Dirac Live Exclamation Mark issue

[ppataki]

I was able to finish Dirac Live calibration on a 10-years old HP laptop running Windows 10 Home Edition (i7-3610QM CPU @ 2.30GHz, 8GB). Had to repeat some measurements a couple of times after getting the "imprecise measurements" notice. I think the message is generated when the signal to noise ratio is low. When the measurement is run, Dirac Live displays a graphical cue in the form of a modulated strip of varying width. The wider the strip the lower the S/N ratio. In those cases I just increased the sweep volume a little and ran the measurent again. Always worked for me. Also, I made sure to turn off the biggest offenders, AC and the fridge.

For me the best thing about DL is the ability to quickly design room curves and store them in presets. I run my Flex in a small office with a lot of reflections and a couple of nasty modes. I can't say the modes are fixed completely, but I can now listen to rock music with basslines that sounded very objectionable before. And the curves made my DBR2s sound warm and unfatiguing.

My setup: laptop/TV/Bluray Player -> Flex DL -> Yamaha A-S801 -> DBR2
-> SVS SB1000

Here is my most often used calibration:
View attachment 366515

Have you made a before Dirac/after Dirac verification measurement with REW in your main listening position for Left, Right and Left+Right channels?

As mentioned in some of the above posts when the measurement seems to be successful the result might still be rubbish in terms of frequency response (especially in the higher frequencies) and phase response as well

 

[ppataki]

I think Dirac has overreached itself with ART & Bass Control and can no longer produce a stable version.

Yes, agreed

 

[JW001]

Have you made a before Dirac/after Dirac verification measurement with REW in your main listening position for Left, Right and Left+Right channels?

As mentioned in some of the above posts when the measurement seems to be successful the result might still be rubbish in terms of frequency response (especially in the higher frequencies) and phase response as well

Acoustically-wise, my office is atrocious: because of its small size (11.5 x 10 x 8 ft), I have very little choice in positioning the speakers:

Having said that, subjectively the calibration improved the sound. It could be for two reasons: 1) by shaping the overall frequency response slope and 2) by equalizing low frequencies. I didn't notice any adverse effects of applying DL, and I listen to a wide selection of music.

I did 13 measurements in a pretty large area about 5 x 2.5 ft around MLP. For a smaller area, the results of equalization were better than what you see below. Mainly, that huge peak at 60 Hz was lower, but I'm not sure which strategy is better.

 

[mstanley117]

Have you made a before Dirac/after Dirac verification measurement with REW in your main listening position for Left, Right and Left+Right channels?

As mentioned in some of the above posts when the measurement seems to be successful the result might still be rubbish in terms of frequency response (especially in the higher frequencies) and phase response as well

You weren't replying to me directly, but I took before & after measurements myself of only my listening position. I've attached the REW .mdat in a zip if you want to check it out. While I'm okay at measuring (okay might be an overstatement haha), I don't have a full understanding of what I'm looking for or looking at. What you will notice is that my left speaker has a little bit of elevated distortion. This is due to the compression driver. I purchased these speakers 2nd hand, and that driver has been giving me issues as of late. The distortion is not audible to me, thankfully!

 

[ppataki]

You weren't replying to me directly, but I took before & after measurements myself of only my listening position. I've attached the REW .mdat in a zip if you want to check it out. While I'm okay at measuring (okay might be an overstatement haha), I don't have a full understanding of what I'm looking for or looking at. What you will notice is that my left speaker has a little bit of elevated distortion. This is due to the compression driver. I purchased these speakers 2nd hand, and that driver has been giving me issues as of late. The distortion is not audible to me, thankfully!

Thanks for sharing your measurements

Good news: Indeed it seems that Dirac works fine for you, I just compared the phase curve from the Left channel:

Post-Dirac (highlighted) looks much nicer

In the frequency response one can clearly see the corrections especially in the low-end; there are no anomalies in the high-end
(usually it is very clearly visible there)

Just out of curiosity, is this response with the subs included or just the JBL S3900 alone?

