TOPIC:

Out of interest to anyone:

Lets assume the OP gets his channel delays spot on through REW, so Dirac doesn't get confused by seeing two slightly out of sync signals which earlier we've likened to a reflection (only in this case a 'reflection' with equally high SPL to the 'direct' signal).

What then happens if the centre speaker and say the left have different effects on room modes? If say the centre causes a high peak at the listening position that the Left does not or vice versa. Will Dirac then struggle to find the balance between the two competing speakers? If it reduces the frequency to one it has to do for all.

Other threads have discussed not using PEQs ahead of Dirac but in this case would it be worth doing PEQ on the centre channel to reduce peaks ahead of running dirac?

@asx77, that's a very good question.

Any peak on the centre compared to left and right will cause the soundstage width to reduce. Dirac will end up with the frequency response requested but because it will maintain the balance of the mains and centre it will preserve the soundstage reduction.

This also means any notch on the centre compared to left and right will increase the soundstage width at those frequencies.

Conversely a relative peak on mains will increase the soundstage width and arelative notch on mains will reduce the soundstage width.

The key point is arguably the bandwidth involved. For sharp resonances only spot frequencies would be affected and it's not obvious that this would be perceived easily. What would be easily perceived would be wider resonances. Speaker's tonality is often defined by such wider resonances at mid and treble frequencies.

So it is important to use a centre speaker from the same manufacturer as the mains using similar technology. This will hopefully result in better matching of tonality. Its also important to use the centre at the lowest level possible compared to the mains which will reduce the effect of any imbalance.

Beyond that I would use wider PEQ to get closer to a tonality match. I would always try to use cuts to cancel peaks, boosts to cancel notches are more problematic with respect of dynamic range and headroom.

It doesn't matter whether its a speaker difference or a room mode difference as far as I can see, they both cause the same effect but room modes are likely to be narrow band and so hopefully less annoying that wideband speaker tonality effects.

A centre speaker will almost certainly excite different room modes. Anything on the centre line has a zero for all odd modes across the width of the room (That's actually one reason I don't like sitting on a centre line)

Fortunately for the OP his huge room will have room modes starting at such low frequencies that the modal density will be far better than a smaller room. It is somewhat unusual to get bad room mode problems in larger rooms compared to smaller rooms.

As I saw some approximations I re-specified the specifications at Dirac and I also took the opportunity to pass your last questions:

"And I will say like you" Let's see if I understand you correctly ":

You tell me: "the left channel will be emitted by both your left speaker (output 1) and your center speaker (output 3) which will do the midrange / treble."
Yes and no, to be perfectly precise: "the left channel (full band) will be emitted by both the left speaker (output 1) and the center speaker (output 3 medium / high and output 4 bass), the central speaker like the left and right speakers is bi-amplified.
The difference in processing between the left (or right) channel and the central channel is at the crossover level. This is the SHD crossover that I use to split the high midrange bass for the center speaker into two parts, while the crossover for the 2 side speakers will be outside the SHD (as it works today). hui - JBL filter before amplification).
I chose this solution because of the limitations of the SHD device (2-way input and 4-way output), but if you have a better way for me to do this I'm totally OK.
The day when there will be a multi-channel device of equivalent or superior quality to SHD and which also performs the same functions (crossover, Dirac, Dac, digital and analog inputs, equalizer, etc.) in a single device or possibly in several connected digitally without loss quality by the way, then I would ask myself again to do everything in this device and get rid of my analog JBL filters.
I think a priori the SHD will align in time the output 1 against 3 and 4 (grouped in the central) as well as the output 2 against 3 and 4. Am I clear?

