Lou Dorren: A new CD-4 Demodulator!!! [ARCHIVE]

[Disclord]

CD-4 isn't a discrete format in the same way as Q4 or Q8 - CD-4 is a discrete matrix format, using a 4:4:4 compatibility matrix, in the same way stereo FM and the 'old' BTSC Zenith/dbx NTSC stereo TV system did with their 2:2:2 discrete matrix. According to the January 1974 issue of High Fidelity, the JVC 4DD-5 had 15db to 20db of separation back to front (depending on frequency) and Right Front leaked into the Right back at a level of only -5db due to the supplied JVC CD-4 set up record being cut wrong, meaning the right side couldn't be properly calibrated - it was the same with the Kenwood CD-4 receiver they tested. In any case, the CD-4 format IS a discrete system but will never (at least with the old demodulators) have the same separation as a non-matrixed discrete format like Q4. Depending on your cartridge and setup (and demodulator), Front/back channel separation anywhere from 15 to 20 db or so can be expected.

 

[Doug G.]

Good write-up Disclord!

I wanted to write something similar but was too lazy.

Anyway, it is generally accepted that if the separation between channels in quad can equal that of regular two channel stereo, it can be considered discrete. So, at 15-20dB, it is.

Also, if one is experiencing noticibly audible crosstalk with CD-4, something is not working correctly.

Doug

 

[Disclord]

Good write-up Disclord!

I wanted to write something similar but was too lazy.

Anyway, it is generally accepted that if the separation between channels in quad can equal that of regular two channel stereo, it can be considered discrete. So, at 15-20dB, it is.

Also, if one is experiencing noticibly audible crosstalk with CD-4, something is not working correctly.

Doug

And there is two kinds of 'discrete' or 'audibly discrete' - there's the kind that CD-4 and SQ via the Tate provide, where the channels are related so 15-20db separation is just fine, and then there is the kind of discrete that's meant to carry two completely unrelated, different signals - like if a CD-4 LP were recorded with one stereo album on the front channels and another different album on the back - that it could not do. It's not a 'fully discrete' source like Q4 tape which can carry unrelated sources.

Also, psychoacoustic tests have proven that for stereo and quad sources, once you reach 20db of separation, any more is a waste - you can't hear any difference between 20db separation and 70db.

 

[Doug G.]

Actually, two different albums could be put on the front and back channels with CD-4. It really wouldn't be any different because the inputs to the modulators are still going to create the sum and difference signals the same as if they were related signals from the same album.

The demodulators would separate the front and back channels the same.

Another way to look at it would be: say the drums from a given source are in the front channels and the lead guitar in the back. That's not any different than feeding, say, the vocals from "Hello, I love you" into the front and the lead guitar from "You're So Vain" into the back (where, in fact, it is). They would still be demodulated correctly.

Doug

 

[surroundme33]

CD-4 isn't a discrete format in the same way as Q4 or Q8 - CD-4 is a discrete matrix format, using a 4:4:4 compatibility matrix, in the same way stereo FM and the 'old' BTSC Zenith/dbx NTSC stereo TV system did with their 2:2:2 discrete matrix. According to the January 1974 issue of High Fidelity, the JVC 4DD-5 had 15db to 20db of separation back to front (depending on frequency) and Right Front leaked into the Right back at a level of only -5db due to the supplied JVC CD-4 set up record being cut wrong, meaning the right side couldn't be properly calibrated - it was the same with the Kenwood CD-4 receiver they tested. In any case, the CD-4 format IS a discrete system but will never (at least with the old demodulators) have the same separation as a non-matrixed discrete format like Q4. Depending on your cartridge and setup (and demodulator), Front/back channel separation anywhere from 15 to 20 db or so can be expected.

Thank you for clearing this up for me. I can accept the term discrete matrix. When i first ventured into CD-4, I was only expecting something along the lines of stereo separation from front to back. Boy, was I wrong... And, lots of money and time invested, i have come to the conclusion that this is NOT possible. Today, i had the chance to download someone's rip of AWB CD-4 LP who used a modded JVC 4DD-5.. It sounded sonically superior to my personal transfer, but still included the crosstalk. Listen to "You've got it" (first cut) and the first verse the vocals should ONLY be in front. However, you can hear them in the rears with a reverberated effect.

