Trinaural decoding equations for 3 speaker stereo matrix ?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
So, are we going to have some measurement results soon ? :D

Well...my friend George and I both own Trinaural. I use my system daily, while George awaits home remodel and new speakers coming before I assemble his system.

Sorry, no scope etc, here.

I've not read read the entire thread, but plan to. I would consider supplying a Trinaural for testing. I'd have to insure EJ @ W4S has replacements available and frankly I could not afford the risk of loss so someone would have to prepay fully refundable deposit equal to my replacement cost (wholesale dealer cost, $2500 msrp last checked a few months ago).

What is the consensus desire for members to replicate Trinaural? Opamps, passive transformers...?

I love the idea of DIY Trinaural but relevant questions arise. What is estimated parts cost for the basic functionality, L/R input, LCR output?

I presume each member desires a different mix of Trinaural's OEM features. IMO at least one non-OEM feature seems critical, a Stereo/Trinaural switch with unity gain (IOW overall system output stays constant when switching between Stereo and Trinaural).

OEM features:

3.1 HT bypass
Balanced outputs
Unbalanced inputs
Noise rotary volume/off switch
Center channel level screw adjustment
Noise generator to balance Center level
Noise routes to LCR via 3-position toggle
LCR toggle switch: Fullrange/active HP crossover
Subwoofer rotary level, significant output @ minimum setting
HP/LP active crossover in-phase 2nd order @ 80 Hz, LP always on

I agree with Kal Rubinson, who does not understand James' strong recommendation against comparing stereo to Trinaural. I disagree with James' warning that it takes a week or two to adjust to Trinaural. Finally, I disagree with James who says leave the C Ch level alone once it is set: I enjoy a remote multi channel line preamp with remote trim of C Ch and Bass level.

A glaring, egregious Trinaural fault is no switch to compare Stereo/Trinaural, preferably with unity system gain between the two.

I tend to agree with Jim Smith on two criticisms, though I solved both to my satisfaction:

As mentioned earlier, C Ch quality seems slightly different and less pure vs. LR. I solved this by raising my C Ch speaker 4cm and no other change (highly refined Dynaudio Esotec 2-way MT above 80 Hz).

Smith suggested that Trinaural processing for spaced omni recordings ala Telarc (RIP) is just wrong. Independently before I read Smith's opinion, my practice was greater attenuation of C Ch level playing Telarc and other spaced omni recordings. The following is also interesting, and goes directly to Smith's criticism: with proper C Ch level (about -4dB on my preamp, system is unity gain LCR including Trinaural C Ch trim pot) Trinaural still blows away Stereo. So Smith was right, but wrong after solution applied.

I very much thank Smith for the above criticisms, which clarified problems and their solutions. OTOH, I personally found his remaining criticisms laughable coming from someone who charges several thousand dollars for home stereo setup and who used to import $100k speakers: Trinaural requires more space (yeah, and those Tannoys you love and your old horns are small?), requires more cost, requires another channel of amplification, etc, etc. Every one of those criticisms apply across the audio board to every single advancement except the extra channel.

If you want free easy solutions to improve performance stop visiting here!
 
Last edited:
So, are we going to have some measurement results soon ? :D

Your website has a great wealth of information and data, even on 3-ch matrix.

Have you built a decoder yet based on your new so-called ideal matrix? How fun would it be to compare to Trinaural!

I'm awful with math while you are obviously quite adept with it. Have you translated the Austrian Trinaural DIY site, and if so: did they post the Trinaural equations? How do they compare to your own ideal matrix?

Very glad to find this thread, and especially your contribution. As mentioned earlier, I've had no interest in stereo for almost seven years now.

If only James was alive. 'Twould be so much fun to show him your math and get his cantankerous replies, expletives and all!

I remember at one CES James insulting a Merdian engineer for their outdated test gear! I quickly slithered away....at least he made up for his outrageous personality with understated clothing...NOT!
 
What is the consensus desire for members to replicate Trinaural? Opamps, passive transformers...?

I love the idea of DIY Trinaural but relevant questions arise. What is estimated parts cost for the basic functionality, L/R input, LCR output?


