channels again. Look for symmetry and any bumps in the suppression region. When everything is symmetric and suppression is uniform, you're done, except for minor magnitude correction in the decoded channels to keep total magnitudes of the whole decoder exactly matching a static decoder (we didn't find the less-than-1dB correction to be audible).

This should illuminate what I mean by "symmetric" ... it's actually a calibration function of the real-world decoder, and it dynamically measures separation across an entire arc of localizations. I'm not sure anyone else every did anything like this; they were all focused on corner separation, with a nod to preserving overall energy through the decoder. As you can see, if the action of the dynamic decoders is accurate, it not only preserves energy levels in that channel (following an ideal cardioid pattern) but preserves overall energy as well.

This is why it sounds more spacious than other decoders, including, I suspect, modern DPL-II (music) and DTS Neo:6. I don't hear equal energy reverberation through these, and based on modern loudspeaker layouts for home theatre, I don't think symmetric distribution of energy is a design goal.

Theatre systems are, by design, asymmetric, and are asymmetric in the home, as well. This is radically different than the design goals of the better quadraphonic systems of the Seventies, which were designed exclusively for home use. They didn't get re-purposed for theatre use until the late Seventies, when the QS system was re-aligned for a L/C/R/surround format, and after that, Dolby Digital, DTS, and the Sony system.

The requirements for a theatre system are different because nearly all the audience is listening off-center ... the "sweet spot" we use at home is less than 1~2% of the seats in a theatre. The result is no phantom imaging; everything jumps into the nearest speaker. This is why a movie theatre MUST use a center speaker for spoken (or sung) dialog; it sounds really weird if the critical dialog seems to come from the L or R speaker for the whole movie; it would be very, very distracting. This is why all theatres, going back to the dawn of sound movies, have 1 speaker, 3 speakers, or 5 speakers directly behind the perforated screen. It's the only system that isn't distracting.

As a result, the special effect of a moving pan requires lots of speakers, overhead as well as the sides and back. It's the only system that works when nearly all the audience is sitting off-center.

Quadraphonic, by contrast, is a refinement of domestic stereophonic sound, and smooth, evenly distributed phantom images are essential to the whole system. A theatre already sounds large ... because it is ... while a domestic system has to reproduce the spatial impression of the recording, which can be any size, and often completely artificial (EMT plates, digital reverb, etc.).
 
Well, you should contact Malcolm Lear. I'm just the old guy telling stories about how it was done in hardware (with the one working and demo-quality prototype). Unlike the TATE DES, it was never reduced to silicon, and never went into production. Weirdly enough, Audionics had TWO different inventors on contract, and they chose the TATE over mine, because the TATE team had Hollywood backers that paid National Semiconductor the umpteen dollars it took to reduce to silicon. All I had was the one working prototype that had been demo'ed to CBS, EMI, and BBC Labs ... and after that demo tour, I was re-assigned in 1975 to designing loudspeakers. (I do that for entertainment value.)

The TATE team found the conversion to silicon turned out to be both challenging and more expensive than anticipated. True to form, National Semiconductor cut corners without telling TATE (I found later at Tektronix they were notorious for this and were sued by the Federal government for shoddy work on Defense contracts) and the real-world chips needed a fair bit of add-on circuitry to actually function as designed in the Audionics Space & Time Composer. (That was a project that went on after I left Audionics in 1979, and I had no connection with the TATE design team.) The years of delay at National Semiconductor, along with all the additional engineering needed by Audionics, moved the TATE SQ decoder past the official death of SQ recordings, and almost into the earliest days of digital. So the market for quadraphonic playback of LP's was much smaller than anticipated.

When I left Audionics for Tektronix in 1979, it was from being a medium-sized frog in a very small pond to a little tiny frog in a very large pond. Tektronix at the time had 22,000 employees, and along with Hewlett-Packard, was the world leader in test and instrumentation. I was a beginning tech writer (since I had written product brochures and an AES article for Audionics), thrown into the deep waters of microprocessor design and test fixtures. After getting out of that job as quick as I could, I went to the main campus at Beaverton, Oregon, to write for the Spectrum Analyzer group, where I was much more comfortable writing about analog technology. That was also my introduction to the DEC VAX 11/780, VMS, Unix, Emacs, and the Arpanet email and group discussions about audio. I didn't really get back into audio until 1990 or so, when I joined the Oregon Triode Society and started writing for their club magazine (which is now Positive Feedback Online).

As you can tell, I still have an affinity for what's now called "surround sound". I'm not a fan of the cinema-centric version we have now, but that's what drives the market. I have a Marantz AV8003 pre/pro and MM8003 combo (with balanced interconnects) after finding most AVR's don't sound very good ... considerably inferior to the old Audionics CC-2 amplifier I used to listen to back when I was designing loudspeakers for Audionics.
 
Last edited:
Malcom, feel free to call your product, or design, a "Shadow Vector" decoder. It deserves to see the light of day after all these years.

If it fits easily into the existing digital paradigm of S/PDIF (2-channel) input, and S/PDIF (multichannel) output, so much the better. My own setup, like most others, is an AVR with a few spare S/PDIF inputs, one set of 6-channel analog inputs, and lots of HDMI. The analog AVR inputs are reserved for SACD multichannel, which is a format that's usually trapped inside the player (by design on the part of Sony). That leaves HDMI, which has annoying licensing fees, or industry-standard S/PDIF, which does support PCM multichannel, and is recognized by just about any AVR made in the last ten years. Yes, there's Toslink too, but that has less bandwidth, and is known for high jitter and not-so-great playback quality.

As for "things I'd like to see", aside from the all-digital S/PDIF input and S/PDIF output features, SQ and QS decoding would be really nice, with an option for EV4 an extra plus, although I admit that's pretty similar to QS. Both QS and EV4 are optimum for stereophonic expansion into high-resolution surround, which is my preferred listening mode for nearly any stereo recording.

Unlike some other posters, I'm not a purist about listening to 2-channel recordings on 2 loudspeakers. True, that's way it was mastered, but if you're a hard-core purist, you really should replicate the mastering speakers used for that recording. For most recordings made in the Fifties and Sixties, that would be 15" Tannoys in the UK, Altec 604 Duplexes in the US, and I don't know what in Europe and Japan. If the recording is an original pressing and not a remaster, you should also listen on the cartridge used to audition the lacquer ... an Ortofon SPU with spherical stylus in the Europe and the UK, and a Stanton 681A (spherical) in the USA. And all-vacuum-tube equipment, of course.

I'm a fan of vacuum-tube gear, classic high-efficiency speakers, and the Ortofon SPU, and by golly, they sound fantastic on vintage recordings. Really, no joke, tone quality is amazing, and they have plenty of "vibe" and the feel of the era.

But we can also play these recording on modern equipment and hear far more than the original mastering session (if the system is good enough). So the "purism" argument goes out the window with modern ultra-flat loudspeakers, modern line-contact stylus cartridges, or digital transfers of the master tape. That sounds nothing like the vintage equipment used when the artists signed off on the master many decades ago. For that matter, an LP "remaster" is not an exact duplicate. The cutterhead amplifier is not the same (a lot different, with much more power), the analog tape deck might use digital preview instead of using a dedicated preview head (this means all of the audio that goes onto the new LP has passed through a 96/24 PCM delay system), and most likely different mastering loudspeakers are used for the final audition and tonal balancing. The master tape is the same (minus the oxide that's flaked off), but all the rest is different.

