PC Based SQ DECODING - ALMOST DONE!

QuadraphonicQuad

Help Support QuadraphonicQuad:

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

QUADradial

Senior Member
Since 2002/2003
Joined
Mar 3, 2002
Messages
238
Location
Near Lexington, KY
OK, this one comes from Andreas Bergbauer...from Germany. I have tried this and it works amazingly well, but it needs final tweaking. Let's get this done, guys! Best of all, it's all done in Adobe Audition with no additional plug-ins.

His instructions are as follows...


1. Phase change the original summary signal by -90 degrees each channel and exchange the channels (left gets right and vice versa)
2. Mix this "temp wave" with the original summary signal 1:1, out comes the
"no logic" decoded signal for the rear speakers.

Now comes the "logic" trick:

3. Remove the front center signal (0 degrees phase), the front L+R parts
(now 90 and 270 degrees phase) from the rear (logic-free) signal by a filter
called "center channel extractor" from Adobe Audition (formerly Cool Edit
Pro). That's it ! Even if the filter is not set too agressive, it is
possible to decrease these signal parts by -30 to -40dB each! Out comes
(after 30 minutes calc time at an P4 3 GHz) a rear only signal as it almost
had to be before encoding!

4. Now I do something that I don't know if it is allowed: I also use these
90 and 270 degrees filters, but not the center filter of course, to remove
the "back" parts from the front signal. At the end I receive four channels
as they used to be before the encoding process, as I think.


OK...that's the basis.
Now, I tried it and it was close...but the rear center information was still leaking into the front.
I tried this modified approach...

Open the wav file in Multi-track view. Ctrl + right click on the mouse and drag the mouse into another track. This creates a 2nd copy of the original wav file.
Open that file in the edit view. Then go to effects/filters/graphic phase shifter and apply a -90 degree shift with the Channel set to "both."
Then open Effect/Amplitude/Channel Mixer and apply a "Swap Channels" to the file.
Return to the multitrack view and create a mixdown to an empty track of both the original AND the newly modified copy. Open that file in the edit view.
Now...go to effects/filters/center channel extractor. Chose "Center" "Full Spectrum". Now...here is where the tweaking comes in. Try dropping as much of the center channels as possible before it starts sounding weird. Try dropping -40db, crossover 25, phase discrimination 1, Spectral Decay 70%, Amplitude Discrimination 3 and Amplitude Band Width 10. These are rough settings and will probably sound a bit watery. Then go back into the center channel extractor and choose custom 90 degrees/full spectrum/ and run it thru again. Go back thru it one more time and choose custom 270 degrees. These are your new rear channels! To me, they sound better than a Tate, but a bit watery. Play with the settings!!!
Save the file as rear.wav

Now...open the original SQ wav file in multitrack view. Open the center channel extractor once again and run the custom setting once at 270...then again at 90. I have tried -20db with a crossover of 35, phase disc of 2, amp disc of 3, spectral rate of 70% on both passes. Probably way too much, but I've just begun experimenting. Now...run the channel extractor one more time and run a custom degree of 180. Now, the rear left, right and center are removed from the front channels.

OK guys...play with this and post your settting results!
 
Last edited:
QUADradial said:
OK guys...play with this and post your settting results!

Wow! I'd be interested in everyone's results who have Audition and can experiement with this. I don't know how you guys know what parameters to tweak. Is it because you know exactly how the TATE II reconfigures the signals to extract rear channel information?

You would think that the computing power capability and sophistication of today's software would be able to emulate or exceed whatever the TATE II and the QS Vario-Matrix does, considering both work off a set of digital instructions. Consider the possibilities.....copy analog matrixed LPs into the digital domain, create multichannel sources that are as close to the original multitrack tapes and preserve them onto DVD-A. :rolleyes:
 
I'll give it a try in the next few days! Thanks Tab!
 
The key is to not overdo the settings. I could only try this at work thru PC speakers. Once I got home, I realised the settings are too strong. Try narrower crossover settings
 
Alas - AA 1.5 does not support ASIO protocols, and it will not recognise my HDSP 9652 as a multichannel device.
I'll give it a go, and let you know how it comes out, but I fear it will not be an easy task.
 
I've ran some test using the last 30sec of "Theme from shaft" by Isaac Hayes - it has a wonderful hihat that is supposed to stay firm on front right and a guitar that swirls around all 4 speakers, while having also a orchestra boast every now and then.
So far these are the best "decoding" setting i've found - ran all against the front and rear channels.
There's a bit of leakage of center front to center back but i'm not sure if it depends of the "carrier" (=LP, which has no discrete separation between L-R).
Another one that decodes well enough is the intro of "For the love of the money" by O'Jays.
Tab, could you post somewhere a 30-sec mp3 of Chase "Open up wide" *from the SQ cd* so L-R separation is not a issue? The part you used for "what's the logic in that?" page.

