Tate DES

QuadraphonicQuad

Help Support QuadraphonicQuad:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
The Shadow Vector experience, with all its ups and downs, provided a good experience for loudspeaker design, although without the daylong discussion at KEF with Laurie Fincham in 1975, it would have been a lot harder when Charlie at Audionics threw me into the deep end of the pool with the order to finish the TL-M200 loudspeaker.

One result was to make me much more aware of the phase angles between pairs of drivers; a lot of speaker designers get kind of sloppy, and let phase angles between midbass and tweeters be 90 degrees or more (at crossover). It looks harmless enough on the frequency-response plot, but sounds very disjointed in practice. Even though you might be sitting 10 feet/3 meters away, it is plainly audible if you know what's wrong with the sound. I try and aim for 5 degrees or less of phase difference - this makes the pair of drivers sound like a single large driver, regardless of listening distance.

When two loudspeakers spaced 10 feet or more apart also have 90-degree phase difference, it helps if the phase difference is the result of a well-constructed phase-shifter, and the pairs have identical drivers, crossover, and polar pattern. Thanks to THX, this is NOT true of HT systems, where the surround speakers are intentionally dissimilar, which guarantees that a dynamic matrix will not provide a uniform soundfield with music.

Even a good discrete mix will sound somewhat disjointed and separated from the front array if the surrounds are dissimilar. Of course, a cohesive soundfield was never a design goal with THX; it was always about replicating the theater experience, not creating a realistic or even plausible soundfield.

It's kind of weird seeing THX using 9 speakers to do what we did with 4 - and we had better results, too, since the soundfield was designed from the get-go to be smooth and uniform, while THX speakers are a weird catchall of at least four different types - Center, the LR pair, dipole or monopole surrounds, and the little guys intended for height or extra-width applications. All different from each other, all with different crossovers, and all with different polar patterns. A circular pan is going to have noticeable timbre shifts - which Audyssey EQ will not be able to correct - while reverb information will fall into several different pools (or trapped in the loudspeaker), instead of joining into a single cohesive soundfield.

My biggest gripe with Audyssey is that is plainly aimed at lo-fi speakers - the frequency response correction in the mids and treble might smooth out the overall room sound, but the first-arrival wave coming from the speaker is then wrong - and it's the first-arrival that determines timbre, not the overall sum from the room. I might use the EQ gizmos if they limited the correction to 300 Hz and below - that's the range where a decently designed speaker is inherently flat, and the room is starting to really go up and down (also, first-arrival and room sounds start merge at those frequencies).

But room correction should not be applied to a flat loudspeaker at frequencies above 500 Hz. The ear works like a cross-correlator, and can easily distinguish between the first-arrival spectra and the first 10 or 20 reflections. It uses the first arrival as a reference, and compares the reflections against it, to try and determine the acoustical properties of the space. If you gum up these 10 or 20 reflections, then the spatial information is degraded, and listening fatigue is increased.
 
Have to agree with Doug G, good to hear the real experts , thanks guys (it's a bit like meeting a rock star from your youth !, I combied both hobbies by getting Rick Wakeman to sign my Q8's of Six wifes and Arthur. We chatted about quad and he recalled havign to sit in the right place and keep dead-still to hear the quad stage)
A question I'd like to ask Martin and Lynn is Have you guys tried out the Audiion SQ and QS decoding scripts on this site?, and if you have how do you think they compare with your hardware inventions?
 
After reading a few pages of the referenced thread, it all seems to come down to the Adobe Audition 1.5 "extract center channel" feature, which apparently leaves stereo intact while extracting a center channel.

I'd be very curious what happens when a test signal (like a speaking voice or pink-noise) is panned from left to right, first slowly, then quickly. Does the center channel disappear progressively, or abruptly? What happens when the panning speed is changed? If the center disappears suddenly, then the AA function would result in a very nonlinear decoder, with intermediate (phantom) localizations jumping around in an abrupt on-off way. If the center channel smoothly fades away to zero as it reaches center, while leaving L&R intact, then that's exactly what you'd want for a digital matrix decoder. Add four wideband 90-degree phase shifters (0 and -90 degrees for each channel), a set of inverters (to get 180 and +90 degrees for each input channel), a bunch of sum-and-difference mixers, and an array of the "extract channel" features, and yes, it could be done.

