TOPIC:

Yes actually as described in another post I’ve had to do that with my left and right being different distances away so they tie up with my sub. It was more to make the point if you’re not equal distance from corners plus your centre is a different distance the OP will have some REW measurements and plugin delays to apply ahead of Dirac. Appreciate though I didn’t really make the point. My mind wanted to but my fingers didn’t!

Ultrasonic wrote:

entripy wrote:

Given that Left/Right 2.0 systems are a bit of a crude way of approximating plane waves with a full width sound stage anyway shoving centre speakers in at -12dB probably won't mess with things as much as you might expect. Its really just inserting the signal that should be there anyway but isn't in this particular case.

Only for a signal that is panned dead-center. For any other signals it is adding an additional/incorrect sound.

If any miniDSP were tried I really don't think it should be a Dirac Live capable one as the algorithm will not even be remotely be designed to handle a situation like this.

I just don't understand that. Do the maths and all it does is reduce the sound stage width. It halves the width for 0dB center and reduces the width slightly for -12dB centre. Nothing else assuming full range mains and centre. If the centre doesn't go quite as low as the mains the soundstage may return to full width at LF but that won't be a problem.

Well OK, that's assuming point sources and flat frequency response ideal sources. In practice unmatched frequency response between mains and centre would cause the image position to wobble between speakers at spot frequencies. How big a wobble is impossible to predict and is the exact same effect you get from image reflections from side walls anyway.

Perfection shouldn't be the enemy of an improvement here.

You are probably already aware of this but Klipsch suggest creating false corners as one solution to the room being too wide. Given the complexity of that approach even if it was acceptable I would still investigate a centre speaker first. Alternately they suggest adding something like a Klipsch La Scala centre to create a three speaker array.

Thank you all for your answers, I will try to answer your questions.

You can't say you hear a sound hole, I haven't done any measurements, but subjectively it seems when listening that the center has less energy ... I would say in the order of -9db. Moreover, Klipsch suggests this solution when the speakers are too far apart (head above 4m).

OP? Excuse me I'm not used to English terms, what is this?

The Klipshorn are loudspeakers designed to be placed in the corners, the bass consists of a folded horn and it is in fact the corners of the room that end the horn.

A solution exists which is to make a wedge to bring the speakers together, but with me it is not applicable.

A music processor could indeed be a solution ... but not complete because in this case it does not do Dirac or if it does there are not all the possibilities of crossing and working on the sound.

I agree with what Entripy is saying.

So yes, the further away we go, the fewer holes there will be, but I don't really have that possibility.
Adding a central, will not make the system monophonic, if we imagine that the central plays as loud as the side speakers (for the example), compared to the listening point, the left (or right) speaker which is located 3m to the left of the listening position will virtually seem to be positioned at 1.5m… the soundstage will be reduced here to 3m.

The 3 speakers are not on an arc but aligned, the timing must be ensured by the delay.

OK, I have taken note of the information about Rew.

“It wouldn't be possible to get an exact left + center time alignment in one place. Dirac relies on several measurement locations, for which the varying delay between the two will cause signal variations that will not at all fit the assumptions of the Dirac model. "
That’s exactly why I asked the question here to try to understand. How do you do the time alignment? There are 2 possibilities either physically based on the location of the speakers - which I believe to be the right solution, because the side speaker will play its score when necessary and the central it will have to have a slight delay because it is closer of the listening point .. But how does Stereo Dirac do in this case ?, ie 2nd possibility, based on the virtual point resulting from the left or right set; in the example cited above (center speaker playing as loud as the side speakers) the soundstage is reduced to 3m, the virtual left speaker would then be 1.5m closer to the center.

“Well, that assumes point sources and ideal sources with flat frequency response. In practice, an unparalleled frequency response between mains and center would oscillate the position of the image between the speakers at point frequencies. It's impossible to predict the magnitude of a wobble and it's exactly the same effect you get from image reflections from side walls anyway.

Perfection shouldn't be the enemy of improvement here. "

Totally agree, it is in this spirit that my central is also a compatible Klipsch (Cornwall) and that the rest of the system (amps, etc.) will also be ... the only point here that bothers me is is that the bass / midrange cutoff frequency is not originally located exactly at the same place between these 2 types of speakers ... since I have the possibility of modifying this parameter by the cross over, I wonder if is it worth trying to make them the same?

Now it seems to me that Klipsch is now advocating the Heresy over the Scala which I already bought on a good opportunity.

I have already done tests with this 3rd center speaker by trying the connections indicated by Klipsch (additional cables between left and center speaker (ditto on the right), there were positive and negative.

