Can Dirac make things worst ?

[PappyBlue]

For the first time, I did measurement to see what Dirac correction did ... Green line is the response without Dirac correction, purple line is with Dirac "Auto target" correction. To get rid of a peak it creates a big dip between 60 - 100 hz.

I tried to manually adjust the Dirac curve like this :

This is the result (in black vs Auto Target in purple) :

What is happening. How can Dirac algorithm do something so wrong ?

 

[RayDunzl]

What is happening. How can Dirac algorithm do something so wrong ?

Consider my "problem" around 48Hz. Blue and Green are Left and Right. Orange is both speakers playing:

 

[RayDunzl]

Why?

Phase.

The left and right, due to the asymmetrical room, go out of phase with each other by 180 degrees (result = null) at the listening position.

My DRC plays Left/Right/Left in the measurement process, never checks "both" together.

Even so the MiniDSP may not have enough taps to do anything about it at those low frequencies.

 

[PappyBlue]

Thank you Ray, your comment gave me a good hint. Dirac also does Left/Right/Left in the measurement process and doesnt seem to check both together. Look at the left (green)-right (blue) measurement.

I tried REW EQ and it gave me similar results to Dirac (huge dip between 60-100 hz). The problem is that they are cutting the lef channel at 65 hz to get rid of the peak but since there is a dip on the right speaker at this frequency, it causes a bigger dip. Same thing around 75 hz, this time on the right speaker. My room is also assymetrical ...

 

[amirm]

Dirac makes multiple measurements and makes judgement based on that. This tends to make things less optimal for one position.

 

[PappyBlue]

I don't think it is caused by Dirac making multiple measurements. I got similar results with REW EQ with only one measurement/channel

 

[RayDunzl]

The good news is that although it measures poorly, being omission, it isn't something you consciously notice.

Psychoacoustic and ERB smoothings say there's nothing there, move on...

 

[kevinh]

I believe that tools like Dirac and REW are great tools that weren't available in the past. They can point out room issues that are a result of speaker placement, room anomalies and so on. Ray's measurements show that. Rather than relying solely on the DRC software, We can use multiple subs and room treatments to optimize the room prior to using the DRC software to get the best results.

 

[Ron Texas]

Psychoacoustic leads to psychotic. On Saturday morning I saw a huge line outside a used LP shop near downtown Houston. I thought they were giving breakfast away.

 

[Absolute]

While Dirac certainly can make things worse, it's usually not the case. When I'm not sure whether or not I like the changes, I re-measure in Dirac and save the filter-file as another preset and create yet another one measured only in the main listening position and create that filter as a comparison to the other two.
If the first two differs in sound quality, something's off with the measurements. If they sound better than the last one-point calibration, I'm happy and start to fiddle with target settings.

Usually I also create EQ-settings in REW and load them into the minidsp to compare against Dirac. The last step I do is to experiment with how high in frequency I let Dirac do the correction.

That being said, it'll always sound better fixing the acoustical problems acoustically, and Dirac will always sound better in a better room.

 

[Krunok]

Thank you Ray, your comment gave me a good hint. Dirac also does Left/Right/Left in the measurement process and doesnt seem to check both together. Look at the left (green)-right (blue) measurement. View attachment 24930
I tried REW EQ and it gave me similar results to Dirac (huge dip between 60-100 hz). The problem is that they are cutting the lef channel at 65 hz to get rid of the peak but since there is a dip on the right speaker at this frequency, it causes a bigger dip. Same thing around 75 hz, this time on the right speaker. My room is also assymetrical ...

As Ray explained problem is in the phase mismatch caused by asymetric position of the speakers. The only way to do it right is to get left and right speakers corrected until their response is reasonably flat (you can do it in REW) and then adjust phase of one or both speakers at problematic frequencies manually with rePhase to get both speakers response right. (REW filters are imported in rePhase and then phase is adjusted)

Btw, when checking Dirac results with REW I suggest you make at least 5 sweeps of each channel covering the same area as you did when measuring with DIrac and then average the response. This would provide you with a more realistic picture of the response than a single sweep.

You should also keep in mind fact that peaks are more problematic than dips and that the sound may actually be better than what graph suggests.

 

[daftcombo]

Hi,
I would use DIrac on only one speaker, from 1m or so (not listening position) and report the same corrections to the other.
Also, I wouldn't use DIrac correction under 500Hz. I don't know its algorithm but suspect it can't take measurement errors into account.