You also tell me, “Also, the left and right channels are completely independent, so I don't see a problem with correcting there. "
To be very precise: Yes of course the left and right channels are totally independent at the input of the SHD, but not quite at the output: channel 1 and 2 (right and left) are completely independent but channel 3 and 4 are each adding the left and right track.
You continue with: “Dirac Live will measure and correct the left and right channels independently, but will also ensure that their respective phases are also aligned. "
And to be more precise ... but I don't know if my reasoning is correct because I am not a specialist in these questions: Dirac Live will measure and correct the left channel (which itself consists of the left speaker and center speaker independently of the right channel (consisting of the right speaker and the center speaker).
Thus the phase of the left channel will be determined by the left speaker + center speaker assembly and Dirac will deduce corrections which will go in a certain direction, when it switches to the right channel it will not take into account the corrections provided for in left ... and the corrections it will apply to the right channel (including the center speaker) may duplicate the center speaker or the opposite ... that's what I understand but maybe I am wrong? I now believe that my reasoning is indeed wrong.

If I take a simplified example in a field that I master better and that Dirac will also analyze / correct: If the left channel (left speaker + center speaker) has a peak of 12db at 3000hz, Dirac will correct by applying a correction of - 12db, it will then go to the right channel (right speaker + center speaker) which for example has a dip at 3000hz of -6db, Dirac will correct by applying a correction of + 6db (I know it does not do exactly that but it is for simplicity).

What will be the total result? -6db At 3000hz for the central speaker ... this may actually be what is needed and which will allow to have a very stable location because at the change of frequency in the center of the stage the left-right distribution will remain stable thanks to the work of Dirac. I believe that by forcing myself to clear my mind, I understand better. Thank you.
But the same reasoning can be held for the phase, the impulse response (on all the points on which Dirac are working) and there I am incompetent and I therefore ask myself the question. Will Dirac be able to do this in this particular case? "

Thank you for your 2 questions which seem very interesting to me.

There is something Entripy that I don't understand in your reasoning:
“Any peak on the center compared to left and right will cause the soundstage width to reduce. Dirac will end up with the frequency response requested but because it will maintain the balance of the mains and center it will preserve the soundstage reduction.

This also means any notch on the center compared to left and right will increase the soundstage width at those frequencies. "

The above I agree, on the other hand less here (unless it is a trick of the automatic translation which is badly done?): "Conversely a relative peak on mains will increase the soundstage width and arelative notch on mains will reduce the soundstage width. "

What is the "mains"? Are these the 2 side speakers?

If this is the case, as I understand it, for me a peak or a dip on one of the side speakers will cause the soundstage to shift to the right or to the left.

“So it is important to use a center speaker from the same manufacturer as the mains using similar technology. This will hopefully result in better matching of tonality. Its also important to use the center at the lowest level possible compared to the mains which will reduce the effect of any imbalance. "
We fully agree, moreover that's why I asked your opinion if we had to try to bring together the low / medium-high cutoff frequencies which are not located exactly at the same frequency between the side speakers (Khorn) and central enclosure (Cornwall).
Regarding the rest of the interventions ... I am not competent enough.
Thank you anyway for your involvement.

Yes, the mains are the main left and right speakers..
Any imbalance between left/right and the centre will cause the soundstage to narrow or widen at those unbalanced frequencies.

As for the crossover for centre speakers. Given that crossovers are intended to blend the drive units above and below the crossover frequency optimally I would think having them at what the Cornwall naturally wants would probably be best. I see no reason why the Cornwall crossover should be matched to the crossovers for the Khorns. But maybe they would be better if they were the same. This is something that you can only be certain about by experimenting. Since the SHD has 4 presets with separate settings it would be simple to set up one preset with natural crossover frequencies for the Cornwall, and another preset with the crossover frequencies for the Cornwalls matching the Khorns. That would allow you to do rapid A/B comparisons.

Since the SHD has 4 presets and its relative simple to do Dirac calibrations its often best to just try things and not overthink things too much.

Thanks for this answer, yes of course I would experiment to determine the best frequencies for the crossover, as well as the power needed for the center speaker, the presets will be used for this.

