Trinaural decoding equations for 3 speaker stereo matrix ?

[markus76]

Originally Posted by Elias

A metal plate hit by a wooden stick In some music there is two metal plates hitting each others in a consecutive on-off manner

Then why is it that cymbals are amongst the instruments that generally can be localized best? They are not "split into the speakers".
The "instrument" you are talking about with "less signal energy at 1 kHz and more energy above 6kHz" is probably a steady-state test signal that doesn't resemble any real instrument.

[Rudolf]

Originally Posted by Elias

A metal plate hit by a wooden stick In some music there is two metal plates hitting each others in a consecutive on-off manner

Those small metal plates at the top end of a vibraphone or marimbaphone are at f³ (english F6), which is nearly 1.4 kHz. Even if they were f4 (which is the highest f on a standard piano), it would be 2.8 kHz only. People often grossly mistake the frequency of high notes.

Rudolf

[markus76]

Elias was probably talking about a hihat. "two metal plates hitting each others in a consecutive on-off manner" would fit the description

Anyway, while Tri-something is interesting, the real question is what such an upmixing approach is supposed to do. Stereo unfortunately presents the listener with ambiguous cues and I don't think there's a benefit in presenting even more ambiguous cues over 3 speakers instead of two

[Rudolf]

Originally Posted by markus76

Elias was probably talking about a hihat. "two metal plates hitting each others in a consecutive on-off manner" would fit the description

Typical hi-hats are usually between 300-3000 Hz dominant frequencies, and can extend up to 10-17k Hz for crispness, “air” and sparkle.

[Elias]

Well,

Percussion Frequencies Part 2 – Cymbals | Easy Ear Training Online

Pedal Hi-Hat: 12kHz
"In this case the upper crispness happens to be the most prevalent. However 6kHz is competing for dominion – it was only by a fraction that 12k won"


Closed Hi-Hat: 10kHz
10kHz dominates loudly on this one.

[markus76]

Originally Posted by Rudolf

Typical hi-hats are usually between 300-3000 Hz dominant frequencies, and can extend up to 10-17k Hz for crispness, “air” and sparkle.

Exactly, hence my comment "Sounds like a constructed case."

[markus76]

Originally Posted by Elias

Well, [...]

The "clang" of cymbals is typically in the hundreds, even as low as 200Hz.

[bzfcocon]

Originally Posted by bzfcocon

Did some quick maths based on the paper.

For low freq (phi = 35 deg) you get:
C3 ~ 0.58(L+R) L3 ~ 0.78L - 0.21R (similar for R3)

For highs ((phi = 55 deg)) you get:
C3 ~ 0.40(L+R) L3 ~ 0.90L - 0.09R (similar for R3)

So you can't achieve exactly the same effect by rolling off the high freq on the center, because the phi also controls the "mix" in the L and R, but one might get away with rolling off the center ~3dB down above 5kHz and still getting a good effect, while avoiding additional high/low pass filtering in the matrix circuit.

Sooo, I gave a try to trinaural decoding, more or less as per Gerzon paper, with a phi around 45 degrees. My "implementation":
- main L and R speakers are dipoles using 6xFRS8 + Neo3PDR, crossover to bass is active
- In the crossover, I added a switchable simple circuit that does the matrixing as per Gerzon paper something like L-0.19R / R-0.19L
- for the center, I simply built another dipole panel with 4xFRS8. 2xFRS are connected in parallel and fed from L channel, the other 2xFRS8 fed from the R channel - they sum acoustically. Overall, given the different series/parallel connections of the FRS8 in the L/R/C, the center gives the acoustic sum of around 0.55(L+R), which is what the decoder should give. I measured the output to be sure. The center has no tweeter and slightly rolls off about 3dB above 5kHz, emulating the frequency dependent effect described in the paper.

Results: not to dismiss the whole thing yet, but am _not_ convinced. Having tried Ambiophonics at some point in time with the same speakers, I must say Ambiophonics did a better job.

Details:
- yes, the center image is there, stable and carved in stone no matter how you turn the head. But unfortunately there's a big "who cares" factor about this: the center image has been always dead sharp stable also without the center with my speakers ! Maybe in contrast to what some people might say about dipoles, mine have an excellent "phantom center".

- it might be that the sound stage is marginally more stable when listening off-center compared to standard stereo, but OTOH it is flatter, narrower and unconvincing(artificial) no matter where you sit compared to listening to normal stereo from the central listening point. So the trade-off is either a "rich" spatial perception in one single listening position(standard stereo setup) or a flat and weak one overall with trinaural, not a fortunate choice Spatial perception is an illusion and if the illusion is bad, then it's not really worth it IMHO.

- something weird seems to happen with the tonal balance, might be related to the combing of the rear radiation. It actually should be better with a center, but it's not, the sound is somewhat colored.

I'll do some more listening tests, including some stereo test tones. Maybe I got it wrong, maybe in needs more tweaking or maybe it simply does not work well with dipoles.

I wonder if the paper authors also used music for testing - did not find it explicitly mentioned. Because, you know, people don't listen to test tones..

[Elias]

Originally Posted by bzfcocon

Sooo, I gave a try to trinaural decoding, more or less as per Gerzon paper, with a phi around 45 degrees. My "implementation":
- main L and R speakers are dipoles using 6xFRS8 + Neo3PDR, crossover to bass is active
- In the crossover, I added a switchable simple circuit that does the matrixing as per Gerzon paper something like L-0.19R / R-0.19L
- for the center, I simply built another dipole panel with 4xFRS8. 2xFRS are connected in parallel and fed from L channel, the other 2xFRS8 fed from the R channel - they sum acoustically. Overall, given the different series/parallel connections of the FRS8 in the L/R/C, the center gives the acoustic sum of around 0.55(L+R), which is what the decoder should give. I measured the output to be sure. The center has no tweeter and slightly rolls off about 3dB above 5kHz, emulating the frequency dependent effect described in the paper.

Results: not to dismiss the whole thing yet, but am _not_ convinced. Having tried Ambiophonics at some point in time with the same speakers, I must say Ambiophonics did a better job.

Details:
- yes, the center image is there, stable and carved in stone no matter how you turn the head. But unfortunately there's a big "who cares" factor about this: the center image has been always dead sharp stable also without the center with my speakers ! Maybe in contrast to what some people might say about dipoles, mine have an excellent "phantom center".

- it might be that the sound stage is marginally more stable when listening off-center compared to standard stereo, but OTOH it is flatter, narrower and unconvincing(artificial) no matter where you sit compared to listening to normal stereo from the central listening point. So the trade-off is either a "rich" spatial perception in one single listening position(standard stereo setup) or a flat and weak one overall with trinaural, not a fortunate choice Spatial perception is an illusion and if the illusion is bad, then it's not really worth it IMHO.

- something weird seems to happen with the tonal balance, might be related to the combing of the rear radiation. It actually should be better with a center, but it's not, the sound is somewhat colored.

I'll do some more listening tests, including some stereo test tones. Maybe I got it wrong, maybe in needs more tweaking or maybe it simply does not work well with dipoles.

I wonder if the paper authors also used music for testing - did not find it explicitly mentioned. Because, you know, people don't listen to test tones..

Interesting results. So you tried Trifield, not Trinaural

Ambiophonics is propably hard to beat by any method, if you are willing to remain sitting in the sweet spot all the time.

Would you try Trinaural next ?

You can wire three speakers Trinaurally in a very simple way to form a matrix (abandon the input resistors in this case)

Ls = L - 0.5*R
Rs = R - 0.5*L
Cs = 0.5*L + 0.5*R




- Elias

[Pano]

Or the Dynaco method from the early 1960s