Again, i don't want to sound like I'm complaining.. Because, I will take matrixed discrete quad hocus-pocus over conventional stereo anyday.

 

[surroundme33]

Anyway, it is generally accepted that if the separation between channels in quad can equal that of regular two channel stereo, it can be considered discrete. So, at 15-20dB, it is.

Doug

With all due respect Doug.. The separation of all frequencies in Stereo between Left and Right channels does not compare to the separation of front and back (I.E. Left Front, Left Rear). And yes, I've done all those twist turns with the carrier level pots and separation pots on all my demodulators. I have even toyed with the Technics SH-400 which has a CCC (Cross Channel Cancelation) pot and still.. The song remains the same.

 

[Doug G.]

One thing to remember is that, although the potential separation is 20dB between all channels (and this was proven long ago), any given mix may be such that there is bleed and it could have been done purposefully.

The vocals you spoke of could have been mainly in the fronts with reverberation in the backs on purpose.

But, say "Let It Happen" by the Jazz Piano Quartet where there is one of four pianos in each corner, I can turn my controls to listen to just one piano at a time in its respective channel and the other pianos are pretty much inaudible except for studio echo.

Doug

 

[surroundme33]

And there is two kinds of 'discrete' or 'audibly discrete' - there's the kind that CD-4 and SQ via the Tate provide, where the channels are related so 15-20db separation is just fine, and then there is the kind of discrete that's meant to carry two completely unrelated, different signals - like if a CD-4 LP were recorded with one stereo album on the front channels and another different album on the back - that it could not do. It's not a 'fully discrete' source like Q4 tape which can carry unrelated sources.

Most CD-4 catalogs usually have mixed musical sources with exaggerated separation (like drums on front left, rhythm guitar on right, bass on Left rear, and lead guitar on right rear). I like this kind of quad actually. This is where CD-4 is most interesting to listen too. Unfortunately, there's still that channel bleed because you are always going to hear the drummers cymbals and high hats bleed to the rears. SQ has some nice quad catalog sources too. There are some very creative SQ mixes done, that you would swear you're listening to a Q4 or Q8 source. One that comes to mind is The soundtrack to Shaft. Lovely fidelity and very interesting quad effects using random 'ping-pong' effects from speaker to speaker.

With CD-4, separation is sometimes better when certain frequencies are involved. When certain musical elements mask the smears of sound to the rears.

 

[surroundme33]

One thing to remember is that, although the potential separation is 20dB between all channels (and this was proven long ago), any given mix may be such that there is bleed and it could have been done purposefully.

The vocals you spoke of could have been mainly in the fronts with reverberation in the backs on purpose.

But, say "Let It Happen" by the Jazz Piano Quartet where there is one of four pianos in each corner, I can turn my controls to listen to just one piano at a time in its respective channel and the other pianos are pretty much inaudible except for studio echo.

Doug

Doug, what you are hearing is a ghosting effect of the smearing of the other pianos into that respective channel. This is not uncommon. I have a couple of Doobie Bros Lps in CD-4 (let's talk crazy separation here, LOL!)... And, you can hear the same ghosting effect when listening to the respective channels. THere is a track that divides bass guitar only on front and drums on back. But, c'mon... You can still hear the drums leak through. Compare it to the Q4 source, and you will see a world of difference.

I recently acquired the CD-4 Lp to john cage and George Flynn 'winter music and wound' .. I believe they recorded this with 2 separate stereo takes. Where they divided the 2 takes using front and rear.. It's hard to hear the smearing effect here, because it's only one instrument (a piano).. So, I would validate this as an acceptable CD-4 record.