Now, until further confirmation about Trinaural equations are revealed, we must rely on what was available from other Diyaudio members. The decoding matrix coefficients were given earlier in this thread as

Ls = L - 0.5 R
Cs = 0.5(L + R)
Rs = R - 0.5 L

Parts cost ? How much a wire costs ? :D
(Of course you'll need three speakers, but that cost does not count, even if it's 50% more expensive than 2 speaker stereo).

This is what I've been using. It fulfills the above matrix coefficients.

An externally hosted image should be here but it was not working when we last tested it.




If you want free easy solutions to improve performance stop visiting here!

I think quite the contrary ;) This is the place to visit for easy solutions to improve stereo performance :D
 
Jim Smith made one other criticism, that Trinaural setup diagram alters LCR spectral balance: C Ch speaker is on-axis, LR cross-fire 12" IFO centered listener. It's possible this did not occur to the inventor, he ignored it in the design, and employed no compensation in the design. The opposite is also possible, that this spectral change occurred to the designer and he employed some compensation in the design.

Possibly the qualitative difference I noticed and mentioned earlier (LR more highly refined sound vs. C) might be related to the above. Possibly slightly more toe-in than recommended would solve the problem (my solution was to elevate the C Ch 4cm).
 
Sorry I'm not yet up to speed. Is your matrix line level or speaker level? Diagram appears to be speaker level, but I just have no idea, sorry!

I can ask my friend who builds countless amps and DAC to test Trinaural. Maybe you could PM or email me specific questions for him to answer.

I'll finish reading and get up to speed to not waste member's time.
 
...Are there actually 3-speakers solutions that do NOT assume that the center speaker is identical to L and R ? The approach sounds promising and I would like to test it, but not if it means another bulky loudspeaker in the middle. How about if the center would only reproduce >100Hz, would the concept still work and would it change the matrixing ?

Interesting thread, btw.

For me the only two bass reproduction systems in domestic space (below 80 Hz) are Distributed Array (four subs) and Double Bass Array. I employ the former. The latter costs at least twice as much and ideally requires subs installed semi-permanent in-wall.

This works perfectly with Trinaural, requiring the main LCR speakers cover only above 80 Hz, hence they do not need to be large full range floor standers.

Trinaural instructions allow C Ch speaker to be different from LR but if so C Ch must be better in quality not worse compared to LR. My experience validates the above claim. When I tried it the speakers employed same drivers but different quantity mid bass and planar mids.
 
I'll confirm later, but meanwhile: is it likely that speaker grounds are common on my 2003-2004 vintage Pioneer VSX-D912 HT receiver, with pure analog chip power amplifier output stage? (Earl Geddes' preferred amp once upon a time...superb into the correct load). Can I confirm by simply turning off power and checking continuity with Ohmmeter?

As soon as I confirm I'll happily compare performance with Elias' circuit vs. Trinaural. I promise to have an open mind. My speakers full range cutoff in the mid 40s.
 
Last edited:
I should mention in passing the comb-filter coloration that appears in 2-speaker playback with centered sources can be alleviated with quite low Center levels, possibly as low as:

Center = Left 0.223 + Right 0.223, which is 10 dB lower than the usual pan-pot level of Center = Left 0.707 + Right 0.707 (assuming Left = 1.0 and Right = 1.0)

If the primary aim of the Center speaker is to reduce coloration on vocals, as opposed to greater image stability, the optimum Center level may indeed be lower than the usual recommendations.

What I am not sure of is the effect of Center level on image depth; this is where three symmetrical loudspeakers with identical diffraction signatures and crossovers may be necessary for image stability and reduced listening fatigue. (The one sure sign of incorrect multichannel setup is more listening fatigue than straight 2-speaker stereo.)

If the depth begins to suffer, that may be a sign that the Center level is too high for that particular recording ... annoyingly, with modern recordings sourced from studios all around the world, it might be necessary to adjust Center levels on a track-by-track basis.

