Trinaural decoding equations for 3 speaker stereo matrix ?

[CLS]

Hi there,

Late to the party.

About the 6dB separation among channels, I'd like to share an experience. The number looks lousy, but to my hearing, it's quite different.

In a 3-ch linear matrix system with x=0.5, when the input signal is panned to the extreme left, then:

Ls= L
Cs= 0.5L
Rs= -0.5L

Listen to each unit closely, they're all playing. You'd hear the center and right are quieter, but not very much. Step back to the normal listening position, it's fun to hear the image goes far left and beyond the left speaker. At this moment, you'd swear there's absolutely no sound from center and right. It's one of the most interesting part of this thing.

-------

And, I've been planning a line level matrix for better flexibility in choice of amps and types of speakers.

Other than OP-amp based mixer circuit, are there any better solutions?

I tried cross-feeding the secondaries of R/L interstage transformers to each other to generate the partial subtractions. They're not buffered, so I'm not sure if it's robust enough to avoid interfereing with the upstream signals. Now I'm using this in bass section of active system only.

[CLS]

The turns ratio of the transformer is 4.5:1+1+1+1, wired as 4.5:2+1, and the "2" section is riding on the "1" of opposite channel (and the "1" is in nagtive polarity).

[Elias]

Originally Posted by CLS

About the 6dB separation among channels, I'd like to share an experience. The number looks lousy, but to my hearing, it's quite different.

I think 6dB is quite a lot of sound difference in perceptual sense.
For example, I have observed that just about 5dB difference between early reflections and direct sound is able to overrule precedence, or Haas, effect. That is, if reflections are 5dB higher in level than direct sound the perceived image is at the direction of the reflections. The literature does not suggest these kind of values but usually refer something like 12dB which is much bigger of course.
This happends in a small room acoustic space, so I guess most psychoacoustic studies have been done either by headphones or anechoic chamber.

Originally Posted by CLS

Other than OP-amp based mixer circuit, are there any better solutions?

What was the problem of passive speaker wiring matrix ?


- Elias

[Elias]

Originally Posted by Pano

Or the Dynaco method from the early 1960s

Is it that Hafler was working in Dynaco ?

[Pano]

Possibly, I could check. That diagram comes from the manual of my Dynakit ST-35, circa 1963.

[CLS]

Originally Posted by Elias

What was the problem of passive speaker wiring matrix ?

I need more flexibility to build different speakers in the center and sides.

Different speakers have different impedance curves and efficiencies, so it's not ideal to take them into passive matrix directly. Compensations in those factors all cause some attenuation, which I don't like.

Also, not all amps are suitable for this task -- those with floating output stage like bridged class D, or high output impedance tube amps... etc.

[Elias]

Originally Posted by CLS

I need more flexibility to build different speakers in the center and sides.

Different speakers have different impedance curves and efficiencies, so it's not ideal to take them into passive matrix directly. Compensations in those factors all cause some attenuation, which I don't like.

Also, not all amps are suitable for this task -- those with floating output stage like bridged class D, or high output impedance tube amps... etc.

Balanced output amps can be a problem with passive matrix, yes.

But why use different kind of speakers ? Ideally they should be identical at least in the bass and midrange. In top range treble I'm not sure if identical speakers are anymore required.


- Elias

[KSTR]

Originally Posted by CLS

Other than OP-amp based mixer circuit, are there any better solutions?

If you have balanced in's and out's at the amps and source, a simple resistor matrix/mixer does the job, provided output impedance is low (<100R) and input impedance is high (>10k)

Feed the same polarity signals with R and the opposite polarity with 2R and tap of the output at the connection to get the "X-0.5Y" type of summing.
So Left_Out+ comes from Left_In+ via 750R and from Right_In via 1500R, and so on for all four combinations.
Center is just summing up like polarities with 2000R resistors, then dividing it down with a paralleled 4300R at the output (can be omitted if you have volume control at the amps).
You may want 200R trimmers at the junctions of the side-channel to trim out errors from non-zero output impedances and replace the 4k3 with 3k9+1k trimmer for a center channel level trim.

Easiest of course to provide the signals already as L-0.5R and R-0.5L (center is sum of it, can be done with resistors and amp volume) in software if your playback is PC-based. Matrix mixing plugins for the VST-Interface are available at no cost.

[Elias]

Originally Posted by Elias

Balanced output amps can be a problem with passive matrix, yes.

Or, can be a benefit sometimes

If you're happy with matrix:
Ls = L
Cs = 0.5(L + R)
Rs = R

then can use this configuration with balanced output amplifier:


Apparently "Originally appeared in the May/June 1990 issue of Car Stereo Review magazine."
How To - Car Stereo - How To Install A Center Channel


However, this matrix may not be optimal.

- Elias

[SEdwards]

I'm just getting into 3-channel. I like listening near field and the stereo collapse was too bad for me with 2 speakers.

What do you guys think about focusing cash on the center channel and treating all others like 'surrounds'? Right now I'm using 3x upfiring 3" speakers in pipes (Demokrit style). I like it. I have some nice drivers lying around that I bought cheap from ebay (e.g. AL200) but not all paired up. Can't really afford to spend more on speakers at the moment. (Besides foil for making an AMT. )

Center is where the action should be, both for music and movies. And when guests come mono is more convenient anyway. (Also when I want to drag a speaker to the kitchen.)

I'm doing all my matrixing in Reaper.