 

[surroundme33]

duplicated this post.. I can't delete :-(

 

[Quadro-Action]

And now my voice again among this discussion. When Lou will read this messages, it may help, that Jon Urban will receive after a long time of development very shortly the prototyp for the test. Announced is here a channel separation (front-back) of 34 dB or similar).
That one can not notice a separaten about 20 dB is of course only the separation in front left to right. By front back one will listen, that there is a differenz of 20 dB or 30dB and more. By an adjustement with the technics adjust-single my SH-400 shows for the right channels 30 dB, and for left 28 dB. For my JVC 1000 I have a technical paper with a manual measurement of the channel separation (we speak always for front-left), which has interesting results. The measurement was not done with such a single, but with electronic sources of an CD-4 signal generator by the engineers. Left and right with little differences. 50 Hz = 37/33 dB, 100= Hz 25/26, 200 Hz 27/30, 1 kHz= 31,3/35, 4 kHz= 25,5/29, 7 kHz 28/22,5, 10 kHz 14,5/14,5. This shows, that most important is the separation in low frequencies and around 1 kHz (similar to UD-4). So the 30 dB of my SH-400 my be also the value near 1 kHz.
This whole discussion was also my consideration, how could work a CD-4 record in real discrete separation. This means only the front channels in the usual listening freqruency area till 15 kHz and only the rear channels with the carrierer 30 kHz in the frequencies 30-45 kHz. One means as an answer, that the fidelity of the back channels would by those working lower. Nevertheless for my an interesting ask with no a real answer, if those considerartions was done also by the ingeneers of the CD-4 development in 1970/71.

The Matrix by CD-4 works a lot different to those for SQ. The CD-4 "matrix" is only a very simple A/B matrix to sort the front plus rear informations minus the different signals. Not so expensive as with mathematc formulars by SQ. So the result is for me better as an usual real matrix. OK, the late demodulators like Space & Image or tate II has better results as those before. The first decoders with a separation of 3 dB have been really a joke. The 26-30 dB of the CD-4 system was from beginning an impressive listening experience.
Now I am curious for further comments.
Dietrich

 

[Disclord]

Actually, two different albums could be put on the front and back channels with CD-4. It really wouldn't be any different because the inputs to the modulators are still going to create the sum and difference signals the same as if they were related signals from the same album.

The demodulators would separate the front and back channels the same.

Another way to look at it would be: say the drums from a given source are in the front channels and the lead guitar in the back. That's not any different than feeding, say, the vocals from "Hello, I love you" into the front and the lead guitar from "You're So Vain" into the back (where, in fact, it is). They would still be demodulated correctly.

Doug

No, with a discrete matrix like CD-4 used, there is never going to be enough separation so that you never hear the other, unrelated, album without any leakage between the front/back channels. That's why in formats that used discrete matrixing, like RCA's CED VideoDisc format, they used a sum-difference matrix (sum on one carrier, difference on the other) for stereo compatibility, but for bilingual discs each channel was encoded discretely on its own carrier without the matrix because you can never get 100% perfect separation from an analog discrete matrix sound format. It would be absolutely impossible to record two completely separate stereo albums on a CD-4 LP (say The Doors on the front two channels and the Rolling Stones on the back two channels) and be able to demodulate them back to where you could shut off the front or back channels and hear the individual stereo recording without ANY audible leakage from the other stereo recording. Analog discrete matrixing like CD-4 used just doesn't work that way.

 

[Disclord]

And now my voice again among this discussion. When Lou will read this messages, it may help, that Jon Urban will receive after a long time of development very shortly the prototyp for the test. Announced is here a channel separation (front-back) of 34 dB or similar).
That one can not notice a separaten about 20 dB is of course only the separation in front left to right. By front back one will listen, that there is a differenz of 20 dB or 30dB and more. By an adjustement with the technics adjust-single my SH-400 shows for the right channels 30 dB, and for left 28 dB. For my JVC 1000 I have a technical paper with a manual measurement of the channel separation (we speak always for front-left), which has interesting results. The measurement was not done with such a single, but with electronic sources of an CD-4 signal generator by the engineers. Left and right with little differences. 50 Hz = 37/33 dB, 100= Hz 25/26, 200 Hz 27/30, 1 kHz= 31,3/35, 4 kHz= 25,5/29, 7 kHz 28/22,5, 10 kHz 14,5/14,5. This shows, that most important is the separation in low frequencies and around 1 kHz (similar to UD-4). So the 30 dB of my SH-400 my be also the value near 1 kHz.
This whole discussion was also my consideration, how could work a CD-4 record in real discrete separation. This means only the front channels in the usual listening freqruency area till 15 kHz and only the rear channels with the carrierer 30 kHz in the frequencies 30-45 kHz. One means as an answer, that the fidelity of the back channels would by those working lower. Nevertheless for my an interesting ask with no a real answer, if those considerartions was done also by the ingeneers of the CD-4 development in 1970/71.