So if the goal is sheer enjoyment rather than archival adherence to recreating what was heard in the studio thirty to sixty years ago, what's wrong with modern gear? This isn't a religion. A really good dynamic decoder isn't "making anything up" or adding reverb or anything like that, it's all there in the 2-channel master. You just can't hear it on 2 speakers, that's all, and you can't hear it on headphones, either. To hear the phase relations between the 2 channels takes a decoder ... headphones discard phase relations (in-phase and out-of-phase sound similar), while out-of-phase content on 2 speakers merely sounds diffuse and non-localized. By contrast, a good, high-resolution QS or EV4 decoder will reveal all kinds of spatial detail that's there in the recording ... it's just not audible in traditional 2-speaker playback.

Why not hear more, rather than less? It's all there in the recording, waiting for you to hear it.
 
Last edited:
By the way, that Surround Master V2 from Involve Audio looks like a lot of fun (great thread here on the forum). Time to send my order in ... although I'm curious about the digital-in/digital-out version, if that happens. If it doesn't, I can work around the all-analog version by using an inexpensive DAC to feed the unit, then feed the 5.1 analog outs to the (only) analog inputs of the Marantz pre/pro.

I guess I shouldn't be surprised that people are still confused about SQ versus QS/EV4/DPLII encoding. Without using Scheiber notation it's difficult to describe the difference. For one thing, an unusual property of quad (and 5.1) matrix systems is the complexity is in the decoder, not the encoder. The encoder is really simple, just a few op-amps and summing or subtracting resistors, and that's it. The mathematical relations are spelled out in the documents.

SQ requires 0 and 90-degree allpass phase shifters on both encode and decode ends. These are optional for QS, and not needed for EV4 or DPLII. Why does SQ require these? Well, Left Back is encoded by a mix of L at 0 degrees and 0.707 magnitude, and R at -90 degrees (delayed) and 0.707 magnitude. Conversely, Right Back is encoded by a mix of R at 0 degrees and 0.707 magnitude, and L at -90 degrees (delayed) and 0.707 magnitude. Note that both 0 degrees and -90 degrees come from allpass phase shifter circuits that cover the entire audio band, and they're exactly 90 degrees apart (at all frequencies). Yes, square waves look pretty weird after they pass through these things ... but phase distortion is actually pretty hard to hear, even for the "golden ear" types.

SQ encoding has the interesting property that Left Back plays back in 2-speaker stereo with excellent compatibility ... the transients tend to appear closer to the Left speaker, with some blurring across the soundstage (this is what 90-degree phase shifts sound like). The same applies to the Right Back speaker ... the transients appear closer to the Right speaker, with some blurring across the soundstage. Without those 0 and -90 degree all pass phase shifters, you can't do SQ .. you can't encode it, and locations in the rear are moved around if played in QS/EV4/DPLII. (To be precise, QS/EV4/DPLII plays SQ-encoded Left Back and Right Back in all 4 speakers at equal magnitudes, just with different phase relationships. On other hand, an SQ decoder plays QS/EV4/DPLII-encoded Left Back and Right Back as something pretty close to Center Back, with no real L/R directionality.)

QS encoding is a different animal. The allpass phase shifters allow a 360-degree spin around a circle, but aren't really needed otherwise. Left Back is encoded by a mix of +L at 0.924 magnitude, and -R (reverse phase) at 0.383 magnitude (these are sine/cosine relationships). Right Back is encoded as +R at 0.924 magnitude, and -L (reverse phase) at 0.383 magnitude. When played in 2-speaker stereo, Left Back appears somewhat outside the Left speaker with a somewhat vague image, and Right Back appears outside the Right speaker with somewhat vague image.

One of the downsides of classical QS encoding is Left Front and Right Front only have 8 dB of separation in stereo playback, which is a pretty serious drawback. EV4 encoding is similar, except that Left Front and Right Front have 20 dB of separation, while the rears are much closer together (almost summed, but not quite). In other words, EV4 is a modified form of QS, except that it preceded QS by a year or so. But they are closely similar and compatible with each other.

It also doesn't help that the names of the SQ and QS systems sound so much alike ... but they're not. SQ is the odd duck with some real advantages, but the SQ dynamic decoder is far more complicated than the QS Vario-Matrix dynamic decoder. Very few ever heard the Shadow Vector, the CBS professional Paramatrix, or the early, all-discrete TATE DES decoders, which showed the real potential of the SQ system. That the Surround Master decoder even exists is nothing short of a miracle ... and I agree with the scope photos that just how poor DPLII is in retrieving surround content from stereo sources.
 
Last edited:
By the way, that Surround Master V2 from Involve Audio looks like a lot of fun (great thread here on the forum). Time to send my order in ... although I'm curious about the digital-in/digital-out version, if that happens. If it doesn't, I can work around the all-analog version by using an inexpensive DAC to feed the unit, then feed the 5.1 analog outs to the (only) analog inputs of the Marantz pre/pro.

I guess I shouldn't be surprised that people are still confused about SQ versus QS/EV4/DPLII encoding. Without using Scheiber notation it's difficult to describe the difference. For one thing, an unusual property of quad (and 5.1) matrix systems is the complexity is in the decoder, not the encoder. The encoder is really simple, just a few op-amps and summing or subtracting resistors, and that's it. The mathematical relations are spelled out in the documents.

SQ requires 0 and 90-degree allpass phase shifters on both encode and decode ends. These are optional for QS, and not needed for EV4 or DPLII. Why does SQ require these? Well, Left Back is encoded by a mix of L at 0 degrees and 0.707 magnitude, and R at -90 degrees (delayed) and 0.707 magnitude. Conversely, Right Back is encoded by a mix of R at 0 degrees and 0.707 magnitude, and L at -90 degrees (delayed) and 0.707 magnitude. Note that both 0 degrees and -90 degrees come from allpass phase shifter circuits that cover the entire audio band, and they're exactly 90 degrees apart (at all frequencies). Yes, square waves look pretty weird after they pass through these things ... but phase distortion is actually pretty hard to hear, even for the "golden ear" types.

SQ encoding has the interesting property that Left Back plays back in 2-speaker stereo with excellent compatibility ... the transients tend to appear closer to the Left speaker, with some blurring across the soundstage (this is what 90-degree phase shifts sound like). The same applies to the Right Back speaker ... the transients appear closer to the Right speaker, with some blurring across the soundstage. Without those 0 and -90 degree all pass phase shifters, you can't do SQ .. you can't encode it, and locations in the rear are moved around if played in QS/EV4/DPLII. (To be precise, QS/EV4/DPLII plays SQ-encoded Left Back and Right Back in all 4 speakers at equal magnitudes, just with different phase relationships. On other hand, an SQ decoder plays QS/EV4/DPLII-encoded Left Back and Right Back as something pretty close to Center Back, with no real L/R directionality.)

QS encoding is a different animal. The allpass phase shifters allow a 360-degree spin around a circle, but aren't really needed otherwise. Left Back is encoded by a mix of +L at 0.924 magnitude, and -R (reverse phase) at 0.383 magnitude (these are sine/cosine relationships). Right Back is encoded as +R at 0.924 magnitude, and -L (reverse phase) at 0.383 magnitude. When played in 2-speaker stereo, Left Back appears somewhat outside the Left speaker with a somewhat vague image, and Right Back appears outside the Right speaker with somewhat vague image.