To fix the ideas I draw the planned diagram of the installation



.

That all looks as I expected.

Active centre should be so much better than passive centre. I can't actually imagine how passive centre could possibly not ruin left and right.

I don't quite understand what you are saying. Maybe this is due to the specifications of the different technical language between French and English?

Anyway, yes of course the bi amplified center should be better than mono amplified, is that what you mean?

Or do you mean the center is "active" because it will also receive Dirac corrections? Yes in this case I also completely agree with you.

I got a response from Dirac after I detailed the connections, here it is:

“Thanks for the detailed answer.

I'm afraid that if the "Left channel" is measured and a certain correction is found necessary it should be preserved while in this configuration the calibration of the "Right channel" could modify it ... in other words, those don't seem to be fully independent channels even if all drivers are time-aligned.
My understanding is that this was your concern and I agree with you.
(it's possible though that in real life the problem will not be relevant) "

My impression is that Dirac didn’t really understand my subject. And it’s a shame, I don’t know if I should go back to them and how?

What do you think ?

Dirac are reasonably making the point that the centre speaker is affecting both the left and right measurement calibrations which as a result are not fully independent. They are concerned that because the left/right calibrations are performed sequentially the second calibration could undo some of the effects of the first calibration because of the shared centre speaker.

They can't actually guarantee Dirac will work for your application but admit it may well do. It seems likely that you probably aren't communicating with the very clever mathematicians that came up with the theory though, first line support normally have a clue but probably not enough knowledge for your unusual application. There should be a definitive answer.

My own belief is that the Superposition Principle should mean it does work. That way of looking at the problem is that the centre speaker is not itself calibrated, rather the left calibration is applied to the left signal going to the centre speaker and the right calibration is applied to the right signal going to the centre speaker. There isn't actually anything that can be undone by sequential calibration as far as I can see. Similarly the left and right measurements are never fully independent because they are affected by being in the same room, clearly Dirac applies separate calibrations for left and right even though they both affect the same room without undoing each other. Is a common speaker mathematically different to a common room? I can't see that it can be. Dirac can't actually tell which part of the measured response is due to the speakers or the room.

It also seems very likely that the left calibration that applies to the centre will be very similar to the right calibration that applies to the centre but even in a pathological case where they differed hugely I can't see that it would actually matter.

I believe Dirac do understand your concerns, they just aren't willing or able to give a definitive answer.
Unfortunately I can't give you a guarantee that it would work but the omens appear to be very good indeed.

Entrypy I totally agree with you;

Throughout the discussion here and the reflection my ideas have become clearer and I come to exactly the same conclusions as you, and also regarding Dirac support.

Thanks again for all your answers.

With the exception of time delays which I think you’ll need to get exact using REW ahead of Dirac I can see how Dirac could deal with the situation.

What I can’t get my head around is how you’re about to mess with the soundstage. I’ve come across a few well mixed albums where instruments or sounds appear to come from miles outside of the speakers. Don’t ask me how it’s done but it’s sometimes there. Maybe they put some sounds out of phase yet more to one side to give the impression. Maybe it’s lucky reflections. Who knows but No doubt the sound engineer has created the mix assuming two discreet channels. I wonder how a mono centre is going to mess things up. I don’t think there is any such thing as the perfect system. There will always be a compromise. Fix the hole but at what price?

@asx77, we covered the fundamental effect of the mono centre near the start of the thread, the effect is reduction of soundstage width. The OP was initially intending to have the centre at -12dB which would minimise the soundstage width reduction significantly, as well as many other secondary effects caused by the mono centre.

As for sounds outside the speakers, it's done by subtracting parts (rather than adding to get mono). Nothing more complex than that. (Well Ok, there are more complex ways of doing it from Zucarelli heads to Roland RSS for more 3D sound)

Hi,
Sorry I read the earlier post but missed the fact that there was agreement that the soundstage would narrow.