The Matrix by CD-4 works a lot different to those for SQ. The CD-4 "matrix" is only a very simple A/B matrix to sort the front plus rear informations minus the different signals. Not so expensive as with mathematc formulars by SQ. So the result is for me better as an usual real matrix. OK, the late demodulators like Space & Image or tate II has better results as those before. The first decoders with a separation of 3 dB have been really a joke. The 26-30 dB of the CD-4 system was from beginning an impressive listening experience.
Now I am curious for further comments.
Dietrich

The channel separation tests done with the CD-4 modulators, and not using a CD-4 LP, will give much greater separation results because you don't have the phase-shifts, poor HF channel separation, tracking distortion, etc... that the LP medium and playback introduces into the CD-4 process. So testing with a CD-4 LP gives "real world" results.

In regards to separation - and how much is needed to localize instruments/voices clearly, we need about 20db Left to Right in front and back. Center Front to Center Back we need only about 12-15db - on the sides, such as Left Front to Left Back, we need 15-20db. The way our hearing works has different separation needs depending on direction. Those numbers I gave are for single sources - when all channels are active, like in a well mixed quad recording with instruments coming from every channel, less separation is needed because we can't hear all the individual instruments at once - we can localize one or at most two dominant sounds and the directions of the other sounds are 'masked' until another one of them becomes dominant - then the previously dominant sounds location is masked. Evolution has 'wired' our brains to pay attention to the dominant sound and ignore lesser sounds. So even with discrete Q4 recordings, we're never localizing all channels at the same time - our brains and hearing just can't do it. The BBC, CBS, Sansui and the National Quadraphonic Radio Committee all discovered this effect.

 

[Quadro-Action]

Another point for this discussion. The channel separation is not only dependent from the technical possibilities of the CD-4 system or the demodulator. It is also dependent from the mixdown. The DVD-Audio and SA-CD's have technical values for a big separation (may be 80 or 90 dB or above). But some surround productions sounds with a less separation as CD-4. That means, the sound engineers make a lower separation by the mixdown, because a too big separation could sound steril and unnatural. So many actual or modern surround sounds with modern sound mediums have sometimes a lower separation as it could be possible.

Dietrich

 

[Disclord]

Another point for this discussion. The channel separation is not only dependent from the technical possibilities of the CD-4 system or the demodulator. It is also dependent from the mixdown. The DVD-Audio and SA-CD's have technical values for a big separation (may be 80 or 90 dB or above). But some surround productions sounds with a less separation as CD-4. That means, the sound engineers make a lower separation by the mixdown, because a too big separation could sound steril and unnatural. So many actual or modern surround sounds with modern sound mediums have sometimes a lower separation as it could be possible.

Dietrich

That's true - even most stereo material doesn't have much separation between channels - most is kind of wide-mono (which is clearly exposed in the "Surround" mode on my Fosgate Tate II 101A) but our CD-4 discussion was more about the capability inherent in the CD-4 format rather than the quad mix itself.

 

[The Quadfather]

CD-4 is discrete in the sense that it has four channels of audio. A matrix record only has two. The method that is used produces channels that are mathematically correct. A thing not achievable with SQ, QS and any other two channel system. The matrix part is not like an SQ matrix where phase shifting is employed, in CD-4 phase shifting is intentionally avoided, and processing delays are compensated for. Where there is lack of separation is where the technology of the day fails to maintain the specs. In matrix quad, it is inherent in the system, because there are only two channels, plus you have the technology failure to maintain specs as well. That is the point of Lou designing a new demodulator, to overcome CD-4's shortcomings with new technology that is superior than what was acheivable back in the 70's. In CD-4 there are no "logic" or any other separation enhancing circuitry, simply because it is not needed. This is evidenced by the result. However, if the resistors in the matrix of the CD-4 circuit have changed value as components some thirty years old sometimes do, you can have these effects: Poor separation front to back, both sides if both sides have bad resistors. Poor separation one side if only one side has a bad resistor. You can even have poor separation in one speaker only, say for example, a rear, while it's counter part (front) has good separation. Incidently, left right separation in CD-4 has nothing to do with the demodulator in CD-4. That is maintained by the cartridge and it's ability to discriminate between the outer and inner groove walls, which are modulated separately with left and right signals and subcarriers. A CD-4 demodulator is actually two demodulators, one deals with the left side and the other deals with the right with each one only providing front to rear separation.

