Dirac Live Exclamation Mark issue

[Ricwa]

It is great indeed if/when it works
The problem is - and that's why I started this whole thread - that it does not work properly (for some users, including me)

In my opinion this is not acceptable (especially given the fact that we are talking about a payable software...) and should have been fixed already like 6 months ago

I too have been having similar problems with Dirac over the past several months. I have a 2 channel setup with 4 subs running on a Mac with DLBC. Couldn’t get the flat curves like I used to and subjectively the sound was off. I have several pairs of bookshelf speakers including LS-50 metas which used to measure extremely flat.

In running Dirac sweeps I noticed in one measurement (bottom front right) the frequency response was rolling off sharply at about 3.5 kHz and that wasn’t right. I switched left and right speakers, rebooted, reinstalled the latest version, changed amplifiers, uninstalled and then installed the older version 3.4.4; nothing fixed the problem.

While still running 3.4.4, I then lowered the sampling rate from 96k to 48k and began running Dirac sweeps again. But this was painfully slow so and wasn’t making any difference so I switched back to 96k and then reinstalled the latest Dirac version again.

FWIW, bottom line is that this sequence of steps fixed my problems. The other thing I now notice is that as Dirac sweeps are made, visually the sweeps from left to right on the display screen occur much faster than they did before. Sounds crazy, I know, but I’ve done a lot these and I’m pretty sure I’m not imagining this.

 

[RenPa]

I have 5.1 system with 2 subs SVS PC-2000 pro connected to Arcam AVR-21. I am used to measure with a timing reference. I will have time to make some measurements to see GD in 1 week from now!

 

[Davide]

While still running 3.4.4, I then lowered the sampling rate from 96k to 48k and began running Dirac sweeps again. But this was painfully slow so and wasn’t making any difference so I switched back to 96k and then reinstalled the latest Dirac version again.

Consider that with an increase in sample rates, if the number of taps of the Dirac processor does not scale, the frequency resolution is reduced.
It would be necessary to verify from what frequency Dirac goes from minimum phase to mixed phase, but I assume in the 300Hz range.
I have no idea how many taps it has, but going from 44.1kHz to 96kHz doubles the frequency at which the correction start to be effective, if the taps remain fixed.
In doubt it is better to keep it at 44 / 48khz. Also because there's nothing to gain with more sample rates, indeed...
In my case if I measure at 96khz Dirac detects a completely wrong IR.

 

[Davide]

current versions of Dirac are computer hardware dependent

I asked this some times ago... They said that it's hardware independent

 

[RenPa]

Consider that with an increase in sample rates, if the number of taps of the Dirac processor does not scale, the frequency resolution is reduced.
It would be necessary to verify from what frequency Dirac goes from minimum phase to mixed phase, but I assume in the 300Hz range.
I have no idea how many taps it has, but going from 44.1kHz to 96kHz doubles the frequency at which the correction start to be effective, if the taps remain fixed.
In doubt it is better to keep it at 44 / 48khz. Also because there's nothing to gain with more sample rates, indeed...
In my case if I measure at 96khz Dirac detects a completely wrong IR.

I am out of my pay grade here but I read somewhere that Dirac works at 48kHz so music file at 96kHz are downsampled at 48 when filters are on. I totally don't know if it makes sense in your discussion subject.

 

[Davide]

I am out of my pay grade here but I read somewhere that Dirac works at 48kHz so music file at 96kHz are downsampled at 48 when filters are on. I totally don't know if it makes sense in your discussion subject.

This should only applies where Dirac works on hardware DSP (AVR/minidsp typically).
I don't think it downsampling on PC / Mac.

 

[Davide]

From Dirac' Tech Support:

We are looking into these results reported by you and others in the community and acknowledge that it is not the expected behaviour. However, there are precautions that need to be taken when solving this in order to avoid unintended side effects. This is why its unfortunately not yet possible to say when this will be addressed; the solution needs to maintain or improve every aspect of Bass Control.

One thing you could try is to click ‘Disable delay/gain’ in the file menu of Dirac Live. I wouldn’t normally recommend this, as it could effect group delay in other ways, but your system is setup such that the phase excess is almost 0 to begin with which is a really good starting point.

So, at least we know that extra GD is not expected nor desired.