 

[Theo]

Is the result actually worse than the original response as far as your listening impression?

 

[Krunok]

Hi,
I would use DIrac on only one speaker, from 1m or so (not listening position) and report the same corrections to the other.
Also, I wouldn't use DIrac correction under 500Hz. I don't know its algorithm but suspect it can't take measurement errors into account.

Both your advices are fundamentally wrong. In the scenario where speakers are not positioned to play in a similar environment the worst thing you can do is forcing the same filter on both speakers, even if it would be possible to do with Dirac (which I think it's not).

Also, room correction is mostly applied for LF up to Schroeder (transition) frequency which is usually around 250-300Hz as this is the range where room affects the sound. Above that frequency room EQ is either not applied or applied very gently to compensate for HF absorption and/or to adjust to the desired target curve.

 

[Krunok]

Here is how it looks with my room & speakers. This was measured with filters created with DRC-FIR tool. You can see problematic phase dip at 83Hz and peak at 340Hz from the left speaker which is positioned in the corner too close to the walls so it behaves suboptimaly. That situation I had to correct manually in rePhase.

 

[daftcombo]

Both your advices are fundamentally wrong. In the scenario where speakers are not positioned to play in a similar environment the worst thing you can do is forcing the same filter on both speakers, even if it would be possible to do with Dirac (which I think it's not).

Also, room correction is mostly applied for LF up to Schroeder (transition) frequency which is usually around 250-300Hz as this is the range where room affects the sound. Above that frequency room EQ is either not applied or applied very gently to compensate for HF absorption and/or to adjust to the desired target curve.

I guess we have completely different ways of setting up our systems.

Everybody is free to try & decide what they like better.

The fact that Dirac is expensive whereas REW+RePhase is free does not imply that the first solution gives better results than the other.

 

[daftcombo]

Here is how it looks with my room & speakers. This was measured with filters created with DRC-FIR tool. You can see problematic phase dip at 83Hz and peak at 340Hz from the left speaker which is positioned in the corner too close to the walls so it behaves suboptimaly. That situation I had to correct manually in rePhase.

View attachment 24936

Correctly differently the phase of the left and right speakers conducted to a total mess in my living room. Once again, try & decide what you like.

 

[Krunok]

Everybody is free to try & decide what they like better.

Sure. We are, for example, free to try poisonous mushrooms as well, but we shouldn't advise other people on forum to do it.

The fact that Dirac is expensive whereas REW+RePhase is free does not imply that the first solution gives better results than the other.

Neither Ray nor I were suggesting that manual EQ is better - we merely stated that this situation is a phase issue and can only be corrected by correcting the phase manually which automatic EQ tools doesn't allow. But I already noticed that you are not a very careful reader and that you like to jump to conclusions.

 

[Krunok]

Correctly differently the phase of the left and right speakers conducted to a total mess in my living room. Once again, try & decide what you like.

If you have read carefully posts were it was stated that phase was the problem you would understand that the point was that left and right speaker should have as similar phase response as possible.

 

[daftcombo]

You have to apply the same phase correction to both speakers.
Trying to have the same phase curve for both speakers, then trying to apply different corrections, will leave you with a weird soundstage. If you like "live" recordings, then you'll feel like all your CDs sound like live recordings and you'll be happy. But forget how the recorded music was intended to be played.

Even for frequency response I wouln't apply different corrections to speakers. At my place, I do it under 100 Hz though, (high-pass filter) because one speaker is in a corner, which gives a +10dB boost in that area. Under 100Hz, you don't know where the music comes from so it is ok. Don't do it elsewhere or you'll have a weird soundstage, with singers singing sometimes in front, sometimes on the left, sometimes on the right. You have to deal with the fact that you are listening in a room, and that it is normal to have some asymmetry.
Correcting the 100Hz-500Hz area by filling the gaps, which are due to measuring errors, could make things better somewhere in the room (I doubt it) but worse 30cm away. It's a no.


As for Pappyblue: Dirac seems to act wisely on the magnitude curve beyond 300Hz, but what it does under 300Hz looks complete bullshit to me. I would desactivate it in that area.
Put a FIR High-pass LR24dB/octave at 40hz instead, and you're done.