Dirac Live Exclamation Mark issue

[RenPa]

All the 8 measurements above were done with Dirac Live correction on (not DLBC but regular Dirac Live)
So this means that I had to perform altogether 8 measurements to have one that is sort of OK-ish (the delay values for the rear channels were totally wrong, hence I am saying OK-ish)


Are you sure that 3.4.4 works fine for you?
I have tried that and although I do not get an exclamation mark issue the correction is still wrong, see post 42 here
I would perform some verification measurements with REW to see if it works fine

OK. Many of the comments in this thread implies new informations for me. For once, I didn't think that phase would be important for FR above 300Hz -issh. ( I am amateur and learned everything by myself on the net).
Second about Arcam issues above 3.4.4 version here are what I talk about:

• Calculated response does not match measured response in DLBC 3.9.7
• IRs between 3.4.4 and 3.9.7 are different enough to indicate either an issue or a change in how the optimization works
• Low-pass slope on the subs seems more like 12dB than 24dB (probably Arcam specific)
• In all versions of DLBC there is no IR plot of the corrected response displayed

Third here some printscreen shot.
Here unwrap phase of NO EQ Front left and 2 subs xo at 80Hz

Here the same with DLBC ON xo 80Hz

This is new for me but if I compare to yours, it seems DLBC optimization filter on helps phase isn't it?

 

[ppataki]

This is new for me but if I compare to yours, it seems DLBC optimization filter on helps phase isn't it?

Yes it should and as far as I can tell from your screenshots it indeed helps
Also the FR seems to be well optimized in your screenshots
So based on these I would say that Dirac works fine for you (at least the version you used on the screenshots)

 

[PeterNL]

Here my measurements with and without Dirac (mdat attached) for a comparison.
PC Intel NUC Windows 10, Dirac Live, Umik-1 (90degree cal file),Acoustic Timing Reference LEFT speaker, Wharfedale Linton.

 

[RenPa]

Also the FR seems to be well optimized in your screenshots

Indeed. I check with REW and MMM RTS measurements and the corrected optimized response forwarded by DLBC is exactly what I get. Not true above 3.4.4

 

[ppataki]

Here my measurements with and without Dirac (mdat attached) for a comparison.
PC Intel NUC Windows 10, Dirac Live, Umik-1 (90degree cal file),Acoustic Timing Reference LEFT speaker, Wharfedale Linton.

It seems to be good! I see no anomalies, both the frequency response and the phase seem to be fine (good for you)

 

[RenPa]

After many measurements with Dirac last update 3.10.1 I can confirm that Arcam owner mest stay with 3.4.4 nothing has been fixed for us.

 

[DWPress]

Can't even open the new Dirac version, maybe it's only for embedded devices? Oh well, 3.9.7 working for me so far.

 

[Mikycoud]

Same boat here, with various Denon and Marantz, and various Umik (1 and 2), and various laptops. (I'm an installer). I personnaly own an AV10 and a Umik2, and a DLBC multisub licence, so I'm in a position where I can fully test things at home.

Not gonna go through all the various and numerous investigations that have been carried out both with and without Dirac (Hi Glenn!) since last year's december, but mostly it's all been recaped here.
There is one exception thought, about a variable that's not been discussed so far, and that's the fact that my succes rate is way more consitent WITHOUT loading the UMIK calibration file than it is WITH. (Or should I rather say that Dirac's measurement failure rate is not 100%). Some measurement points that always get the infamous message actually get a pass without the cal file loaded.
Observable and repeatable on various permutations of laptops, processors and Umiks.
All my findings has been forwarded to Dirac, and, as others have reported, they are aware of the issue, have confirmed it, and are "working on it for a solution in an upcoming release"
Who wants to hold their breath with me?

 

[ppataki]

Some measurement points that always get the infamous message actually get a pass without the cal file loaded.

Have you performed verification measurements to see if the overall Dirac correction was OK?

 

[Mikycoud]

Have you performed verification measurements to see if the overall Dirac correction was OK?

I find that when all measurement points are (fully) successfull , the Dirac algorithm does its job as intended. This can be confirmed in REW by a before/after.
When the measurement are faulty on too many points, the REw measurements show that the correction is crap.
When just one or a few points are "imprecise", well, it's hit or miss, as you have correctly described earlier.
@ppataki Have you tried WITHOUT a cal file loaded? (as fas as the Umik versions are concerned, it doesn't matter which one, the 0° and 90° both introduce a higher "measurement failure" rate.)

 

[ppataki]

@ppataki Have you tried WITHOUT a cal file loaded? (as fas as the Umik versions are concerned, it doesn't matter which one, the 0° and 90° both introduce a higher "measurement failure" rate.)

I have tried it with older versions but not yet with 3.10.1 beta
With older versions it did not solve the issue for me but the occurrence got lower indeed

 

[ppataki]

I wonder if there is anybody here with a multichannel pro audio interface such as the Motu M4 or the Motu Ultralite mk5 using XLR measurement microphones?
I would love to know if they have ever encountered the issue we are having...
I heard that one potential issue could be that the Umik 1-2 USB microphone's clock is drifting from the audio playback device's clock and that can cause these issues we are facing
So if anybody could confirm the situation using an XLR microphone with a multichannel interface that would be really great

It would not explain the phenomenon though that ever since using Dirac Live 2.1 (and a UMIK-1 mic) I have never ever had these issues up until like 9 months ago...
This is what I cannot comprehend at all....