Unfortunately I don't have time to check by deactivating the delay / gain compensation. If anyone can try it would be very welcome.
(PS: they mean with option in the menu of Dirac Live tool, not the flag in the processor).

 

[ppataki]

I also have some updates:

Just out of curiosity I made some measurements with Dirac for both my systems (home office and the living room) yesterday.
I made 4-4 measurements; 1-1 measurement out of 4-4 resulted in the exclamation mark error - the other 3-3 were fine (or to be precise seemed to be fine)
Then I made verification measurements with REW to see what actually happened:
Usually none of them were OK in the past but now I am happy to report that 1-1 measurements were actually fine out of 4-4!

This is from my home office system:

Both the frequency and the phase curves look great and it also sounds great by listening to it
(that huge dip is a room mode, can't do much about it now)

I have no idea why it works now, maybe it is just sheer luck - I have used exactly the same versions (the latest from both Live and the Processor)
Maybe Dirac have done something in the backend (= the connection to their server?) Not sure but I am happy that finally I could make it work, even if I had to perform many measurements + separate verifications

Needless to say this is not expected from a 'normal' user who just wants Dirac to work fine....he/she will not perform n number of measurements and verifications
So there is still a lot to be fixed

 

[Davide]

I also have some updates:

Just out of curiosity I made some measurements with Dirac for both my systems (home office and the living room) yesterday.
I made 4-4 measurements; 1-1 measurement out of 4-4 resulted in the exclamation mark error - the other 3-3 were fine (or to be precise seemed to be fine)
Then I made verification measurements with REW to see what actually happened:
Usually none of them were OK in the past but now I am happy to report that 1-1 measurements were actually fine out of 4-4!

This is from my home office system:

View attachment 369961

Both the frequency and the phase curves look great and it also sounds great by listening to it
(that huge dip is a room mode, can't do much about it now)

I have no idea why it works now, maybe it is just sheer luck - I have used exactly the same versions (the latest from both Live and the Processor)
Maybe Dirac have done something in the backend (= the connection to their server?) Not sure but I am happy that finally I could make it work, even if I had to perform many measurements + separate verifications

Needless to say this is not expected from a 'normal' user who just wants Dirac to work fine....he/she will not perform n number of measurements and verifications
So there is still a lot to be fixed

Good for you.
The calculation algorithm is local, not remote, so there can be no difference in that. The only thing that the tool does via the Internet is licensing.
This behavior that you report does not surprise me as in my experience with each new measurement set I have always achieved different results (but as mentioned above even with various filter creation iterations).
The only variables are:
- microphone placement
- sound pressure (but this is easily controllable)
- background noise
- furnishings/ layout (ditto)
Among the options I believe that the most variable in low freq is probably the microphone positioning.
I don't know how Dirac works in filtering measurements. If it were to use FDW it would therefore be too large.
But I think I have read that their measurement filtering method is more complex.
This is the disadvantage of having a highly automatic and not very adjustable system.
Even with Audiolense and Acourate there is the difficulty of filtering the measurements in order to make them representative and not lead to overcorrections.
But there there is the possibility of working on it.

I am a fan of intelligent and learner algorithms, so I hope Dirac experts will be able to achieve the goal.
Not too far ahead in time though.....

 

[Keith_W]

I am a fan of intelligent and learner algorithms, so I hope Dirac experts will be able to achieve the goal.
Not too far ahead in time though.....

The algorithm will never be as smart as you. You sound like you are not happy with what it is doing. You have no way of over-riding it, except for repeating the whole process and hope that it does something different. Time to upgrade to software which gives you more control

 

[Davide]

The algorithm will never be as smart as you. You sound like you are not happy with what it is doing. You have no way of over-riding it, except for repeating the whole process and hope that it does something different. Time to upgrade to software which gives you more control

I disagree ... I know/read that Acourate and its creator are very inclined to manual optimization, because it is more precise.
But in 2024, with the level reached by the AI, it is foolish to say that an algorithm cannot do better than a human in the optimization of an audio system. It's just a matter of correct programming.
An algorithm can simulate N scenarios and determine which one is best (which is what Dirac Bass Control already does).
A person can do the same for a single point at most.
Then that in practice Dirac is not (yet) reliable is another matter...