The simplest way of explaining the soundstage width reduction is to consider left, centre and right speakers.
If the centre is disconnected the image is full width between left and right, its a simple 2.0 system.
If left and right are disconnected the image is at the centre with no width, its an archaic 1.0 system.

If left, centre and right are all at the same level the width is half the full stereo width. This is because we perceive the location of sound coming from two sources at equal levels as midway between the two sources. So left appears to come from midway between left and centre and right appears to come from midway between centre and right.

Its not difficult to imagine that if the centre then gradually reduces in level the image gradually moves from midway between speakers out towards left and right. This is likely where the OP will need to be, some soundstage width reduction but some filling in of the centre hole.

This all relies on proper time alignment of the centre with left and right so that image position is only controlled by relative amplitudes and not relative timings. It also relies on similar tonalities from all speakers.

Anyone with a 2.0 system can actually experiment with this. pretend either your left or right speaker is a centre speaker and sit in front of it. Then reduce its level and see where the image goes. Then pretend the other speaker is the centre speaker and sit in front of that. Reduce its level and see where the image goes. A full 3.0 system is the superposition of these two test systems. To do it properly you should really sort the time alignments out differently for each test but the effect should be clearly audible without doing this.

I told Dirac support that I was sorry that a company like theirs could not give me a substantiated and definitive answer.
To my surprise, support told me to pass the ticket to one of their colleagues so that I have a second opinion, which I find very professional. Well done to them.
I thanked them and took the opportunity to present what I intend to do in a slightly different way, hoping it would be more understandable to them:
"Thank you, in fact the system that I am creating with 3 speakers and as it will be powered, should ensure that the center speaker will virtually disappear as well as the side ones and there will be virtually only one left channel made up of the left speaker and the center speaker and ditto for the right.
Thanks for the opportunity to get a second opinion. "

Entripy your explanations are crystal clear.

Just a small adjustment, the power coming into the center speaker will be determined by testing, I don't think I would go down to -12db, but from the first tests it will vary I think between -6 -9db for music and -9 -12db for a movie for example. Because as you indicate if the 3 speakers play at the same power the scene is already reduced by half, which is huge

In one of your previous interventions if I understood correctly you indicate
“Any peak on the center compared to left and right will cause the soundstage width to reduce. Dirac will end up with the frequency response requested but because it will maintain the balance of the mains and center it will preserve the soundstage reduction. "
You indicate here that Dirac in the case of a peak in the center will try to preserve the balance between the center and the sides ... can we suppose that he does the opposite during a "hole" in the center ?

If this were the case, perhaps the mere presence of Dirac could be sufficient to remove the central hole without a physical enclosure there and put the system back in a more standard case?

In the case where Dirac does not do this kind of thing, in any case in these circumstances an idea comes to me… which goes against the grain of what we usually have to do in stereo… Please do not cry scandal because this will sound like a "hack"

What is the role of the central speaker in my diagram ?, it is used to make the soundstage narrower, how does this work: The central speaker will play the left and right channels, which means that the left lane will end up shifting to the right and the right lane shifting to the left.
In the standard case (2 ways) for a stereo sound at 90 db on the left and 75 db on the right, this sound will be felt as not being in the center but between the left and the center, at home in 2 strict ways it will will be the same, but this sound will appear to come even further to the left.

Isn't there a way to make this sound take a less left-hand location (like with the center speaker in place)? Well I have a "hack" means, just inject into the left speaker a little bit of the sound from the right at -12db or -30db .. To test, which would give for example:



In the case of a perfectly centered sound, the fact of proceeding in this way this sound would be reinforced in the center (since present as much on the right as on the left), in fact the more a sound would be centered the more it would be reinforced ... is what this kind of hack could make the hole disappear without of course causing other problems, that is the question.

From my understanding the depth of the soundstage with such a medium would be adequately reduced to the width of the stage.

These are my last thoughts of the night !!!