CD-4 (Quadradisc) software demodulator project

[TimDog73]

OK so I am not a Lathe Troll but I do appreciate your art!

Not so long ago I discovered one of my friends had a collection of Quadradisc vinyl and a turntable equipped with a Shibata stylus but to my dismay I discovered that his demodulator had broken some yeas ago and he had not replaced it.

Due to the short supply of demodulators in the UK I set about investigating the possibility of using a PC to capture and demodulate the the signal. I do believe this is possible and we are getting some results although they sound a rough. However I am confident we should be able to improve on these if I can get some more detail on the CD-4 standard.

We have managed to find articles and manuals on on the subject including Lou Dorrens "New CD-4 Demodulator Project" but so far there is not enough detail to get a result good enough to listen too.

The sort of information I require is detailed filter characteristics for the PLL and ANRS circuits e.g. filter roll off rates, PLL Gain etc.

If there is anyone on this forum who has contacts in JVC who developed this technology or people from record company who might have access to detailed documentation about the CD-4 standard.

I would most appreciate it if you could put your hand up or get me in touch with them.

Many thanks in advance,
Tim

[markrob]

Hi,

I've played with a developing a software based solution and had some luck. I'm reasonably sure that a fully DSP based solution can be developed.

You are correct that the ANRS parameters need to be determined to decode properly. The JVC ANRS patent seems to have all of the needed info. Have you looked at US patent 3757254?

Is Lou Dorren alive? I was following his project some time ago, but haven't kept up. Did he ever deliver his updated decoder?

The JVC 4DD5 schematic shows the ANRS expander to be a simple FET based design. It seems like it would not be that hard to determine the proper time constants and expansion curve from the schematic along with some bench testing.

On the cutting side, it would be interesting to develop a CD-4 encoder so that new disks could be cut.

Mark

[TimDog73]

Mark,

I'd be interested to know what tools you used for your experiments? If you visit http://forum.audacityteam.org/viewtopic.php?p=220106#p220106 you can eavesdrop on what we're doing.

Thanks for the US Patent tip we're looking at it now; also Lou Dorren has a patient on this subject too which we are also looking at.

I can not be 100% sure if Lou is still alive but he last posted on QuadraphonicQuads forum on 21st November 2012 http://www.quadraphonicquad.com/forums/showthread.php?9035-Lou-Dorren-A-new-CD-4-Demodulator!!!&p=165519&highlight=#post165519 and other members of the forum still talk as though he is. I think the updated demodulator project stalled, there's speculation on the forum around that, Lou's health and employment status. The schematics have also been published, I wonder if anyone has built a working example?

Unfortunately we don't have access to a CD-4 demodulator so bench testing at the moment is not possible for us; maybe someone out there would like to help with that? We are using some short clips from a test disc recorded before and after demodulation by someone in the US and some samples captured by my self with out CD-4 demodulation.

With regard to your interest in the development of a CD-4 modulator for cutting new disks, that had also crossed my mind. We are already experimenting with software generated signals to test and calibrate our demodulator as they are devoid of interference and distortion inherent found in the analogue system; it's use could also be extended for calibrating the analogue ones still in existence too.

Best Regards,
Tim

[markrob]

Hi,

I have a JVC decoder that I purchased on ebay some time ago for that purpose. You should be able to find one on ebay.

I've just been so swamped at work over the last year that I've put the project on the back burner. I was developing my DSP based decoder with Synthmaker (now Flowstone). I had custom coded some modules to do the dirty work. I was running at 96Khz sample rate, but it would be better to run at 192Khz. I need a audio interface that runs at that sample rate. I also do not have a proper pickup with a Shibata stylus, so I probably not recovering the carrier and sidebands with enough accuracy. I've considered capturing at 1/2 speed for development purposes, but never got around to setting that up. I have a setup and cal disk, but its pretty beat, so I was also thinking about writing some software to generate clean test signals.

http://www.dsprobotics.com/

I'll check out the forum in the near future.

Mark

[diamone]

more info here
https://lathetrolls.com/viewtopic.php?f=15&t=4096&p=27814#p24918

and here
https://lathetrolls.com/viewtopic.php?f=15&t=4350

FYI Even 192/32 only gives you nineteen samples of the carrier wave at the BOTTOM of its' range of 30KHz and less than FOUR samples at its' top range of 50KHz. The only people I have ever heard of that have been able to capture the VARYING i.e. FM modulated carrier wave with any kind of integrity is those people who are experimenting with PCM on HDTV.

