TOPIC:

A very interesting last point you make. You are quite right that just bleeding the other speaker in using the SHD Plugin output matrix would reduce the soundstage width exactly like a centre speaker would. It seems likely that since your left/right speaker physical separation is only just more than recommended, reducing the apparent separation may well fill the centre hole just as well as an actual centre speaker. Logically at one extreme of bleeding the other channel in at 0dB the image reduces to mono in which case there can't be any hole in the centre, there just isn't any sound from left/right. Its not hard to imagine that at some lower bleed level you both fill the centre hole and preseve left/right to a substantial extent.

Well OK, I just can't think of any reason that wouldn't achieve the same thing. A brilliant idea.

Thanks for thinking I'm awesome… it feels good !!!

More seriously, my first solution (without a center speaker and without passing a small part of sound from right to left and vice versa), don't you think that Dirac will be able to correct-compensate the sound hole on its own?

Well in any case if I understood correctly I just have to make tests to "see" what it gives, because in addition to the icing on the cake if the result is good, it will allow me to use the SHD to make the bass / treble separation on the 2 side speakers and to bi amplify them directly without going through an external filter making this separation. A French expression says "it's butter and money" ... wins on all sides !!!

A priori proceeding in this way would fit the case of a standard stereo for Dirac and so in theory I would not have any more questions for them on this specific subject, do you agree?

I can't see that Dirac can do anything to fill the centre hole. Only reducing soundstage width can do it I think.

OK thank you for your answer.

However I think if the left/right bleed idea works you won't need the centre speaker and amplifiers.
So all 4 channels will be available for left/right bass/treble biamplification.
So the SHD Plugin routing matrix will just duplicate the bleed for left/right bass and left/right treble.
So you should be able to biamplify left and right and avoid an external filter to seperate bass/treble each side.

Put another way, if you duplicate the exact same bleed for left/right bass as for left/right treble you get the same soundstage width reduction for bass and treble. (there may be reasons why you wouldn't want bass and treble soundstage width exactly the same)

I haven't used the SHD Plugin for quite some time but I believe you can change the names of the input and output channels in the routing matrix. The outputs could then be named something like Left Bass, Right Bass, Left Treble, Right Treble. The inputs could be called something like Left and Right. It should then be obvious where the four bleeds go in the matrix, essentially everywhere you don't have a 0dB.

Apply suitable crossovers between Left Bass and Left Treble and between Right Bass and Right Treble and send the four outputs directly to amplifiers. Apart from the fact that there's some left/right bleed in the matrix its a fairly standard 2.0 two way speaker system then.

We completely agree, yes I would apply the same settings in order to get the same size of the stage to be the bass and treble, I see no reason to do otherwise.

Thanks again for helping me hopefully found the right solution.

Here is a new answer from another Dirac advisor, this answer concerns the question with the central speaker in place:

"Hi there,
I'd like to pass along some advice from one of our developers about your setup:
Measuring Dirac Live through a 3.0 system using only two discrete Dirac Live channels could work. The resulting impulse responses from "Left" & "Right" will include the impulse response from the "Center" channel. The problem now is that the Center impulse peak will essentially be as strong as the Left & Right peak. For Dirac Live to give decent resulting filters in this case it MUST either choose the Left impulse peak for Left speaker analysis & Right impulse peak for Right speaker analysis or choose center impulse peak for both speakers. Our algorithm should be smart enough to figure this out but It could also go wrong. If wrong impulses are chosen the staging will probably be terrible.
In this case, I would say that Dirac Live could help, but this is a nonstandard setup that does not follow our official guidelines, which state that each speaker must belong to its own logical channel for proper calibration.
This being the case, I cannot provide further advice about your proposed implementation about Dirac Live. If you need any more technical support, please do not hesitate to contact us again or open a new ticket. "

My comment: A priori the central channel will play much less loudly than the side ones, so we should be able to use Dirac

I don't understand the reasons about impulse peaks unsurprisingly.
However left/right bleed would also not meet their official guidelines for the same reason.
Which suggests left/right bleed should be minimised just as centre speaker level should be.
Which should be fine for slightly reducing soundstage width to fill a centre hole.
But probably bleed at 0dB which theoretically gives mono wouldn't be a good idea.

Not bad support for a use case well outside their guidelines though.

Thank you for your comment, but this time I don't think I agree with you.

In fact, by removing the center speaker and replacing it with the injection of sound (reduced in power) from right to left and vice versa, we find ourselves in a classic case of stereo, at least for Dirac.

In the case of a sound located in the center (with or without sound injection) we have the 2 side speakers that play it and at the same time, I do not see the difference from this point of view when we inject sound between the 2 ways.