But the overall quality would critically depend on the panning linearity of the "extract center" function. A sharp cutoff would yield a high-separation decoder (at the cardinal points of SQ), but the intermediate localizations, or any kind of pan would sound not sound right. The image would have noticeable "kinks" in the panorama, and reverb would not sound natural at all, with noticeable point-sourcing at the cardinal locations.

Aside from the unknown properties of the AA channel-extraction function, I'm curious how the phase-shifters really work. In the analog domain, these little devils were one of the primary sources of sonic degradation (along with the VCAs). CBS used something like 8 op-amps (the 301, which was pretty bad) in a cascade to implement their 8-pole shifter in the professional SQ encoder, and I suspect this had a lot to with the "CBS sound" of their SQ records. By contrast, EMI used a parallel-array passive phase shifter for their encoder (I asked when I visited their research staff), and this had a much cleaner signal path (I believe they used discrete transistors instead of op-amps in the buffers that preceded and followed the array).

At Audionics we avoided the CBS approach of a cascade of phase-shifting op-amps because we felt the sound-quality degradation was not acceptable (the same reasoning as EMI Studios). At the cost of some careful component selection (similar to building an RIAA network) to 0.1% precision, we used a parallel array, and made up the gain following the network.

Multiple poles of phase-shifting in the digital domain would have to be done with great care to avoid distortion in the LSB's - 32-bit or higher precision, and very good algorithms to inject correctly shaped digital dither to avoid distortion buildup at low levels. In addition, it is strongly preferable for dither to be independent and non-correlated for all channels, otherwise there is a "monoing" effect as signals fade down to zero and have common dither content between channels. Resampling, equalization, and phase-shifting at studio-quality levels is not trivial to implement - but studio-quality analog phase-shifting isn't cheap, either.

I am more wary of the sonic penalties of 8-pole (or higher) wideband phase-shifting than I used to be. Back then, I found it difficult to hear, but our speakers and amplifiers weren't as revealing and transparent as they are now. I can now hear subtle tonal differences when a single all-pass shifter is inserted in the signal path in the 500 Hz range - not huge, but on direct A/B comparison, audible. And that's just from a single all-pass filter. I'm not so sure that 8 poles on the encode end, another 8 poles on the playback end, and another 2 to 4 poles in the allpass crossover network of the loudspeaker (adding up to 18 to 20 poles of phase distortion between microphone and listening room) doesn't have a sonic impact on the sound. (The encode-decode poles do not cancel - they add. Additional poles from the crossover network in the loudspeaker are in the same direction, so they add as well.)

By contrast, plain old stereo has very little phase distortion until it hits the playback loudspeaker crossover network (microphones, studio equalization, tape recorders, phonograph records, and the RIAA emphasis/deemphasis process are minimum-phase). If a square-wave hits the microphone in the studio, it will still look that way when it emerges from a (fullrange or linear-phase) loudspeaker in the listening room. This is not true for an SQ, QS, or UHJ encode/decode system.

So for all the difficulties and challenges of SQ, I sometimes wonder if it didn't turn out for the best that we ended up with a sophisticated form of the Electro-Voice EV4 encoding matrix - which we now call Dolby Pro-Logic II. It doesn't require any phase-shifting in the encoding process, unlike SQ, QS, or UHJ. And phase-shifting in the playback end is optional - the only reason to use it is to retrieve a center-back image, something I don't find all that pleasant anyway.

Minor digression on loudspeaker design: It is possible to design loudspeakers with 6 dB/octave crossover networks, and staggered tweeter/woofer positions (required to compensate for the acoustic time-delay of the woofer), but 6 dB/octave networks put a lot of excursion stress on the tweeter, which significantly increases IM distortion. Higher-slope networks, like 12 dB/octave, are acoustically a single-pole allpass network, which rotates phase through 180 degrees at the crossover frequency. The more modern Linkwitz-Riley 4th-order network is an acoustic 24 dB/octave crossover, which is effectively a 2-pole allpass filter with a 360-degree phase rotation at the crossover frequency. So a modern 3-way speaker will have 3 to 4 poles of allpass phase rotation, depending on the choice of filter type. There is still much controversy in the loudspeaker fraternity on the audibility of a few poles of phase rotation - some say no phase rotation is acceptable, and are willing to live with more tweeter distortion as a consequence, while others are willing to live with more phase distortion and receive the benefit of better tweeter performance. All of the hot air and hoo-haa in the loudspeaker world over the last 40 years has been over the audibility of zero poles of phase distortion versus 1 to 4 poles. I would think the prospect of adding 16 poles would really get these guys going.
 
