The Legacy of Dr Bauer

QuadraphonicQuad

Help Support QuadraphonicQuad:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

Bill Brent

300 Club - QQ All-Star
Joined
Apr 13, 2004
Messages
354
First let's accept that the mind interprets what we hear (and see) in ways we might not expect. we see a white box - take a photograph of it and it looks like an orange box. We put two speakers in front of us, play a record, and hear sounds behind us (as well as centered between the speakers). The photograph's color shift happens because we do not see objects - we see light reflected by objects - the sun provides a light source outside and along the Kelvin scale (or the scale by which we measure the relative temperature of light) it is quite cool. Indoor light can be provided by a number of sources, each with their own temperature. Incandescent light is very warm, thus the orange cast. Film, or the photographic process in any form, does not adjust for the temperature of light sources - so on digital cameras you use the white balance, and with film you use filters, or film specifically created for the light source. The brain - the human brain, takes the reflected light in and instantly adjusts for the temperature of the source. This is why you can pick up that white box in your bedroom, illuminated with florescent lights, take it into the kitchen with all the light being incandescent and outside - and you always see it as white. And it remains white if the sun is shining or if its deeply overcast. - Amazing thing this brain. But what does this all have to do with Australia?

Well, just as the brain takes in visual stimuli and allows, well, itself to interpret it in wondrous, though predictable ways. It does quite the same thing with sound. How we interpret sound relies on many elements - starting with anatomy. While our ears are on the side of our heads, they are cupped in such a way as to favor sounds coming from in front of us.

There are 127 (or more) elements that impact our mind's dealings with sound. For example; sounds that come from behind us, that are reflected - that is, echoes of the sound coming from in front, reach us "out of phase". This is why we can listen to a two channel recording, that has some information not "in Phase" - and think those elements are coming from behind us - the brain is used to telling itself; “out of phase sounds come from behind”.

When I'm ready to write my "History of Quadraphonic Sound in the Modern World" - I'll detail the development of recorded sound from the tin foil sheets of Edison's cylinder to the Super Audio Compact Disc (none of those French experiments from the 1860s). There will be a chapter on the thought behind quadraphonic sound from 2 channel sources, and a chapter on each system developed to accomplish stuffing 4 signals onto a stereo LP, and reconstituting it. Until that time I'll assume you know all of that, or at least the relevant portions, and we can jump right into CBS's SQ system.

At its simplest - the front channels are untouched, the rear channels, 90 degrees out of phase, are mixed into the front, 3 decibels lower in amplitude. They sound to the stereo listener and vaguely mono, somewhere in the center, between the two speakers. Because the signals intended for the rear ARE out of phase, the brain will place them there - particularly when listening through stereo headphones. The SQ system relies on that - simple decoders reverse the "mix" - the front are untouched - the rear now has those signals meant to be in front - 90 degrees out of phase, mixed 3 decibels lower in amplitude (all this, while the sounds meant for the rear are now restored to the left and right). That 3db is a bit of an issue - it is the “threshold of perception” - this means with 3db of "separation" between front and rear - is just enough to perceive, but only just. The problem it introduces - is noise. In order to make the rear channels balanced with the front (ie; as loud as) - you need to increase the levels by 3db - when you do that, you increase the noise as well (basically lowering the signal to noise ratio by, you guessed it - 3db).

So the problem before us - and the engineers from the 70's onward - has been to extract the 4 channels back out from the "encoded" two channels.

Whether SQ was the best of the 1/2 dozen or so (matrix) encoding systems is for another day - it does not impact us here (you'll see these discussions in chapters 8 & 9 of my book).Further, we won't agonize over the evolution of the SQ decoder (that will be chapter 6). Today we look at two systems. One, long considered the state of the art, but no longer in production - the other in production, not as widely known.

The first system, the one no longer in production, but highly thought of and sought after - is the Fosgate Tate II 101a.At the heart of this was the DES (The Directional Enhancement System), also known as the Tate. DES, was an advanced decoder for SQ and enhanced the directionality of the basic SQ matrix. It first performed a straight SQ decode – and took the four outputs of the SQ decoder to derive additional signals, then compared their envelopes to detect the predominant direction and degree of dominance.

