miniDSP Dirac Live causing imaging problems

[RickyC34]

10-4
In Dirac you select which mic before setting levels. Under the mic you selected, the cal file name should appear if previously loaded. If no cal file is loaded this could cause issues. Other than that sounds like you have your bases covered.

 

[Anthony LoFi]

10-4
In Dirac you select which mic before setting levels. Under the mic you selected, the cal file name should appear if previously loaded. If no cal file is loaded this could cause issues. Other than that sounds like you have your bases covered.

I couldn't agree more with Ricky. It is critical you use the correct correction file that came with your USB UMIK. I believe DIRAC recommend (and other speaker testers) the 90 degree position for testing at the listening position.

It would seem that you may already have this correct as you indicated previous use of DIRAC on other speakers have produced the desired results?

I have a similar setup and found DIRAC wanting. I too have separate subs used with a SHD and the only benefit in using the DIRAC config was the lower bass was more controlled.
I emailed DIRAC and minidsp over the same imaging problem and only received a generic response stating I need to be very precise when performing the microphone placement tests in DIRAC.

So I went through the whole process again, firstly setting up my speaker and listening position without DIRAC using a Chesky Test disc with 3 D audio tracks. I obtained a very good image with many adjustments to the angle and distance between speakers. I then proceeded with the DIRAC test procedure making sure the first test ( at LP) was accurate and making sure background noise was dead quiet. I used more than the standard number of tests measurements (you can manually add more) and tried my best to ensure the front /rear positions where equally distant from the first LP position.
All but to no avail.
I have given up on DIRAC, used other methods to calibrate my mains and subs ( see OCA https://www.youtube.com/@ocaudiophile )

I believe the reason why my DIRAC imaging problems is due to my non symmetrical room. I have no technical proof.
I sincerely sympathize with your dilemma.
You are not alone.

 

[theREALdotnet]

The problem is the center image was not good and Left/Right imaging was off. I use the Roger Waters - Amused to Death - The Ballad of Bill Hubbard to test if speakers are set up correctly. If they are, the dogs barking at the start of the song will be to the far right (nearly behind you) and the radio man talking will be to the far left (nearly behind you).

Thanks for mentioning this track, don’t you just love tracks that reliably show when things are out of order?

Some of my go-to tracks for extreme stereo imaging are from Madonna’s Immaculate Collection album (tracks like Express Yourself or Vogue). This album was mastered in analogue Q-Sound and contains these kinds of stereo effects. Now, I’m no fan of Madonna’s music, or pop music generally, but this album has been a fixture in my collection of test material since the 1990s, alongside test/demo recordings from Chesky, Delos, Denon and others.

Today they were both between the speakers. Wrong!!!!

I turned Dirac Live off in the miniDSP screen on my laptop and the imaging problems all went away. Had to readjust subs volume levels since no EQ was applied via Dirac Live. Has anyone else run into this issue using miniDSP/Dirac Live??

In my case, Dirac Live Bass Control does not interfere with the stereo imaging, so I guess you hit upon a bug or some unfortunate microphone placement. I hope you find a fix or work-around.

 

[ReDFoX]

Have you tried disabling delay in Dirac software? I started doing such because Dirac a lot of times gets delay clearly wrong

 

[bmwr75]

I figured out what was going on. Bought this miniDSP 2x4HD and later upgraded it to a DDRC-24. This is a 2.2 stereo listening system. The two subs sit up against the front wall. The main left/right speakers are about 8 feet away from the front wall. I manually inputted the main speaker delay time into the 2x4HD to account for the distance between them and the subs behind them.

After upgrading to a DDRC-24, didn't think about Dirac Live 3 also calculating and saving delays and gains. Its calculated delay was about twice the number of milliseconds I had already saved in the 2x4HD setup. This was the problem. After I told Dirac Live 3 to not save the delay/gain settings to a slot in the DDRC-24. All the imaging problems went away.

One aggravating thing though is while I was doing all this the DDR-24 Master Volume level setting was erased. Took me a while to realize why I was getting no sound from my system. After resetting the Master Volume to -0.0 dB, all was well now.