Lynn, one thing I wanted to add to what you wrote about phase shifting... All Dolby MP encoders (as well as Circle Surround and Lexicon Logic-7 encoders) Send the 3 front signals through reference all-pass networks - the surround is sent through a +90 degree shift then it's output is split and polarity inverted to create the 180 difference needed. None are used in the decoders, however. Dolby's MP encoding manual has a section on problems the shifts can cause, in certain situations, in tonal quality and how to deal with it.

So, we're still dealing with phase shifting. Just digitally now.

BTW, did you get the Pro-Logic II AES paper I emailed to you?
 
No, sorry, I didn't get the AES paper. I'll recheck my spam filters to make sure it didn't end there. Please resend, if possible.

Sounds like the Dolby MP, Circle Surround, and Logic-7 (who used that?) encoders are similar to Sansui QS encoders, with the fronts at a reference 0 degrees (after passing through an all-pass network), while the rears are +90 and -90 shifted relative to the fronts. That would allow full-circle panning and four, five, and seven-channel "sum" signals with no trouble. No diagonal splits, of course - that's for SQ only.

Been enjoying one of the very first matrix quad recordings, the Beach Boys' Surf's Up, recorded in 1971 with the EV4 matrix. Sounds really good through Dolby Pro Logic II Music mode on the Marantz AV8003/MM8003 combo. Pretty accurate decoding too, with left and right surrounds nicely distinguished from each other, and a spatial quality on "Feel Flows", "Till I Die", and "Surf's Up" that is as good as Pink Floyd's DSOTM SQ album.

Out of curiosity, I tried a more recent recording, Ulrich Schnauss' Far Away Trains Passing By, and that has superb matrixing as well. A lot of undiscovered gems out there.
 
Logic-7 is from Lexicon and its encoder, while not used much, is a fairly bizzare active encoder meant more for pair-wise-mixed 'spotlighting' than creating a 360 soundfield. The Logic-7 decoder has changed a lot over the years and in fact, is rarely the same algorithm from product to product.

I'll resend the PL-II AES paper right now.
 
Just received the AES paper, much appreciated. Regarding the question asked in the e-mail about the 1979 AES Journal Tech Note on the "Stereo-180" microphone, no, I never built it. But the math was perfectly good, and the proposition to build a stereo microphone with a limacon pattern and the mentioned spacing - similar but a bit different than ORTF - was perfectly sound. I was contacted much later, in 1991, by a mysterious phone call out of the blue. I had no idea how they tracked me down in Oregon, but they did, and the engineer worked for one of the major NPR affiliates, in Washington DC as I recall, and had actually built one and was quite pleased with the results.

I was frankly dumbfounded that somebody had found one of the more obscure AES articles, built the microphone, used it to make recordings for NPR, and tracked me down in the Oregon to call me up and thank me! This was one of the more remarkable coincidences I've ever come across in my life - I had almost totally forgotten I had written that article in my last days at Audionics as a final nod to the field of quadraphonics. The timing was especially fortuitous - I had been summarily fired from a rather politicized job as IT director for a local school district, and getting that call as a bolt out of the blue was a reminder to get back into audio, and set these miserable computer jobs aside. So I did. Out of that came the articles I wrote for Positive Feedback magazine, the Ariel and ME2 loudspeakers, conversion of the PF articles to the early days of the Web in 1995, the Amity and Karna triode amplifiers, and so on. So that phone call had quite an effect on me.

Regarding the AES article - which I really need to scan and post on the www.nutshellhifi.com website - it was derived from a prior article that analyzed the phasors needed for a microphone to capture a 270-degree or wider soundfield. Since all the phasors were at 500 Hz, I realized in a flash of insight that the phase domain was the wrong way of looking at the problem - convert phase into time delays, and the microphone didn't need a complex external phase-encoding system, but could be simply realized through physical spacing and the appropriate selection of polar pattern. Thus, a self-encoding microphone that also had a high degree of stereophonic playback compatibility, as well as good playback with quadraphonic matrices and headphone listening.

