SQ decoding on the computer

QuadraphonicQuad

Help Support QuadraphonicQuad:

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

ArmyOfQuad

Guest
I think it's possible to decode an SQ LP in the computer. And I think with enough work, someday a software decoder could be built to be as good or better than a tate, because a software decoder would have the advantage of not having to be realtime. Anyways, I've started work on this possibility, and I'm getting results of something like a very bad non logic decoder. By studying test tones from a test record, I figured out that the rear info is on both channels, the right channel lagging by about .01s. In phase for right, out of phase for left. so, by adding .01s to the right speaker (this has to be just exact, it's not exactly .01, but it's not .011 or .009, I got the exact by measuring it with the testtone, and cutting and pasting that time), then using dartpro I added and subtracted, thus getting rear channels. Then, on a whim, I added and subtracted the rears and surprise surprise, I got fronts. The left to right seperation is decent, but the front to back seperation is about 3dB at the most on the test tones, so this is probably even worse with music. I tested it with a song, and it's kinda seperated, but not enough to be worth anything. This is probably what the first non logic decoders sounded like. Anyways, I've got a start here, if I could just get my hands on technical info on the SQ system, how a tate works so well, and the like, I could probably improve this method. I think we can all agree a software tate would be a great thing for all, except those who own one and want to make their money back someday on ebay. Anyone with suggestions, please post. Thanks.
 
You would think that a modern computer could blow the doors off of the best Tate II from the '70s if the software was written properly.

Good luck and keep us posted. Maybe QB could help.

:-jon
 
lol....adding and subtracting the rears just got me the original left and rights again. I should've done the math in my head earlier before posting something so stupid like that here. Anyways, Yeah, still sucks. Basically, adding and subtracting them boosts the proper rear and front info about 3 dB, but doesn't eliminate anything that shouldn't be there, so you get a sense of what sorta should be where. So, I have something far less then a non logic decoder. A non logic decoder would probably at least do a decent job on the test tones, I dunno, I don't have a non logic decoder. Anyways, I just wanted to point out my mistake on basically undoing what I did to get front channels that are really just undecoded SQ, there is much more to this format than what I've already discovered.
 
In theory Jon is right......it could be done.8o

But, it is not nearly as simple as what you've attempted Army.

You would need a program that can run algebraic equations and can also identify shifts in phasing of an audio signal.

Then you could program the equation, and run the SQ decode program on the signal. After that, it becomes a matter of dealing with cross-talk cancellation, and figuring out a means to increase the separation from the basic equation (ala the TATE process).

Unfortunately, there is very little information available on the TATE process.........basically nothing from Fosgate. Although I've had some success with chasing the patents.

As to the SQ equation......just let me know and I'll post it if you'd like. But, step one is being able to identify and separate the rear channels which are 90 deg. lagging both front channel signals in SQ......and it isn't "fixed" as a time constant.....it "rides" the front signals 90deg. out of phase.

QuadBob:cool:
 
QuadBob, sure, please post or send me all info you have on SQ. I guess I was way off. Anyways, thanks for your advice.
 
I may be off for a mile, but doesn't the Tate work by enhancing the separation after the four signals have been decoded by a basic decoder?

Maybe the right approach is to use a digitized four channel analog signal as the starting point, not the stereo tracks.
 
Well you're pretty much at least half right Proufo!

<blockquote><strong><em>Quote:</em></strong><hr>I may be off for a mile, but doesn't the Tate work by enhancing the separation after the four signals have been decoded by a basic decoder?[/quote]

Yes, the TATE process is a separation enhancement.

<blockquote><strong><em>Quote:</em></strong><hr>Maybe the right approach is to use a digitized four channel analog signal as the starting point, not the stereo tracks. [/quote]

Here, you miss the point that we're starting with an already "SQ" encoded source. So, the key is to figure out how to decode that original SQ material equal to or better than a TATE decoder. And, all SQ encoded material is in fact "stereo tracks" on whatever format it is......lps, cds, reels, etc.

QuadBob:cool:


 
<blockquote><strong><em>Quote:</em></strong><hr>Here, you miss the point that we're starting with an already "SQ" encoded source. So, the key is to figure out how to decode that original SQ material equal to or better than a TATE decoder. And, all SQ encoded material is in fact "stereo tracks" on whatever format it is......lps, cds, reels, etc.[/quote]

Hello QB.

What I meant is that once you have the four signals digitized inside the computer, you (I guess) don't need to worry about phase relationships. At that point you "just" need to check the content and play with volume levels.

I have somewhre scans of the Tate patent. I someone wants them I can e-mail them. Wil try to upload them to the files section in the Yahoo group.
 
To further clarify, what I propose is to digitize the four channels coming out of a hardware-based basic SQ decoder. As if you were making a DTS CD but using a basic decoder.
 
<blockquote><strong><em>Quote:</em></strong><hr>To further clarify, what I propose is to digitize the four channels coming out of a hardware-based basic SQ decoder. As if you were making a DTS CD but using a basic decoder[/quote]

The key here Proufo is that if you were to do it like this.....why use a "basic" decoder. You should use a TATE II........which is what everyone already does to do a DTS conversion from an SQ source.

The goal at the start of this thread was to "do away" or improve on the TATE II in the PC and decode the SQ material in the PC. So, you've walked in a circle right back to where we started.<img src=http://www.ezboard.com/intl/aenglish/images/emoticons/grin.gif ALT=" >D">

QuadBob:cool:
 
Well, if you say so.

