Talk:Matrix decoder

Question
does matrix decoding involve utilizing 2 stereo channels being "mirrored" and separating offsets on the other channels? (unsigned comment from anon) —Preceding unsigned comment added by 69.209.157.150 (talk) 13:41, 4 Febuary 2006 (UTC)


 * No. For example, UHJ decoding produces W, X, Y channels from Left, Right. Martin.leese 22:33, 6 January 2007 (UTC)

Merging with Matrix (sound recording)
The two pages are complementary in that one should describe encoding and the other decoding. (Although that doesn't seem to be the way they have been written.) I don't have strong views on merging, but feel strongly that the merged article, if there be one, should retain the "2:4" notation used in Matrix decoder. Martin.leese 07:52, 7 January 2007 (UTC)


 * I am of the opinion that there should be two pages, namely 'audio matrix decoding' (not 'decoder') and 'audio matrix encoding', each of which should have proper links to the other. References to applications for the technology (such as sound recording) should appear on the page(s). Note that pages already exist for encoder and decoder. Novatek 19:56, 20 March 2007 (UTC)


 * HairyWombat, you undid my addition of SQ Matrix Encoding, which, in hindsight, I agree with - we should either split the topics out into matrix encoding and decoding or just do a general 'surround matrixing' topic with separate sections on the complete encoding/decoding equations for individual systems. Ty Chamberlain (talk) 13:31, 18 May 2009 (UTC)

If you had changed all the headings then I would not have undone. As for what to do with this article, please see my comments under below. Note that when I wrote this, the SQ section was as uninformative as the rest. Why don't you carve out SQ into its own article (called Stereo quadraphonic) and delete Matrix decoder? HairyWombat (talk) 15:49, 18 May 2009 (UTC)


 * Ah, OK, gotcha - thanks for telling me "why" - I'll remember that in the future. I'm working on, but haven't posted yet, an in-depth article on Lynn Olson's (never produced) advanced Shadow Vector SQ Decoder, so I'll start working on SQ too so it can be a really complete source of info on the SQ system -encoding/decoding/history/development, etc... - there is SO MUCH incorrect info on the net about SQ (and quad/surround history in general) - some of it from people who were 'there' at the time and should know better.  BTW, HairyWombat, how much do you know about quad/surround history and formats? I ask because your additions/deletions about the Ghent microphone made me dig deeper and (initially thinking you were wrong) I found that you were 100% correct - development of the SoundField and Ghent is quite "obscure" really, from a standard quadraphiles knowledge, so I've been really impressed.  —Preceding unsigned comment added by Disclord (talk • contribs) 17:24, 18 May 2009 (UTC)

Although I was around in the early 1970s, and was interested in the various quad systems, I have no special knowledge of them. I did not purchase one, making do with a Hafler Circuit until the format war had been resolved. (Smart move on my part.) I do know Ambisonics (which is not a quad system); I created and maintain the FAQ on this. Hence, I know well the development history of the Soundfield mic, and knew it was entirely separate from SQ. However, I did not buy an Ambisonic decoder until 1994. The detail of the Gwent mic I had to research to edit your contribution on it. I was never a fan of SQ; its locus on the Scheiber Sphere was too twisted. Note that Ambisonics, unlike SQ, is alive and well with over 200 contemporary pieces available for free download at Ambsionia.com.

If you get your SQ article up and running then I will insert a link to it at Quadraphonic sound using. At that time, I will also trim further the SQ section. People keep trying to expand it and I keep trimming it because the article is not about SQ, so the SQ section needs to be about the same length as the other sections. A separate SQ article will give such people more freedom. By the way, don't forget to always sign contributions to Talk pages using ~, like this HairyWombat (talk) 23:59, 18 May 2009 (UTC)