In practice, both user adjustable Center level and moderate degree of Center-channel EQ may desirable (preferably on a remote). For example, a slight degree of HF rolloff starting at 5~8 kHz might be necessary to subjectively timbre-match the Left and Right speakers; even though all speakers might measure identically, remember that your ears have a different spectral response for sounds arriving directly in front versus phantom images created from speakers that are 20 to 30 degrees off-center. What you need is a subjective tonal match, preferably made with a pink-noise source, for all L, C, and R image locations, and all intermediate locations in between.
 
Last edited:
I should mention in passing the comb-filter coloration that appears in 2-speaker playback with centered sources can be alleviated with quite low Center levels, possibly as low as:

Center = Left 0.223 + Right 0.223, which is 10 dB lower than the usual pan-pot level of Center = Left 0.707 + Right 0.707

If the primary aim of the Center speaker is to reduce coloration on vocals, as opposed to greater image stability, the optimum Center level may indeed be lower than the usual recommendations.

What I am not sure of is the effect of Center level on image depth; this is where three symmetrical loudspeakers with identical diffraction signatures and crossovers may be necessary for image stability and low listening fatigue.

If the depth begins to suffer, that may be a sign that the Center level is too high for that particular recording ... annoyingly, with modern recordings sourced from studios all around the world, it might be necessary to adjust Center levels on a track-by-track basis.

As dedicated Trinaural devotee for 6+ years, I agree with the last sentence. Strangely, the inventor specifically instructs users to never alter C Ch level once it is "properly" set.

I remotely trim C Ch and Bass level (below 80 Hz, Distributed Array with proprietary setup instructions available for public use), and presume such trimming really must be done with a remote for efficiency and convenience. I feel quite sure the dearth of affordable transparent remote 3-ch (or more) line preamps contributed to low Trinaural acceptance. The inventor's instructions (above) and also the lack of Stereo/Trinaural A-B switch also contributed.

I think it noble of Kalman Rubinson to recently recommend Trinaural to stereo devotees at AA.

Sorry to repeat, but my center image and spatial effects, even in stereo, are so good that every time I switch off the Trinaural and toss a blanket over the C Ch speaker, I still have to physically walk up to the C Ch speaker to confirm it is off. Listeners would likely qualify performance as exemplary in that specific area.

But still, switching back to Trinaural instantly shames the stereo format. Compared in this way, the label "phantom" most definitely applies to C Ch image and spatial effects in stereo.
 
Last edited:
I'm curious, ro9397, about your listening experience with a well-tuned 3-channel system (using 2-channel sources). The Ariels project a phantom center that is quite noticeably elevated, by about 3 feet, or well above the height of the 48" high Ariels. When I engage the semi-mid-fi Marantz receiver and use the Center speaker, the image elevation ceases, and is pinned on a midline between the three speakers. Do this happen with Trinaural, or does the Center image remain as elevated as it does in 2-speaker playback?

Also curious about depth distortion. I found that depth perception seems to be critically dependent on Center level, with too-high Center levels reducing the impression of depth behind the speaker array. It hardly matters for movie soundtracks, since they have pretty poor spatial mixing anyway, but it makes a big difference for 2-channel music sources, where depth cues are an important part of the mix.

I have heard superb mixes from discrete 3-channel sources and matched speakers and amplifiers, but only have limited experience with static 3-channel decoders. Back in the day, I did plenty of listening to dynamic decoders like the Shadow Vector, CBS Paramatrix, Tate DES, and Sansui QS, and have learned the artifacts of dynamic decoders (odd shifting impressions, listening fatigue, unwanted detenting at loudspeaker locations, etc.)

The usual downside with most static matrix decoders is blurred localization, although this can skillfully arranged with crosstalk terms ... for example, a moderate degree of antiphase crosstalk in the L and R speakers, which offsets the image narrowing of a too-simple L = 1.0, C = L 0.5, R 0.5, and R = 1.0 decoder.

My limited experience with 3-speaker playback is much greater clarity and transparency for central and near-central sounds, which is very welcome with nearly any recording, but level setting and subjective timbre-matching can be a little tricky. How's it been for you?
 
Last edited:
In the mid 80's I tried 3 identical speakers wired as Elias shows them and found imaging too vague (for me). Seemed everything was stretched across the stage. Admittedly, I am an imaging junkie so those who aren't may really like it. (I also remember a slightly distorted center, but chalked that up to stylus mistracking.)
 