I've always seen quadraphonics as an extension of stereophonic sound, using a decoder to discriminate the phase relations of a two-channel source into a natural-sounding front-to-back spatiality. It's not easy to hear partially out-of-phase content with two speakers (well, I can, but I have to listen for it), but a good decoder easily discriminates the phase relations and assigns localizations based on that. This upward compatibility is one of the most important features of any well-designed decoder - most important of all, it should not generate artifacts that are not present in 2-speaker playback, but while retaining the musical values of the original recording.
 
So the mic was kind of like the Ghent Microphone but without needing the external phase shifters for encoding? I have a DTS CD made with the Ghent and it's incredible sounding - the most realistic 360 sound I've ever heard.

I'm sensitive to wide phase shifts in stereo playback - they make me physically uncomfortable - it's one of the reasons I hated 2-channel UHJ Ambisonic.
 
I agree 100% about the Center Front phase-shift problem with UHJ - the only reason it was done was to make Center Back audible on monophonic playback, which always struck me as pretty retarded reasoning. I said as much to the BBC when I met them in 1975 - anyone listening in mono doesn't give a ***t about fidelity, so why cater to them? And the whole neurotic emphasis in Center Back is wrong anyway - it's a very unpleasant localization, mostly suitable for horror-movie effects, or very quick pans from Left Back to Right Back.

EMI was quite right to favor Front SQ encoding over the CBS style of 360-degree encoding - and Front encoding is a natural feature of the Ghent microphone anyway. For all my quibbles about symmetry, a decision to use Front encoding solves them at a stroke - both diagonals are symmetric, and pans across the allowed 270 degrees are very smooth and symmetric. It also sounded more spacious than the 360-degree encoder, probably due to the absence of phase pileups with random-phase microphone and EMT-reverb sources.

The smooth distribution of random-phase information around the room was a key design feature of Shadow Vector - other decoders let it pile up at the speaker locations, or pulse or shimmer around the room thanks to unsteady decoder gains. With Shadow Vector, the reverberant information sounded the same as a static matrix or discrete. This was not true with many other dynamic-decoder designs, which tended to "dry out" the spaciousness of the original mix. One thing that's annoying about Dolby Digital and other high-compression-ratio lossy codecs is the lack of spaciousness compared to the original mix, which is probably why a Dolby technician is present at movie mixdowns - to restore what the lossy codec takes out. I was very surprised to discover that a 2-channel Dolby Digital codec actually discards phase information, converting stereo into little more than steered mono, which gives a matrix decoder very little to do on playback.
 
Taking a larger perspective, the systems that had the best compatibility with 2-channel stereo were EV-4, SQ, and DPL II. All of them leave the frontal arc intact, and this is a good thing, since most of the market for music from 1958 through the present has been 2-channel stereo, with surround still a very small portion of the market. Mono, as mentioned before, is mostly the domain of AM radio (since AM stereo failed) and very old (pre-1985) analog televisions.

The touchstone of an upwards and downwards compatible surround system is how well it integrates with the installed base of 2-channel systems ranging from car stereos, iPods, up to exotic high-end systems. First off, there should be no downgrade in audio quality when moving from 2 to more channels (the tradeoff imposed by CD-4 LP's, the Compact Disc proposal using 32 kHz as a sampling rate, and Dolby Digital). Part of the reason DVD-A failed in the marketplace was a user-unfriendly television interface in order just select the high-quality tracks of a DVD-A disc - instead of just hitting the "play" button, which customers expected.

Second, transcoding between formats should not "lose" any spatial information. 2-channel difference information should be translated into front-to-back depth when played back in surround, and surround front-to-back depth should be translated into extra-width information in 2-channel stereo. This is where a well-designed matrix system comes to the fore. A crude folding-together of the front and rear channels destroys front-to-back information, and is NOT the way to translate between systems. Even an ideal CD-4 system, with no mistracking, perfectly flat group delay in the 15 to 50 kHz region, and a distortionless FM demodulator, still suffered from front and rear channels laying on top of each other when played back on a conventional stereo system.