 

[mstanley117]

Hey thanks for doing an analysis on the measurements, I really appreciate it! My ears confirm that Dirac certainly helped as well with a number of issues not just with the bass, but also the mid range. Purely subjectively, the presence of the mids are not as forward and up-front, and the entire frequency range is overall sounds more even. Overall, the results of Dirac is everything sounds more even. Sound stage didn't change too much other than maybe the center image is a bit narrower on some tracks. Again, the mids are the most dramatic change. Similarly, it's not too different from how Dirac impacts the Ohm Walsh 2000's in my room, which also have "elevated" mids (for lack of better explanation).

Just out of curiosity, is this response with the subs included or just the JBL S3900 alone?

The measurements are with subs included crossed at 35 Hz. The null at 80 Hz is a room mode, and I'm considering doing a full MSO optimization maybe up to that frequency area with the 2 subs which would let me play with the S3900 placement as well. The S3900's have an F6 of 33 Hz, which is where I chose the subwoofer crossover. Subjectively, the S3900's have a very pleasant-sounding bass as well. Not sure if there would be a big tonal difference crossing that high, so I might experiment with both and see what I prefer.

Like many, I have a difficult room to deal with. I have the subwoofers aligned with one in the front just left of a center console, and one behind the listening position off-center. I used the REW RTA to find the most even subwoofer locations a couple years ago. I don't want to take this thread too off-topic, but I thought I'd address this and provide some context.

The S3900's play pretty low even with the subs muted, but you can tell that the subs add a bit of extra "oomph" sub 60 Hz.

As for Dirac, I'm limited to only using a separate laptop due to the USB issues with the UMIK-1 and Dirac. Those REW measurements were taken using the USB extension cable directly to the mini-PC, interestingly enough.

 

[mstanley117]

I have an update. I tried to run Dirac 3.9.7 with an "old" laptop (circa 2019 HP Envy x360 also with a Ryzen 7 3700u), and it also did not work (kept getting the imprecise measurement error regardless of measurement locations). For my situation, I have to use the Lenovo legion laptop I referenced earlier.

I also did Dirac measurements with the Ohm Walsh 2000's if you want to see those, I've attached the REW files as a zip. It looks like the phase is mostly correct on those. I've posted in other Ohm Walsh threads, but I think the way they work being semi-omnidirectional speakers might not fully work correctly with how measurement mics work currently. That being said, Dirac definitely improves their sound!

If you want, go ahead and take a look. I'm not expecting an analysis, but again, based on prior responses I know more about what I'm looking at. These are crossed at 50hz using 24db Butterworth filters on the MiniDSP SHD with my Rythmik subs.

I'm not sure what you've tried using, but maybe try something with some very, very good hardware in it for Dirac. It seems processing power has an impact.

 

[Davide]

Hi @Flak
Thank you for confirming that you are aware of this issue and doing the utmost to fix it

Actually version 3.4.4 does not work properly either. It just seems to work fine since it won't throw the Exclamation mark/Imprecise measurement error message
I have informed Dirac Support Team about this several times already

Let me show you what happens:

I measure with version 3.4.4

View attachment 354532

It seems to be working fine

View attachment 354533

At filter creation it shows you a nice simulated frequency response curve

View attachment 354534

But when I check it with REW it turns out that Dirac actually did not correct the FR to flat and what's more it destroyed the phase curve too

Without Dirac
View attachment 354535

With Dirac
View attachment 354536

Just to be on the safe side I have tried version 3.4.4 on my other system too and the same happened: huge dips in the frequency response and a much worse phase curve vs the original

When listening to it I can clearly hear that the phantom center is not in the center anymore with Dirac (whereas without it it is definitely in the center) - this shows that the delay values were measured/fixed incorrectly (and probably the same root cause of the poor phase curve too)

This is certainly not the expected behavior and it clearly shows that this version does not work properly either

I hope that after more than 6 months now your team will be able to come up with a proper solution that actually works fine

I also add my case of dizzying increase in GD at low frequencies. And other issues here.
I also opened a ticket to Dirac to get support on the matter. Let's see if we solve something.
@Flak in the meantime, can you give some feedback about this phase/GD issue? Is this a problem you are analyzing or it is a normal behaviour (to get better freq response)?
Thanks

 

[ppataki]

I also add my case of dizzying increase in GD at low frequencies. And other issues here.
I also opened a ticket to Dirac to get support on the matter. Let's see if we solve something.
@Flak in the meantime, can you give some feedback about this phase/GD issue? Is this a problem you are analyzing or it is a normal behaviour (to get better freq response)?
Thanks

There were already some people suggesting having a setting in Dirac to optimize for frequency vs GD
I am currently doing it manually which is fine but would be much better to have such an option built-in

There is still zero update on the exclamation mark issue - now it has been 3 months since the last update
I find this rather unacceptable as a user (and beta tester) when it comes to an app that we pay for...