A processor section, implemented outside of the Tate IC chips, applied variable attack/decay timing to the control signals and determined the coefficients of the "B" (Blend) matrices needed to enhance the directionality. These were acted upon by true analog multipliers in the Matrix Multiplier IC's, to multiply the incoming matrix by the "B" matrices and produce outputs in which the directionality of all predominant sounds were enhanced.

Since the DES could recognize all three directions of the Energy Sphere simultaneously, and enhance the separation, it had a very open and 'discrete' sounding sound-field. While there were two consumer decoders that employed these chips; the Audionics Space & Image Composer and the Fosgate Tate II 101A, The Fosgate used a faster, updated version of the IC, called the Tate II, and additional circuitry that provided for separation enhancement around the full 360° sound-field, using the Haas effect.

In order to maintain the highest quality levels, Fosgate used hand-sorted ICs and 1% -tolerance components, and each decoder was hand-optimized. Unlike the earlier Full Wave-matching Logic decoders for SQ, that varied the output levels to enhance directionality, the Tate DES cancelled SQ signal crosstalk as a function of the predominant directionality, keeping non-dominant sounds and reverberation in their proper spatial locations at their correct level. In addition, the enhancement was done with sufficient additional complexity that all non-dominant sounds were kept at their proper levels.

All seems pretty impressive, no? In fact up until 1986 or so, it was the fosgate DES that was the core engine used in Dolby Surround. - but now I'm skipping ahead to chapter 9. The fosgate remains impressive - not only for it's ability to separate out the SQ encoded signals into 4 - but it's ability to simulate surround (or quad) from stereo sources with it's cinema setting - something very closely resembling RM/QS.

If you've read my stuff, you know I'm not about measurements - for me it's all about the sound - yet you need to start somewhere. For my tests I've created my own SQ files. I used wide spectrum pink noise (or white noise if you prefer) and then a series of single tones. The sequence was:(all 0db VU) - left front, right front, left rear, right rear, front center, rear center, left center, right center, center channel (all 4 channels at equal volume). It should be noted that the last one there - didn't amount to much. Actually - it amounted to nothing at all, you see in SQ there can be no center channel. Because of the phase shifts for a signal to exist in all for channels at equal volume - it would need to exist 180 degrees out of phase with itself - which it can not do. I took VU readings of all four channels during decode, for each signal. The Tate provided at least 24db of separation. This means that when a signal was played - it was at least at -24db in the corresponding channel(s). A 0db signal in Left Front was between -24 and -30 in the left rear. I picked -24 as this was the worst reading (-32 being the best) - different frequencies provided different levels of separation. There was no perceivable "pumping" as the decodes "kicked - in". The next was a music test - the results are purely subjective but in an attempt to be comprehensive - I used a rock track, an orchestral track, a jazz track, a vocal (all with fixed placements) and an opera track with stage movement and a novelty track with surround movement.

For straight listening - they were all very good, in a roughly square room the speakers were in each corner - with the listener 2/3 of the way toward the back of the room. I was not looking to compare the Fosgate's output to that of discrete playback of the same material. Nor was I checking speaker by speaker to judge the crosstalk between channels - simply, how good a listening experience it provided.

Next up, the unit that remains in production, created by a little company from "down-under" called Involve Audio. This device is humbly called The Surround Master. It should be pointed out that SQ decoding is not what the company, or device, is all about. They have created a system of encoding that is very similar to the Regular Matrix (RM/QS) system - and they have produced a decoder to go along with it. I've covered these encoder/decoders extensively in chapter 2-3 but suffice it to say that the surround master basic decoder does a heart stopping job with QS material - and their own encoded programs. But if you want the engineering that went into that unit applied to the sq system - it's gonna cost you extra. Further, there is no SQ encoder available from that outfit. Guess they figure, if you're gonna encode - do it with their system, and, as it's compatible with QS and Dolby PLII - it seems fair. Now - if all you want to do is decode your old SQ records, and you want the best - do you go the on-line auction route and grab yourself a Tate, or do you send some coins to the land of the walkabouts? Well now we come to the reason for this tome. I ran the SM/QS circuit through the same paces as with the tate, and the results kind-of surprised me; on two levels. First - the measurements. The SM produced results within a DB of the tate. - not enough of a difference to make more exacting measurements needed. Then the sound tests - and damn! - if the SM didn't sound better,,,,quite noticeably better. And yes, I do not put much faith in measurements vs listening - soooo there just had to be something I was missing.