Here's the settings used:
Center channel level: -48
Crossover: 100 (max narrow)
Phase discrimination: 6
Amplitude discrimination: 0
Spectral decay rate: 90%
Amplitude band width: 16db
FFT: 16384
Overlay: 8


Avoid at all any value different by 0 on Amplitude discrimination - the "watery" effect comes mainly from that.
 
Last edited:
I will have to encode it and send it to you. I'll have to dig around for the chase CD. :)

What do you think of the results from the process?
 
A different request: instead of mp3, a 30-sec FLAC file will be better.
 
QUADradial said:
I will have to encode it and send it to you. I'll have to dig around for the chase CD. :)
What do you think of the results from the process?

The setting you postes were giving slightly waterly sound... the one i've experimented and found, afer wading thru some "ocean eleven" effects :) are a decent improvement.
 
Aldo,
Your settings are very good, but I would use 70 - 80 for the crossover. What just came out of my speakers makes the Tate sound bad!!! This is really something.
How can I get the Chase file to you. The file will be 20 to 30 megs or more. Can I upload it to the alt.binaries.sounds.dts area in the newsgroups?
 
QUADradial said:
Aldo,
Your settings are very good, but I would use 70 - 80 for the crossover. What just came out of my speakers makes the Tate sound bad!!! This is really something.
How can I get the Chase file to you. The file will be 20 to 30 megs or more. Can I upload it to the alt.binaries.sounds.dts area in the newsgroups?
Hello this sound vary promising will it run on a MAC?
Thanks for the work in this area.
 
hey,

what version of Adobe Audition are you using?

how much memory do you have?

what speed chip?

what multichannel audio card are you using?

thanks for all the work and info.

terry
 
QUADradial said:
Aldo,
Your settings are very good, but I would use 70 - 80 for the crossover. What just came out of my speakers makes the Tate sound bad!!! This is really something.
How can I get the Chase file to you. The file will be 20 to 30 megs or more. Can I upload it to the alt.binaries.sounds.dts area in the newsgroups?

I can't access newsgroup from here. Got to find out some web space... i'll keep you posted.
I've tried a couple more files and what suprises me is that i can't hear any form of "pumping". That seems to be good.
 
Hi folks,

first I have to introduce myself:

I am that Andreas Bergbauer QUADradial mentioned in his post.

Some additional ideas came up in me, I have to get rid of them...:

SQ mostly comes from vinyl sources that have no complete channel separation. That's why it is not possible to remove all leaking parts without influencing the remaining ones.
The quality of the playback equipment for the old SQ records and quality of the records themselves is the most unstable factor in the decoding process, that is why EVERYBODY HERE WILL FIND OUT DIFFERENT FILTER SETTINGS.
The more separation is desired, the more the settings must be tweaked individually for each used source.

This is why THE TATE HAS KNOBS!

Regarding the "logic" of SQ: THERE IS NO LOGIC in all analog decoders in my opinion! There is (was) only an analog approach to separate those phase shifted signals and only maximizing the separation when absolutely necessary by loosing sound quality because both in one time was not possible.
So the logic is only balancing between channel separation and sound quality!
That is where the "pumping" of old decoders came from: It can not decide what to do, separate or not.
One big problem was the basic SQ decoding: Shifting an analog stereo audio signal by 90 degrees linear over the whole audible frequency range is not that easy by technology of the 70's! With that basis the following processing can not be perfect!
The famous Tate seems to have much better phase shifters and phase filters than the rest of the common decoders of the 70's what proofs that also analog decoding was almost perfectly possible! For that a high quality manufacturing process with narrow selection of the parts is needed what let the prices explode.

With today's digital filters the decoding process is even possible with a standard PC and a good program like Adobe Audition. Does anybody know a simular plugin of annother program to remove signals at a specified phase? It may even work better!

So: Is there anyboy out there with the capabilities of programming a VST plugin that does the job, now after the "secrets" are reveiled ? One with knobs? A kind of Tate rebirth?

I decoded the complete DSOTM from Pink Floyd and fed my Marantz 4300 with the discrete channels from my PC-after 5 minutes I decided never to use my SQA-1 decoder again!

Once infected, I now have to decode everything I have in SQ. That is the virus!

Thanks for following me so far and thanks to QUADradial to invoke this forum with the topic.
And thanks for reading my rough English....

So my last provokation for today:

Only today's high end analog equipment should be used for the playback of the old SQ vinyl records if the intention is to convert them to modern digital formats. If the basis is bad, the result can not be perfect!
 
Thanks for joining and posting Andreas. This afternoon, I am going to try this myself.

I was thinking along the lines of your posts, how do we know how well these SQ discs are encoded, or how discrete the mix was in the first place? I never liked SQ back in the '70s, I was always a CD-4 guy, because the seperation was there. Even when the Tate appeared, the Q8 of an SQ title would always have better seperation, but not fidelity.