 * I guess I had a 'brain fart' yesterday because I thought I did sign my 'talk' contribution using the four tildes - of course, I was up all night before that with flu, so it probably affected my memory... I have a truly beautiful sounding Ambisonic G-Format DTS CD (a Jeff Silberman Soundfield recording done for DTS in 1996), but for the life of me, I've never understood the attraction of the 2-channel UHJ "BHJ" format - in either undecoded 2 channel or fully decoded playback, I find UHJ to be completely unlistenable - to me, decoded UHJ is as bad as non-logic SQ in terms of phasiness and lack of localization - and the CF phase error makes it even worse than non-logic SQ.  This is all completely personal opinion, of course, but I'm very familiar with all the psychoacoustic advantages of the Ambisonic 2-channel format (and have read the various Gerzon, Sommerwerck, Leese, Fellget - and others - Ambisonic UHJ papers) but I've never found the 2-channel format to 'work' in real-life - now, in the 'discrete' B-Format/G-Format, Ambisonics works very well and is stunning - but UHJ, forget it. Please don't take that as an attack on you - UHJ advocates can be very rabid in their defense of the format.  I wish some company had produced a variable-preference UHJ decoder - I believe the BBC strongly advocated for their use, but none of the UHJ patent owners ever bothered to produce one. I apologize for asking about your quad knowledge - I didn't realize who you were until you mentioned the Ambisonic FAQ, which I've read many, many times.  Where in the world does the name "HairyWombat" come from?  "Disclord" comes from the name of a VHD Video Disc player I got from Japan - the VHD format was never released in the USA and Panasonic called their VHD players "Disclord" - JVC called theirs "DiscWorld." Maybe I should have kept my 'old' AOL screen name of "Stereoboy" for Wikipedia?  They (Wikipedia) need to add a user preference to automatically sign all "talk" contributions. Ty Chamberlain (talk) 14:39, 19 May 2009 (UTC)

UHJ was optimised for the average listener using the results on perception of phaseyness carried out by the BBC. Unfortunately, nobody is average so the results vary. There were plans for a "logic" UHJ decoder in the 1970s, but Ambisonics died before it could be brought to fruition. These days, given multi-channel DVDs, nobody is interested in developments for UHJ. "HairyWombat" came about because I used to work in satellite image processing. Somebody published a paper on the detection of hairy-nosed wombats from space, and the name stuck in my mind. HairyWombat (talk) 17:44, 19 May 2009 (UTC)

See also Discretedored?
Is Discretedored even a word? I did a google search and the only page it returns is this wiki entry, not even one other suggestion. Possible mistype of some form of audio encoding or strange vandalism? —Preceding unsigned comment added by 69.241.82.132 (talk) 09:19, 4 January 2008 (UTC)

Dolby Surround and Pro Logic should be the same
The Dolby Surround matrix and the Dolby Pro Logic matrix should be the same. I don't recall which of the two presented is the correct one, and so cannot fix the article. But they should definately be the same matrix; one of them is definately wrong. HairyWombat (talk) 13:50, 17 April 2009 (UTC)


 * I think the second set of equations is correct (i.e. sqrt(1/2)). Dolby Laboratories received US Patent 5,291,557 (http://www.google.com/patents?vid=USPAT5291557), which has this equation in claim 6.  —Preceding unsigned comment added by Beavioso (talk • contribs) 23:05, 12 May 2009 (UTC)

Yes, the encoding for both Dolby Surround and Dolby Pro-Logic is exactly the same - both use the "Dolby MP" Matrix for surround encoding. Dolby MP employs the same matrix encoding equations as Peter Scheiber's original matrix patent that started the "matrix quad wars" (US Patent #3,632,886). While Dolby MP was developed independently of Scheiber's matrix, they later paid royalties to him which stopped when the patents expired (Scheiber now collects no royalties from Dolby). Dolby MP is a pure amplitude matrix using anti-polarity signals to encode the surround channel (with no phase encoding used), unlike, say, CBS SQ which is a true phase matrix and has accurate encoding/decoding only over the bandwidth and precision of its 90 degree phase shifts. Dolby Pro-Logic II encoding is also a pure amplitude matrix that uses anti-polarity encoding and slight level differences to encode the Lb and Rb channels (Circle Surround uses basically the same surround encoding technique as PL-II, as does Lexicon Logic-7) - again, no phase encoding being used. Both Dolby MP and PL-II are "Great Circle" encoding systems on the Scheiber Sphere (Energy Sphere). '' (apparently, no one has done an entry for the Scheiber Sphere on Wikipedia yet) '' —Preceding Ty Chamberlain (talk) 19:37, 17 May 2009 (UTC) • contribs) 19:16, 17 May 2009 (UTC)

Why ENcoding matricies in Matrix DEcoder?
This article is titled Matrix DEcoder, and yet it contains little except the ENcoding matricies for the various schemes. Surely it should contain the DEcoding matricies. I have no desire to correct this; if I had my way the whole article would be deleted as it contains only information better expressed elsewhere, and I do not see what purpose it serves. HairyWombat (talk) 19:23, 20 April 2009 (UTC)