One of the downsides of classical QS encoding is Left Front and Right Front only have 8 dB of separation in stereo playback, which is a pretty serious drawback. EV4 encoding is similar, except that Left Front and Right Front have 20 dB of separation, while the rears are much closer together (almost summed, but not quite). In other words, EV4 is a modified form of QS, except that it preceded QS by a year or so. But they are closely similar and compatible with each other.

It also doesn't help that the names of the SQ and QS systems sound so much alike ... but they're not. SQ is the odd duck with some real advantages, but the SQ dynamic decoder is far more complicated than the QS Vario-Matrix dynamic decoder. Very few ever heard the Shadow Vector, the CBS professional Paramatrix, or the early, all-discrete TATE DES decoders, which showed the real potential of the SQ system. That the Surround Master decoder even exists is nothing short of a miracle ... and I agree with the scope photos that just how poor DPLII is in retrieving surround content from stereo sources.


Hi Lynn

I have read with interest this thread but have stayed out. I think we have used many similar techniques but having said that I really do not fully understand shadow vector. May I suggest that if you are just curious about Involve that you get the encode and decode evaluation kits at some stage (its way cheaper and really is technically the same). At the very least think about the encoder board as it has Involve encode which is a twist on QS in that it is actually a triband variable matrix encode (I know it sounds philosophically wrong) but it maintains a full stereo separation on the encode stream and also full surround. I can send you privately some information about it if you want.

I have attached a press release on the units.

Keep up the great work!

Chucky
 

Attachments

  • Involve evaluation module press release_V1.4(1).pdf
    725.5 KB
Seriously, I'm quite interested in buying one of your units ... the price is entirely fair, particularly considering the development costs, and the versatility of it. My preference is for S/PDIF in (2-channel) and S/PDIF out (multichannel), but that's because all of my sources are digital, and the fairly decent-sounding AVR is a Marantz with bass management, etc. I don't use the Audessy EQ, which doesn't sound good at all in my system, nor do I use distance correction. So most of the receiver processing functions are bypassed, which is fine by me.

I am strongly contemplating getting a NAD T777 v2 to replace the Marantz AV8003, not because of the HDMI foolishness, but because it supports BluOS streaming (I have full-res TIDAL subscription) and Dirac Live time/frequency equalization. As an old-timer speaker designer, I don't really believe in room equalization except to resolve a few peaks in the bass region, but I do see unwrapping crossover phase distortion as a good thing, and Diract Live does that, along with offering the user a selection of custom EQ curves.

There are actually two systems in the living room; they meet at the crossovers of the Ariel speakers, which have changeover switches with isolated grounds, so the grounds of the two systems are independent and don't touch at any point. One system is the Marantz AV8003 and MM8003, and the other system is a pretty exotic 2-channel all-triode, push-pull Class A 300B amplifier and vacuum-tube DAC with Burr-Brown PCM-63K converters. The signal path in that system is a current-mode converter, passive resistor I/V conversion with passive lowpass filter, 6DJ8 SRPP amplification, volume control, 6DJ8 linestage, and the Karna amplifiers, with an input transformer (which isolates the RFI from the DAC and the outside world), 5687 input stage, transformer-coupled PP 45 triode driver stage, and transformer-coupled PP Class A 300B output stage.

So I can have either all-triode, zero-feedback sound, or all-transistor, all-feedback sound at the flip of a few switches at the crossover. But ... no, I'm not going to build an all-tube quad system. I have to draw the line somewhere, and besides, the sound out of the Marantz is actually half-decent, unlike many AVR's. It probably helps that the pre/pro and power amps are on separate chassis with separate power supplies.

Adapting the Involve unit would be a moderate hassle, but there are pretty decent little DACs in the $100~150 range these days, and that could take care of the input side of the unit, while the output feeds the Marantz or NAD set of discrete analog inputs. As luck would have it, my Marantz blu-ray player can send SACD multichannel through HDMI, or, failing that, convert DSD into PCM multichannel and send it through the S/PDIF output. Either is satisfactory for the handful of SACD discs I have.

Listening to surround sound from old SQ recordings or contemporary stereo recordings is a sentimental journey for me. I spent a half-decade with the Shadow Vector and all the arcana of dynamic decoders, but that was back in the early Seventies. I briefly auditioned UHJ at the BBC labs in 1975 and was duly impressed with the natural, effortless spatiality, but the one recording I have never forgotten was a 4-channel discrete recording made on a Soundfield microphone, recording on a Studier professional 4-track recorder, and played through four BBC monitor loudspeakers. It was Beethoven's 9th, with full chorus, and not only was it 100% spatially realistic, but there was ZERO audible distortion, even during the climaxes with full orchestra and chorus. I have never before or since heard anything like that. A carefully recorded first-generation tape master doesn't hurt, of course, and the Soundfield microphone was stunningly realistic.

Anyway, I want to thank you again for all the heavy lifting and putting the Involve unit into production! Send me a PM, and put me on the waiting list, particularly if there is a stripped-down unit with no analog, just digital interfaces (I vote for S/PDIF, not HDMI). I don't have a working turntable at present, even though I have a stash of about 30 SQ and QS records back from the Shadow Vector days. Maybe I'll get a friend to digitize them at 96/24 PCM encoding, and have them as FLAC files on my computer.
 
As for the theory underlying the Shadow Vector, this is back in the all-analog days, when the cheapest computer would be a DEC "minicomputer" that took up one rack, cost $150,000, and required an on-staff FORTRAN programmer. That's the setup I saw at KEF loudspeakers when I visited them in 1975. I found it pretty intimidating, and there was no way a little company like Audionics was going to spend that kind of money on R&D costs. So the Shadow Vector is very much an all-analog system, using 3-axis linear direction-sensing that controlled three VCA arrays per channel. These see-saw pairs of VCA's swung the decode points in 2 possible directions, or a mixture of both, depending on the control signals. Unlike the CBS Paramatrix or the several variants of TATE DES, there was no "corner" optimization for the SQ cardinal points, so the separation was the same whether the encode localization was a hard corner (Left Front, Left Back, Right Front, or Right Back) or an intermediate location. The vectors of the suppressed channels tracked the dominant sound and did their best to keep 180 degrees away from it, and there was a soft (not hard) handoff as the suppression effect smoothly decreased at the intermediate locations.

I suspect the comments about the TATE DES occasional "spittiness" are the results of the decode vectors suddenly jumping into the corner locations, which is something that never happens with QS Vario-Matrix or Shadow Vector. Neither of these has any optimization for corner locations, and no tendency to jump into them. One difference between Vario-Matrix and Shadow Vector is that SV is much faster, in the 1~3 mSec attack, and 20~50 mSec release, range. The last time I looked at a Vario-Matrix, it used opto-couplers for the VCA's, which are much slower in their action, and also not very accurate in dynamic tracking (the logic can't command a 75% VCA response .. you get an approximate value that is time and level-dependent).

