Excess group delay and mismatched phase response

[RenPa]

Thank you
So there are 9 measurements there, can you please explain them?
Which are the two that you would want to compare?

Sorry I should have done it first without you asking for it.
Take this mdat instead.

GD for ASR thread..mdat

1 file sent via WeTransfer, the simplest way to send your files around the world

we.tl

Speakers G=L, D=R, C, ArrG= Surround left and Arrd= surround right and they are measured in that order.

Measurements 1 and 2 are front left and right spl large Dirac off and no bass management, speakers set to large.
In the following measurements 82 is the xo for the calculated DLBC filters
Measurements 3 -7 with the name «Tooles 2V 82» are the 5 speakers with DLBC filter on that sound better for me (first image in my previous post).
Measurements 8-12 with the name «Tooles 82» are the 5 speakers with the other DLBC filter less good (second image in my post).

So I compared «G mlp Tooles 2V 82» vs «G mlp Tooles 82»
and «D mlp Tooles 2V 82» vs «G mlp Tooles 82»

 

[ppataki]

OK, clear, thank you

So when comparing number 3 and 8:

Frequency response - number 8 does not have that dip

Distortion:

Number 8 has way lower distortion

Group delay:

There are ranges where nr8 is better and others where nr3 is better but the fact that nr8 has no GD peak at around 90Hz (where the frequency response dip is) would make me vote on that

The Wavelet diagrams both look pretty horrible to be honest

The fact that they look different would make me believe that they sound different too
Nevertheless, these wavelets show an immense amount of delay in the lows - meaning that starting from 200Hz downwards you have a huge group delay (even up to 95ms)
The GD curves confirm the same for both measurements

I am not entirely sure why that is happening, I would guess you have ported speakers but even then this level of delay would be way too much... (with sealed speakers the max. peak energy delay value I saw in this very same graph was 10-15ms, even when considerable (+10dB) low shelf gain was added)
I would experiment with placing the speakers (sub) elsewhere in the room just to see if this huge delay is still there

I would also let other members to comment

 

[pollock0424]

@ppataki

Can you please explain the wavelet diagram ? How to interpret it? Is it related to spectrogram?

I don't have access to a PC with REW at the moment, can you please look at Filtered IR response at questionable frequencies w.r.t GD ?

 

[LIΟN]

I haven't seen the mdat file of OP, but if you're using DLBC, excessive group delay is known in the issue of DLBC.

Neumann MA1 vs. Dirac Live Bass Control (DLBC) - Dual KH750 subs

You are avoiding my question. I asked those question to creator of Audiolense, Acourate and Deqx. I dont own Neumann Systems but I am Pręty confidential it is no diffrent to mentioned above. You may prove me wrong.

audiosciencereview.com

 

[RenPa]

I would guess you have ported speakers

Good guess. ALL (2 subs included) are ported. Should seal them all and run Dirac?
My 2 subs are at the only place possible and the best integration possible afte trying 9 places to put them.
This wavelet analysis is new for me. I read John help and thoughrt that I understood what it is; apparently not!
To help me to understand can you tell me if this would:

measure L and R large no EQ
measure L and R small with xo 80 no EQ
measure l and R with DLBC on
compare wavelet

 

[RenPa]

I haven't seen the mdat file of OP, but if you're using DLBC, excessive group delay is known in the issue of DLBC.

Neumann MA1 vs. Dirac Live Bass Control (DLBC) - Dual KH750 subs

You are avoiding my question. I asked those question to creator of Audiolense, Acourate and Deqx. I dont own Neumann Systems but I am Pręty confidential it is no diffrent to mentioned above. You may prove me wrong.

audiosciencereview.com

here the link

GD for ASR thread..mdat

1 file sent via WeTransfer, the simplest way to send your files around the world

we.tl

It is difficult to us mere mortal to understand how is it possible for top acoustician as Dirac to sell a software with what seems fondamental flaws.

 

[ppataki]

Can you please explain the wavelet diagram ? How to interpret it? Is it related to spectrogram?