 

[Mikycoud]

I have tried it with older versions but not yet with 3.10.1 beta
With older versions it did not solve the issue for me but the occurrence got lower indeed

Thanks, that's the whole point: the success rate increases indeed, which should be a valuable pointer for a reliable troubleshooting from Dirac.

 

[PeterNL]

I wonder if there is anybody here with a multichannel pro audio interface such as the Motu M4 or the Motu Ultralite mk5 using XLR measurement microphones?
I would love to know if they have ever encountered the issue we are having...
I heard that one potential issue could be that the Umik 1-2 USB microphone's clock is drifting from the audio playback device's clock and that can cause these issues we are facing
So if anybody could confirm the situation using an XLR microphone with a multichannel interface that would be really great

It would not explain the phenomenon though that ever since using Dirac Live 2.1 (and a UMIK-1 mic) I have never ever had these issues up until like 9 months ago...
This is what I cannot comprehend at all....

"I heard that one potential issue could be that the Umik 1-2 USB microphone's clock is drifting from the audio playback device's clock and that can cause these issues we are facing"
Could be true, but so far I know Umik-2 has internal clock.
I have "old" Umik-1 with micro USB instead of new usb-c, and never experienced any problems.
Maybe usb-c is causing this issues, who knows.

 

[RenPa]

Could be true, but so far I know Umik-2 has internal clock.
I have "old" Umik-1 with micro USB instead of new usb-c, and never experienced any problems.

I read somewhere that Dirac has less problem with Umik-1 than Umik-2 for the "internal clock" reason.

Same boat here, with various Denon and Marantz, and various Umik (1 and 2), and various laptops. (I'm an installer). I personnaly own an AV10 and a Umik2, and a DLBC multisub licence, so I'm in a position where I can fully test things at home.

So you are the person I"m searching for a while. Can you tell me if for DENON\MARANTZ the calculated response ALWAYS does match REW measured response in DLBC 3.9.7 or 3.10.1 or any above 3.4.4?

 

[theREALdotnet]

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.

The problems with this approach are:
- you have to become an expert (certainly not impossible given enough time and motivation),
- you have to put in a lot of effort each and every time you take a measurement after a change in your setup (change of speakers, their placement or toe-in, number of subs and their placement, room treatments, furnishings, etc.)
- the process is sufficiently complex that it is easy to make mistakes

That doesn’t make the approach invalid or inferior, you just need to understand what you’re getting into.

 

[RenPa]

the process is sufficiently complex that it is easy to make mistakes

Exactly why I couldn't choose MSO. Too complex for my brain! Though maybe the best of all. So I choose (poorly!) DLBC...

 

[Mikycoud]

So you are the person I"m searching for a while. Can you tell me if for DENON\MARANTZ the calculated response ALWAYS does match REW measured response in DLBC 3.9.7 or 3.10.1 or any above 3.4.4?

Yeah, the tax inspector told me the exact same thing once and that wasn't good news lol.
As far as your question is concerned, are you talking in general (when all works OK in Dirac) or specifically for this "imprecise measurements" situation?

 

[Keith_W]

The problems with this approach are:
- you have to become an expert (certainly not impossible given enough time and motivation),
- you have to put in a lot of effort each and every time you take a measurement after a change in your setup (change of speakers, their placement or toe-in, number of subs and their placement, room treatments, furnishings, etc.)
- the process is sufficiently complex that it is easy to make mistakes

That doesn’t make the approach invalid or inferior, you just need to understand what you’re getting into.

Yes, I agree. I gave my friend an analogy that went something like this: DSP is a process, just like frying onions is a process. Tools have to be used and procedures followed in the right order. Onions have to be peeled and diced before they are fried. You can try frying the whole onion and dicing it later, but the result will not be the same.

- Acourate: here is your knife, here is your pan. It will happily let you fry the whole onion and dice it later if that's what you want to do. What's bad (or maybe good): you have to learn the entire process and dice the onion yourself. The outcome depends on your skill in dicing onions and frying. What's good: you can use these tools for anything you like. Can make deep fried onion rings or any onion dish that requires a knife and frying pan.
- Audiolense: give me your onion and I will dice and fry it for you. Tell me how you want the onion diced, and how long you want it fried for. What's good: you don't have to learn the process. Much easier for beginners. The outcome is the same. What's bad: knives can be used for other things besides dicing onions. Less versatility and flexibility. Can't make deep fried onion rings if the option isn't included.
- Dirac: give me your onion, and I will dice and fry it for you. Give me a general idea of what you want (caramelized? translucent?). What's good: very easy to use. What's bad: less control over the outcome. Definitely no deep fried onion rings for you.

There are others.

- MATLAB/Octave: here is some metal. Forge your own knife and pan. Then go dice it yourself and fry it. Literally true, you have to make your own tools by typing in equations.

 

[Miniyouuuu]

Dirac allow you to define sweet spots of measurement, target curve and range of correction. You are able to set automated settings of curve and curtain, or you can set them manually.
My first setting with Dirac (automated process) didn’t fulfill my expectations. Sound is perfectly centered but dull and with barriere soundstage. After manually playing with the target curve and curtain and narrowing the spots of measurement, everything improved. Soundstage is affected, as Dirac is allways killing some echoes due to perfect alignment of speakers, but not enough to kill the experience.
I try to replicate the measured room curve, avoiding huge differences between original and target, specially above 1000hz. Thus you get a more natural sound.

And after Dirac I always check REW compared to original sound in order to see RT and waterfall. Sometimes Dirac get worse results by forcing the speakers to achive the target.