Multi-Matrix Quadraphonic Decoder for Plogue Bidule

[S T Riddle]

Obviously a tough work !
Have you tried to work with the guys at VLC ?
They usually implement any codec they can in their universal software...
It would make a very well spread matrix decoder engine, for all

Sorry for slow reply but due to vacation, birthday celebrations, bad internet connection, and current rework of Multi-Matrix I've been pretty busy.
I'm not familiar with VLC but since the design is mostly pure math a clever C+ programmer could probably implement it. However it would probably be prudent to wait for some redesign and efficiency improvement before this took place. Currently the "logic' functions are only implemented in SQ and QS and may be extended to other matrices in a later build.

 

[S T Riddle]

Intriguing...so if I digitized some SQ vinyl, the pre-processor would allow the stereo file to decode better via the Surround Master as well as your decoder?

Would there be an advantage to using the pre-processor even for standard stereo captures?

Thanks!

The inventor of "Shadow Vector" SQ, Lynn Olson, said that he created a pre-processor to inject a controlled amount of positive or negative feedback to cancel crosstalk between channels and found it "necessary" to achieve good separation results even when using CBS's own SQ test records. The pre-processor of my layout is an attempt to add that function. Using it I have been able to achieve around 20dB separation using a noisy, popping, and poorly recorded short channel ID "Test Record" file circulating around the internet. Before pre-processing the separation was terrible.

It also added the ability to synch and balance the 2 channels for a file which you didn't create that may have been improperly recorded or aligned in time.

Used carefully it is possible to improve the performance of any decoder, SM included. Be aware that when adjusting the timing between the channels the measurements are in milliseconds and at 10,000 Hz a change of 0.1 mS will change it a full cycle and the adjustable range is + or - 0.4mS. Obviously unless the file is REALLY messed up a very slight change would be necessary. Pay more attention to to the lower frequencies (eliminating a center bass or vocal from the rear) first then fine tune for higher (snare or cymbal).

If you make your own rips of SQ material here's what I would suggest. Take a good mono record, play it while monitoring the rear channels and carefully adjust for no output from the rears. Note these settings and use as a good starting point for your next SQ rip. The resulting file should then be pretty well balanced and phase aligned (at least in time, not phase angles of each channel - that's a different problem).

Give it a try and let me know if it helps.

 

[atrocity]

If you make your own rips of SQ material here's what I would suggest. Take a good mono record, play it while monitoring the rear channels and carefully adjust for no output from the rears. Note these settings and use as a good starting point for your next SQ rip. The resulting file should then be pretty well balanced and phase aligned (at least in time, not phase angles of each channel - that's a different problem).

Give it a try and let me know if it helps.

Great, thank you! I'm back at work after a week off, so I'm not sure when I'll get to this, but I'm very interested.

 

[along47416]

Hello S T Riddle,
Thank you so much for the wonderful work you have done on your multi matrix decoder. This is truly an exciting development.
This is my first time posting to any QQ thread since I joined the group a few years ago. I am rather surprised about the small activity on this thread, so I've decided to answer your call for suggestions.
First a little about myself, I am a retired former IT professional from business processing. I have no scientific programming experience in DSP. Also, I have no experience in the recording industry. I became interest in the surround sound technologies in the early 70s and have been following the development of these technologies since. I have listening experience with several of the components described on this forum including the Tate II and lately the Surround Master. I also have experience with the Adobe Audition scripts. Like you, I have been intrigued by the ideas of Lynn Olsen and have read through the original patent papers for this technology long ago.
Let me walk you through the results of a couple of tests I have completed with your decoder. First of all I have obtained the SQ and QS Test Tones from the website...
http://www.quadraphonic.info/SQandQS/
and after learning the basics of the Plogue Bidule platform, have processed the test tones through your decoder. I am assuming that these tones are accurate as is claimed on the website.
The first test was of your version of QS Vario Matrix. The results are shown from the screen print using Audacity.

The test shows the signal processed through 8 points of the circle. Viewing from left to right the test tone is decoded for CF, LF, LS, LB, CB, RB, RS, RF, and finally returning to CF. I did not change any parameters, which you can adjust, but used the default settings. The channels shown from top to bottom are LF, RF, LB, and RB.
Some wonderful things are happening here.
1. CF power is reduced to around .707 , that magic number, in each of the LF and RF channels. There is no crosstalk in the rear channels.
2. LF is completely isolated at full power.
3. LS is shown at a reduced power level with some cross-talk to the channels of the right side.
4. LB is completely isolated.
5. CB is shown at the reduced power level with no crosstalk to the front channels.
6. RB is completely isolated.
7. RS is shown at a reduced power level with some crosstalk to the channels of the left side.
8. RF is completely isolated.
9. CF is shown similar to its original description (see number 1 above).
Bravo. This was only one test and involved no listening on my part. I know from other listings on this forum, that the phase of the output is also important in establishing a clean image. I have no knowledge on how to test that. Perhaps one the other readers will step forward.
I would hope to see similar results for the SQ test. A screen print for that test follows.