Remember the original PCM standard was German - based on their 25 frame 50 Hz standard, three samples per NTSC line = 44.100 KHz. In the US (for those of you old enough to remember PCM LaserDiscs in the 70s and early 80's) three samples per line on NTSC = 44.056 KHz.

So you can go on the ATSC HDTV engineering forums and see how they're capturing the ultra-high frequencies without the compression (which is why you can't record an F1 two channel PCM recording made on a VHS or Beta in the 70's onto a DVD and decode it back to audio in the 2010's).

But you are absolutely going to need something A) professional/broadcast quality and B) capable of capturing video/instrumentation tape signals which are also in the same frequency range. Only THEN will you be able to get enough samples of the carrier wave to demodulate it in the software.

And even if you DO - the data storage for the by-necessity uncompressed signal going out to the MHz - is going to be ridiculous.

[Steve E.]

Remember the original PCM standard was German - based on their 25 frame 50 Hz standard, three samples per NTSC line = 44.100 KHz. In the US (for those of you old enough to remember PCM LaserDiscs in the 70s and early 80's) three samples per line on NTSC = 44.056 KHz.

Interesting!!!!

Hey, it's interesting as well to look at the factors in the two most common PCM sampling standards:

2x2 x 3x3 x 5x5 x 7x7 = 44100
2x2x2x2x2x2x2 x 3 x 5x5x5 = 48000

(4 x 9 x 25 x 49 = 44100)
(128 x 3 x 125 = 48000)

(I don't know how to render exponents here)...

[markrob]

more info here
https://www.lathetrolls.com/viewtopic.php?f=15&t=4096&p=27814#p24918

and here
https://www.lathetrolls.com/viewtopic.php?f=15&t=4350

FYI Even 192/32 only gives you nineteen samples of the carrier wave at the BOTTOM of its' range of 30KHz and less than FOUR samples at its' top range of 50KHz. The only people I have ever heard of that have been able to capture the VARYING i.e. FM modulated carrier wave with any kind of integrity is those people who are experimenting with PCM on HDTV.

Remember the original PCM standard was German - based on their 25 frame 50 Hz standard, three samples per NTSC line = 44.100 KHz. In the US (for those of you old enough to remember PCM LaserDiscs in the 70s and early 80's) three samples per line on NTSC = 44.056 KHz.

So you can go on the ATSC HDTV engineering forums and see how they're capturing the ultra-high frequencies without the compression (which is why you can't record an F1 two channel PCM recording made on a VHS or Beta in the 70's onto a DVD and decode it back to audio in the 2010's).

But you are absolutely going to need something A) professional/broadcast quality and B) capable of capturing video/instrumentation tape signals which are also in the same frequency range. Only THEN will you be able to get enough samples of the carrier wave to demodulate it in the software.

And even if you DO - the data storage for the by-necessity uncompressed signal going out to the MHz - is going to be ridiculous.

Not sure I agree with your statement. My digital FM demod seems to work very well at 96Khz witht the testing I've done. I don't think all of the direct digital radio projects out there would work if what you are saying is the case. I've attached the paper I based my demod on. Fig2 in the paper shows the block diagram I used for my implementation. Basically, you hetrodyne the 30Khz carrier down to a baseband signal and extract the I and Q quadrature signals. Then, cross multiply I and Q with the the derivative of the the cross term. Finally, you remove the AM component by calculating the magnitude (I^2 + Q^2) and you are good to go.

My entire decoder including 64 tap FIR filters (low pass and high pass) to isolate the FM and baseband audio, RIAA, FM,demod, AM/PM filter, and de-matrix only uses about 9% of CPU running on a i7 3Ghz pc in realtime in Flowstone. I still need to get the ANRS compandor designed and debugged, but it sure looks to be very doable. My biggest issue is getting a good program source. I currently have a Marantz CD-4 cal disc, but its not in good shape. I also have a few CD-4 releases as well. I don't have a good Shibata equipped stylus and a true CD-4 approved pickup. That, I feel is giving me some bad results. I also still need to evaluate the group delay between the two audio paths and, compensate for the difference.