I think I'll try a Paul Simon cut and see what happens. I'll report back here later today.

Thanks again to you and Tab for this insight!
 
JonUrban said:
I was thinking along the lines of your posts, how do we know how well these SQ discs are encoded, or how discrete the mix was in the first place?

That is the most intriguing question. Some of the SQ mix can be questionable; so far i've tried, with the setting i've posted, these:
Isaac Hayes: Theme from Shaft (SQ track from LP)
Tower of Power: Ain't no stoppin us now (SQ track from LP)
Santana: Lotus (SQ-encoded CD)
Nomadi: Interpretano Guccini (SQ-encoded CD)

Results:
Hayes: slightly unbalanced L/R and thus decoding has some blurry on the -30dB range. Nice separation
Tower of Power: that was a better recording and the separation and balance between the 4 channels is stunning.
Santana: Lotus: you can't believe how good this can be... try it.
Nomadi: that was a long shot, because in origin (1974) this album was released as SQ-Only, so i tried the CD edition... and it decodes fairly well, especially track 2 which has a long intro with only strings in front *and* back. If there were phase problems the sound would had come like crap. This is not the case. However, there is some weird sound on the back, to me sign that the sq encoding has been poorly done from the start. Nevertheless good, and THE FIRST TIME i can listen something italian in QUAD as it was.

For testing purposes, i've tried a couple of other non-sq tracks with the SQ settings to check out how they sounded. Very very very phasey back channels. These settings are NOT good for a non-SQ recording because they relies strongly and nearly only on phase relationship, so there's still room for creating a software QSD-2!
With this settings, if the encoding was ok, you must get clean sound on the back channels - the artifacts should be only on low level and they depends IMHO of the small separation at the start with the LP format, so YMMV.
That's why i asked Tab for the OpenUpWide sample from the SQ-encoded CD. This should decode perfectly - at least in theory... :)

I'm on Sq nirvana now... Lotus quad is sooooooo grooooooovy....
 
34 sec of Chase, decoded.
 
Last edited:
JonUrban said:
Thanks for joining and posting Andreas. This afternoon, I am going to try this myself.

I was thinking along the lines of your posts, how do we know how well these SQ discs are encoded, or how discrete the mix was in the first place? I never liked SQ back in the '70s, I was always a CD-4 guy, because the seperation was there. Even when the Tate appeared, the Q8 of an SQ title would always have better seperation, but not fidelity.

I think I'll try a Paul Simon cut and see what happens. I'll report back here later today.

Thanks again to you and Tab for this insight!


The problem is, that back in the '70s as SQ was released, I was just born that time...
Of course a complete discrete source was(is) the best solution because the recording engineer did not get crazy about all the phase relations between the channels and the bad result that can come out if mixed wrong. On discrete material there is also the possibility of producing virtual stereo images between every pair of channels which SQ can not!
Q8 players were not common here in Germany, they are hard to get.

But:
SQ was the most common Quad format ant the most recordings exist in SQ. This was the reason, why I tried to decode that first.
QS and Dolby Surround should be also no problem at all, I think.

So, try the thing, it is worth it, believe me!
 
Guys,

Can you "dumb down" the directions a little? I am not sure where to start. I have a recorded encoded SQ tune in AA 1.5, but I am not sure what you mean in steps 1 and 2. I looked and searched for "Phase Change", and all I came up with is "generate tones"!

Although I have CEP2.1, AA1.0, and AA1.5, I usually use SF7, just because that is the program that I started with. So I need a bit of hand holding here with AA.

If you get a chance, could you post a bit of "step by step"?

THANKS

:-jon
 
JonUrban said:
Guys,

Can you "dumb down" the directions a little? I am not sure where to start. I have a recorded encoded SQ tune in AA 1.5, but I am not sure what you mean in steps 1 and 2. I looked and searched for "Phase Change", and all I came up with is "generate tones"!

Although I have CEP2.1, AA1.0, and AA1.5, I usually use SF7, just because that is the program that I started with. So I need a bit of hand holding here with AA.

If you get a chance, could you post a bit of "step by step"?

THANKS

:-jon

Jon,
I scanned the princip of the basic SQ decoding (what is done in the steps 1 and 2) that You can see how the SQ princip works.

It means that a special helper file has to be mixed with the original SQ summary file to get the basic file for the rear channels BEFORE removing the phase related things.
This helper file has to be a -90 degrees (each channel!) phase shifted original signal with swapped right and left channels.
This file has to be mixed down with the original summary SQ file as You can see from the schema I attached here.
I will try to make a complete documentation with screenshots from AA1.5, but they will be in German (I have no AA1.5 in English).
This will need some time...
Now I will try decode the Tubular Bells boxed version first, I need motivation...

Andreas
 

Attachments

  • BasicSQDecoding.jpg
    BasicSQDecoding.jpg
    30 KB · Views: 1,313
Back
Top