 

[Davide]

There were already some people suggesting having a setting in Dirac to optimize for frequency vs GD
I am currently doing it manually which is fine but would be much better to have such an option built-in

There is still zero update on the exclamation mark issue - now it has been 3 months since the last update
I find this rather unacceptable as a user (and beta tester) when it comes to an app that we pay for...

Sure, we pay lots of money for this toy and it's not good to see we are all beta testers...
They should include a "Pro" option as with smartphone cameras, where the user is allowed to set the various parameters manually.
What would currently be the modifiable parameters to improve FR and GD sorry?

 

[ppataki]

What would currently be the modifiable parameters to improve FR and GD sorry?

It is not in Dirac - in Dirac you can't do that currently
I do it manually outside of Dirac
In REW you can simulate what happens to the GD curve once you start playing around with the EQ - see attached an animation about how to do it

What I do is that apply Dirac then I apply the changes to the GD curve
It sounds totally counter-intuitive since it will 'destroy' your nice frequency curve - but when you listen to it the bass will sound much better (tighter, 'faster')

They should include a "Pro" option as with smartphone cameras, where the user is allowed to set the various parameters manually.

Fully agreed (btw this has been requested for ages too but never implemented...)

 

[Davide]

It is not in Dirac - in Dirac you can't do that currently
I do it manually outside of Dirac
In REW you can simulate what happens to the GD curve once you start playing around with the EQ - see attached an animation about how to do it

What I do is that apply Dirac then I apply the changes to the GD curve
It sounds totally counter-intuitive since it will 'destroy' your nice frequency curve - but when you listen to it the bass will sound much better (tighter, 'faster')



Fully agreed (btw this has been requested for ages too but never implemented...)

Thanks.
I too thought of correcting the GD in retrospect with a FIR filter but it is honestly nonsense... I paid over 400€ to avoid having to do the job. Furthermore, a further processing step means further latency. If it's not a problem with music, with videos it is.
It does not seem to me that Audiolense, that has the same automation level, creates all these problems ... and its developer is essentially one person. Dirac is a company...
I understand that a universal algorithm is not possible to create it, and Uli of Audiovero affirms it, but today AI is everywhere, even in Dirac BC apparently.
I'm sure they can provide a smarter algorithm, as long as it is not intentionally done like this (for some compromise as better frequency response, low latency, low pre-ringing).
Of course 100 ms of LF GD is not a ruin... in home theatre they are fine, and I think it's their main business... but in studio it is unacceptable value to me.
I would like to have their feedback before expressing myself further.

 

[Davide]

Dirac replied to the ticket by saying that they submit my case to the investigation team.

In the meantime, a doubt is arising.
Almost all calibrated microphones are only supplied with magnitude values.
From some readings that I have done online, which I cannot find now, it would seem that the measurement software can estimate the phase response based on the frequency one (probably assuming a minimum phase behavior).

In your opinion, is it possible that REW and DIRAC estimate the phase in a different way, or that one of the two does not compensate for it, and that therefore these differences arise in the measurements with and without Dirac' filters?
In fact, a few degrees of phase are many milliseconds at low frequencies.

 

[ppataki]

In your opinion, is it possible that REW and DIRAC estimate the phase in a different way, or that one of the two does not compensate for it, and that therefore these differences arise in the measurements with and without Dirac' filters?

To be honest I have no idea
However there is one thing for sure: up until like 8 months ago Dirac used to work absolutely fine with no issues whatsoever (I have done literally hundreds of measurements with it + confirmation measurements with REW to see what Dirac did)
Then something happened and these issues started to happen....
Very strangely if I revert to older versions (such as 3.4.4) the issue is still there, however at the time they were released they used to work fine
They still seem to have no idea about the issue whatsoever... and apparently they don't care either

 

[Davide]

To be honest I have no idea
However there is one thing for sure: up until like 8 months ago Dirac used to work absolutely fine with no issues whatsoever (I have done literally hundreds of measurements with it + confirmation measurements with REW to see what Dirac did)
Then something happened and these issues started to happen....
Very strangely if I revert to older versions (such as 3.4.4) the issue is still there, however at the time they were released they used to work fine
They still seem to have no idea about the issue whatsoever... and apparently they don't care either

Have you tried using an old version of the processor too?
Honestly, I have also used Dirac for some time (since ver. 2) but I had never started to verify the response in the time domain, because I was not competent in the matter.
I would like to try not to use bass control to see what happens, but I first have to find a way to integrate Main and Subs manually.
I will keep you informed.