If the Tate does include a basic SQ decoder as a starting point, one can try to emulate and improve the onward separation-enhancement analog algorithms in the digital domain, having a four-track recording as the starting point.

One can envision a method that would detect, measure and diminish pumping and swishing, by doing trials and correcting until optimum parameters are found. And this would be based on signal content --as the Tate presumably does and perhaps with the same "logic"-- but with the advantage of not needing to do it in real time, and being able to perform trial and error.

There could be easy ways do detect phase delay in the digital world, and then the basic SQ decoder wouldn't be needed at all.

But form the initial post, it seems it may be a task better suited to analog circuits.

My two cents.
 
I dunno, this seems too complicated for me to be able to do anything with it. Besides, ever since I got a Q8 conversion of Toys in the Attic and compared it to the Tate conversion, I can't help but think of SQ as a joke. I wonder who could've thought this could ever be a serious format. Anyways, I guess I give up for now.
 
I think you would want to do the basic decoding in the computer as well, as some of the pumping is an artifact of the slowness of the matrix circuit.

I can't help but think an improved hardware SQ decoder would be possible. The Tate was a first generation DES (directional enhancement system), and later Fosgate designs are just improved DES with a different matrix. It seems that someone with electronics knowlege could frankenstein a modern Fosgate unit with a SQ matrix or change the matrix already in the Fosgate to SQ spec instead of the stereo to surround matrix already in the unit. The DPL-II is the best DES system yet. I have a RFQ-5000, the only DPL-II unit made from discrete components. If I only knew enough about electronics it sure seems like it could be modified into the ultimate SQ decoder. Quadbob?

Dan in Spokane
 
<blockquote><strong><em>Quote:</em></strong><hr>I think you would want to do the basic decoding in the computer as well, as some of the pumping is an artifact of the slowness of the matrix circuit.[/quote]

Hello DS.

I don't believe there's any slowness in the basic matrix circuit, as there's no logic involved.

As you point out, a non-real-time digital logic can go back and compensate for the slowness of the logic and its parameters.

Re-reading the initial post, getting 3 dB of front-back separation is what is expected (or achievable) from non-logic circuitry. So perhaps a basic SQ decoder might be not useful after all. AOQ may have already all that you need as a pre-enhancement step.
 
<blockquote><strong><em>Quote:</em></strong><hr>Re-reading the initial post, getting 3 dB of front-back separation is what is expected (or achievable) from non-logic circuitry. So perhaps a basic SQ decoder might be not useful after all. AOQ may have already all that you need as a pre-enhancement step.[/quote]

I didn't really have 3dB. I didn't do a phase shift in my decoding experiment, I did a time shift, which worked on the test tones since they were steady tones. But, that's not how SQ works, so if I were to try it with music, it would do very little.
 
Cool Edit Pro version 2 features an exceptional wide bandwidth phase shifter, perfect for decoding matrixed Quad and Dolby Surround

my only problem is, every time i attempt to find the CORRECT formula for SQ encoding/decoding i find different equations!!!!!

does anyone know for CERTAIN the correct phase shifting formula for SQ???

with the correct formula, you could also digitally create NEW SQ recordings from 4 discrete tracks, as well as your own Dolby Surround encoded recordings.

Dolby Surround is easy to passively decode using any wave editor:

Lf = Lt
Rf = Rt
C = Lt+Rt
Surr = Lt-Rt

note: you will get much better results from a Dolby ProLogic setup with steering logic

Dolby Surround is very simple to encode:

Lt = Lf + [+90] Surr + (-3dB)C
Rt = Rf + [-90] Surr + (-3dB)C

note: the surround channel is mono, present in both L and R but with a +/- 90 deg phase shift

those of you with Cool Edit Pro 2.0 who produce your own multitrack recordings might like to try encoding your mixes in Dolby Surround:

first make a normal stereo mix. (don't mix it down tho)
then select the tracks you want present in Lf and Rf and mix them down in stereo.
then select the tracks you want present in the center channel and mix them down in mono.
then select the tracks you want present in the surround channel and mix them down in mono.
then convert the mono surround track to stereo and phase shift the left channel +90 degrees and the right channel -90 degrees.
finally, create a new multitrack session and import the Lf/Rf track, the Center track and the Surround track
then all you need to do is set the volume of the center track to -3dB and mix the lot down in stereo!!!



 
Well, you people are braver than me, that's for certain!
The correct (as far as I know) encoding equations for SQ was:

Lt = Lf + 0Rf - j0.707Lb + 0.707Rb

Rt = 0Lf + Rf - 0.707Lb + j0.707 Rb

(j indicating that rear channel is phase shifted by +90 degrees in the once channel, -90 in the other)

There's no reason why a modern decoder shouldn't better the Tate II -technology isn't a problem, but SQ was tricky to get the best out of. You never knew what information would be lost in the encoding for a start. That said, with logic, it still gives Pro Logic II a run for it's money in synthing surround from stereo. After all, that's all it ever did. Whatever the encoding, the material is still stereo.
Oh, and bear in mind that the Tate II didn't ride gain at all -unlike other logic decoders. Rather, it constantly shifted the matrix co-efficients while decoding.
 
Back
Top