I sent about 20 recordings to 18 fellow quad enthusiasts - and 15 picked the SM as producing a better experience - the other 3 said they could hear a difference, but didn't have a preference. How could this be? Well it seemed to come down to what we call the “precision of imaging” and smears (plus a fair bit of cross bandwidth pumping at a micro level). The Tate remains a single band unit. This means for example that low frequency information (with no real directionality) can dominate the math's of the resultant decode and you can get conflict situations where the very high frequencies want one direction yet the midrange wants another direction and the bass – well it just sits there like the .1 it was meant to be, but hampering the decode of the upper frequencies.

This creates poor imaging and smear. The SM is a true tri-band circuit and thus is, in effect, 3 parallel decoders working independently. Involve also spent a lot of time massaging the attack and decay time constants for each band to ensure instruments were picked up as a whole and not smeared elsewhere. So in effect - they made the SM "listen" to music in the same way the ear/brain does - adjusting all mathematics for the logarithmic nature of hearing and allowing for the Fletcher Munsen sensitivity (frequency curve).

This ensures that the direction of steering is not dominated disproportionately by signals that may be higher in magnitude but lower in apparent level. For example in pink noise and music a 100 Hz noise has a way higher magnitude than a 3 kHz tone yet it is the 3 kHz tone that will greatly dominate.

The complexity to do all this is significant and is not easily accomplished with conventional analogue circuitry (due to tolerance drift in components etc). Of course Involve did cheat - some of the software, and circuit designs simply didn't exist in 1979.

At the end of day, it's difficult to find fault with the SM unit. If you already have a Tate would I suggest the SM be your next purchase - no - especially as there is no new SQ material being produced, and the Fosgate/Tate II is really beyond reproach (it even has a joystick type remote ---- it's a guy thing) - but if you’re just getting into surround and are picking up some old vinyl - the surround master with the sq option - well - that's the way to go. If that box had existed in 1972, quad would never have died out - but then the queen would be king if...well, you know the rest.

<note: all spelling errors due to line noise>
 
One key advantage of any digital implementation is "lookahead", which isn't practical with 1970's analog technology.

It takes a finite amount of time for the directional-detection system to work ... usually one or more wavelengths. At 1 kHz, this is obviously 1 mS, and in practice, the dynamic decoders typically have 2 to 3 mS attack times, with release times 5 to 10 times longer. This means the suppressed channels are always a little late, again by 2 to 3 mS, and very brief artifacts may appear. It gets much more complicated with multi-band decoding, because the attack time for each section has to be different ... slow for the LF section, and fast for the HF section.

Digital decoding solves all these problems at once ... basically, everything gets dumped in a buffer that's deep enough for the slowest portion of the decoder, and the main audio signal is then delayed so the matrix-shifting happens *before* the signal transient arrives. Thus, no artifacts, and the direction-sensing section has an effectively unlimited amount of time to decide where the signal should be coming from. The amount of DSP is trivial by modern standards ... while by comparison, low-distortion variable-gain analog amplifiers are very expensive studio-grade parts usually found in mixing desks. So it all gets much easier in the digital domain, along with any desired amount of DSP signal enhancement (which can very subtle if desired).

The biggest problem is outwitting dodgy multichannel mixes with spaced mikes with a fair amount of acoustic crosstalk between mikes. This creates random phase information in the final mix that has no effective directional content at all, and the decoder has to be smart enough to know when this is happening ... DSP spectral analysis of the control signals is where this comes in useful. (The DSP can be trained to enhance transients, treat steady-state signals a little differently, and ignore random-phase information.)
 
a digital, or software based decoder seems like a great idea - at least on paper. I haven't come accross one that has an end result approaching the Tate - no less this little SM unit. I'd like to see one, but would be much more interested in a software method of demodulating cd-4 - where accomodations can be made for sub-carrier, and analog artifacts in general.
 
a digital, or software based decoder seems like a great idea - at least on paper. I haven't come accross one that has an end result approaching the Tate - no less this little SM unit. I'd like to see one, but would be much more interested in a software method of demodulating cd-4 - where accomodations can be made for sub-carrier, and analog artifacts in general.