On the positive side: indeed we can make the central hole disappear, the soundstage is much more precise .. and by playing on the power of the central one we play on the width of the soundstage.
On the negative side: this passage through cables connecting the 3 speakers causes that we have residue left to right and vice versa .. and that's not good ... that's why I'm looking to m 'orienting towards a completely independent “electrically” speaking power station, the processor doing the job perfectly… remains to be seen how exactly Dirac will work, which I also care about.

OP is Original Poster, or Original Post, depending on context, so shorthand for either yourself or your original message.

www.minidsp.com/applications/acoustic-me...ts/292-rew-timing-ht

That article explains how to use REW to measure the delays of each speaker in a system. In practice REW appears to give more accurate results for delays compared to calculations based on physical measurements, possibly because the effective acoustic centre of drive units can be somewhere counter-intuitive, this is true for small speakers and not likely to be less true for large horns.

I don't see any reason Dirac won't achieve a decent result. All it does is measure left and right outputs seperately and do clever maths. very clever maths but it still sees however many speakers you have as some response when a left signal is output and some response when a right signal is output. It doesn't know or care that several speakers may give output for left or right. As long as the system sounds reasonable Dirac should be able to cope. The one key point with multiple speakers is that you must perform the time alignment prior to using Dirac, Dirac is clever but not that clever that it can compensate for differently delayed signals in one channel. Once Dirac has measured the seperate delays for Left, Centre and Right you plug these delays into the SHD plugin channels 1, 2, 3 and 4 to properly compensate for the delays and achieve full time alignment between speakers.

You also have to remember that the measurement microphone is monophonic. It doesn't know where the sound is coming from when a signal is output to the left, it just knows how loud it is. Dirac has two channels of control but that's effectively why Dirac doesn't care how many speakers you have, it can't tell where they are, just how loud they are in combination.

I'd suggest contacting Dirac Live directly. I suspect they'll advise against using their product in this instance but they are the people best placed to be advise here.

Yes you are probably right, but I do not know Rew, it only remains for me to study him, and indeed in the large pavilions the same problems arise and sometimes worse, Rew should be of great help here. .

It all seems clear to me, indeed Dirac should be able to cope with this situation..hope.

AlainHorn wrote:

It all seems clear to me, indeed Dirac should be able to cope with this situation..hope.

You will have measurements with two interfering impulse response contributions, with varying time delay between the two depending on the measurement location. This to me is a big reason to think that Dirac Live will NOT work as intended/desired.

Ultrasonic wrote:

AlainHorn wrote:

It all seems clear to me, indeed Dirac should be able to cope with this situation..hope.

You will have measurements with two interfering impulse response contributions, with varying time delay between the two depending on the measurement location. This to me is a big reason to think that Dirac Live will NOT work as intended/desired.

It's not even clear to me that Dirac does much to suppress echos anyway. I vaguely recall a claim that it may halve the level of early echos. Even with a single speaker the impulse response will vary at the different measurement positions. Two different source positions concurrently appears to double the complexity of the impulse responses but doesn't obviously make it beyond the amount of compensation that Dirac attempts. At any given measurement position the impulse response from two sources is static, just as with a single source. In either case the impulse responses differ at the individual measurement positions which is perhaps why it's such a hard problem for Dirac to deal with effectively.

My personal experience is that although I could hear effects of reflections in my room Dirac did not obviously help but acoustic treatment did so I remained unconvinced of Dirac's abilities in this area.

entripy wrote:

My personal experience is that although I could hear effects of reflections in my room Dirac did not obviously help but acoustic treatment did so I remained unconvinced of Dirac's abilities in this area.

Although I think I was lucky not to have reflections, my seating position is against the rear wall. The impulse response was showing a second blip shortly after the main. Dirac didn't get rid of it for me, instead while listening I place a couple of cushions on the back of the sofa. Problem solved.

Doesn't mean that Dirac didn't improve the impulse response, but it was cushions that got rid of the reflection.

entripy wrote:

It's not even clear to me that Dirac does much to suppress echos anyway.

It doesn't, as it's not possible. My concern is more that the the second signal may be treated as a reflection from the point of view how the Dirac algorithm 'decides' what to 'correct' or not. It could even do something as simple as apply a time window that may end up excluding the signal that arrives second.

This is just so different to what the algorithm was designed to handle that I see more than enough reason to actually ask Dirac Live themselves rather than speculate about this.

New speakers in better positions would still be the route I'd go for what it's worth.

Ultrasonic wrote:

entripy wrote:

It's not even clear to me that Dirac does much to suppress echos anyway.

It doesn't, as it's not possible. My concern is more that the the second signal may be treated as a reflection from the point of view how the Dirac algorithm 'decides' what to 'correct' or not. It could even do something as simple as apply a time window that may end up excluding the signal that arrives second.

This is just so different to what the algorithm was designed to handle that I see more than enough reason to actually ask Dirac Live themselves rather than speculate about this.