 

[Doug G.]

I didn't mean to imply that it would be perfect if you recorded one album for the front channels and one for the back channels.

I just meant that you COULD do it and if you were listening to all channels at once, you would hear one album in the front and the other in the back.

But yes, I'm sure if you turned the controls to only listen to one or the other, you would be able to hear some bleed.

Doug

 

[Disclord]

CD-4 is discrete in the sense that it has four channels of audio. A matrix record only has two. The method that is used produces channels that are mathematically correct. A thing not achievable with SQ, QS and any other two channel system. The matrix part is not like an SQ matrix where phase shifting is employed, in CD-4 phase shifting is intentionally avoided, and processing delays are compensated for. Where there is lack of separation is where the technology of the day fails to maintain the specs. In matrix quad, it is inherent in the system, because there are only two channels, plus you have the technology failure to maintain specs as well. That is the point of Lou designing a new demodulator, to overcome CD-4's shortcomings with new technology that is superior than what was acheivable back in the 70's. In CD-4 there are no "logic" or any other separation enhancing circuitry, simply because it is not needed. This is evidenced by the result. However, if the resistors in the matrix of the CD-4 circuit have changed value as components some thirty years old sometimes do, you can have these effects: Poor separation front to back, both sides if both sides have bad resistors. Poor separation one side if only one side has a bad resistor. You can even have poor separation in one speaker only, say for example, a rear, while it's counter part (front) has good separation. Incidently, left right separation in CD-4 has nothing to do with the demodulator in CD-4. That is maintained by the cartridge and it's ability to discriminate between the outer and inner groove walls, which are modulated separately with left and right signals and subcarriers. A CD-4 demodulator is actually two demodulators, one deals with the left side and the other deals with the right with each one only providing front to rear separation.

But it's not a discrete system in the same way Q4 and Q8 are - it can't carry four unrelated signals that must stay totally unrelated. It's a discrete system using 4:4:4 matrixing for compatibility, thus, it's a discrete matrix, just as stereo FM is a 2:2:2 discrete matrix system. No analog discrete matrix can maintain the same channel separation as a truly discrete 4:4 system like Q4. The matrixing of SQ and QS is different only in that they 'throw away' half the channels, meaning you have to introduce phase-shifting in the matrix to cover the full 360 compass and decoding involves solving for four unknowns with only two linear equations, thus original independent channels can never ever be recovered. But, since the musical dominance is constantly changing, the only accurate decoder is one that can change its structure to follow the channel dominance - hence the Tate, Paramatrix and Shadow-Vector. (power-transfer gain-riding decoders don't change their decoding so are not accurate) The CD-4 system has four unknowns with four equations, so the channels can be recovered, limited by the LP medium and analog gear of the day, which wasn't very good. The 1974 High Fidelity tests of JVC's 4-DD5, Kenwood and Heath CD-4 demodulators were not very impressive separation-wise. JVC cut and pressed the High Fidelity test record using their half-speed lathes (CBS did the same for the SQ tests using the Position Encoder - Sansui chose not to participate). Hopefully, Lou's new demod will make CD-4 work much better. It's a hell of a flaky format. (BTW, according to High Fidelity, JVC repressed the CD-4 set up record that was packed with demodulators since the first one was mastered wrong. Apparently, the corrected repressed set up disc had a #2 on the inner-wax.)

I'll be scanning the entire High Fidelity article in the next few days and posting it as a PDF since it's so interesting - plus the updates that are in later issues.

 

[Doug G.]

Of course CD-4 isn't the same as an analog tape system in that the tape systems maintain 4 distinct channels that remain in the audio band throughout whereas CD-4 "manipulates" the four signals with FM and PM modulation and such.

