Trinaural decoding equations for 3 speaker stereo matrix ?

[WithTarragon]

Originally Posted by Elias

It is well known optimum linear matrix provides 6dB electrical separation between the three front channels.

Is that 6dB between L & C (or R & C) or is that 6dB between L & R ?

And is that true when the level of L (or R) is the same as C?

[KSTR]

What is the ideal directivity pattern for stereo speakers?

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.

[pos]

Originally Posted by KSTR

C=L+R, L'=L-0.5R, R'=R-0.5L

With this approach you are reducing comb filtering for a correlated signal by 6dB, but then you also get comb filtering on uncorrelated signals. Is it worth it?
I ask this because I find comb filtering between the two speakers to be the major drawback of stereo reproduction: it makes me aware of the two loudspeakers and takes me out of the stereo illusion all the time, resulting in a diffuse phantom image...

I have never tried Trinaural, but if it adds some more comb filtering how can it produce a better illusion?

[Elias]

Originally Posted by KSTR

Bongiorno's Trinaural is the same as Miles' Optimum Linear Matrix with k=0.5

Now we are getting somewhere ! Does there exist a reference for this information ?

Originally Posted by KSTR

and also one of Gerzon's TriField variants, if I gather up things correctly.

I don't remember Gerzon had a matrix with 0.5 ? Which particular one this could be ?


- Elias

[Elias]

Originally Posted by WithTarragon

Is that 6dB between L & C (or R & C) or is that 6dB between L & R ?

And is that true when the level of L (or R) is the same as C?

I should have said three front speakers. There is only two channels, of course. But yes, between three speakers maximum separation is 6 dB, L <-> C and L <-> R.

[KSTR]

That will remain to personal taste I guess. Some people are more sensitive to subtle tonal coulouring etc, others are more after the most realistic spatial impression. Trinaural is more something for the second group, one could say. Also it may depend on the genre and style of music you play.

Trinaural makes a different sound field around your head and quite more a difference in the room sound, and, if only for this, it is different. In fact Tri positions and sizes different kinds of phantom sources in a slightly different way than 2 speakers, and this not always in a "better" way. But more often it does, for me.

[Elias]

Originally Posted by pos

With this approach you are reducing comb filtering for a correlated signal by 6dB, but then you also get comb filtering on uncorrelated signals.

Uncorrelated signals cannot comb filter. If there is comb filtering then the signals are correlated to some degree.

Originally Posted by pos

Is it worth it?
I ask this because I find comb filtering between the two speakers to be the major drawback of stereo reproduction: it makes me aware of the two loudspeakers and takes me out of the stereo illusion all the time, resulting in a diffuse phantom image...

I have never tried Trinaural, but if it adds some more comb filtering how can it produce a better illusion?

It is worth it.

I've been listening a couple of weeks 3 speaker stereo matrix with coefficient of 0.5, the optimum matrix. I have three identical prototype speakers, and I'm doing extensive comparisons with standard 2 speaker stereo. I can change the configuration from the listening position by use of a switch so A/B comparison is immediate.

So far I've listened about 100 songs and there is no single song I prefer with 2 speaker over the 3 speakers. That should tell some direction already


- Elias

[KSTR]

Originally Posted by Elias

I don't remember Gerzon had a matrix with 0.5 ? Which particular one this could be ?

The basic TriField equations are :
L' = 0.5[(sin(phi) + w)L + (sin(phi) - w)R],
R' correspondingly, and
C = sqrt(2)cos(phi)(L+R)

A certain selection of phi and w should yield the trinaural amplitude relationships but I haven't tried to solve the eq's
EDIT: Once I progress to try a 3-way full TriField matrix with the coeffs that Gerzon recommends, and with DSP linear phase, I'll know....

[Elias]

Originally Posted by KSTR

EDIT: Once I progress to try a 3-way full TriField matrix with the coeffs that Gerzon recommends, and with DSP linear phase, I'll know....

It could be interesting. Is it known what order the low-high pass filters are in Trifield decoder ? I've allways assumed 1st order.

[pos]

Originally Posted by Elias

Uncorrelated signals cannot comb filter. If there is comb filtering then the signals are correlated to some degree.

An uncorrelated L/R stereo signal split into 3 speakers following this matrix will result in correlated sound between speakers, as they all have sums or differences of L and R...