First I would like to draw your attention to the fact that the test tones travel in the following order: CF, RF, RS, RB, CB, LB, LS, LF and finally CF again. (Note this is the opposite direction from the QS example)
You have produced a fine start for the SQ shadow vector. Again I used the default settings for the parameters in your component. Here are some comments.
1. CF is produced at full power in both the left and right channels. It needs to be reduced to the .707 is each channel.
2. RF is full power with little crosstalk, excellent!
3. RS test tone is produced reduced (to .707?) in the RF channel, however the RB is still at full power and also needs to be reduced to .707.
4. RB is full power with little crosstalk, excellent!
5. CB is produced at .707 power in both LB and RB channels, excellent!
6. LB is full power with little crosstalk, excellent!
7. LS test tone is produced reduced (to .707?) in the LF channel, however the LB is still at full power and also needs to be reduced to .707.
8. LF is full power with little crosstalk, excellent!
9. CF is produced at full power in both the left and right channels. It needs to be reduced to the .707 is each channel.
All in all, an excellent effort. You are well on your way to a fantastic implementation. There have been many theories and products tackling this idea and are you are providing an excellent solution. Of course a lot of further testing is desired. Together with the insight of others, I know you will conquer this dream. Thank you.

 

[JonUrban]

Welcome to the forum along47416. I love it when members post info like this. It's great. And on your first post as well. Great job!!

I was going to do something like this as well when I got more time, and I still might, but what you did is just what the doctor ordered, so to speak. Very nice!

 

[Colin Dunn]

Welcome to the forum along47416. I love it when members post info like this. It's great. And on your first post as well. Great job!!

I was going to do something like this as well when I got more time, and I still might, but what you did is just what the doctor ordered, so to speak. Very nice!

I don't see the post you referenced here. Did it go missing?

 

[JonUrban]

Ooops! You should see it now

 

[Colin Dunn]

As a point of comparison with the Adobe Audition script methods, I ran those same test tone files through the best Audition scripts that have been made available in the past. For QS, that is the QS Lucanu 2013 script. For SQ, it is the OD SQ*Final (/I) script.

At first, when looking these over, I was confused until I realized the order of channel panning is different between the SQ and QS test tones! I will recap the channel order in my analysis.

It appears some of these issues of 70.7% power for center images (vs. 100% in both channels) also occurs in the Audition script decoding methods. I'm not sure if this is a limitation of the two matrix formats, or a decoding error. Perhaps others who know more about this than I do can help interpret the results? Attachments are screen shots of the decoded files shown in Audacity, similar to what along47416 did earlier.

QS Lucanu 2013:

A general comment: Something about the decoding process adds a slight flutter/warble artifact to the test tone. I'm not sure what causes this, though I think it is also audible with recordings that expose wow/flutter artifacts.

Tone Sequence is:
Center Front - Looks correct with no crosstalk.
Left Front - Correct, but with crosstalk of approx. -15 dB in the front right and surround right channels, and -12 dB in the surround left channel.
Center Left - Should be 70.7% power in front-left and surround-left channels? Crosstalk of approx. -12 dB in the front right and surround right channels.
Rear Left - Correct, but with crosstalk of approx. -15 dB in the front right channel, and -12dB in the front left and surround right channels.
Center Rear - Looks correct. No crosstalk.
Right Rear - Correct, but with crosstalk of approx. -12dB in the front right channel, and -15dB in the front left and surround left channels.
Center Right - Should be 70.7% power in front-right and surround-right channels? Crosstalk of approx. -12dB in the front left and surround left channels.
Right Front - Correct, but with crosstalk of approx. -15dB in the front left and surround left channels, and -12dB in the surround right channel.
Center Front - Looks correct with no crosstalk.

SQ*Final (/I)

Tone sequence is:
Center Front - Should be 70.7% power instead of 100%? Looks correct with no crosstalk to the surrounds.
Right Front - Looks correct. No crosstalk to the surrounds or front-left channel.
Center Right - Should be 70.7% power instead of 100%? Crosstalk of about -10dB to the front-left and surround-left channels.
Rear Right - Slightly less than 100% power but this is likely negligible. No crosstalk to the fronts or surround left.
Center Rear - Looks correct (70.7% power) with no crosstalk to the fronts.
Left Rear - Looks correct with no crosstalk to the fronts or surround right.
Center Left - Should be 70.7% power instead of 100%? Crosstalk of about -10dB to the front-right and surround-right channels.
Left Front - Looks correct with no crosstalk to the surrounds or front-right channel.
Center Front - Should be 70.7% power instead of 100%? Looks correct with no crosstalk to the surrounds.

 

[GregHouse]

Hi.
Is it possible to use this layout in realtime using Bidule's VST in foobar, for example?