I am not an engineer but the point of the Wavelet is essentially the same as the point of Group delay
The more delay there is in the lows the more it will skew to the right - with the help of the colors it can also show where the peaks will be in terms of frequency and delay

Ideally it shall look something like this:
(this is from one of my projects)

The lower the delay starts (shown by the Peak Energy Time dashed line) the better it is (=delay only occurs with lower frequencies) and the less it skews again the better it is (=less delay)

This graph also shows pre-ringing, for example when using linear phase filters - you will see the lows start skewing to the left (like in the image above around 30Hz)

 

[ppataki]

Good guess. ALL (2 subs included) are ported. Should seal them all and run Dirac?
My 2 subs are at the only place possible and the best integration possible afte trying 9 places to put them.
This wavelet analysis is new for me. I read John help and thoughrt that I understood what it is; apparently not!
To help me to understand can you tell me if this would:

measure L and R large no EQ
measure L and R small with xo 80 no EQ
measure l and R with DLBC on
compare wavelet

I would first measure one sub just on its own then measure the second sub, again on its own and send those two measurements here so we could check if those group delay values are there already (without Dirac or any other DSP)

 

[RenPa]

I would first measure one sub just on its own then measure the second sub, again on its own and send those two measurements here so we could check if those group delay values are there already (without Dirac or any other DSP)

In the following mdat you have my 5 satellites set at large measurements number 1-5
measurement 6 is the left sub woofer (see photo).
measurement 7 is the right sub woofer on the right side of the coach (not seen in the photo)
measureemnt 8 2 subs together same polarity
measurement 9 sub left reverse polarity 2 subs tobether.
don't bother measurements 10 and 11

A 51.5cm 58 et C 68 grille mic 101 312ctr 22.5.mdat

1 file sent via WeTransfer, the simplest way to send your files around the world

we.tl

thanks a lot

 

[DaveBoswell]

If you want to know the relative delay between your subs and your main speakers, easiest way is the plot the "Excess Group Delay Plot" for a test run that includes the subs and the mains together (not the raw Group Delay) per the instructions earlier in this thread and then compare the delay in the flatish portion of the plot in the 50-100Hz range to that of the 400Hz-1000Hz Range (see example below).

In you case, your subs are running about 35.5ms delayed relative to your mains (that's a lot!). You can fix this issue by simply adding distance to your subs in your AVR to balance that by the simple formula Distance to add = delay (in Seconds) x (Speed of Sound). The slope up of the GD plot below 50Hz is normal for ported subs and does not look unusual so don't plug those ports or you will loose the benefit of ported subs.

So for you, the distance to add to your sub in the AVR is = (0.0355s) x (343 m/s) = 12.2m. This is an unusually high sub distance tweak so also look for anything else in the signal path to the sub that might be adding this much delay (DSP or other processor?).

 

[ppataki]

In the following mdat you have my 5 satellites set at large measurements number 1-5
measurement 6 is the left sub woofer (see photo).
measurement 7 is the right sub woofer on the right side of the coach (not seen in the photo)
measureemnt 8 2 subs together same polarity
measurement 9 sub left reverse polarity 2 subs tobether.
don't bother measurements 10 and 11

A 51.5cm 58 et C 68 grille mic 101 312ctr 22.5.mdat

1 file sent via WeTransfer, the simplest way to send your files around the world

we.tl

thanks a lot

The good news is that your subs look much better individually

Left:

Right:

These graphs correspond better to a ported sub

So the conclusion is that the sub integration is not fully optimal (actually far from optimal), resulting in the elevated low frequency delay
I very strongly suspect that this is due to the fact that your measured delay values are unrealistic due to the UMIK-1 mic - and that is the same reason why DLBC cannot optimize your subs correctly (UMIK-1 measures inconsistent and incorrect delay values)

You can very simply test this: place your UMIK-1 in your main listening position and perform acoustic timing reference measurements between one of your subs and your front left channel

(just replace R channel with one of your subs)

Perform this measurement 5 times and then check the measured delay values:

They shall be consistent up to two decimals - if you see inconsistencies you will need to replace your mic with a UMIK-2 (that can indeed measure delay correctly and consistently)

 

[RenPa]

So for you, the distance to add to your sub in the AVR is = (0.0355s) x (343 m/s) = 12.2m. This is an unusually high sub distance tweak so also look for anything else in the signal path to the sub that might be adding this much delay (DSP or other processor?).