New speakers in better positions would still be the route I'd go for what it's worth.

What second signal? The centre speaker is time aligned with the mains so the direct sound is received at the same time as that from the mains. The reverberant field would be more complex but would necessarily occur after the direct sound.

I still don't see that its so different to a 2.0 system. I also don't see that Dirac would even know it was a 3.0 system. It still puts out left and right signals, receives back left and right responses, balances the frequency response for each and may or may not reduce the level of certain reflections. I've long since read the Dirac technical paper and can't recall anything that would cause it problems.

Let's face it, anyone who has Klipschorn speakers really really wants them, you don't buy them by accident or on a whim, so suggesting different speakers probably isn't realistic For similar reasons I avoided suggesting a different room.

Echo cancellation is theoretically possible, up to the latency of the processor, so maybe up to 25ms or so for the SHD. It's problematic to fully cancel though as it likely to create echos in some listening positions as much as cancel them in others. The echos are also likely to have a different spectral content compared to the direct sound so you often only tend to cancel certain frequency components. That is possibly why Dirac is modest in its ambitions in this area.

The question is complex and not knowing exactly how Dirac works, I am not sure that it works…

so I will ask them my question on my case.

I will send you Dirac's answer.

Yes, for sure I wouldn't change my Klipschorns, I'm too in love with them !!!

Hello everyone, and thank you again for your contributions which have kept me moving forward and clearing up the subject.

So I made the request to Dirac.

At first, they did not quite understand the subject and the answer was:
“I understand that miniDSP offers a US $ 10 plugin deriving a Center channel from the Left and Right channels:
www.minidsp.com/products/plugins/rear-center-channel-detail
I don't know if it will work with the SHD, that's a question for miniDSP to answer: "
I am giving you this answer because it might also interest me.

So I repeated my question at Dirac, here is the answer:

"Thanks for the update ... yes, Dirac Live will be able to correct those two channels,
there might be some (small I expect) compromise though as Dirac will "see" outputs from two different sources at different distances ... b.t.w. I understand that you expect that already.
Indipendent channels for Front, Center and Left would be ideal and, in my opinion, that's more relevant than the differences on paper between the SHD and a DDRC-88 but your case is somehow special as you need the best possible SNR because of your ( very nice) high efficiency speakers.
So I can't give a definitve opinion on the best choice but I can say that I had several reports of excellent improvements using Dirac Live with the Khorns "

Dirac's answer is interesting, but not entirely satisfactory to me, so I completed my question with:

“Thank you very much for your reassuring answer.
But there is something that I do not understand well in your answer… probably because I do not know well how Dirac works.
You tell me "there might be some (small I expect) compromised though as Dirac will" see "outputs from two different sources at different distances ... b.t.w. I understand that you expect that already."

For me with the SHD I will indicate exactly where the left channel is located in distance (or time) from the listening point and I will also indicate where exactly the central channel is, I think it's the right one solution. I also point out that the 3 speakers are located along a wall - the center speaker is therefore closer to the listening point.

So when Dirac will run for his calculations and the left channel will be activated by the Dirac “sound”, if we suppose (this is an example to simplify the reasoning) that the central channel is set so as to play as loud as the left lane, then there will be only one lane that Dirac will “see”.
This resulting left channel will be located in my example halfway between the left speaker and the center speaker on a plane slightly behind the center speaker (the latter being physically closer to the listening point) and slightly forward of the left speaker.
Same reasoning for the right lane. So in my opinion if we limit ourselves to that, it should work.

But the problem that I see, if I understood correctly to make his calculations Dirac analyzes one way after the other, I do not know if he also does a combined analysis of both at the same time ... thus the risk in my case is that once the left channel is analyzed Dirac foresees a certain number of modifications concerning the left channel (including in my case also the central speaker)… then by switching to the right it will not take into account the modifications envisaged on the central channel (planned by Dirac following analysis of the left channel) and he will repeat a passage as if nothing had happened for the right, in turn leading to modifications on the right and on the central speaker ...
All of his modifications on the central channel from the left and the right can be duplicated or on the contrary canceled. And I don't know if this can be seen on the different graphs and possibly be corrected manually?
Am I making a mistake in reasoning?"

This is where we are, so I'm waiting for another response from Dirac ... more in the next issue.
If you have any comments on my questions and Dirac answer please do not hesitate.

Re: the problem you see...

Dirac will make one set of compensations for the left signal to the centre speaker and another set of compensations for the right signal to the centre speaker. Neither will cancel out the other. When you play audio, both compensations will be used at the same time. This is all correct and proper.

Put another way, Dirac never actually compensates the centre speaker. The SHD derives the output for the centre speaker from the Dirac Left and Right compensated signals. Nothing cancels or is missed.