Should be an article on advanced decoding
There should be an article on Wikipedia on advanced matrix surround decoding, such as Variable De-matrixing (Sansui QS Vario-Matrix, SQ Shadow Vector), Matrix Multiplication/Vector Cancellation (Tate DES for SQ, CBS Paramatrix for SQ, Scheiber-Sonics 360 Spatial Decoder for SQ, Dolby Pro-Logic & PL-II), Gain-Riding Logic (CBS SQ Full-Wavematching Logic, Circle Surround, Peter Scheiber's 1970's quad patents), plus the various methods of detecting channel dominance and directionality, (such as Axis-Crossings, wavematching, envelope comparison, etc...) I've been talking with Lynn Olson, the inventor of the Shadow Vector Signal Analysis Unit for SQ decoding (US Patent #4,018,992) and while it never reached the market because Audionics set its development aside to work on the Tate DES, I think Wikipedia would be the perfect place for info on it and other advanced forms of matrix surround decoding. Hopefully, we could get the inventors such as Lynn Olson or Martin Willcocks (inventor of the Tate DES) to add info. Or Jim Fosgate, since he's the inventor of Dolby Pro-Logic II as well as the designer of the vast majority of award winning matrix decoders, such as the legendary Fosgate 101A Tate II Surround Stereo SQ Decoder - it was also the first advanced cancellation decoder to incorporate a high-separation Cinema mode for Dolby Stereo encoded films. (Dolby theater decoders at the time were based on the Tate II DES chip that the 101A used) Ty Chamberlain (talk) 19:37, 17 May 2009 (UTC) —Preceding unsigned comment added by Disclord (talk • contribs) 19:34, 17 May 2009 (UTC)

Pro-Logic VS Pro*Logic VS Pro Logic
''This section is in the History for this Talk page, but seems to have vanished into the ether. I have therefore extracted and recreated it.'' HairyWombat (talk)

HairyWombat, I'm glad to have the lack of a hyphen in Pro Logic cleared up - I've always either used a hyphen or an asterisk since the 'official' logo uses a "dot" between Pro and Logic. Dolby seems never to have addressed this on their site or in my conversations with Roger Dressler of Dolby, so knowing they told you that it's just "Pro Logic" finally gives the subject 'closure' for me since I like to be precise with stuff like that. -- ~ —Preceding unsigned comment added by Disclord (talk • contribs) 14:45, 19 May 2009 (UTC)


 * The person who told me "no hyphen" was Lily Cheung, Dolby's Intellectual Property Manager. This would have been back in 1995. She also told me that is was "Pro Logic" and not "Dolby Pro Logic". Looking on Dolby's website, they seem to have changed this. I didn't see "Pro.Logic" (with a dot), but it is a big site. The only version Dolby have trademarked in the US is "Pro Logic" so I suggest you go with that. According to their website "Dolby Pro Logic" is a trademark, but this is incorrect. HairyWombat (talk) 19:25, 19 May 2009 (UTC)

Article to be deleted
I thought the gameplan was to carve out the SQ section into an article of its own (titled Stereo quadraphonic), and to then delete this article. This was because all the information in it will then be replicated elsewhere (which is almost the case now). If this is the plan then adding more content to this article is futile. HairyWombat (talk) 19:15, 25 May 2009 (UTC)


 * It is, I was just adding bits of relevant info for the interim until I had it 'together' enough to start the SQ section - we could do that now, but what I have so far isn't complete... I figured all of the info here would be 'pulled over' to the SQ article when it was ready and then this deleted. I don't know how to start a totally new topic, so if you would create it, I'll go ahead and put what I do have ready now.  I'm currently in contact with Martin Willcocks, inventor of the Tate DES for SQ, and he wants to help out so we should be able to get the SQ topic into a reference-quality topic.  —Preceding unsigned comment added by Disclord (talk • contribs) 12:57, 26 May 2009 (UTC)

When I suggested that adding more content was futile, I was thinking of the paragraph you added to the UHJ section. Somebody called Rhamphorynchus has already created the article Stereo quadraphonic, but as a #REDIRECT. To change this, click on the link then scroll to the top and click on "(Redirected from Stereo quadraphonic)". You can then replace the #REDIRECT with content in the usual way via the "edit this page" tag. Good luck. HairyWombat (talk) 15:13, 26 May 2009 (UTC)