Mark

[diamone]

My digital FM demod seems to work very well at 96Khz with the testing I've done. I don't think all of the direct digital radio projects out there would work if what you are saying is the case. I've attached the paper I based my demod on. Fig2 in the paper shows the block diagram I used for my implementation. Basically, you hetrodyne the 30Khz carrier down to a baseband signal and extract the I and Q quadrature signals.

Not a mathematician here, but I recall from my days in analog NTSC engineering that anybody who tried to de-matrix the I and Q (peach and purple for the uninitialted) or to even EXTRACT it e.g. to enhance the accuracy of skin tones since unlike PAL it was Never Twice the Same Color - had problems - and I would have thought the same de-matrixing issue would apply here i.e. to quote the title of a famous 80's album by The Police.

The only reason I could think of where that WOULDN'T apply is because the signal wasn't video but was a low-band audio instead - but then THAT doesn't make any sense EITHER because otherwise why would you need rotating heads and videotape (or in the early days - instrumentation tape) to record PCM.

Then, cross multiply I and Q with the the derivative of the the cross term. Finally, you remove the AM component by calculating the magnitude (I^2 + Q^2) and you are good to go.

Can you do the same to retrieve the bias signals off of tape and be able to use that as sync to genlock the playback deck and resolve the miniscule speed variations of the original analog recording?

http://www.plangentprocesses.com Engineer Jamie Howarth has been doing it for awhile - but of course it only works on original session tapes because the original bias tone gets erased and a new one is applied everytime a tape is dubbed across or down to make a mixdown or safety.

My entire decoder including 64 tap FIR filters (low pass and high pass) to isolate the FM and baseband audio, RIAA, FM,demod, AM/PM filter, and de-matrix only uses about 9% of CPU running on a i7 3Ghz pc in realtime in Flowstone.

I don't have half of that running to do my own conversions and what I DO have is a resource hog. You'd think I was trying to record 32K digital on an 8086 so I kind of backed off the conversions for awhile - much to the annoyance of everybody at the QuadraphonicQuad forums.

I still need to get the ANRS compandor designed and debugged, but it sure looks to be very doable.

Yeh I'm sure besides me all kinds of guys on QQ would like to be able to get better CD-4 and UD-4 transfers than we get now. Send me the review copy once it's all done.

My biggest issue is getting a good program source. I currently have a Marantz CD-4 cal disc, but its not in good shape.

The Fisher and the Enoch Light are both CONSIDERABLY better for that than the Marantz - both in mastering quality as well as pressing quality - but only if you can get something OTHER than the Columbia Record Plant pressings that are ubiquitous for both. (No they don't say Columbia on them any more than the Star Wars soundtrack on 20th Century Records - but just as you can tell that the vast majority of SW and Close Encounters and etc records were pressed at the CRP - you can tell just as easily with these what was pressed at the CRP and what was pressed at Capitol or WB - both of which are considerably better than the CRP editions.

And - they both come on Q-4 reel to reel - of which it is advisable to get copies so you have an idea what the original quadraphonic program was supposed to have sounded like once it has been recovered.

I also have a few CD-4 releases as well.

The only one I ever heard of that gives TERRIFFIC separation is the Elvis: Aloha From Hawaii on a BLACK `SPECIAL PRODUCTS' label - because - unlike the orange labels - the black SP label is pressed on 135-160 gram vinyl instead of the 70-90 grams of the orange edition the same as the non-CRP editions of the Fisher or the Enoch Light - the advantage of the EL being that it - like the Elvis - also has a Q-4 edition where the Fisher does not.

I went over five years looking both online and off before I found one - and then another three years before I found one in mint condition - and the difference even on an ELLIPTICAL stylus between a Dynaflex and this is noticeable even in STEREO nevermind trying to extract the quad on my DD5 demodulator. A 135-160 gram black-label CD-4 pressing of The Doors Greatest Hits or The Best of Bread also do alright as does the WIDE-GROOVE-PITCHED black label pressing of Judy Collins: Colors of the Day. As expected - the far-more-common narrow-pitched versions (that only go halfway to the label and have a long runoff groove) have a pinchy quality to them.

But as mentioned above - in all instances - also get a copy of the Q-4 so you can hear for what you are supposed to be listening when you demodulate.

I don't have a good Shibata equipped stylus and a true CD-4 approved pickup. That, I feel is giving me some bad results.