:censored::SG:phones
 
Hmm ... in principle, a 96/24 PCM non-RIAA capture of the output of a wideband modern moving-coil cartridge, and a ton of DSP, could re-create the JVC CD-4 decoder. I suspect the fidelity would be low by modern standards, since the distortion on the two recovered FM carriers is partly baked into the recording itself, and unfolding that FM distortion would be a difficult-to-model high-order function.

A software SQ/RM/QS decoder is easy in comparison. Malcolm is doing one in the UK, with no less than 8 bands, the Surround Master is up and running and a commercial product, and best of all, a digital decoder *does not* add distortion, which is more than can be said of an analog SQ/RM/QS decoder or a CD-4 FM demodulator.

Does that mean that some generous person is going to offer the SQ/RM/QS code for free? Probably not. It took me two years of hard work to build the Shadow Vector prototype, Malcolm has spent an equivalent amount of time on his digital version, and from I've heard of the development of the Surround Master, the coder nearly lost his mind re-creating the TATE DES in software. These gizmos are pretty complicated, at least conceptually. And Dolby and DTS "upmixing" technology is a closely held secret, protected by multiple layers of lawyers and code baked into a proprietary chipset that is only sold to Dolby and DTS licensees.

I only heard about the development of the TATE/Audionics product through a series of rumors at the time. The National Semiconductor TATE chipset was partially inoperative, and National had zero interest in re-doing them correctly, so Audionics was compelled to develop what they called a "band-aid" circuit to get a functioning decoder. That added years of development and additional cost to the project. I never knew exactly what the external "band-aid" circuit did ... was it the variable-gain amplifiers, the direction-sensing, internal control logic, or what.
 
Fascinating, not just the effort that is being put in, but that it's been put in at all. All in and for a technology that the mainstream has so little appreciation for.
 
I tried the PSpatialAudio (StereoLab) software to decode a CD-4 LP. One setting that mimicked the JVC demodulator had some nasty distortion / artifacts while the other algorithm was easier on the ears but had less separation. To be fair, I had no means to defeat the RIAA curve on LP playback, so I had to apply RIAA pre-emphasis to the recorded track before running it through the software. Maybe this added some distortion to the final result.
 
I always cringe when I read that the Involve SQ is as good as or better than a Tate! The comparison is usually with the Fosgate unit, which INHO does have a few sonic issues. The Audionics Space and Image Composer is still the best for SQ and I'll tell you why.

Audionics vs Fosgate, the Audionics uses TL074 op-amps run in class A. Fosgate uses TL084 (higher noise version) not in class A. Audionics uses 10μFd coupling capacitors throughout. While some of the coupling capacitors in the Fosgate unit are film types they are of a much lower value. The Audionics has noticeably better bass than any other decoder unit. Audionics uses an eight pole all-pass network, Fosgate a six pole one. Audionics added their "Axial Tilt" circuit and proper adjustment noticeably improves the decoding. Audionics includes a soft clipping circuit (should not be necessary if you are careful not to overdrive the input). Fosgate uses Cmos Analog switches in the signal path, they are somewhat non-linear devices. Audionics uses regular mechanical switches. Audionics has a fully adjustable Stereo Enhance control and a Separation control.

The idea that the Fosgate was better than the Audionics perhaps came from the fact that the Fosgate interface ran faster than the Audionics, and that they used Tate II (Exar chips). Audionics also switched to the Exar chipset from the National ones. As I recall Steve Kennedy in one post says that the Audionics decoding action was smoother than the Fosgate.