I'm curious, ro9397, about your listening experience with a well-tuned 3-channel system (using 2-channel sources). The Ariels project a phantom center that is quite noticeably elevated, by about 3 feet, or well above the height of the 48" high Ariels.

Within 48 hours I'll experiment and take notes. Till then I'll go by memory. Last stereo audition was a couple days ago.

Trinaural inserts between source and multi channel preamp. For stereo, Trinaural is turned off, CDP directly drives the preamp, and a thick winter comforter completely covers the C Ch speaker.

LCR mains are matched improved clones of the stellar $5k/pr French ASA Pro Monitors designed by a recording engineer: Dynaudio Esotec, every detail highly refined, electrical 1st order in-phase, PIO series tweeter caps trimmed to 2nd decimal or better, litz series mid bass coil, Zobel network, time-aligned with mid bass at ear level/above tweeter with slight tilt back, 1.25" panels (old growth Baltic Birch ply over wood by product), 38# each. Stands are 40# refined/overbuilt DIY "Stubby" with sand filled 5" square vertical support.

CDP is Paul Weitzel/TRL modified Sony DVP-NS900V (DVD deleted, SACD stereo output only). In direct A-B over several weeks, with unbalanced CDP the multi-ch line preamp was indistinguishable from my $7500 Bongiorno designed SST Ambrosia. Amps are pure analog chips, 100W but rated only for 8 Ohm load (think of them like a SS OTL) with long list of passive mods...on the right load the only thing I miss about my Hypex Ncore NC400 mono blocs is wider/deeper stage.

Rooom is well treated, OC703 with ultra high tech dispersion film (newspaper, cryo treated...sarcasm last two words). Normal function (Trinaural) is four piece Distributed Sub Array employing Trinaural internal active high pass for main speakers, in-phase 2nd order @ 80 Hz. Stereo/Trinaural A-B requires subs turned off and mains run full range (toggle switch on Trinaural). Mains cutoff mid to high 40 Hz anechoic, lower in room.

Whew...to your point: yes, definitely in stereo center image is elevated, possibly even as high as with your lovely classic Ariel, which I have long wanted to audition.

When I engage the semi-mid-fi Marantz receiver and use the Center speaker, the image elevation ceases, and is pinned on a midline between the three speakers. Do this happen with Trinaural, or does the Center image remain as elevated as it does in 2-speaker playback?

Open to review, but first thought is Trinaural definitely does indeed lower the image, again, about same as you describe.

Which Processing function to you employ to derive the C Ch? (I once compared Trinaural to the DSP in my Pioneer receiver, the latter which sounded unbelievably awful.)


Also curious about depth distortion. I found that depth perception seems to be critically dependent on Center level, with too-high Center levels reducing the impression of depth behind the speaker array.

Again, ditto. My experience is that C Ch level is critical and requires adjustment. You earlier suggested one might benefit by adjusting C Ch level per track. For some CD's such as live solo piano one setting is fine while other CDs benefit by adjusting per track. Still, cost to benefit ratio favors Trinaural over stereo.

It hardly matters for movie soundtracks, since they have pretty poor spatial mixing anyway, but it makes a big difference for 2-channel music sources, where depth cues are an important part of the mix.

Agreed, with clarification. Comparing my CDP > Trinaural vs. Pioneer HT DAC/DSP grossly favors the former. OTOH, large world class orchestra recorded at Abbey Road and/or Air Lyndhurst studios in London, such as The Hobbit score, played through the Pioneer, is among the best reproduced sound I've heard. Black Hawk Down is another absolutely stellar sound track. I can not overstate how incredibly huge is the stage of such sound tracks, epic scale and layering of depth and breadth, like your eyes are attached to the mic boom overseeing a sea of orchestra members.

Dual use music/HT employs perforated retractable 92" screen, dead quiet Mitsubishi 1080P projector, and total light control (basement with only one light well).

