TOPIC:

Don't twist my words please, I only said your example makes no sense. It seems to me that perhaps the thing you would find most illuminating is doing the nine measurements (post filtering) as Flavio suggested. Please report back when you do, I'm sure we'd all find it educational.

No need to get all defensive here, I didn't twist any words.

And the reason I've stated that in my experience (!) the Dirac predictions were far too optimistic in the first place, is exactly because I've actually done these before and after measurements.

OK, let's see them then.

That was a lot of measurements and I didn't keep them, what for? Their purpose was to check what Dirac does in real life and thus find the best measurement points to do the Dirac measurements.

All I'm saying is that the predicted frequency response by Dirac is far too optimistic for any 'real' spot (not talking about some theoretical average). I'm not putting Dirac down, I'm not doubting its usefulness.

There's really no need to 'challenge' me to show measurements, just relax..

It's not a challenge, it's to try to help you learn something.

Get off your high horse. I have done the experiments and the measurements. Have you?

Based on my actual and factual experience, I've tried to give some advice and what I get is condescending remarks. So far you've yet to show any factual knowledge and just resort to the unfortunately oh so common internet bla bla, without any backup.

I really do not understand why you argue like this.

It is fundamental to know how a sound correction system works in oder to predict its accurency.
Currently, we do not have enough white paper on Dirac Live to get enough knowledge, besides, we do not have anything technical from Dirac either.

As for correction accurency, it can be realistic only at sweet spot with few measures.
With multi measures any odds can happen, and without manual control or advanced analysis algorithm sweet spot correction accurency can be compromized.

In a typical environnement one can check with counter measures and observe accurate reasults compared with the predicted FR curve.
If the environnement show many singular accidents at different measure points, then counter measures will show anomalies compared to the predicted FR curve. This cannot be avoided.
We can relate to other automatic correction systems, which many people would find that the predicted correction is false.
It is always due to bad acoustic environnement and multi measuring.
A single measuring always work and the predicted corrected FR is accurate.
In a "good" acoustic environnement, multi measure "should" work out correctly, so does the predicted correction.

I'm not arguing at all. Just saying that Dirac's predicted response might be true for the sweet spot (like measurement no.1), but not for the rest of the 'optimized' area. Not that big of a deal really, it's just exaggerated, that's all.

It's all good really..

Beauvais wrote:

I'm not arguing at all. Just saying that Dirac's predicted response might be true for the sweet spot (like measurement no.1), but not for the rest of the 'optimized' area. Not that big of a deal really, it's just exaggerated, that's all.

It's all good really..

I think it is quite normal and it depends mostely on the the room .
Besides it is quite reassuring, because it means that the sweet spot is not affected by the other measurements.

Maybe the multi channel Dirac works differently as it is designed for larger listening areas in which sweet spot smoothing makes more sense.

It would be very interesting to get more details about how Dirac deals multi measurements and how it is used by the correction algorithm.

Yes, actually I think there is agreement among all of us... Dirac Live shows the average of the nine measurements (the only version which shows the nine individual curves also is the full standalone PC/Mac multichannel version).
Nonetheless the magnitude response correction is not based on that information.
In order to validate the predicted curve in a single point we should use a single point measurement (which is possible) while the validation of the shown average curve should be done by averaging all of the measurements.
The logic behind the Dirac Live algorithm is discussed here:
www.dirac.se/media/12044/on_room_correction.pdf

Flavio

Is it just me or do others here find the backwards pictures of Couch and Chair confusing to work with?
DD

Actually, it's not just you It took me a while to get used to the mirror image. However, I don't believe the Dirac algorithm cares at all which way around you do it.

john.reekie wrote:

Actually, it's not just you It took me a while to get used to the mirror image. However, I don't believe the Dirac algorithm cares at all which way around you do it.

And it should not !
The first measure point is very important though, also how you point the mic at the speakers middle axis.