"our official guidelines, which state that each speaker must belong to its own logical channel for proper calibration." appears to be equally violated by left/right bleed as for 3.0.

For a pure 2.0 system, Dirac sends a signal out the left channel and gets a single impulse response from the left speaker.
For a 3.0 system, Dirac sends a signal out the left channel and gets a double impulse response from the left speaker and the centre speaker.
For a left/right bleed 2.0 system, Dirac sends a signal out the left channel and gets a double impulse response from the left speaker and the right speaker.

Both 3.0 and left/right bleed cause the same double impulse response from the left speaker and another speaker. Dirac doesn't care or even know that the second impulse response is in one case from the centre speaker and in the other case from the right speaker. So in both cases the solution to the second impulse response is to minimise its level.

The only benefit of the left/right bleed approach is that the impulse reponse from the right speaker should be almost indistinguishable from the impulse response from the left speaker, certainly more similar than the impulse response from a centre speaker. Assuming left/right room/speaker symmetry. That would certainly be true for the centre measurement position, but not quite as true for the other eight measurement positions.

Thank you for these explanations on how Dirac works, it's clearer like that, and I understand better.

But that doesn't stop us from having another great idea.

Indeed what matters: It is that each left and right speaker is correctly adjusted after passage-calculation "Dirac", in this case it will be enough not to make the injection of the sound between the left and the right at the time Dirac readings, and restore them for listening after calibration is complete.

But is this possible in the logic of operation of the SHD?
• Correction of these parameters in the SHD after the Dirac analysis and before listening?
• or possibly by modifying the files resulting from the Dirac analysis before injecting them into the SHD?

Matrix after Dirac analysis:

The matrix before passage for Dirac analysis is the same except the -9b replaced by off

If you use Dirac first on a pure 2.0 setup and later do the left/right bleed it may work well. It just feels a bit wrong with the Dirac calibration not being applied to the final SHD setup. But the bleed shouldn't change the frequency response so it shouldn't affect that aspect of Dirac. It will change the impulse response and its difficult to know what affect that will have in Dirac. No more difficult to know than if Dirac is calibrated with the double impulse response initially though as I discussed in my last post. The situations are different for Dirac but whether either are correct or one is better than the other I have no idea.

What I would do is setup the SHD as a pure 2.0 system with no bleed in preset 1. Then do a Dirac calibration and listen.
Then apply bleed and listen.
Then copy the SHD routing to a second preset, do a Dirac calibration for that second preset and listen.

The two presets can then be used to A/B the two orders. A is Dirac followed by bleed. B is bleed followed by Dirac. Hopefully one or both would sound correct.

This is really the main benefit of the presets to my mind, they allow you to A/B (or A/B/C/D) different setups to find out which work best. When I find the best setup I only used a single preset, before that multiple are great.

As for modyinfing Dirac files, I'm not sure thats even possible, let alone sensible.

Thank you for your well-documented answer, and I will do as you tell me to do the tests, it is indeed the best way to make up your own mind.

Regarding this case "use Dirac first on a pure 2.0 setup and later do the left / right bleed it may work well", I agree with you that when you say "But the bleed shouldn't change the frequency response so it shouldn't affect that aspect of Dirac ”,

But I don't understand why this shouldn't be the case with impulse response, I don't have such a good understanding of this aspect of sound, but I can't see why this would affect Dirac.

Indeed the impulse response of the right and the left will be corrected by Dirac in pure stereo, the fact of adding sound to the right for the left and vice versa (after Dirac calibration) will make that for a sound coming exclusively from the right played normally in pure stereo exclusively by the right will be played this time by the right and the left… like a sound located between the right and the center in pure stereo… and the corrected impusional responses of Dirac are good in this case.

In short I do not see why this can cause problems .. But here I am reaching the limits of my knowledge and skills.

No matter how far away the bled speaker is the frequency response remains the same but may have an echo.
There are certain of the 8 non-central microphone positions needed for Dirac calibration where the bled echo will actually arrive at the microphone before the direct sound from the main speaker.

I have no idea which microphone positions Dirac needs impulse timing from, possibly just the centre position where it will be identical because time-aligned. But if it uses any of the other 8 it could conceivably get confused by the 4 with pre-echoes. Pre-echoes can't occur with a pure system without multiple speakers per channel so Dirac could conceivably get confused by something nobody ever expected it to see. Or it may not.

So it comes down to what particular magic Dirac uses. Because we're not certain of that we don't know how Dirac will respond and so can't reason about whether using it before or after bleed will have any difference. Hence you're pretty much at the stage where only experimentation can answer these questions. Could I be "worrying too much" about theory? Most definitely.

OK thanks for the additional information, this is not all obvious, and unless we know exactly how Dirac works, we can only imagine.

We just have to experiment and see how the magic will work.