You can get away with an AT-14S with a burgundy Tetrahedral stylus in a pinch - rebuilt versions are all over eBay for a pittance compared to what the salons and boutiques charge.

Far more common - but a little more expensive - is the Shure V or Shure Φ both of which RELIABLY go out to 50K and both of which either some with a MicroLine/Nude Square Shank Hyper-Elliptic stylus (two of several modern names for a Shibata) - or some New-Old Store Stock versions thereof - can be FAIRLY easily located compared to the Burgundy 14S stylus - even though the latter often gives a better result for the same genre of mid-line cartridges.

It's all a matter of taste, but in my opinion - as well as those of a good many quadraphiles - the Shure - being built for strength and not speed - has a little bit of a tubbiness to the bass and gives a little bit of a megaphone-ish sound to the midrange - on ex-recording-studio Tannoy main monitors (i.e. not nearfield - which in my room are KEF tri-ax - i.e. the midrange is mounted inside the woofer and the tweeter mounted inside of that. This of course can be corrected with the proper tweaking - but even on my other set (Altec studio monitor A2 Voice of the Theatre being driven by McIntosh monoblock tube amps) minimizes that effect - it can still be heard even on period material.

I also still need to evaluate the group delay between the two audio paths and, compensate for the difference.

You have a four-channel Tektronix scope I imagine. The one I have can be set up so that the offset between any and any other is displayed on an LED screen at the bottom of the scope - from 1° to 359° - or on the newer ones either 1°-179° left or right - which can be switched to read `top and bottom'.

Looking forward to further developments.

[diamone]

I don't know how to render exponents here...

2² + 3² + 5² + 7² = 44100
2⁷ x 3 x 5³ = 48000

Well that didn't work. Maybe this.

http://thefiringline.com/forums/misc.php?do=bbcode

[diamone]

Although in order to save the LP I HAVE been able to record CD-4 real-time including the carrier wave onto half-inch helical-scan instrumentation tape (Ampex 777) on an old Viking from NASA I inherited - if I disabled the onboard audio-frequency modulation and basically told it that the AFM was coming in with the program data from offsite - just like a couple of NASA Whitepapers instruct.

Meaning you could probably do the same thing on an old 3/4 inch U-Matic 1630 by bypassing the digital encoder and the onboard AFM circuitry - and for that matter you could probably re-engineer a Betacam SP or maybe even a DA-98 to do the same thing since both the deck as well as the media is a hell of a lot easier to come across.

[whelm]

I've given this some thought, also, but so far been too busy with other things to go far. If you send me a privage E-Mail, I will give you some information I have collected, particularly related to ANRS. Look at the Lou Dorren thread mentioned previously as he has some information about the way he implemented it to avoid phase distortion problems.

I'm quite confident this can be done and done well. ANRS is the trickiest, but should be doable. I'm also in favor of sampling at 192 KHz, which is not hard with some of the Audio Codec chips out there today. I was considering a PLL approach to the FM demod, since that was what the analog solution used, but quadrature arctangent is also viable. I have used that before for rapidly measuring the error on sine wave tone frequencies (7 Khz to 0.1 hz resolution in about 2 ms).

contact me by my user name given here at compuserve dot com

[Dub Bull]

I strongly recommend you stay analog with your music signals - 100%.

Monogroove mono vinyl is hard enough to get right, so there's still a noble challenge awaiting enthusiasts, and this is where I feel we should focus our Helium-less futures.

Fear God; not the Church!
Father Jose'

[markrob]

Although in order to save the LP I HAVE been able to record CD-4 real-time including the carrier wave onto half-inch helical-scan instrumentation tape (Ampex 777) on an old Viking from NASA I inherited - if I disabled the onboard audio-frequency modulation and basically told it that the AFM was coming in with the program data from offsite - just like a couple of NASA Whitepapers instruct.

Meaning you could probably do the same thing on an old 3/4 inch U-Matic 1630 by bypassing the digital encoder and the onboard AFM circuitry - and for that matter you could probably re-engineer a Betacam SP or maybe even a DA-98 to do the same thing since both the deck as well as the media is a hell of a lot easier to come across.

Hi,

I haven't been able to respond to your PM's as you seem to have that function disabled. Thanks for the heads up and the additional info.

Mark