Excess group delay and mismatched phase response

[RenPa]

View attachment 407172

These are the best measuring picks of pairs in your mdat. That step response growing back after 60ms is very odd, points to a non-zero DC offset in the impulse response.

Maybe Dirac hears you. Here a DLBC filter calculated with firmware update 3.12.00. (Although again this update doesn't fix any of the important bugs with Harman/Arcam/JBl receivers )

 

[ppataki]

Maybe Dirac hears you. Here a DLBC filter calculated with firmware update 3.12.00. (Although again this update doesn't fix any of the important bugs with Harman/Arcam/JBl receivers )
View attachment 420067

Not sure what we need to see here - is it that the step response graphs are more or less aligned post 25ms?

To be honest that huge oscillation would worry me; I have always been under the impression that the less oscillation the better but please correct me if I am wrong

This is how my current step response looks like with two subs (manually integrated) and Dirac Live (which hence 'sees' the subs already integrated)

And this is the Wavelet

 

[RenPa]

Not sure what we need to see here - is it that the step response graphs are more or less aligned post 25ms?

It was a follow up of @OCA post https://www.audiosciencereview.com/...-mismatched-phase-response.50523/post-2140825
only to tell that 3.12 seems tohave fixed what he saw.
Your wavelet are TOTALLY impossible to get with DLBC .
Here my last DLBC 3.4.4 filter front left+ subs DLBC on:

And the wavelet

we aren't in the same world.
Seeing the @OCA 's blast is doing with Neuron and Audy and his «à la Dirac filters» I don't know what fondamental principles he applies that Dirac doesn't to get his supposed huge extraordinary fantastic super results...a mystery...

 

[pollock0424]

I have given up on Dirac correction for a 2 ch system. Instead I'm just resorting to PEQ correction on my RME ADI-2 Pro FS. I have a separate 2 ch system. I still use Dirac on Integra DRX-8.4 for HT...


I'm tired of it and I don't have time to mess around.

 

[RenPa]

My system is for HT and stereo music. I am asking myself if it would be better to put my old miniDSP 2X4Hd in the chain, learn MSO, EQ my 2 subs for 2 seat positions and present only one sub to DLBC. Would DLBC messes less with GD in this configuration?

 

[ppataki]

My system is for HT and stereo music. I am asking myself if it would be better to put my old miniDSP 2X4Hd in the chain, learn MSO, EQ my 2 subs for 2 seat positions and present only one sub to DLBC. Would DLBC messes less with GD in this configuration?

You might not even need MSO, just do a level match for your fronts and two subs (e.g. use REW's SPL Align tool) measure the sub distances, enter the corresponding delay values into your DSP (whatever DSP you have) then set up a crossover (make sure it is Linkwitz-Riley, 24dB/octave, if possible linear phase, around 80Hz or whatever works for you based on your measurements) and then run Dirac Live (non-DLBC) to optimize your system which it will 'see' without the subs since you have integrated them already.

This is what I do as well and it works really really nicely

 

[RenPa]

I have an other question. I see that for only one of my speaker the surround left Dirac put a dotted line on the wrong peak in the measured impulse response as if it was choosing the wrong polarity which is not the case of course .
See the left picture is REW measured impulse of the surround left and the right side of the picture show how it was measured by Dirac Infortunatly I can't change the color code for that speaker in Dirac the doted line is at 0,72 mark

All my speakers are at the same polarity in REW, the first great spike up.
With DLBC filter on the corrected impulse response for the left surround is positive too as all others too. Is the starting «mistake» before calculate a DLBC filter important or not?

 

[OCA]

I have an other question. I see that for only one of my speaker the surround left Dirac put a dotted line on the wrong peak in the measured impulse response as if it was choosing the wrong polarity which is not the case of course .
See the left picture is REW measured impulse of the surround left and the right side of the picture show how it was measured by Dirac Infortunatly I can't change the color code for that speaker in Dirac the doted line is at 0,72 mark

View attachment 424268
All my speakers are at the same polarity in REW, the first great spike up.
With DLBC filter on the corrected impulse response for the left surround is positive too as all others too. Is the starting «mistake» before calculate a DLBC filter important or not?
View attachment 424265

It can be tough to get speaker polarity right, took me a lot of time to develop a working function for that in Neuron and it can still give false positives with external amps and dolby enabled speakers. I don't know how Dirac handles it but a dirty solution could be swapping cables of this speaker during measurements and then swap them back after Dirac calibration. Otherwise bass will be extinct from your SL/SR combo and distance of SL will be somewhat off.

 

[ppataki]

Is the starting «mistake» before calculate a DLBC filter important or not?

Yes, I believe it is a mistake, indeed
The dotted line shall be on the first peak
You can try what @OCA recommended above

Actually this is funny: last night I was also playing around with the polarity of my two subs in the living room and I managed to actually get rid of a ringing that I could not remove in the last 6 years......

 

[RenPa]

a dirty solution could be swapping cables of this speaker during measurements

Yes indeed. I measured the timing difference between the up and down peak; 13 milimeters. DLBC doesn't change satellites polarity, only the subs. Does it matter?

 

[OCA]

Yes indeed. I measured the timing difference between the up and down peak; 13 milimeters. DLBC doesn't change satellites polarity, only the subs. Does it matter?
View attachment 424401

No, that's fine. It will get the distance and also the EQ filters when crossing over that speaker with sub right. Just don't forget to run that calibration with speaker's polarity physically swapped back to normal.

 

[RenPa]

No, that's fine. It will get the distance and also the EQ filters when crossing over that speaker with sub right. Just don't forget to run that calibration with speaker's polarity physically swapped back to normal.

