Some questions about AL

[Leo11]

I'm considering to buy AL to correct my system and I'm experimenting a bit to see what can I do.

Some things are not clear to me.

When I load multiple measurements for multiseat correction and then I create the filter, AL warns me that the original measurement file will be modified to integrate new information.
And this happens every time I generate the filter.
Checking the files of the various measurements are all modified at the same time (seeing the date / time of the last modification).
The question is whether these measurements are irreversibly altered and if they are no longer faithful to the original.

The other thing that is not clear to me is if the noise reduction function must be applied singly to all the measurements for multiseat correction or it's sufficient to the main.

Regarding TTD correction, it's just an all-pass filter that corrects the impulse in the time domain or even in magnitude?
What does the correction of the room do different from the TTD instead?

Finally, experimenting with the TTD window I don't understand the way the filtered measurement approximates the GD.
Where in the raw measurement there are GD peaks, these are blunted when filtering with low FDW values.
However, by progressively increasing the width of the window, the filtered GD tends to assume at first an opposite trend to the raw one (which is what I no longer understand).
Then continuing to increase the window, in the filtered GD a peak appears in the same direction as the original one, but much higher and sharp.
Only by bringing the window values very high at the end does the filtered measurement take on the trend of the original measurement (always in terms of GD).

The problem, besides not understanding how the GD deviation should ideally be approximated respect to the raw meas, is that FDW values are unique to all speakers. But for each speaker a certain FDW value works better or worse in approximating the raw one.
It's hard to compromise for everyone using the same FDW.

Can anyone give me some tips?
@juicehifi ?

 

[Leo11]

Below are some explanatory screenshots.


Starting with TTD 2/1 the measured GD peak at 75hz is not included in the filtered measure.

With TTD 2/2 the filtered measure takes the opposite direction.

With TTD 3/2 three measures go in the right direction while one worsens on the opposite side.

What is the criterion for finding a compromise?

Also, how much should the filtered measure include such GD peaks?
If they are, as we assume, room effects, they should be left to correct it with the appropriate window (Room Correction FDW), right?

 

[juicehifi]

I’m afraid the question doesn’t have a simple answer.

From time to time the size of the TTD window makes a significant difference, other times not so much.

So this is a bit of qualified trial and error.

1st priority is a best possible step-response, but with a reasonably clean entrance.

2nd pri is correction filters that does not have substantial pre-ringing.

The TTD window that most often behave well when I am troubleshooting is 6-4. It may sound like a lot, but it often leads to the best interpretation and a healthy ttd overall.

 

[Leo11]

Thank you Bernt for your contribution.

I will do some experiments.

What do you think about determining the amplitude of the FDW using the spectrogram in REW?

Shouldn't Peak Energy Time, excluding large room resonances, be a good indicator of the necessary window?

I also take this opportunity to ask you another question: how do I improve the IACC?
The IACC post Audiolense is lower than the original one or the one I get with Dirac Live...

 

[juicehifi]

Iacc should usually be very good after correction. So I recommend that you find out where the difference is and take it from there.

To your first question: What do you mean by FDW amplitude?

 

[Leo11]

Iacc should usually be very good after correction. So I recommend that you find out where the difference is and take it from there.

To your first question: What do you mean by FDW amplitude?

Thanks.

When I can, I will try to post an image of Rew's IACC measurements.

However, I mean the time length (millisec) of the frequency dependent window.
From the Rew spectrogram you can see more or less the distribution of the Peak Energy Time. Leaving aside the sudden peaks due to reflections, you could see which window length to set to intercept the right amount of information for the TTD, without entering too much "room".
Clearly, using the same zero time between REW and Audiolense.

 

[juicehifi]

I haven’t tried to use the REW spectrogram for this, but it doesn’t hurt try it out.

 

[Leo11]

Bernt, take a look at the attached graphs showing full, early and late IACC coefficients for raw speakers (in this case excluding subs) and speakers + subs corrected with Audiolense and Dirac.
Audiolense seems to worsening IACC over 4kHz compared to raw system.
All measurements were taken in the main listening position and then were aligned in REW with cross correlation between L and R, so no error should occurs.

Also, look at the digital loopback measurements of AL filters compared to Dirac. For unknown (to me) reasons, with AL there is 1 cycle difference between L and R filter.
What's the cause?

 

[juicehifi]

The phase plot doesn’t look normal. I suspect that something has gone wrong somewhere.

 

[Leo11]

I am investigating and indeed it seems that the cause of the phase shift is the convolver (MCFX) or REW. But it looks more like an artifact of measurements. Let's leave this question out for a moment then.

As for the IACC, what do you think?
Let's say that I would also expect a general improvement, but instead there is a trend similar to that of incorrect speakers and then a worsening above 4khz.
What can I do to investigate?

 

[juicehifi]

If the simulations in Audiolense look good perhaps there is a problem with the measurement quality. What is your gear … microphone, dac and speakers?

 

[Leo11]

Mac M2 Mini
REW latest version
Kushview Element DAW
MCFX Convolver (VST)
Motu Ultralite Mk5 (USB)
Beyerdynamic MM-1 with level calibration file (Ultralite is the mic pre so no clock drift between in and out).
Hypex NC252MP amplifier
Dali Opticon 2 MK2 speakers

Everything at 44.1kHz.
Measurements were taken with acoustic timing reference.
To loopback the REW signal in the DAW I use the Ultralite digital loopback channel.
Before to calculate IACC in REW I have performed the cross correlation alignment between L and R measurements (to avoid mic positioning error).

 

[juicehifi]

I am investigating and indeed it seems that the cause of the phase shift is the convolver (MCFX) or REW. But it looks more like an artifact of measurements. Let's leave this question out for a moment then.

As for the IACC, what do you think?
Let's say that I would also expect a general improvement, but instead there is a trend similar to that of incorrect speakers and then a worsening above 4khz.
What can I do to investigate?

You should expect a very tight and precise sound stage and a very good timbre, so tour expectations are on the mark.

The most frequent problems are measurement problems, mostly glitches in the data stream. But I would expect your your mic and dac combo to
be a good one in that regard.

Could you post some screen dumps from Audiolense, close-up IRs of measured and simulation? And full range frequency plots of smoothed measurement and correction and target?