 

[theREALdotnet]

After upgrading to a DDRC-24, didn't think about Dirac Live 3 also calculating and saving delays and gains. Its calculated delay was about twice the number of milliseconds I had already saved in the 2x4HD setup. This was the problem. After I told Dirac Live 3 to not save the delay/gain settings to a slot in the DDRC-24. All the imaging problems went away.

Interesting that the wrong sub delay would affect imaging so much. What is your crossover frequency?

 

[bmwr75]

Interesting that the wrong sub delay would affect imaging so much. What is your crossover frequency?

The delay in on the main speakers. They are closer to where I sit than the subs. The mains are Magnepan LRS+ and crossover is set at 80 Hz.

 

[bmwr75]

Have you tried disabling delay in Dirac software? I started doing such because Dirac a lot of times gets delay clearly wrong

Yes, mentioned earlier that disabling Dirac made the imaging issue go away.

 

[theREALdotnet]

The delay in on the main speakers. They are closer to where I sit than the subs. The mains are Magnepan LRS+ and crossover is set at 80 Hz.

Well, relatively speaking, I meant. The delay is always zero at the farthest speaker, and progressively larger the closer to the listening position they are (also taking into account their internal delays, if any).

Still surprising. I cannot make the soundstage collapse in the way you describe by toggling the delay on or off in Dirac Live Processor. I’m crossing over anywhere between 50 and 90Hz, depending on the filter.

I suppose there are some complex calculations involved there, that can be thrown off track when starting from wrong assumptions.

 

[ReDFoX]

Yes, mentioned earlier that disabling Dirac made the imaging issue go away.

I meant disabling delay compensation in Dirac, not just the entire calibration

 

[Davide]

I too must say that with Dirac I have had questionable results over time.
I think I have used it on 3 different systems, and in common there has always been the soundstage reduction effect. It really seems that the scene recedes and is limited between the two speakers.
I've never understood whether to consider it a flaw or a consequence of the more accurate reproduction, but I am more convinced of the second.
Uncorrected delay and phase discrepancies between speakers should actually contribute to the creation of a greater stereo effect than is actually recorded.
At the same time, however, they blur the details and clarity of the sound, in fact with Dirac active everything is much more tangible.
However, all this is valid between 100 and 500Hz. Above there is no big change other than the obvious magnitude correction.
A separate chapter however is for low frequencies. Dirac was made for that, and the result is night day. The correction is always much more "correct" and appreciable than the raw response.
On the other hand, it is where the room plays a decisive role and the alterations are significant.
I would add that a big variable on the quality of the perceived sound is the target curve. Changing the curve often results in a very different perception.
The fact that the ideal curve does not exist and that the preferred one has a boost at the bottom and a decreasing tilt towards the highs, indicates that there are factors that are mitigated in this way in order to have a natural result on the ear.
As a preference, I honestly don't like soundstage collapse... in fact I'm using a plugin to increase it and the result is similar to the not correct one.
For the record, these phenomena are not typical of Dirac but also of all other correction systems. Just read around the web. Evidently the correction is technically correct, but the result to the ear is, for not well defined reasons, more or less credible.

 

[TheLastGerman]

I feel the same way as you do. In addition, if I switch off the delay compensation, the sound stage expands, but not to the same extent as without Dirac. Limiting the correction to below the Schröder frequency has the same effect. I think that the latter is actually a malfunction.

I suspected that Dirac misinterpreted the indirect sound of my dipole tweeters (ESS AMT-1), but even if I attach a closed rear volume during the measurement, the soundstage is reduced even if I remove the rear volume after the measurement.

Which plugin do you use to expand the soundstage?

 

[Davide]

I feel the same way as you do. In addition, if I switch off the delay compensation, the sound stage expands, but not to the same extent as without Dirac. Limiting the correction to below the Schröder frequency has the same effect. I think that the latter is actually a malfunction.

I suspected that Dirac misinterpreted the indirect sound of my dipole tweeters (ESS AMT-1), but even if I attach a closed rear volume during the measurement, the soundstage is reduced even if I remove the rear speaker after the measurement.

Which plugin do you use to expand the soundstage?