I have heard superb mixes from discrete 3-channel sources and matched speakers and amplifiers, but only have limited experience with static 3-channel decoders. Back in the day, I did plenty of listening to dynamic decoders like the Shadow Vector, CBS Paramatrix, Tate DES, and Sansui QS, and have learned the artifacts of dynamic decoders (odd shifting impressions, listening fatigue, unwanted detenting at loudspeaker locations, etc.)

I have plenty of experience with systems that make one daydream and want to do something else.

This system invites one to listen for several hours at a stretch. The only thing that makes me want to do something else is running out of good music. (My vinyl collection is much better than my CD collection.)


The usual downside with most static matrix decoders is blurred localization, although this can skillfully arranged with crosstalk terms ... for example, a moderate degree of antiphase crosstalk in the L and R speakers, which offsets the image narrowing of a too-simple L = 1.0, C = L 0.5, R 0.5, and R = 1.0 decoder.

My limited experience with 3-speaker playback is much greater clarity and transparency for central and near-central sounds, which is very welcome with nearly any recording, but level setting and subjective timbre-matching can be a little tricky. How's it been for you?

Yes, my experience is similar. As I mentioned earlier, I find the C Ch level control virtually indispensable. Without it I don't really think it works.

For instance, take a solo ultra high end classical guitar like the one played by John Williams on a CD. I sell such guitars and play guitar, so I know the correct image "size" and proper blend of center image with the "reverb" apparently mixed in by the engineer. Proper C Ch level gets this blend just perfect where one hears what appears to be a live guitarist playing on a bench or chair and proper mixture of late arriving boundary effects in the original room, which in reality is an artificial construct of the engineer. It the C Ch level is wrong, there's either too much LR speaker reverberant effect with the direct guitar sound blurred, or the opposite which is too much direct guitar sound in the C Ch and too little reverberant sound from the LR.

With orchestra such as spaced omni Telarc, when it's right you just can not believe the arc of the orchestra in every plane. Likely the best stereo playback I ever heard was the late Bob Crump's last CES display prior to his untimely demise (RIP). That system may be the only one to equal or approach a properly tuned Trinaural system (unfortunately I only played one recording being a well engineered solo live jazz/boogie piano). When Trinaural C Ch level is wrong the depth is flat.

I love to play Telarc's Tchaikovsky 4th Symphony for visitors. This system allows listeners to hear every counterpoint line and still enjoy the whole, keeping them keenly interested in the composers intent as he carries them on an epic and brilliant musical journey. The atmosphere around each orchestra group is incredible, as is the layering in all planes. On lesser systems this is impossible and this is one of the reasons people can't enjoy more complex pieces.
 
Last edited:
I should mention in passing the comb-filter coloration that appears in 2-speaker playback with centered sources can be alleviated with quite low Center levels, possibly as low as:

Center = Left 0.223 + Right 0.223, which is 10 dB lower than the usual pan-pot level of Center = Left 0.707 + Right 0.707 (assuming Left = 1.0 and Right = 1.0)

If the primary aim of the Center speaker is to reduce coloration on vocals, as opposed to greater image stability, the optimum Center level may indeed be lower than the usual recommendations.

Maybe, what you state, it is a sort of lower level limit of center speaker signal, and one should not go under it. While even small level at center speaker reduces comb filtering, I think the full benefit of center speaker comes from enhancing center image especially for non center listeners.



What I am not sure of is the effect of Center level on image depth; this is where three symmetrical loudspeakers with identical diffraction signatures and crossovers may be necessary for image stability and reduced listening fatigue. (The one sure sign of incorrect multichannel setup is more listening fatigue than straight 2-speaker stereo.)

If the depth begins to suffer, that may be a sign that the Center level is too high for that particular recording ... annoyingly, with modern recordings sourced from studios all around the world, it might be necessary to adjust Center levels on a track-by-track basis.

I think there are more to image depth than center speaker level only. More inportant is what is the decoder 'side signal'.