However, it must be recognized that tape systems are not perfect either being subjected to crosstalk from mistracking of the tape heads on the tape, particularly with the Q8 system due to the rather imprecise nature of the tape path.

In fact, it is difficult to get an 8-track system to maintain acceptable separation from the top tracks to the bottom with different tapes. Hence the old matchbook trick people used to use.

Doug

 

[Quad Linda]

Disclord, I take exception with your criticism of '70's analog and LP technology. Having owned and sold some high-end Quad and 2ch back in the day, I'll stand by 70's fidelity being great. In fact, despite the quantum leaps that have been made in the last 40 years, I'll put nearly everyone's old Quad gear against 90% of what's sold as surround today. The vintage gear sounds better. That seems diametrically opposed to my comments on the "What's Your Quad Receiver" thread. It isn't. I would rather own an upper line modern surround receiver than any 2ch or Quad '70's gear. I feel even more strongly about modern surround preamps and amps.

Play a 70's LP, homemade reel or cassette, or a 7 1/2 ips prerecorded reel, and you'll find some great analog fidelity. Things like 1/2 speed mastered LP's and dbx II n/r upped the ante for analog, but the technology was pretty great without it. In fact, beyond preamps/amps, early and mid-70's gear sounded way better than the late '70's stuff. It certainly had a better build quality.

There were very high fidelity recordings and components in the late '50's. Sadly, most people had console systems that couldn't realize it. I'd put my 7 1/2" Bell & Howell mono reel, circa 1961, against most modern stuff. My Dad gave it to me back then so I woudn't buy 45's anymore.

I'm gonna invoke my advanced age here: most people under 50 think fidelity began in the '80's with Pro-logic and/or CD. It began in the 50's with a concept called high fidelity. The term mono doesn't adequately describe the transition from tinny limited fidelity to full frequency response. From there, everything is simply a refinement: stereo, Quad, CD, digital, 5.1 or whatever technological breakthrough. Don't get me wrong, I don't care to listen in 2ch or mono if I don't have to. I wouldn't own 1000+ Quad and 5.1 titles if I felt differently. If you think I'm wrong, buy any Blue Note mono or 2ch SACD. Is there anything on them that wasn't on the master tape to begin with? If Steve Hoffman and co. magically made wheat from chaffe, then why couldn't they create that out of Armstrong's Hot Fives & Sevens?

70's hardware and software. It's all there, simply listen.

Linda
Have a Nice Decade

But it's not a discrete system in the same way Q4 and Q8 are - it can't carry four unrelated signals that must stay totally unrelated. It's a discrete system using 4:4:4 matrixing for compatibility, thus, it's a discrete matrix, just as stereo FM is a 2:2:2 discrete matrix system. No analog discrete matrix can maintain the same channel separation as a truly discrete 4:4 system like Q4. The matrixing of SQ and QS is different only in that they 'throw away' half the channels, meaning you have to introduce phase-shifting in the matrix to cover the full 360 compass and decoding involves solving for four unknowns with only two linear equations, thus original independent channels can never ever be recovered. But, since the musical dominance is constantly changing, the only accurate decoder is one that can change its structure to follow the channel dominance - hence the Tate, Paramatrix and Shadow-Vector. (power-transfer gain-riding decoders don't change their decoding so are not accurate) The CD-4 system has four unknowns with four equations, so the channels can be recovered, limited by the LP medium and analog gear of the day, which wasn't very good. The 1974 High Fidelity tests of JVC's 4-DD5, Kenwood and Heath CD-4 demodulators were not very impressive separation-wise. JVC cut and pressed the High Fidelity test record using their half-speed lathes (CBS did the same for the SQ tests using the Position Encoder - Sansui chose not to participate). Hopefully, Lou's new demod will make CD-4 work much better. It's a hell of a flaky format. (BTW, according to High Fidelity, JVC repressed the CD-4 set up record that was packed with demodulators since the first one was mastered wrong. Apparently, the corrected repressed set up disc had a #2 on the inner-wax.)

I'll be scanning the entire High Fidelity article in the next few days and posting it as a PDF since it's so interesting - plus the updates that are in later issues.