Thanks for the comment. The example you take is with DLBC filter ON. Delays between the 2 subs are fixed by Dirac . In my case to tintegrate the 2 swubs Dirac add 1,73ms to the right sub. Now if I understand your comment we see that to integrate the 2 subs (integrated together before) with the fronts DLBC add more than 30ms of delay. I agree it is a lot.
I can't modify those delays in DLBC.

 

[RenPa]

They shall be consistent up to two decimals - if you see inconsistencies you will need to replace your mic with a UMIK-2 (that can indeed measure delay correctly and consistently)

Good clear instructions!
Only change I made (because OCA said it...) End fr to 24 000Hz.
It varies between this

and this:

With -0,8332 ; 0,8330 ; 0,8212 in between.

So what do you say? Too high inconsistencies?

 

[ppataki]

Good clear instructions!
Only change I made (because OCA said it...) End fr to 24 000Hz.
It varies between this
View attachment 406825

and this:

View attachment 406828

With -0,8332 ; 0,8330 ; 0,8212 in between.

So what do you say? Too high inconsistencies?

I think these are OK-ish variations
So this means that your microphone shall be fine

Well, in this case there are a few more options:
- you try what @DaveBoswell was suggesting (but you already confirmed that you are not able to change those values)
- I am not familiar with other DSP options you have within your AVR but could you maybe just add delay values manually? I mean not into Dirac but into the AVR directly, to get around Dirac; that way you might indeed try Dave's point. If you could somehow just test with one sub with a delay value entered correctly you could check if it is DLBC that messes everything up or something else

The issue is clearly with sub integration (because your subs themselves just work/measure fine) so we need to understand what the root cause is

I hope others here will be able to come up with further suggestions

 

[RenPa]

I hope others here will be able to come up with further suggestions

I hope too. But in AVSForum Dirac thread I read many times that DLBC messes with group delay. I also showned that DLBC firmware update after 3.4.4 can't get reliable predicted response with Harman receivers.
I will remeasure the DLBC filter . I would have like to understand why the filters I called 2V sound so better that all I had before. In the predicted response for those filters DLBC change only the response of the fronts between 50Hz and 90Hz and it makes all the differnce to my ears.

 

[ppataki]

I will remeasure the DLBC filter .

Do it with the latest version, 3.11 just came out this week: https://artifacts.connect.dirac.com/public/accord/release/win64/diraclive-v3.11.0-setup.exe

 

[RenPa]

Do it with the latest version, 3.11 just came out this week: https://artifacts.connect.dirac.com/public/accord/release/win64/diraclive-v3.11.0-setup.exe

Here the new mdat but not with 3.11 because I am the one who discovered 20 months ago that since update 3.4.4 no update above can be trusted after you calculate a filter with an Arcam receiver and others Harman related brands.. My results for the latest updat 3.11 are on the Dirac thread of AvsForum under FargateOne. To do what you ask I should make 10 or 12 filters attempts and check all of them with REW before I found one good. I made the exercise early in January this year with 3.5.7 update and will not do again until Dirac fix tis major bug.

gd tooles 2v 80.mdat

1 file sent via WeTransfer, the simplest way to send your files around the world

we.tl

 

[DaveBoswell]

What do you have you Sub-mains cross over set to? My comment above assumes that the x-over is set to 80Hz or higher.

 

[RenPa]

But I can show you in 3.4.4 predicted the difference between the 2 filters.
this the not so good sounding predicted filter for fronts speakers confirmed with REW measurements

and this the so called 2V filter for the fronts:

You see what the allpass filter did for the dips in the response. See the average of 13 positions measurements:

 

[RenPa]

the night brings advice. I reacted a little fast yesterday becaus of my frustration against Dirac bugs and flaws. But I think I have what you asked for if a one position measurement can do the job for analysis. I give one example here. Name of channels are the same ecept that surrounds are Bl and BR. Also I tried the filters in DLBC 3.11.00 . The first 5 measurements are for the same xo at 80 for all speakers. The last 5 measurements is for xo 80 for fronts group, 80 for the center channel and 120 for the surrounds.

hre mdat link: https://we.tl/t-XybxW56Urw