I remember seeing a video by Dirac relating to the measurement procedure. They said that it's better to measure a bigger volume than actual listening one, because in this way the correction is less strong. I also verified this myself, in fact now I tend to always make the measurements very large and random in terms of positioning. I did the first ones on a very small volume (which is where the head really is...) and with precise mic positioning, but the effect of the correction was really bad... sterile sound.
Mathematically it makes sense, because reducing the variance of measurements means that certain things can be corrected very precisely and effectively, without worsening elsewhere. But in practice something much more complex happens, at least depending on what hearing perceives.
I think it has to do with the reflection cancellation waves introduced by Dirac, which become less and less effective as the frequency rises (or at least, they work on very narrow points).
Or with the fact that our hearing gets used to the reflections of the room, and their cancellation appears unnatural.
But these things have been debated since the beginning and are linked to the operating principle of the correction... and I still can't say whether it is good or bad to apply the correction beyond Schroeder.
Certainly leveling the magnitude brings a notable balance overall.

Anyway, the plugin I'm using to widen the soundstage is Nugen Stereoizer Element, which in my opinion has the best algorithm overall in terms of naturalness (is based precisely on the principles of psychoacoustics).

 

[keks8430]

Evidently the correction is technically correct, but the result to the ear is, for not well defined reasons, more or less credible.

Dirac and other time-domain based methods reduce early reflections (they showcase the clean impulse response). With a single-point measurement you would theoretically completely eliminate the room, with multi-point measurements less so.
Now the imaging or stage is often "enhanced" by the lateral reflections.
Difficult for correction systems to distinguish between good or bad reflections.

 

[ReDFoX]

Dirac and other time-domain based methods reduce early reflections (they showcase the clean impulse response). With a single-point measurement you would theoretically completely eliminate the room, with multi-point measurements less so.
Now the imaging or stage is often "enhanced" by the lateral reflections.
Difficult for correction systems to distinguish between good or bad reflections.

How can FIR based correction alter loudspeaker's directivity? We can't equalize reflections only or reduce radiation just applying filter on the input

 

[keks8430]

How can FIR based correction alter loudspeaker's directivity?

It can't, where did you read that?
Not to split hairs, one can distinguish how a filter is derived and how a filter is implemented.
"FIR based" refers to how a filter is implemented (vs. IIR).

We can't equalize reflections only or reduce radiation just applying filter on the input.

In fact we can.
Depends on how the filter is derived. In the impulse response (time domain), reflections arrive later and thus can be spotted.
https://www.dirac.com/4-types-of-distortion/
For exactly a single position, you could derive filters that create a sound as if you listened to your speakers in an anechoic chamber..
To this degree, it's neither practical nor pleasant.
So time-domain algorithms can distinguish between direct and reflected sound.
But not easily between good (spatiality) and bad (timbre, room modes) reflections.

 

[ReDFoX]

So time-domain algorithms can distinguish between direct and reflected sound.

So... For example, Dirac Live can alter early reflections

 

[Keith_W]

For exactly a single position, you could derive filters that create a sound as if you listened to your speakers in an anechoic chamber..

How do you do that? By outputting a cancellation signal to precisely coincide with the arrival of the reflection?

 

[keks8430]

How do you do that? By outputting a cancellation signal to precisely coincide with the arrival of the reflection?

This is what is done in noise-cancelling headphones (subtractive correction).
Works nicely because the outside noise can be considered only slowly changing so the algorithm can adapt, the speaker-ear position, so its phase, is fixed..

In digital room correction, the inverse loudspeaker-room transfer function is applied (multiplicative correction). See
https://github.com/TheBigW/DRC
https://drc-fir.sourceforge.net
https://github.com/pierreaubert/spinorama/blob/master/generate_peqs.py
for theory and code.
Then there is the difference what we recognize as sound (the direct one) and what we measure (omnidirectional mic captures mix of direct and off-axis, reflected sound with often non-ideal characteristic).
Then more psychoacoustics, mostly supporting the illusion.
https://www.harman.com/documents/AudioScience_0.pdf

 

[Eldus]

I will just add that it is important to realize that the measurement positions in the diagram are "flipped" for the home arrangements. The person is facing you. The dots on the right are the listeners left.

The studio option is the opposite.

Confused me today. =)