The most brutal tests for dynamic decoders are CBS-style multimiked classical recordings, with lots of spillover on the many microphones they used, and of course the engineer is always "spotlighting" the various instruments they think are important. The result is a very unstable mix with a lot of random phase content that is bouncing all over the place as the engineer keeps twiddling with the mike gains during the session. The best the decoder can do when confronted with this onslaught of chaotic input is to briefly emphasis the transients, and gently fade back to no enhancement much of the time ... basically, a dynamic highpass filter if the logic suspects it's a multimiked classical recording. I never got to this stage of control intelligence, because this was in the all-analog days, and twiddling with the dynamic response of the control logic was something I was rather careful about. These days, it's just a few lines of code (easy for me to say, I'm not much of a programmer).
 
I remember I sent $75 deposit to get the Tate decoder from Audionics
They sent me details of the Shadow Vector decoder they were working on
Then it went away
Then sent me a couple of letters saying if you wanted your money back
They would send it
I held on for a couple of years and I received it for $300
 
Hi Lynn,
Thanks so very much for allowing the name 'Shadow Vector' to be placed on this decoder, indeed its been long overdue. It took some time to determine the best way to implement the method in software since I had decided from the outset that it must be done without approximations in a mathematically correct way. To this end, the most valuable part of the patent was the modification diagrams. To me, a visual representation is by far the most effective way of imparting a concept and those diagrams were key to the success of the software development. Its hard to consider that some methods of enhanced decoding didn't make use of Scheiber notation and no doubt suffered for it. One of the additions in my decoder possibly not in the patent was overall gain correction applied when the system moves towards a standard decode in response to a complex incoherent source (quite common on some albums). Another difference is a symmetrical attack/release, which is possible due to the use of look ahead. The look ahead allows a more conservative limiting to the vector rate of change, which eliminates low frequency modulation artifacts. Lynn's analogy with colour television and the chroma sub-carrier is so very real. The issue of the black and white video component being mistaken for colour information can be compared with to close audio frequencies being fed into a quad matrix encoder which creates a position 'spin' around the phase axis' in quad decoders. This rotation goes largely unnoticed in basic matrix decoding but can become a major issue with fast separation enchantment. I've found a vast difference in quality between SQ recordings, no doubt due to many reasons, but the shadow vector method always sounds pleasant, filling the room with a superbly immersive image.
 
The digital lookahead allows a more pleasant choice of options for the control logic ballistics ... the logic can swing the matrix parameters around before the sound arrives at the matrix itself, so no pops or bumps in the suppressed channels as the logic activates in different directions. This is similar to analog color television in having the chroma advanced (at the encode end) to allow for the delay in the receiver chroma section, so the chroma and luminance arrive at the display at the same time. When you see color streaking in old videotapes, this was not done, and is pretty obvious if you know what to look for.

In an ideal world, we'd see AVR receivers with "Involve" and "Shadow Vector" decode options, but sadly, the big Japanese companies only want to do business with Dolby Labs and THX. The technically superior "Dirac Live" is only on AVR's made by Cambridge and NAD and some very expensive 16-channel units for professional use. However ... the more units that have the Involve or Shadow Vector algorithms, and the more people hear them, well, it's always possible they might end up in the mainstream.

To continue the metaphor, Involve and SV are both swimming upstream against the current, since most AV reviewers are in love with 10, 12, and 14-speaker systems, the more the better, apparently. On the other side of the culture war, audiophile reviewers despise AV systems (with good reason, thanks to wretched sonics) and insist on 2-channel playback only. In other words, the things we enjoy here are out of the mainstream for both AV and audiophile audio.

But we're not going away ... and unlike niche products like SACD and DVD-A, the sound of a musically excellent surround system is obvious to ANY listener. What really draws a line under that experience is hearing what a well-known stereo recording sounds like with a first-class decoder, one that's far better than DPL-II or DTS Neo:6. And ... we're not pushing a 12-speaker Dolby Atmos system, which is a truly niche market that only makes sense for people who have very expensive professional installation and dedicated movie-only theatre rooms. We're talking about existing 5.1 systems sounding better on recordings we all know and love.

This point needs to be amplified. There are probably tens of millions of perfectly good 5.1 systems, using AVR's of varying quality. However .... the room auto-EQ is so poor it doesn't sound good on music, and to be brutally frank, DPL-II (music) is pretty bad surround sound, with only a hint of ambient impression, and vague localization at best. Not surprising all the hoo-ha is about discrete sound mixes on modern media, but let's be honest here, there aren't a lot of recordings in these formats. Probably the most common are a slew of multichannel SACD's, but the problem there is the sound is trapped in the player with very few exceptions (there are a few that support DSD over HDMI), and that in turns requires the nuisance of using the 6 analog inputs of the AVR .... if the AVR even has those inputs in this era of creeping HDMI replacing all the other inputs.

We have two sets of purists in completely opposing camps. The 2-channel crowd (that I've been with for the last 25 years) only consider multichannel suitable for low-end home theater use, due to mediocre sound from the AVR's, and mediocre sound from the Center speaker and the surround speakers. The A/V reviewers mostly care about the sound effects of movies based on comic books, not musical fidelity per se, and not particularly spatial realism, except for panning effects. This may sound harsh, but how many A/V reviews have you read that didn't mention the SFX of a comic-book movie?

And there's the handful of people who still love quad, and the newer folks who are enthused by the potential of surround sound for music ... and there are concerts where this is the big feature. This is the market that can be grown, particularly if the interface is to digital sources ... streaming music from TIDAL and Qobuz, FLAC files on one's own hard disc, and of course the surround recordings on physical media. If this is available at the flip of a switch ... and after all, we're just talking about one algorithm or another on a dedicated DSP system ... there's a lot of room for growth. DPL-II (music) and DTS Neo:6 are very old and anyone that's listened to a modern decoder knows there's a tremendous amount of enjoyment from conventional stereo recordings ... and this includes the very earliest stereo recordings going back to the mid-fifties. All of these sound pretty impressive when decoded as QS or EV4. That's a far larger catalog than the few hundred, or maybe a few thousand, official "quad" recordings. That's where the real potential is, although turning people on to something recorded in 1970, like the Beach Boys "Surf's Up", which is EV4 encoded, is a lot of fun.
 
Last edited:
The digital lookahead allows a more pleasant choice of options for the control logic ballistics ... the logic can swing the matrix parameters around before the sound arrives at the matrix itself, so no pops or bumps in the suppressed channels as the logic activates in different directions. This is similar to analog color television in having the chroma advanced (at the encode end) to allow for the delay in the receiver chroma section, so the chroma and luminance arrive at the display at the same time. When you see color streaking in old videotapes, this was not done, and is pretty obvious if you know what to look for.

In an ideal world, we'd see AVR receivers with "Involve" and "Shadow Vector" decode options, but sadly, the big Japanese companies only want to do business with Dolby Labs and THX. The technically superior "Dirac Live" is only on AVR's made by Cambridge and NAD and some very expensive 16-channel units for professional use. However ... the more units that have the Involve or Shadow Vector algorithms, and the more people hear them, well, it's always possible they might end up in the mainstream.

To continue the metaphor, Involve and SV are both swimming upstream against the current, since most AV reviewers are in love with 10, 12, and 14-speaker systems, the more the better, apparently. On the other side of the culture war, audiophile reviewers despise AV systems (with good reason, thanks to wretched sonics) and insist on 2-channel playback only. In other words, the things we enjoy here are out of the mainstream for both AV and audiophile audio.