In terms of MS-stereo (Mid, Side), some part of side signal S should be sent to side speakers. In MS stereo M and S are
M = 1/sqrt(2)*(L + R)
S = 1/sqrt(2)*(L - R)

In this case the matrixed output can be written in MS format as
Ls = 1/sqrt(2)*(0.5 M + 1.5 S)
Cs = sqrt(2)*0.5 M
Rs = 1/sqrt(2)*(0.5 M - 1.5 S)


we can simplify
M = L + R
S = L - R

and

Ls = 0.5 M + 1.5 S
Cs = M
Rs = 0.5 M - 1.5 S


So in this matrix the 'side signal' at the side speaker is 3 times bigger than 'mid signal'


In recordings the 'side signal' contains clues of the recording space. It is beneficial to present the signal S through side speakers.



In practice, both user adjustable Center level and moderate degree of Center-channel EQ may desirable (preferably on a remote). For example, a slight degree of HF rolloff starting at 5~8 kHz might be necessary to subjectively timbre-match the Left and Right speakers; even though all speakers might measure identically, remember that your ears have a different spectral response for sounds arriving directly in front versus phantom images created from speakers that are 20 to 30 degrees off-center. What you need is a subjective tonal match, preferably made with a pink-noise source, for all L, C, and R image locations, and all intermediate locations in between.

I'm not sure if there should be EQ or not. Because it is natural perception to have tonal balance varying with the incidence angle. We can think that tonal balance is correctly perceived when dominating energy comes from the intended direction.

Only that 2 speaker stereo cannot do center image properly because all the sound comes from two side locations only.
 
In the mid 80's I tried

Don't you think it would already be time to try again ? :D Everyone should try 3 speaker stereo at least once in a century.


3 identical speakers wired as Elias shows them and found imaging too vague (for me). Seemed everything was stretched across the stage. Admittedly, I am an imaging junkie so those who aren't may really like it. (I also remember a slightly distorted center, but chalked that up to stylus mistracking.)

Maybe it's the conventional confrontation: Unrealistically small sized pin-point image vs. realistically sized image with real ASW.
 
Don't you think it would already be time to try again ? :D Everyone should try 3 speaker stereo at least once in a century.

Maybe it's the conventional confrontation: Unrealistically small sized pin-point image vs. realistically sized image with real ASW.


Current system is 11.1 with a center but, you're right, I should try that setup again.
 
Apology if I mentioned this earlier. IMO a remote "trim"/level control for the C Channel is critical. Jim Smith (former importer Avant Garde horns) advised that Trinaural performance suffers with spaced omni recordings ala Telarc and others.

He's correct, but only if users follow Trinaural inventor Bongiorno's advice to set and forget C Channel level.

I set the Trinaural C Channel level pot (slotted screw driver) for LCR unity gain with a tone generator input/volt meter @ output. (Interestingly, unity gain requires the C Channel pot about +15 degrees above maximum counterclockwise rotation. IMO the lack of preset unity gain is a Trinaural failure, though one can work around it as I did. Another failure is no stereo/Trinaural switch to instantly compare.)

Signal path:

Stereo analog source (TRL/Paul Weitzel modified Sony DVP-NS900V SACD stereo only, Custom Empire 509 TT/Strainguage demodulator) >
Hard wired gold contact toggle switch box >
Trinaural Processor 3.1 analog output, 80 Hz split >
Custom multi-channel line preamp (indistinguishable from my $7500 Bongiorno SSI Ambrosia/2012 TAS Component of the Year) >
Three matched LCR amps (also 950W sub amp powers four distributed subs/proprietary setup, +/-3 dB 100 to 20 Hz throughout the room sans EQ, -1.5 dB @ 20 Hz, conservative 113 dB clean output) >
Three matched LCR custom Dynaudio Esotec monitors (six improved clones of stellar French ASA Pro Monitors, two boxes per channel, patent-pending Late Ceiling Splash radiation pattern)

When the preamp LCD reads "C Ch 0.0 dB" front LCR levels are matched. The large room is symmetrical.

Spaced omni recordings requires C Ch -4 to -6 dB. Lest one think this C Ch level so weak that stereo would be preferred, Trinaural still is preferred.

Other than spaced omni recordings: I estimate 1/3rd sound ideal with C Ch level flat, 1/3rd -1 to -3 dB, 1/3rd +1 to +3 dB.

I never heard it, but NuForce recently released an attractive 8-ch line preamp, balanced/unbalanced inputs/outputs, only $1k MSRP.
 
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.