My original Composer ran the Exar chips while the unit that I'm currently using has the National chips. As the original unit developed a front to rear problem so I can't directly compare them however I have found that at times I have to reduce the Separation a bit (I never had to with the original unit, unless I wanted mono to reproduce through the back speakers as well). So Tate II might well be a bit smoother than Tate I.

Multibanding IMHO is not really needed, it can actually cause the problem (smeared images) that it's supposed to prevent. I had that happen with my QSD-1.

All that being said the Involve is still a real bargain, yes digital processing has very real advantages over analog, separation wise. My issue with it is that subjectively it's not as good as the S&IC, it has a harshness to it's sound, sort of an upper midrange glare. I'm sure that future versions will improve on that, though,
 
I always cringe when I read that the Involve SQ is as good as or better than a Tate! The comparison is usually with the Fosgate unit, which INHO does have a few sonic issues. The Audionics Space and Image Composer is still the best for SQ and I'll tell you why.

Audionics vs Fosgate, the Audionics uses TL074 op-amps run in class A. Fosgate uses TL084 (higher noise version) not in class A. Audionics uses 10μFd coupling capacitors throughout. While some of the coupling capacitors in the Fosgate unit are film types they are of a much lower value. The Audionics has noticeably better bass than any other decoder unit. Audionics uses an eight pole all-pass network, Fosgate a six pole one. Audionics added their "Axial Tilt" circuit and proper adjustment noticeably improves the decoding. Audionics includes a soft clipping circuit (should not be necessary if you are careful not to overdrive the input). Fosgate uses Cmos Analog switches in the signal path, they are somewhat non-linear devices. Audionics uses regular mechanical switches. Audionics has a fully adjustable Stereo Enhance control and a Separation control.

The idea that the Fosgate was better than the Audionics perhaps came from the fact that the Fosgate interface ran faster than the Audionics, and that they used Tate II (Exar chips). Audionics also switched to the Exar chipset from the National ones. As I recall Steve Kennedy in one post says that the Audionics decoding action was smoother than the Fosgate.

My original Composer ran the Exar chips while the unit that I'm currently using has the National chips. As the original unit developed a front to rear problem so I can't directly compare them however I have found that at times I have to reduce the Separation a bit (I never had to with the original unit, unless I wanted mono to reproduce through the back speakers as well). So Tate II might well be a bit smoother than Tate I.

Multibanding IMHO is not really needed, it can actually cause the problem (smeared images) that it's supposed to prevent. I had that happen with my QSD-1.

All that being said the Involve is still a real bargain, yes digital processing has very real advantages over analog, separation wise. My issue with it is that subjectively it's not as good as the S&IC, it has a harshness to it's sound, sort of an upper midrange glare. I'm sure that future versions will improve on that, though,

Ken , did you ever purchase a Fosgate TATE II to give comparison to the Audionics Space and Image Decoder ?
Not at all doubting your conclusions , just want to know what and how many SQ Decoders you purchased & have for test and Demonstrations.

For myself , I had a Sony SQA 2030 Decoder , and my Fosgate Tate II . FWIW My format of choice in the 70's/80's was SQ . Still is as far as matrix is concerned , although I adored my Sansui QSD 2 (QS) , even with the lack of (at the time) available QS software. For those who don't know , QS Imports , most especially from Japan , were next to impossible to obtain in North America.

I myself hope Malcom comes across with Lynn's Shadow Vector SQ , and hopefully at a reasonable pricing.
 
Who is Dr Bauer?

I read this whole thread & didn't see anything about him, other than the title?

Ben Bauer was head of CBS Labs and invented SQ matrix . His awards and accolades for his work are envious of his peers.
It was a tragedy that he passed on back in April 1979 . Much To early to see the new developments in quad matrix , mainly SQ . Nor could he further promote his "quad baby" , the SQ system and all of it's improvements.
 
Who is Dr Bauer?