But we're not going away ... and unlike niche products like SACD and DVD-A, the sound of a musically excellent surround system is obvious to ANY listener. What really draws a line under that experience is hearing what a well-known stereo recording sounds like with a first-class decoder, one that's far better than DPL-II or DTS Neo:6. And ... we're not pushing a 12-speaker Dolby Atmos system, which is a truly niche market that only makes sense for people who have very expensive professional installation and dedicated movie-only theatre rooms. We're talking about existing 5.1 systems sounding better on recordings we all know and love.

This point needs to be amplified. There are probably tens of millions of perfectly good 5.1 systems, using AVR's of varying quality. However .... the room auto-EQ is so poor it doesn't sound good on music, and to be brutally frank, DPL-II (music) is pretty bad surround sound, with only a hint of ambient impression, and vague localization at best. Not surprising all the hoo-ha is about discrete sound mixes on modern media, but let's be honest here, there aren't a lot of recordings in these formats. Probably the most common are a slew of multichannel SACD's, but the problem there is the sound is trapped in the player with very few exceptions (there are a few that support DSD over HDMI), and that in turns requires the nuisance of using the 6 analog inputs of the AVR .... if the AVR even has those inputs in this era of creeping HDMI replacing all the other inputs.

We have two sets of purists in completely opposing camps. The 2-channel crowd (that I've been with for the last 25 years) only consider multichannel suitable for low-end home theater use, due to mediocre sound from the AVR's, and mediocre sound from the Center speaker and the surround speakers. The A/V reviewers mostly care about the sound effects of movies based on comic books, not musical fidelity per se, and not particularly spatial realism, except for panning effects. This may sound harsh, but how many A/V reviews have you read that didn't mention the SFX of a comic-book movie?

And there's the handful of people who still love quad, and the newer folks who are enthused by the potential of surround sound for music ... and there are concerts where this is the big feature. This is the market that can be grown, particularly if the interface is to digital sources ... streaming music from TIDAL and Qobuz, FLAC files on one's own hard disc, and of course the surround recordings on physical media. If this is available at the flip of a switch ... and after all, we're just talking about one algorithm or another on a dedicated DSP system ... there's a lot of room for growth. DPL-II (music) and DTS Neo:6 are very old and anyone that's listened to a modern decoder knows there's a tremendous amount of enjoyment from conventional stereo recordings ... and this includes the very earliest stereo recordings going back to the mid-fifties. All of these sound pretty impressive when decoded as QS or EV4. That's a far larger catalog than the few hundred, or maybe a few thousand, official "quad" recordings. That's where the real potential is, although turning people on to something recorded in 1970, like the Beach Boys "Surf's Up", which is EV4 encoded, is a lot of fun.

Firstly to Malcolm congratulations on a great project in progress. I check this thread specifically everyday to make sure I don't miss anything. Please continue to keep everyone updated.

Secondly since Mr Olson is a significant contributor to this thread I thought this might be a good place to post this link. It's the kind of thing that maybe everybody on this forum allready knows about except me. But it's totally new to me so I would be negligent not to post it. A New York Public Radio interview circa 1974 interviewing Lynn Olson about the Shadow Vector Decoder
 
Wow, that was a blast from the past! How strange to hear myself as the idealistic 24-year-old that I was. Little did I know that we'd demonstrate the prototype (an ungainly black box with a hand-wired backplane and 10 circuit boards) to the BBC and EMI Records next year, and after getting back to Portland, being told the entire Shadow Vector project was being cancelled in favor of TATE DES. Not only that, if I wanted to keep my job, I'd have to complete a giant 4-driver speaker system that was abandoned by the original designer (he left town without a forwarding address). Charlie's reasoning was that I had spent a couple of hours chatting with Laurie Fincham at KEF Loudspeakers, and that's all anyone should need to learn about speaker design. Either finish the transmission-line behemoth or hit the streets, Jack.

So I forgot about Shadow Vector (I have no idea where the hand-wired prototype ended up) and spent six months getting the big turkey (called TLM-200) up and running. Very complex crossover because the KEF drivers were all over the place response-wise (many peaks and dips), and I made a resolution after that nightmare to NEVER finish somebody else's speaker ... I'd rather make my own mistakes, thank you very much. The speakers that followed (which were more successful) were the Audionics TL30 (a simpler 2-way transmission line with improved crossover), the M32 monitor (another 2-way in bookshelf format), the T52 (a vented-box 3-way with a Bessel alignment in the bass section), and the LO-2 linear-phase, low-diffraction sub/sat combination. Those were my last speakers at Audionics. The Ariel was designed much later, in 1992, as a project for a friend with stacked Quad ESL57's, and that became an article for Positive Feedback magazine for folks who owned triode amplifiers. The as-yet-unnamed "Beyond the Ariel" high-efficiency project in the diyAudio speaker forum has been built by a friend of mine in Washington State, Gary Dahl, and he's very pleased with it, the last I heard.

Although the sudden end of Shadow Vector was kind of a bitter experience at the time, I've been reflecting on it over the last few days. Even though I was getting not much more than minimum wage, that was still for two years, and Cliff Moulton, Audionics chief engineer, spent about 1/4 to 1/3 of his free time on building the prototype. When you add the cost of the patent, the whole project probably cost Audionics $30,000 to $40,000, a big investment for a little company with 12~15 employees. I don't think Audionics actually turned a profit until Bob Sickler came along and designed the very successful CC-2 amplifier (Audionics sold more than 2,000 of them, and they were very reliable). The main reason I left in 1979 for the Tech Writer job at Tektronix was the pay was 3 times what I made at Audionics, and frankly, I was burned out on audio. I made a resolution to never, ever, work as an employee for anyone else in the audio business. Consultant, of course, but not as an employee. Too little pay, and too much deadline pressure.

After that, I never had any contact with Charles Wood (president of Audionics) or Jim Fosgate. I only heard about the Space & Time Composer, and the many problems getting the National Semiconductor TATE DES chipset to work, though friends of friends of the people still at Audionics. I did have a minor contact with Wesley Ruggles, who along with Martin Willcocks, was one of the co-inventors of the TATE DES system. I spent a day at Venice Beach (a posh hipster suburb of Los Angeles) advising Mr. Ruggles how to convert a professional Sansui Vario-Matrix decoder to a strange new matrix with a peculiar diamond pattern that we now call Left, Center, Right, and Surround. The adapter circuit was a simple re-matrix with a few op-amps and some associated matrix math, and I got paid airfare and $500 for my time. It wasn't until I heard my first movie in Dolby Surround that I discovered where my little re-matrix diagram for the modified professional Vario-Matrix ended up.

Don't know why, but Charlie and Wesley must have had their reasons for keeping me at arm's length from the TATE project or any follw-on contact with Jim Fosgate in the years that followed. I was at Tektronix then, and didn't think much about audio in the early and mid Eighties. It wasn't until the big vacuum-tube revival in the late Eighties and early Nineties that I got back in again.
 
Last edited:
Sorry if it got a little emotional in the previous post. This whole Shadow Vector episode has brought up a lot of memories for me ... I considered deleting the previous post, but on reflection, it should stand as-is. This is what happened, both good and bad, and the experience of hearing myself as that starry-eyed 24-year-old brought back a lot of memories and emotions. So much water under the bridge. (I had completely forgotten I had given that interview.)