Mathematical definition of matrix
It is not clear from the article what mathematical operation is represented by the phase shift. Phase-shifting an arbitrary complex-valued function f(t) by 90° involves multiplying each frequency component by i, resulting in a phase-shifted function f'(t):
 * $$f(t) \rightarrow if(t).$$

However, for audio signals, we have to deal with real-valued signals. If you do it by taking the real part, you get
 * $$\Re\left( i f(t) \right) = 0,$$

which is of course not very useful. Another approach could be that we assume that the real-valued function is actually
 * $$f(t) = \Re\left(g(t)\right),$$

for some complex-valued function
 * $$g(t) = \int_0^\infty G(\omega) e^{i\omega t} d\omega.$$

(Note that the Fourier transform does not cover negative frequencies) The phase-shifted function would then be
 * $$f'(t) = \Re\left( i g(t) \right) = \Re\left( i \int_0^\infty G(\omega) e^{i\omega t} d\omega \right).\qquad(*)$$

Although this works for sine waves, with $$G(\omega) = \delta(\omega-\omega_0)$$, it becomes rather ugly for arbitrary non-periodic signals. So could someone knowledgeable please expand the article with a mathematical description of the signal transformations? Han-Kwang (t) 22:44, 23 May 2012 (UTC)
 * You have to look at it this way. To code the the surround information in the stereo pair of signals they introduced a phase shift of +/-90 degrees over the entire frequency range. It means that one of the two channels was shifted 90 degrees relative to the other one. (This a rather odd signal processing, because it does not represent anything natural, like f.i. a time delay. Still, it is hard to hear for the untrained ear.) For a single-frequency sine-wave it would result in a circular movement of the needle in the stereo track, for a composition of multi-frequency sine-waves, the circular movement is not recognizable, but it still is there if you would analyze the signal per frequency. It doesn't really matter for decoding, though. After applying the phase shift and some subtracting or adding of signals, it is possibly to decode the surround signals, but most of them with a channel separation limited to 3 dB. Jan Arkesteijn (talk) 17:14, 23 June 2012 (UTC)

Can you express this in mathematical notation? I think that your explanation is more or less equivalent with expression (*) above, but as I said, it is rather ugly. For instance, if you do this for a $\delta$ pulse in the time,
 * $$f(t) = \delta(t) = 2\int_0^{\infty} \cos(\omega t)\, d\omega,$$

it would transform into
 * $$f'(t) = 2\int_0^{\infty} \sin(\omega t)\, d\omega. $$

I don't know what the name is of such a function, but it looks quite ugly when integrated over a finite range or via a discrete Fourier Transform. It oscillates wildly. In the frequency domain, it is equivalent to multiplying by
 * $$h(\omega) = \left\{ \begin{array}{cc} i & (\omega \ge 0) \\ -i & (\omega < 0) \end{array} \right.$$

In Octave/Matlab: I don't have much experience with DSP implementations, but I think that if you want to do this for frequencies down to 20 Hz, you would probably need to work with time slices of at least 100 milliseconds and stitch them together. That would be yet another delay in the playback chain.

You write: "it is hard to hear for the untrained ear". That is probably correct for periodic signals, but for transients, it could generate pre-echoes, wouldn't it?

Han-Kwang (t) 20:15, 23 June 2012 (UTC)
 * There used to be day that I could help you there, but that is a long time ago. Looking at your questions I realize that even knowledge wears when not nourished. But I feel that you are trying to solve a design problem here, instead of improving this article, am I right? If so, that problem has been solved,already. If you want I can give you a direction. But back in the 70s when everything was processed with analog equipment, there was no pre-echo, and in that respect I meant it was hard to hear.. Pre-echo is a side effect that came with DSP. Jan Arkesteijn (talk) 21:00, 23 June 2012 (UTC)

I am not interested in building my own surround decoder; I was curious how one could encode five channels from two independent inputs, and found this description here to be lacking. In such cases, I usually edit the article if I can figure out the missing bits (from math, physics, references, Google, or common sense) but here, neither of those brought me any further.