I read this whole thread & didn't see anything about him, other than the title?
As others have already pointed out Benjamin Bauer was the creator & dedicated promoter of the CBS SQ matrix system. But I see what you are mainly trying to point out... that the entire original post rambled on for length about stuff that didn't have anything to do with Bauer's legacy. And so much could have been said. Off the top of my head the Ghent microphone for recording live & making an immediate SQ encode, the USQ or Universal SQ system that reconciled many of the needs in basic SQ, his New Orleans matrix that was extremely close to UHJ Ambisonics (and he never used), patents on with height that preceded ATMOS by decades. In fairness Ambisonics considered the height aspect from the beginning and Sansui had a scheme for that as well. CBS & Sansui never took it farther than patents. In fact a patent search on Bauer as the inventor yields a treasure trove of info. Most of his best writings are for AES which means you have to pay for them, but even a Google/Wiki search brings loads of interesting reading.
 
Last edited:
I always cringe when I read that the Involve SQ is as good as or better than a Tate! The comparison is usually with the Fosgate unit, which INHO does have a few sonic issues. The Audionics Space and Image Composer is still the best for SQ and I'll tell you why.

Audionics vs Fosgate, the Audionics uses TL074 op-amps run in class A. Fosgate uses TL084 (higher noise version) not in class A. Audionics uses 10μFd coupling capacitors throughout. While some of the coupling capacitors in the Fosgate unit are film types they are of a much lower value. The Audionics has noticeably better bass than any other decoder unit. Audionics uses an eight pole all-pass network, Fosgate a six pole one. Audionics added their "Axial Tilt" circuit and proper adjustment noticeably improves the decoding. Audionics includes a soft clipping circuit (should not be necessary if you are careful not to overdrive the input). Fosgate uses Cmos Analog switches in the signal path, they are somewhat non-linear devices. Audionics uses regular mechanical switches. Audionics has a fully adjustable Stereo Enhance control and a Separation control.

The idea that the Fosgate was better than the Audionics perhaps came from the fact that the Fosgate interface ran faster than the Audionics, and that they used Tate II (Exar chips). Audionics also switched to the Exar chipset from the National ones. As I recall Steve Kennedy in one post says that the Audionics decoding action was smoother than the Fosgate.

My original Composer ran the Exar chips while the unit that I'm currently using has the National chips. As the original unit developed a front to rear problem so I can't directly compare them however I have found that at times I have to reduce the Separation a bit (I never had to with the original unit, unless I wanted mono to reproduce through the back speakers as well). So Tate II might well be a bit smoother than Tate I.

Multibanding IMHO is not really needed, it can actually cause the problem (smeared images) that it's supposed to prevent. I had that happen with my QSD-1.

All that being said the Involve is still a real bargain, yes digital processing has very real advantages over analog, separation wise. My issue with it is that subjectively it's not as good as the S&IC, it has a harshness to it's sound, sort of an upper midrange glare. I'm sure that future versions will improve on that, though,
I've had a thought recently(not serious at all though) to combine the 101a with the S&IC into one big Frankenstein decoder. Install switchable connections so that the 8-pole phase matrix from the S&IC can be fed into to Tate II DES, you could have the 8-pole and use the Tate Automatic Dimension Control at the same time! theoretically
 
As others have already pointed out Benjamin Bauer was the creator & dedicated promoter of the CBS SQ matrix system. But I see what you are mainly trying to point out... that the entire original post rambled on for length about stuff that didn't have anything to do with Bauer's legacy. And so much could have been said. Off the top of my head the Ghent microphone for recording live & making an immediate SQ encode, the USQ or Universal SQ system that reconciled many of the needs in basic SQ, his New Orleans matrix that was extremely close to UHJ Ambisonics (and he never used), patents on with height that preceded ATMOS by decades. In fairness Ambisonics considered the height aspect from the beginning and Sansui had a scheme for that as well. CBS & Sansui never took it farther than patents. In fact a patent search on Bauer as the inventor yields a treasure trove of info. Most of his best writings are for AES which means you have to pay for them, but even a Google/Wiki search brings loads of interesting reading.


Excellent portfolio of Ben Bauer , Sonik Wiz.

BTW was'nt he also heavily involved in the "London Box" addition for SQ encoding ?

As for the Ghent SQ microphone which automatically gave the SQ encoding to any recording in use , I always wondered how many wee actually made ?
I wonder if anyone knows the answer to my question.
 
Back
Top