 

[ppataki]

Have you tried using an old version of the processor too?

No, but it is a good idea, I will give it a try

I would like to try not to use bass control to see what happens, but I first have to find a way to integrate Main and Subs manually.

Regarding Bass Control, I have tried it earlier (before the issues started to happen) and for me it was absolutely no good - sounded like the integration of the sub was totally out of balance (i.e. bad timing). If you search that topic here on ASR you will find multiple people having had the same experience...
If I integrate my sub manually I got way better results, actually you can check those out here

 

[Davide]

No, but it is a good idea, I will give it a try


Regarding Bass Control, I have tried it earlier (before the issues started to happen) and for me it was absolutely no good - sounded like the integration of the sub was totally out of balance (i.e. bad timing). If you search that topic here on ASR you will find multiple people having had the same experience...
If I integrate my sub manually I got way better results, actually you can check those out here

I want to try putting a delay on the mains to see how the filter is calculated (remaking the measurements).
Since the processor delay setting did not have any effect, I tried to put a delay plugin in my DAW with 100 ms setting.
Unfortunately I found that Dirac has some problems with the measurements via plugins, instead of the processor, regardless of the delay plugin.
In practice, after each sweep a strong "thump" starts, ruining the measurement... It is not clear from which speaker comes out.
Has anyone else tried to take measurements via plugins?

 

[Dexter_prog]

This issue is so weird, and more their lack of proper action. This is putting me off from buying a Dirac source (looking at NAD and minidsp stuff). I'm also surprised this hasn't had more repercussion.

Could you guys confirm which version of Dirac (live, full, bass control) you are using? Maybe it's specific to one version?

 

[Davide]

This issue is so weird, and more their lack of proper action. This is putting me off from buying a Dirac source (looking at NAD and minidsp stuff). I'm also surprised this hasn't had more repercussion.

Could you guys confirm which version of Dirac (live, full, bass control) you are using? Maybe it's specific to one version?

I'm using Dirac Live Studio Stereo Bass Control.

Reading the experiences of users on the various forums I am doubting that the problems are rather widespread, and that many are not even identified as such.
To identify them clearly, measurements and a minimum of competence are needed. Only a few macro problems can be detected by ear.
It must be said that a phenomenon often declared by users is the collapse / shrinkage of the soundstage, o sound that comes out exactly from the speakers (without phantom center).
This could be a consequence of the "right" correction, however it is a phenomenon that (from experience) varies greatly to the remaking of the measurements. This is honestly unclear how to measure it, but perhaps the IACC is relevant. Anyway this suggests the lack of robustness and reliability of the algorithm.

To leverage Dirac to give feedback it would be useful to make a nice thread here and on AVS or AVNirvana to collects standardized measurements with / without its correction, but I recognize that it is difficult to achieve significant participation...

I honestly don't think identifying the cause of a problem is very difficult. The difficult thing can be to develop a smart algorithm that remains effective in any case...

For now I have faith in the Dirac team and admire their work, or at least their goals and innovations.
But they have to take into account the feedback from users to know how much they are hitting the goal, otherwise technology is a pure exercise in mathematics and computer science.
So let's give resonance to what we find and use the appropriate feedback channels, namely their technical support.

To return for a moment to the technical question. Has anyone tried to see if using the "normal" algorithm or "bass management" instead of "bass control" is there any difference?
I believe I will try but it takes time for these things and unfortunately I have little.

 

[Kal Rubinson]

Unfortunately I found that Dirac has some problems with the measurements via plugins, instead of the processor, regardless of the delay plugin.
In practice, after each sweep a strong "thump" starts, ruining the measurement... It is not clear from which speaker comes out.
Has anyone else tried to take measurements via plugins?

I have been running Dirac for years via a plug-in (using Jriver) and never encountered a thump such as you describe.