Phase-shifting each frequency component of a signal by 90 degrees (as you propose) is mathematically impossible without knowledge of what the signal is going to do in the future, which also implies that such a phase shifter must generate pre-echos. If these encodings that the article is about can be implemented in analog circuits, then the encoding is actually something different than what is suggested in this article. For example, the transfer function may have a frequency-dependent phase shift and amplitude. If you have pointers to any documentation that is not in the references, I'd like to hear of them.

Han-Kwang (t) 21:54, 23 June 2012 (UTC)
 * First of all, I do not propose anything. An invention of the 70s is described here. Secondly, shifting a signal 90 degrees forward, seems to be a impossibility, because it seems to be a prediction in time. But you have to remember that back then, they were treating this as continuous signals in the frequency domain, and phase shifting was relative between the two stereo signals. So phase shifting one channel 90 degrees forward was the same as phase shifting the other 270 degrees backward. And even 270 degrees backwards was not done by actually shifting 270 degrees (in time), but shifting 90 degrees of the inverted signal backwards, which mathematically is the same in the frequency domain. I know it messes up transients, but as I said, for the untrained ear it was ok. Jan Arkesteijn (talk) 10:25, 24 June 2012 (UTC)


 * No, the article in general is correct, (give or take a minus sign I suppose). This is the way they described the technology. I must admit, the person who wrote it using the $$k = -i$$ made it to complex. I am used to use $$-i$$ notation. Jan Arkesteijn (talk) 10:31, 24 June 2012 (UTC)

It does not make any difference whether the signal is shifted 90 degrees forward or backward; both violate causality in the same way (if you have one of them, the other one can be generated by multiplying by -1) I have been searching and reading up; the signal operation is called Hilbert transform, which has the following impulse response: Analog filters with such a phase response cannot be constructed, although it is possible to generate two signals from one input that are 90 degrees shifted relative to each other, both of them having a frequency-dependent phase-shift relative to their common input. The filter that is actually used in audio matrix encoding must therefore be something different and the statement that the encoder/decoder does/did a 90 degree phase shift relative to the input using analog filtering is incorrect and misleading. Han-Kwang (t) 16:58, 28 June 2012 (UTC)

Update: US patent 4,115,666 describes a surround recording/playback system, based on 90-degree phase shifters, but it does not explain how to build such a phase shifter. By the way, I asked two colleagues with an EE degree, but they did not know of an analog phase-shifting circuit. Han-Kwang (t) 17:43, 28 June 2012 (UTC)
 * Impossible, misleading. I don't like the tone of this discussion. Look in all the documentation still available and you will see that this is the way the working of a matrix encoder/decoder was explained. If you think that is impossible and misleading, I guess you take your complaints to the people who designed this in the 70s. Maybe they bow to you in retrospect. But what is impossible in theory, is possible within limits, the limits being the frequency range to work in and the maximum error in the phase shift to obtain. It isn't necessary to know how the phase shift is realized, in order to explain the en/decoding matrices. Jan Arkesteijn (talk) 19:43, 28 June 2012 (UTC)


 * Matrix encoding/decoding is defined in terms of 90° phase shifts, but this is completely separate from how it is implemented. The article does not discuss implementation, nor does it need to. (However, as Han-Kwang has discovered unearthed, in the analogue domain a pair of signals is produced with a relative phase shift of 90° between them. In the digital domain usually a convolution filter is used, as in this example.) HairyWombat 00:51, 29 June 2012 (UTC)
 * Discovered is a big word, the technology goes back as least as far as quadraphony. Jan Arkesteijn (talk) 12:31, 29 June 2012 (UTC)
 * As you can see, I have changed it. I was thinking of how archaeologists "discover" stuff. (I was around for quadraphonics.) HairyWombat 19:00, 29 June 2012 (UTC)

External links modified
Hello fellow Wikipedians,

I have just modified one external link on Matrix decoder. Please take a moment to review my edit. If you have any questions, or need the bot to ignore the links, or the page altogether, please visit this simple FaQ for additional information. I made the following changes:
 * Added archive https://web.archive.org/web/20120128010154/http://pacificav.com/library/Dolby%20Surround%20Pro%20Logic%20Operation.pdf to http://pacificav.com/library/Dolby%20Surround%20Pro%20Logic%20Operation.pdf

When you have finished reviewing my changes, you may follow the instructions on the template below to fix any issues with the URLs.

Cheers.— InternetArchiveBot  (Report bug) 17:57, 5 June 2017 (UTC)