In practice, you are right at the moment, but in theory you are not.
No offense, it's just my opinion.
And I'm pretty sure your system is currently more correct than mine!

 

[Esprit]

Time to upgrade to software which gives you more control

Which?

 

[RenPa]

For DLBC, you just need to measure a single channel (example L), without disconnecting anything.
To judge the behavior of the filter, however, you must also measure the sub(s) without Dirac, so you must be able to make the measurement by sending the signal from REW directly to the sub(s).
To compare the measurements correctly you must use the same acoustic reference.

What configuration do you have?

If you are willing to help I would appreciate your comments if the measurements are correct. Would you say that I have a GD problem with my DLBC filters. I think not but need to learn.

Red is the 2 sus together DLBC off Orange is DLBC on at xo 97Hz and House curve

 

[ppataki]

If you are willing to help I would appreciate your comments if the measurements are correct. Would you say that I have a GD problem with my DLBC filters. I think not but need to learn.

Red is the 2 sus together DLBC off Orange is DLBC on at xo 97Hz and House curve
View attachment 369975

Sorry I know you weren't asking me but just FYI this is my GD curve with no smoothing or FDW

I have pretty nasty random peaks that I handle with notch filters but overall the curve is around 10ms
The reason why it sky rockets below 25Hz is that I have my bass roll-off there

You have no peaks but your curve reaches 50ms already around 70Hz and reaches 100ms below 30Hz
I would consider that (very) high but I will let others comment on it too

 

[Davide]

If you are willing to help I would appreciate your comments if the measurements are correct. Would you say that I have a GD problem with my DLBC filters. I think not but need to learn.

Red is the 2 sus together DLBC off Orange is DLBC on at xo 97Hz and House curve
View attachment 369975

It looks a lot like my result.
That GD in theory exceeds the audibility threshold.
I clearly hear the difference, there is a tail effect in my bass.

I advise you to open a ticket to Dirac so that they collect an additional report.
The more we are, the more they should have an interest in intervening.

 

[Keith_W]

Which?

Take your pick. Audiolense, Acourate, Focus Fidelity, REW + DRC-FIR, REW + MSO, etc. Almost everything else on the market gives the user more control than Dirac. If you want the most control, then it's Acourate or REW with one of the options.

 

[Esprit]

Which ones are compatible with macOS?

 

[levimax]

But in 2024, with the level reached by the AI, it is foolish to say that an algorithm cannot do better than a human in the optimization of an audio system. It's just a matter of correct programming.

I am not so sure. Kind of like "self driving vehicles" which were supposed to be everywhere by now but instead they are still not reliable.

See example below of the impulse response of one of my subs taken with a loopback and acoustic MIC which shows an accurate timing reference. When I tried this with a USB Mic without loopback I could not get consistent results and it was very hard to interpret. With the accurate measurement I see the that "first peak" is actually "pre-ringing" from the sub (a SVS 3000 which also had a 6 ms delay). In the case of DIRAC and a USB MIC with variable timing I am not sure how it could account for all the possibilities to calculate where a sub actually is in the room. When it tries to figure it out how does it know if it is looking at "distance from MIC", DSP delay, Pre-ringing, or USB timing errors? In some scenarios I am sure it will work fine but in others it is not hard to imaging the algorithm being "confused" which would show up as either an "!" error or a sub timing error or both (which is what I got). I hope they get it working as I paid $500 for DLBC and while it works OK on the mains FR it is currently useless for sub integration.

 

[RenPa]

It looks a lot like my result.
That GD in theory exceeds the audibility threshold.
I clearly hear the difference, there is a tail effect in my bass.

I advise you to open a ticket to Dirac so that they collect an additional report.
The more we are, the more they should have an interest in intervening.

To be sure to understand, the orange curve should be more horizontal or nearer the 0 ms as the red one. So you interpret the result as if DLBC produce to much delays below 70Hz instead of helping?. And you say this difference is audible. ? I must find how to hear it.
But is it true for only the MLP measurement that I took? If I put the mic elsewhere, the GD should be different isn't it? But I have taken 13 measurements. Does an "average GD achievable?
If I open a ticket what should I explain or point out?

 

[Keith_W]

Which ones are compatible with macOS?

Focus Fidelity and REW. Not so sure about MSO, DRC-FIR, etc. Acourate and Audiolense are Windows only.