( I don't know how you manage your time in only 24 hours/day... The Entity... I said) Anyway, I wasn't clear. Is the difference caused by the mistake made by Dirac matters? In other words: is it enough to try the dirty solution you suggested? I am willing to try but to my amateur eyes 13 milimeters is not so important isn't?

 

[RenPa]

DL 3.12.1 surrounds positive polarity impulses Blue is the LEFT surround polarity positive when measured doted line on down peak

DL 3.12.1 surrounds mixed polarity impulses Blue is the LEFT surround polarity negative when measured doted lines are reversed now it is the RIGHT yellow doted line on the down peak
LEFT surround is BLUE

Dl 3.4.4 left surround polarity positive and right surround polarity positive
2 Doted lines are on down peaks for the 2 speakers impulses

DL 3.4.4 with left speakers polarity NEGATIVE right speaker polarity positive. Left surround is BLUE. 2 doted lines on the downward peaks

IN ALL CASES DL or DLBC 3.12.1 or 3.4.4 CORRECTED polarities are ALL POSITIVES

 

[RenPa]

I must say that since Arcam updated their firmware for v1.62_2057 I get a little bit better results with Dirac DLBC v3.4.4 filters.

 

[RenPa]

It can be tough to get speaker polarity right

The new DLBC update 3.13.2 seems having corrected the problem and seem to get speaker polarity better

I have a question about spectrogram and GD etc. while we are waiting to see if Dirac will change their philosophy about it in DLBC @ppataki . Some DLBC filters give large differences in spectrogram of each channels and some give a more even pattern from one channel to an other. For ex imagine one filter give 5 different timing for FR 40Hz on the x axis for each channel : 50m to 100m. Suppose that an other filter gives a more coherent timing for the same FR say between 62m and 75m. Is one better than the other spectrogram only in't enough to decide?

 

[ppataki]

Some DLBC filters give large differences in spectrogram of each channels and some give a more even pattern from one channel to an other. For ex imagine one filter give 5 different timing for FR 40Hz on the x axis for each channel : 50m to 100m. Suppose that an other filter gives a more coherent timing for the same FR say between 62m and 75m. Is one better than the other spectrogram only in't enough to decide?

I would go for the one with more consistent delay values across the channels
That should provide a more coherent experience

 

[RenPa]

I would go for the one with more consistent delay values across the channels
That should provide a more coherent experience

@OCA
All my questions were skewed by false results from the DLBC filters. I discovered that the software in my device (and in others) measures a peak between 0Hz and 20Hz in the front left channel and sometimes in other channels. The peak in the front left channel is always accompanied by a popping sound that I don't hear when the peak is measured in the other channels. It is physically impossible for a speaker to produce sound at these frequencies. We don't know the formula used by Dirac to calculate the filters in DLBC. We don't know whether these suspicious peaks are taken into account or not in the calculation of the filters.

All my questions were based on measurement data without taking the problem into account. Here is the best GD obtained from 5 channels with the DLBC filter:

Then I remeasured deleting all data with a peak and I got this blue GD for the same 5 channels and the same DLBC filter

Then I choose an hdmi input, measures deleting all data with a peak beetween 0Hz and 20Hz and got this yellow GD 5 channels

To compare:

So IMHO this line of receiver has problems somewhere.

Translated with DeepL.com (free version)

 

[OCA]

@OCA
All my questions were skewed by false results from the DLBC filters. I discovered that the software in my device (and in others) measures a peak between 0Hz and 20Hz in the front left channel and sometimes in other channels. The peak in the front left channel is always accompanied by a popping sound that I don't hear when the peak is measured in the other channels. It is physically impossible for a speaker to produce sound at these frequencies. We don't know the formula used by Dirac to calculate the filters in DLBC. We don't know whether these suspicious peaks are taken into account or not in the calculation of the filters.

All my questions were based on measurement data without taking the problem into account. Here is the best GD obtained from 5 channels with the DLBC filter:


View attachment 513678

Then I remeasured deleting all data with a peak and I got this blue GD for the same 5 channels and the same DLBC filter


View attachment 513679

Then I choose an hdmi input, measures deleting all data with a peak beetween 0Hz and 20Hz and got this yellow GD 5 channels
View attachment 513681

To compare:

View attachment 513688

So IMHO this line of receiver has problems somewhere.

Translated with DeepL.com (free version)

I am not sure I fully understand the problem here but below 20Hz there will be lots of peaks/dips and discontinuities due to room modes/resonances/device-speaker filters/even the fft operation itself. All your GD graphs are pretty descent with a steady rise towards 200ms in the low bass region. You cannot go below that without active room treatment.

 

[RenPa]

I am not sure I fully understand the problem here

I have problems with my Arcam receiver and while I wait for answers from Dirac and Harman Luxury UK, I try to understand myself. The only thing I know for sure is what my ears have “told” me. The sound from my 5.1 system and DLBC filter keeps getting better and better according to the three GDs presented above, which makes no sense to me since what distinguishes them are the measurements received by the software, which vary depending on whether or not there is a “pop” heard in the front left channel.
I pay not much attention to distortion graph because I am not very familiar with the scientific concepts surrounding this data, but here is an illustration of what I have observed. Yellow sound the best and «looks» better.

https://www.avsforum.com/posts/64444939/

https://www.avsforum.com/posts/64430787/

 

[OCA]

The distortion levels seem ok and consistent to me except for one measurement during which there was some environmental noise with around 1kHz frequency but if you are hearing pops then it's probably a more serious issue.