The mono-to-stereo transition was handled a lot more gracefully than the stereo-to-surround transition, which is still a work in progress 40 years after EV-4 and SQ records made their appearance in the marketplace. We're pinning our hopes on Blu-ray discs, but it's not like the marketplace is flooded with music releases on that format. Unfortunately, the lossy-compression codecs of the iTunes and Amazon stores degrade the phase information of 2-channel recordings, so playback through an matrix decoder is spatially degraded. On the download front, we're actually going backwards from the earliest days of stereo.
 
I tend to be in and out of the quad / surround sound loop so its no surprise I missed this thread until now. Not that I understand all the particulars, but a fascinating peak behind the curtain nonetheless. It really is a shame there isn't an easier solution to extracting accurate SQ information, particularly with programs like Adobe Audition. I've used the available scripts (with the exception of the newest process by OD) and the results are ok, but problematic as is well documented. I also have a Lafayette LR 5000 with the SQ Full Logic decoder, though it tends to sum the back channels into mono (the front / back separation is superb, however).

Outside of finding a Tate II the options for exceptional SQ decoding are practically nil. Especially frustrating considering the enormous wealth of SQ titles still available. All software and darned little hardware - terrible circumstance!

Anyway, thanks for sharing all the inside info with us technically-challenged folks. Cheers!
----------- Chris
 
All software and darned little hardware - terrible circumstance!
If a significant market existed for a new-technology decoder there would be also a significant market for reissues in digital discs using current technology, and the new hardware would be then unnecessary.
 
Outside of finding a Tate II the options for exceptional SQ decoding are practically nil. Especially frustrating considering the enormous wealth of SQ titles still available. All software and darned little hardware - terrible circumstance!

Do you know of the
Surround Master?
If you want the best decoding for SQ
and QS this is the one
 
Outside of finding a Tate II the options for exceptional SQ decoding are practically nil. Especially frustrating considering the enormous wealth of SQ titles still available. All software and darned little hardware - terrible circumstance!

Do you know of the
Surround Master?
If you want the best decoding for SQ
and QS this is the one

Hey rustyandi! I was just reading about that! Holy Smokes! Like I mentioned, I've been away from the quad stuff for a little while - I had no idea this was developing, let alone being made available. I am most definitely going to buy the Surround Master. Haven't been this excited about quad since I did my first conversion with AA scripts. Wow. Thanks for the heads up as well! Cheers!
---------- Chris
 
Hey rustyandi! I was just reading about that! Holy Smokes! Like I mentioned, I've been away from the quad stuff for a little while - I had no idea this was developing, let alone being made available. I am most definitely going to buy the Surround Master. Haven't been this excited about quad since I did my first conversion with AA scripts. Wow. Thanks for the heads up as well! Cheers!
---------- Chris

You'll be glad you did. The Surround Master is hard to beat for stereo to surround synthesis, not to mention matrix decoding !
 
Well 8 years later, a 10 band shadow vector program decoding in real time on an embedded ARM processor is up and running. A fast vector modification speed of 25ms was used across all 10 decode bands. Also the ability to look ahead made quite an improvement to the decode on large vector changes. I'll have to obtain Bridge Over Troubled Waters and give that a decode. I like the alignment method used by Lynn. I'll post waveform results from such a test soon.
 
Well 8 years later, a 10 band shadow vector program decoding in real time on an embedded ARM processor is up and running. A fast vector modification speed of 25ms was used across all 10 decode bands. Also the ability to look ahead made quite an improvement to the decode on large vector changes. I'll have to obtain Bridge Over Troubled Waters and give that a decode. I like the alignment method used by Lynn. I'll post waveform results from such a test soon.

can help out with the BOTW SQ, start a conversation anytime!
 
Hi All

Has anyone heard of Disclord recently, last time I spoke to him was a few years ago and he was quite ill. Hope he is OK.

Regards

Chucky
 
Sadly Ty, aka Disclord, passed away a few years ago. I was fortunate enough to know him from early days as Stereoboy on AOL & he lived in my area. I value the few times Ty, Tony & I could get together. He had an encyclopedic knowledge of surround sound & just damn fun to be around. He is missed by myself & everyone else on this forum. Interesting that his posts continue to be kept alive.
 
Back
Top