My casual comments about "psychoacoustics" and "more separation than discrete" was based on my readings of the Gerzon Ambisonics system, which I'd read about in the AES Journal, but hadn't heard for myself. Gerzon's main point in his multiple AES articles was that pairwise mixing was sub-optimal for 4-channel playback ... and I had to agree there. The side images are vague and diffuse, the image across the rear is hardly an image at all, and Center Back is not only unpleasant and a little spooky-sounding, it can jump into Center Front quite easily. All true, and all fairly noticeable with a "discrete" 4-channel mix from tape or CD-4 LP playback. Gerzon claimed that carefully chosen amounts of antiphase crosstalk in the adjacent speakers (or was it opposite? I don't remember) would firm up the soundfield and make it more realistic. But ... I hadn't heard Ambisonics yet, but the underlying principles seemed good to me.

I was a little bit disappointed when I heard UHJ for the first time at the BBC a year later. The ambience was indeed smooth and very realistic and symmetric (good design!), but the images coming from the static UHJ decoder were on the blurry side compared to what I was used to ... and it also seemed sensitive to head movements. Head rotation was no problem, and indeed better than stereo or discrete pairwise mixing. But left-to-right motions seemed a little "swimmy" compared to discrete or Shadow Vector, so I was left with mixed feelings. In other words, and I think I said this to the BBC engineers at the time, I was curious what UHJ would sound like with a good dynamic decoder. Since UHJ is pretty close to both QS and EV4, I'd expect Vario-Matrix to sound pretty similar, just with more favorable phase angles between the 4 speakers for the UHJ system. (AFIAK, there's a 90-degree phase angle across the LF/LB pair and the RF/RB pair, while the LF/RF quadrant is at 0 degrees, while the LB/RB is also at 0 degrees. This makes the QS sidewalls a bit phasey sounding compared to the UHJ phase angles. Minor point, but it is audible.)

Listening to the interview, I think I was a bit evasive or maybe even misleading about hemispheric sound. There is NO possible localization in SQ for all 4 speakers at once. You get two opposing diagonal splits, if you want such a weird thing, but there is no encoding for 4-speaker sum. This rules out overhead sounds, which is straightforward with QS or UHJ, which are basically "square" encoding layouts, with the top portion of the Scheiber sphere available for overhead pans.

Since I didn't want to offend CBS (for which Audionics was a licensee), I didn't mention that QS could do some things that SQ couldn't (like overhead sounds), and the off-the-shelf QS Vario-Matrix was quite a good decoder (more pleasant than the CBS logic decoders). On the other hand, QS has pretty poor stereo compatibility, with only 8 dB separation across the front when played in 2-speaker stereo. This would not sound good on most stereo systems. I intentionally avoided these topics, although I certainly knew about them at the time of the interview.

I also didn't mention that the one existing Shadow Vector prototype did not have QS decoding. The next-generation commercial version would definitely have had the internal switching (probably with relays) to switch between SQ and QS, and the planned variable blend control would easily sweep between a 90-degree stage and a 270-degree stage (oddly named "dimension" in other decoders). The mention of a $1400 price tag really surprised me. I guess Charley was telling me the Shadow Vector was really going to be an Audionics product with a real feature set and a planned front panel with 4 big meters and some rotary mode switches.

I'm still in agreement with most of what was said in the interview. A really excellent decoder does wonders for a decently-mixed stereo recording, since all the decoder is really doing is mapping the phase angles of the 2-channel source into the Z axis of the listening room. Nothing is made up; you can just hear the phase angles directly, instead of a diffuse impression spread between the speakers (90 degree angle) or a ghostly extra-width impression beyond the speakers (180 degree angle). Old-school 3 spaced microphone recordings (from the Fifties) are fun too, with a crisp center localization, and a swimming-pool reverb from the widely spaced L and R microphones. The wide spacing generates random-phase between the channels, which the decoder cheerfully spreads across all four speakers. (This might be the one way SQ can provide a 4-speaker spread.)

Decoders also reveal which recordings are multimiked mono, with no real stereo anywhere (except reverb). The instrument localizations are trapped across the frontal arc, and all that comes from the other 270 degrees is a dull roar of reverberation, but no actual instruments. If the recording has even one stereo microphone, or a spaced pair, then the soundstage opens up to the whole room. The recording technique is laid bare, which for better or worse, is a good thing. I think most people would enjoy hearing the whole recording.
 
Last edited:
Sorry if it got a little emotional in the previous post. This whole Shadow Vector episode has brought up a lot of memories for me ... I considered deleting the previous post, but on reflection, it should stand as-is. This is what happened, both good and bad, and the experience of hearing myself as that starry-eyed 24-year-old brought back a lot of memories and emotions. So much water under the bridge. (I had completely forgotten I had given that interview.)

I'm glad you didn't delete that post Lynn as it gives considerable insight into the pressures you must have been under at that time, and given those pressures and your obvious dedication to the art, your emotions are fully understandable.

If you'd forgotten about that interview - had you also forgotten about this one from a few weeks later where you discuss 4 ch psychoacoustics?
https://www.wnyc.org/story/psychoacoustics-4-channelIn this particular interview, I was interested to hear what you had to say about having the opportunity to see the BBC's evaluation set up - they probably did more independent assessment of quad than any other organisation and it must have been fascinating.

I was also intrigued by your talk of the prospect of an 'add on box' to precede the SQ Shadow Matrix decoder to give it the capability to decode QS and to provide stereo - quad synthesis which would surely have been a unique alternative approach to the relay switching of modes within the decoder itself you mention above.
 
I'm glad you didn't delete that post Lynn as it gives considerable insight into the pressures you must have been under at that time, and given those pressures and your obvious dedication to the art, your emotions are fully understandable.

If you'd forgotten about that interview - had you also forgotten about this one from a few weeks later where you discuss 4 ch psychoacoustics?
https://www.wnyc.org/story/psychoacoustics-4-channelIn this particular interview, I was interested to hear what you had to say about having the opportunity to see the BBC's evaluation set up - they probably did more independent assessment of quad than any other organisation and it must have been fascinating.

I was also intrigued by your talk of the prospect of an 'add on box' to precede the SQ Shadow Matrix decoder to give it the capability to decode QS and to provide stereo - quad synthesis which would surely have been a unique alternative approach to the relay switching of modes within the decoder itself you mention above.

Wowsers! Another great find! That was a good listen. Like you I am interested in what sort of add on box could be used to decode QS and other formats. In Scheiber's article Transformations on the Energy Sphere he shows how easy it is to switch between QS & BBC Matrix H. But I am left with the impression that it can not be done external to a decoder for QS>SQ only internally similar to Sansui's approach.
Lynn?
 
Sorry if it got a little emotional in the previous post. This whole Shadow Vector episode has brought up a lot of memories for me ... I considered deleting the previous post, but on reflection, it should stand as-is. This is what happened, both good and bad, and the experience of hearing myself as that starry-eyed 24-year-old brought back a lot of memories and emotions. So much water under the bridge. (I had completely forgotten I had given that interview.)

