Trinaural decoding equations for 3 speaker stereo matrix ?

Elias · 2013-01-06 1:07 pm

markus76 said:
Interesting observation: the center doesn't seem to be particularly stable with head movements.

If your system is like in your pic in post #75, you may note the center speaker is having about 6dB less direct sound in the midrange than the side speakers since the baffle step occurs in considerable higher in frequency. In addition your heavily absorbed front wall makes the center speaker a 4pi radiator while side speakers are radiating in 2pi. This gain mismatch is the propable reason to unstable center phantoms you experience.

Have you tried with speakers being all identical ?

- Elias

Elias · 2013-01-06 1:14 pm

markus76 said:
Gerzon states in his 1991 AES convention paper #3180 that "loudspeakers [that] lie along a straight line [...] generally give ~~less good results".~~

I think, and it looks like that in theory, if the listening is performed in a strict miniature spot without lateral movements, the speaker arch with identical distances to the listening position may be better. However for any off-center listener the speaker line looks like a much better alternative. And here I'm not thinking about tiny head movements of few cm but large shifts like up to 1m.

Personally I dont like tiny listening spots. Currently I have my 3 speakers in a line and it works well in almost everywhere in my room.

- Elias

Elias · 2013-01-06 6:09 pm

KSTR said:
Bongiorno's Trinaural is the same as Miles' Optimum Linear Matrix with k=0.5 and also one of Gerzon's TriField variants, if I gather up things correctly.

I used the search function and found this post from KSTR:

http://www.diyaudio.com/forums/multi-way/123426-horn-vs-waveguide-18.html#post1842463
post #884

If anyone is interested, I have some detailed insight of James Bongiorno's "Trinaural" (a subset of the "Optimum Linear Matrix", as described by Michael Miles). Could even give you the detailed schematic (but I won't, of course. I respect IP).

Basically it's an an ultra-simply analog matrix, completely LTI (if you have balanced connections, 13 resistors will do it damn close to the real thing). And it works like charms... I swear. It creates an artifical but very supportive (for the illusion that stereo is all about, anyway) sensation of a different vertical size/expansion/pattern, giving way more seperation between instruments and between direct and reverberant sounds (as encoded in the recording). It not easy to describe that in words (for me as a non-native speaker, at least).

And the sweet spot is huge (laterally). Toe-In of speakers (that is, "trading" as the effect in stereo psychoacoustics is called, compensating dT with dL) doesn't work for me at all to widen the sweet spot. It still is pretty balanced stereo in the lower registers, but all sensation of space is lost. Sondstage collapses to a straight 1D line between the speakers.

But not in this thread. Better we go here :
Linear 2-->3 Rematrixing for Speaker Signals

- Klaus

If this is indeed correct, then Trinaural equations are based on coefficient of 0.5. And my question in original post has been answered

Thank you.

- Elias

Soundtrackmixer · 2013-01-06 9:03 pm

markus76 said:
Gerzon states in his 1991 AES convention paper #3180 that "loudspeakers [that] lie along a straight line [...] generally give ~~less good results".~~

That is because of a time arrival issue. If you are sitting along the center line between the L/R speakers, the center speaker is actually closer to you than the L/R are. This will perceptively reduce the stereo seperation of the L/R mains in favor of a strong mono center image.

markus76 · 2013-01-06 9:26 pm

Elias said:
If this is indeed correct, then Trinaural equations are based on coefficient of 0.5.

Which is a forerunner of the "energy-preserving matrix decoder" which in turn is a forerunner of the "frequency-dependent energy-preserving matrix decoder" to which I'm listening right now

(From Gerzon's AES convention paper #3180)

lolo · 2013-01-07 9:19 am

What about directivity? I finally had 2 hours for myself last night and experimented with three quickly made pipe speakers, matrix from my old denon receiver and dayton RS100. It works! More so on hard panned effects than orchestras but it's a good start. Only two drivers at the opposite end of a horizontal pipe worked well too, sounds a bit like a corner setup in some ways.
Is omnidirectionality good in the mids? Wouldn't a narrower CD all the way down to 100-200hz be better? I know, not an easy thing to achieve.. but I feel there is considerable leak in that range going straight to the ear and not towards the wall..

markus76 · 2013-01-07 9:24 am

lolo said:
What about directivity? I finally had 2 hours for myself last night and experimented with three quickly made pipe speakers, matrix from my old denon receiver and dayton RS100. It works! More so on hard panned effects than orchestras but it's a good start.
But is omnidirectionality good in the mids? Wouldn't CD all the way down to 100-200hz be better? I know, not an easy thing to achieve..

It's the old question for which we still don't have a good answer, "How much room is needed?", i.e. what reflection pattern is optimal.

lolo · 2013-01-07 9:40 am

Sure, but for this specific application it seems to me you ONLY want the room!

BTW, a corner setup sounds good too. It shows that stereo needs quite a large spread to works well, in my opinion..

Elias · 2013-01-07 10:17 am

lolo said:
Only two drivers at the opposite end of a horizontal pipe worked well too, sounds a bit like a corner setup in some ways.

So you have done something similar to Stereolith ?

lolo · 2013-01-07 10:38 am

maybe?

I am very keen to know what you have to say about the directivity issue.

SEdwards · 2013-01-07 2:28 pm

Last week I was listening to a 2-> 3 matrix with coeff 0.5 in Reaper (I'll upgrade to Markus' script, thanks for posting that). For the afternoon I'm using Dolby PLIIx from my soundcard (Xonar) output.

What exactly does it do frequency wise? It cuts up processing band by band, but how? The Wikipedia article doesn't really say. Or am I totally mistaken and it is just a sqrt(2) center if I'm using 3 speakers.

Elias · 2013-01-07 3:39 pm

lolo said:
What about directivity? I finally had 2 hours for myself last night and experimented with three quickly made pipe speakers, matrix from my old denon receiver and dayton RS100. It works!

I also started with cardboard tubes (diam about 12cm, length about 40cm) and FRS8 full range elements

I lifted the tubes so that elements are on ear height. First listened with elements aimed to the listener, then aimed the elements towards the ceiling. This was with passive matrix wiring with coefficient of 0.5. So they were kind of omnis.

Then I placed a large (60*80cm) baffles just behind each tube to make them 2pi radiators, and it was much better !

Omni midrange ? I don't like it very much. It sounds too ambigouos and not very clear. Half space midrange is better, also with matrix.

- Elias

Elias · 2013-01-07 3:43 pm

lolo said:
maybe?

Is it anywhere like this you made ?

a_tewinkel · 2013-01-07 3:49 pm

Elias said:
Omni midrange ? I don't like it very much. It sounds too ambigouos and not very clear. Half space midrange is better, also with matrix.

I currently have cardioid mids. Much better and more clarity than with narrow boxes or dipoles too close to the wall.

markus76 · 2013-01-07 5:37 pm

SEdwards said:
I'll upgrade to Markus' script, thanks for posting that

If you use the energy-preserving code then remove the 3dB attenuation for the center. Simply replace

Code:

c_out = 0.7071*m_c;

with

Code:

c_out = m_c;

lolo · 2013-01-08 8:58 am

Yep, something like that, but with three separate pipes and a 90 deg angle so they come out as three flowers. I really have to redo it there is just too much omni leakage in the mids, will report back.

Elias said:
Is it anywhere like this you made ?

markus76 · 2013-02-19 11:54 am

Just for the sake of completeness here's the "final" Reaper code for a energy-preserving matrix decoder after Gerzon:
http://mehlau.net/audio/matrix_decoder/

Elias · 2013-03-11 8:45 am

I found some more confirmation on Trinaural matrix coefficients from here

http://www.nubert-forum.de/nuforum/viewtopic.php?f=15&t=25770&sid=21d578b674333aa78642a5023fc0a493

And the matrix coefficients according to them are:

L -> L - R/2
R -> R - L/2
C -> (L+R)/2

Which are the same as referenced earlier from another source.

The thing I don't know is how they know it

- Elias

markus76 · 2013-03-11 9:15 am

That's TriSonic (AES Preprint: "An Optimum Linear-Matrix Stereo Imaging System") with a matrix factor of 0.5.
Just a sub-version of Gerzon's energy-preserving decoder.

markus76 · 2013-03-11 10:20 am

A post by Elias just got lost?

Trinaural decoding equations for 3 speaker stereo matrix ?

Member

Member

Member

Member

Banned

Attachments

Member

Banned

Member

Member

Member

Member

Member

Member

Member

Banned

Member

Banned

Member

Banned

Banned