My casual comments about "psychoacoustics" and "more separation than discrete" was based on my readings of the Gerzon Ambisonics system, which I'd read about in the AES Journal, but hadn't heard for myself. Gerzon's main point in his multiple AES articles was that pairwise mixing was sub-optimal for 4-channel playback ... and I had to agree there. The side images are vague and diffuse, the image across the rear is hardly an image at all, and Center Back is not only unpleasant and a little spooky-sounding, it can jump into Center Front quite easily. All true, and all fairly noticeable with a "discrete" 4-channel mix from tape or CD-4 LP playback. Gerzon claimed that carefully chosen amounts of antiphase crosstalk in the adjacent speakers (or was it opposite? I don't remember) would firm up the soundfield and make it more realistic. But ... I hadn't heard Ambisonics yet, but the underlying principles seemed good to me.

I was a little bit disappointed when I heard UHJ for the first time at the BBC a year later. The ambience was indeed smooth and very realistic and symmetric (good design!), but the images coming from the static UHJ decoder were on the blurry side compared to what I was used to ... and it also seemed sensitive to head movements. Head rotation was no problem, and indeed better than stereo or discrete pairwise mixing. But left-to-right motions seemed a little "swimmy" compared to discrete or Shadow Vector, so I was left with mixed feelings. In other words, and I think I said this to the BBC engineers at the time, I was curious what UHJ would sound like with a good dynamic decoder. Since UHJ is pretty close to both QS and EV4, I'd expect Vario-Matrix to sound pretty similar, just with more favorable phase angles between the 4 speakers for the UHJ system. (AFIAK, there's a 90-degree phase angle across the LF/LB pair and the RF/RB pair, while the LF/RF quadrant is at 0 degrees, while the LB/RB is also at 0 degrees. This makes the QS sidewalls a bit phasey sounding compared to the UHJ phase angles. Minor point, but it is audible.)

Listening to the interview, I think I was a bit evasive or maybe even misleading about hemispheric sound. There is NO possible localization in SQ for all 4 speakers at once. You get two opposing diagonal splits, if you want such a weird thing, but there is no encoding for 4-speaker sum. This rules out overhead sounds, which is straightforward with QS or UHJ, which are basically "square" encoding layouts, with the top portion of the Scheiber sphere available for overhead pans.

Since I didn't want to offend CBS (for which Audionics was a licensee), I didn't mention that QS could do some things that SQ couldn't (like overhead sounds), and the off-the-shelf QS Vario-Matrix was quite a good decoder (more pleasant than the CBS logic decoders). On the other hand, QS has pretty poor stereo compatibility, with only 8 dB separation across the front when played in 2-speaker stereo. This would not sound good on most stereo systems. I intentionally avoided these topics, although I certainly knew about them at the time of the interview.

I also didn't mention that the one existing Shadow Vector prototype did not have QS decoding. The next-generation commercial version would definitely have had the internal switching (probably with relays) to switch between SQ and QS, and the planned variable blend control would easily sweep between a 90-degree stage and a 270-degree stage (oddly named "dimension" in other decoders). The mention of a $1400 price tag really surprised me. I guess Charley was telling me the Shadow Vector was really going to be an Audionics product with a real feature set and a planned front panel with 4 big meters and some rotary mode switches.

I'm still in agreement with most of what was said in the interview. A really excellent decoder does wonders for a decently-mixed stereo recording, since all the decoder is really doing is mapping the phase angles of the 2-channel source into the Z axis of the listening room. Nothing is made up; you can just hear the phase angles directly, instead of a diffuse impression spread between the speakers (90 degree angle) or a ghostly extra-width impression beyond the speakers (180 degree angle). Old-school 3 spaced microphone recordings (from the Fifties) are fun too, with a crisp center localization, and a swimming-pool reverb from the widely spaced L and R microphones. The wide spacing generates random-phase between the channels, which the decoder cheerfully spreads across all four speakers. (This might be the one way SQ can provide a 4-speaker spread.)

Decoders also reveal which recordings are multimiked mono, with no real stereo anywhere (except reverb). The instrument localizations are trapped across the frontal arc, and all that comes from the other 270 degrees is a dull roar of reverberation, but no actual instruments. If the recording has even one stereo microphone, or a spaced pair, then the soundstage opens up to the whole room. The recording technique is laid bare, which for better or worse, is a good thing. I think most people would enjoy hearing the whole recording.

Good reading both posts, thanks!
If you were 24 in 1974 then that puts you only 1 year older than me. And when I was 24 it really was all about ***, drugs, rock & roll. Meanwhile you were creating the Shadow Vector Decoder. Which BTW is the coolest decoder name ever.

When the plug was pulled on your project at Audionics, you still had control over the patents, right? Did try or even consider going elsewhere for a business partnership?
 
No, the patents were assigned to Audionics, not me. Although it sounds appalling, this is the normal course in the tech industry, going back to the days of Edison's lab. In return for sinking time and development money into the invention (which remains a paper patent until the prototype is built), the inventor assigns the rights to the corporation that sponsored it. Getting a patent does not require building a prototype; that's a myth. But ... a physical, working prototype is necessary for assessing proof-of-concept and a rough sense of what it will cost to put in to production, or if it's even possible to put in production. Many inventions, probably most, cannot be put into mass production, so they never see the light of day.

The Shadow Vector (yes, I came up with the name, not Audionics) was a pretty complicated prototype. As I recall, it had 10 circuit boards and a hand-wired backplane they plugged in to.

*There was one board for input buffering and the Axial Tilt Corrector circuit for optimizing phono-cartridge separation. (The magnetic coils of phono cartridges are not exactly at 90 degrees to each other, the phono cartridge is often mounted with a slight rotation in the headshell, and the stylus itself is often twisted by 2 to 3 degrees. This circuit electrically corrected for rotation of each channel ... just play a test record and twist the knobs, set and forget. I also discovered that thanks to high-order modes in the stylus cantilever, the correction at 10 kHz was not at all the same as at 1 kHz or lower. This was one of the subtle defects of the CD-4 system, which relied on perfect performance and full channel isolation in the 20~40 kHz region. A future version of the Shadow Vector was going to offer dual-band correction, at 1 kHz and 10 kHz. For tape playback, we just centered the knobs, which was good enough. In the production version, there was going to be a bypass switch for the ATC section.)

* There were 2 boards for the 6-pole allpass phase shifters, which gave 0, +90, 180, and -90 degree outputs for each incoming channel. These 8 signals can provide any decoding vector when summed by the VCAs. Both boards were identical.

* There were 3 direction-sensing (logic) boards for each axis of the Scheiber sphere (CF/CB, LF/RF, and LB/RB). Each sensor produced a +/- control signal with about 1~3 mSec attack and 20~50 mSec release times. The audio input was filtered between 500 Hz and 8 kHz, with filter shaping following the inverse of the Fletcher-Munson curve, with peak sensitivity around 3 kHz. There was a fast-acting AGC circuit that gave a 40 dB dynamic range for the direction sensors.

* There were 4 boards for each decoded channel (LF, RF, LB, RB), each with 3 VCA's controlled by the +/- control lines (this was the "business end" of the decoder that actually did the dynamic decoding of SQ content). Each board is the same, just with different connections to the 8 incoming audio signals and the 3 control lines.

That's 10 circuit boards, along with a 4-gang volume control on the front panel. The four meters and SQ/QS mode-switching was intended for the production version. The box was completely plain, not beautiful like the CBS Paramatrix, which was a work of art in comparison. (Then again, CBS Labs had an in-house metal shop for making pretty boxes for pro use. We didn't.)

Differences between the TATE DES and the Shadow Vector: I don't have a good understanding of the TATE system, almost no communication with Wesley Ruggles, Martin Willcocks, or Jim Fosgate, so I'll do my best based on a limited understanding of the TATE patent. My guess is that Chuck probably knows more (probably much more) about the TATE system than I do. Anyway, the TATE system has a static SQ decoder at the heart of it, and then adds various VCA signals to the four outputs of the static decoder to make it do what it's supposed to do. Depending on what's fed to the outputs of the static SQ decoder, this could be anything from gain-riding to full dynamic decoding, or anything in between, or maybe something completely different. It's not clear to me what it does, because it is buried under a layer of matrix math I have not unravelled.

The Shadow Vector has no static matrix anywhere in it. When all 3 control lines are sitting at zero, with no modulation, the 3 VCA's on each channel board behave like a static SQ matrix ... it's their default, zero-modulation state. The control signals are linear, with no transitions around zero or approaching full corner separation, so separation is maintained in all directions, including intermediate localizations. There is no "cancellation" per se, unlike TATE DES, just dynamic decoding vectors. In that respect, it's similar to QS Vario-Matrix, just in 3 dimensions instead of 2.

To give you an idea of what Malcolm is doing, imagine this whole thing in software, except with multiband (done several times) and a lookahead feature that delays the audio going to the virtual VCAs so the control signal can swing the vectors around before the audio directional shift arrives. In the physical Shadow Vector, the control signals arrive just a little late, by about 1~3 milliseconds (no digital delay back then).
 
Last edited:
I imagine a lot of this discussion is going over the heads of most readers, although I'm sure that Chuck and Malcolm are enjoying every minute of it. But I think that a review of how matrix, and SQ, systems work is probably a good idea, so here goes.

SQ was invented by CBS Labs to have good compatibility with stereo records and stereo playback systems. This is probably SQ's greatest strength, since LF and RF in the SQ system maps to good old Left and Right on the stereo disc and FM broadcast. So, full separation in the front, not 20 dB like EV4, or worse, 7.7 dB in QS.

Where it gets a little strange is the rear-channel encoding, which requires 90-degree allpass phase shifters (these are required in the front, as well, and also needed for both encoding and decoding). Left Back is encoded as L at 0.707 modulation, 0 degrees phase shift, and R at 0.707 modulation and -90 degrees phase shift. Right Back is encoded at R at 0.707 modulation and 0 degrees phase shift, and L at 0.707 modulation and -90 degrees phase shift.

CBS had already invented a gizmo for broadcasters that gave a quasi-stereo effect by shifting the phase 90 degrees between the channels, so they were already using phase-shifters for other uses. A 90-degree allpass phase shift is also useful for AM broadcasting, since it spreads out and reduces the size of peaks going to the AM transmitter (which lets the broadcaster raise average power without exceeding FCC peak modulation limits). Broadcasting power sets the coverage map and the size of the listening audience, which in turn drives advertising revenue. Anything that increases revenue with nothing more than a signal conditioner is very welcome in the broadcast biz.

CBS made the SQ marketing materials in the record sleeve bizarre and opaque (an old JBL trick) by calling the rear channels "spiral" or "cylindrical" on the LP groove. This is true, but a pretty opaque description. In reality, it's a signal with a 90-degree shift between L and R. LB has the shift going in one direction, and RB has the shift going in the other direction. That's all there is to it. A signal like this is also called "quadrature modulation" and was routinely used in analog color TV. (I and Q in NTSC color, and U and V in PAL color. SECAM used 2 FM carriers in sequence instead of quadrature modulation.)

So far, so good. The SQ decoder is where the weirdness starts. Static decoders, which are essentially passive devices with no logic sensing, have a characteristic of the main channel playing at 0 dB level, one channel completely quiet, and the other two at -3 dB, which isn't much separation. In a QS static decoder, LF will play at 0 dB, RF at -3 dB, LB at -3 dB, and RB at zero (infinite separation for the diagonally opposite channel). In practice, this blurs the image across the 3 speakers that are playing.

The SQ static decoder is different. If there's a 0 dB signal at LF, the -3 dB channels will be LB and RB, and the silent channel will be RF. Well, that doesn't sound too bad ... just a different pattern of crosstalk. Where it gets strange is CF or CB encoding. They play back with equal energy on all four channels! In fact, there's no difference between CF and CB! Well, there is a phase difference. A CF signal will be in-phase across the front channels (just as it would be in good ol' stereo), and out-of-phase across the rear channels (with all 4 at the same level). The situation reverses for CB encoding. So there's an odd bow-tie pattern of separation in a static SQ decoder ... 3 dB at the corners, and zero dB between CF and CB. This is where we see the strange tradeoff made to retain full separation for stereo playback ... a rather peculiar and unexpected behavior for the static decoder.

The CBS engineers came up with an ingenious solution for the "half-logic" decoder. They added a CF/CB sensor (logic) circuit, and one opto-coupler between the two front channels and another opto-coupler between the rear channels. This blends together the two front channels when CB is detected, or blends together the two rear channels when CF is detected (which is frequent, like a solo singer). This is effectively a simple dynamic decoder, operating on one axis, and raises the CF/CB separation from 0 dB to 14~20 dB ... not a bad result at all.

The Full-Logic SQ decoder operates completely differently. There is 3-axis sensing, but the corner channels are simply gain-suppressed when an unwanted crosstalk product appears. The active channels have their gain raised slightly to keep the overall decoder gain at unity, but the individual channel gains go up and down, depending on sensor commands. This is how the Sony SQD-2020 "full logic" decoder operates, and I never much liked, since I could hear the gain-riding as ambience suppression and soundstage movement. To me, this pretty much defeated the entire purpose of quadraphonic sound. Yes, there's 20 dB of separation (by measurement on sine wave test signals), but the ambient impression is not well served by the gain manipulation.

Returning to QS for a moment, Sansui was able to extend the principle of matrix steering from one dimension to two, by blending together pairs of channels as commanded by a 2-axis sensor (LF/RB and RF/LB). This is the Sansui Vario-Matrix QS decoder, which is famous for lack of artifacts and smooth decoding, and later re-purposed for the earliest versions of Dolby Surround in movie theaters.

Shadow Vector extends the dimensionality into 3 dimensions of dynamic decoding, which is not surprisingly more complex than either Vario-Matrix or CBS Full Logic. But it does work, and has a sound quite different than either, as Malcolm can attest. It's also not that hard to shuffle the decoding elements so they decode QS the same way Vario-Matrix does, just faster, since it uses VCAs instead of slower and less accurate opto-couplers.

Opto-coupling is smooth and low distortion (no active electronics are involved except for the driver), but it's not that accurate or quick. If the requirement is for 0.707 modulation, the coupler will gradually move somewhere towards 70%, maybe overshoot it, or maybe fall short. There's no feedback system to make sure it hits its target, and the speed isn't all that fast. This means channel separation is kind of hit-and-miss, and transients might be too fast for the system. VCAs are precise, but getting them to also have low distortion isn't easy. And you need a lot. Shadow Vector used 12 of them, and I think TATE uses more.
 
Last edited:
Back
Top