Why does DIRAC completely remove bass response (KEF R3 + sub + minidsp flex)

[Geertidow]

Dear fellow forum members,

First of this forum has helped me a lot with choosing my speakers.
Hopefully you can aid me in my DIRAC problems.
I have used DIRAC extensively in the past on my NAD C658, I'm familiar with how everything works.
Hence my suprise to what happened when I applied DIRAC to my new KEF R3 with custom build subwoofer.

Setup:
Speakers: KEF R3 + Push/Pull subwoofer (10").
Amp: NC252MP for the R3 + NC500 mono for the sub.
Source: MiniDSP flex (balanced) with USB attachted to computer for measurment.
Mic: UMIK-1 (with cal file)
Crossover: 60 hz with a 24 db/oct roloff
DIRAC in shown measurements: +2db harman kardan with a slight -1 dB roloff in the higher region (standard starting curtain), full range.

You see what happens to the spl curve and waterfal with and without dirac.
The final images are a waterfal and spl curve when I apply some mild EQ manually (max Q = 3, max gain = 3 dB) and a HK + 4db curve from 200 hz.

Why does DIRAC completely cutoff the suwoofer and lower part of the KEF's?
This did not happen when I used DIRAC on a pair of genelecs 8030c in the same setup.

Thanks for your advice!

 

[Geertidow]

Dirac target response, it seems way off

 

[abdo123]

my recommendation for the best crossover for the R3 is at 60Hz (BW2 12db/oct highpasss, LR4 24db/oct lowpass with any crossovers disabled on the subwoofer).

Adjust the delay on the subwoofer so that there is no null in the response at 60Hz with both speakers. leave the speakers delay alone.

I would adjust the target curve to match the speaker response between 20KHz to 1KHz, and from there, with 1KHz being the 0 dB point, put the 30Hz point to be +5.8dB with a linear slope between 1KHz and 30Hz. The current slope is exponential for some reason.

Adjust the delay on the subwoofer so that there is no null in the response at 60Hz and re-run the calibration. Hopefully it should be fine.

 

[Geertidow]

It really seems that using stereo dirac leads to problems with a 2.1 setup

 

[Mr. Widget]

I don't think your conclusion is correct. I don't think that Dirac or any DSP inherently leads to problems with a 2.1 system. I think the issue is that with the simplicity of a 2.1 system, any setup errors are more easily discovered. One of my pet peeves is the assumption that anyone can toss in a DSP and solve their problems... yeah, maybe sometimes... but it isn't easy.

As @abdo123 suggests you should make sure your delays and phase response are correct. One way to set the delay is to play a 60Hz sine wave through the system (since this is your crossover frequency) and reverse the phase of the subwoofer. You then adjust the delay until the 60Hz sine wave is fully minimized and almost cancelled out between your mains and the sub. Then you reverse the phase of the sub and they are now properly aligned.

Also have you tried moving the subwoofer to other locations in the room?

 

[Geertidow]

I have time alligned the subwoofer and speakers perfectly using req. Impulse response, step and phase are perfectly aligned. Furthermore, xo problems should be worse at the xo region and not over the whole 30 - 200 hz region and should persist without DIRAC. Neither is te case. DIRAC somehow does not even remotely adhere to the target curve the whole section below 200 hz gets pulled down a lot (3 db under the 2 kHz mark). It's just strange that this didn't happen with the genelecs 8030c in the same setup. Luckily the R3 sounds great without DIRAC and some minor PEQ but I would still like to find out why DIRAC seems to completely mess up.

 

[Geertidow]

I have also implemented all the suggestions by @abdo123, which helped a little, but not nearly enough. My solution so far is apply dirac and then peq the lower end back to a normal curve. But that seems ridiculous.

 

[Mr. Widget]

I am not a DIRAC user, so I may be off base here, but can you take a cloud of measurements and average them? Particularly when dealing with bass, taking multiple measurements with the mic within a 24" sphere around the listening position will yield much more accurate data than a single mic located presumably at your listening position.

 

[Geertidow]

DIRAC uses multiple measurements.
A pink noise rta measurement yield similar results.

 

[abdo123]

DIRAC uses multiple measurements.
A pink noise rta measurement yield similar results.

If that’s the case i would contact Dirac support, they’re very responsive and helpful.

 

[Aerith Gainsborough]

Dirac target response, it seems way off

Not really. The yellow line is fairly realistic as to what we would set and the scan looks similar to your green Pre-Dirac "R3+Sub" control measurement.
It seems really strange that you do get such a weird post-Dirac control measurement (assuming the brown curve in your first post is a control measurement with Dirac enabled).

I second @abdo123 sentiment: something went wrong Dirac-internally, so give the Customer Support a ring.

 

[Geertidow]

Not really. The yellow line is fairly realistic as to what we would set and the scan looks similar to your green Pre-Dirac "R3+Sub" control measurement.
It seems really strange that you do get such a weird post-Dirac control measurement (assuming the brown curve in your first post is a control measurement with Dirac enabled).

I second @abdo123 sentiment: something went wrong Dirac-internally, so give the Customer Support a ring.

I will contact support what's going on here. Thanks for your response and I will update what the problem is.

 

[Geertidow]

It seems that just restoring default config and reuploading the same filters works.
Apparently there was some kind of software error.
The problem is now that DIRAC does not seem to make it much better after al ;-)
Without DIRAC the curve is pretty nice.
I managed to get away that 80 hz dip by increasing the xo from 60 to 80 hz.

 

[abdo123]

Upload the REW file of a single point sweep for each speaker and i will give it a look.

Raising the crossover often gives you a better response but worse sound stage. If you can tell where the subwoofer is when you turn off the amplifier for the speakers then i would go back to 60.

 

[Geertidow]

You want a sweep without xo in place for the subs and the speakers (one speaker per measurment?)

 

[abdo123]

You want a sweep without xo in place for the subs and the speakers (one speaker per measurment?)

No i’m not going to look at frequencies above 100Hz. Those are pretty much straight forward to fix.

 

[Geertidow]

So a sweep with each XO (60, 80, 100 on 24 db/oct and 80 with 12 db/oct) in place. Correct?

 

[abdo123]

So a sweep with each XO (60, 80, 100 on 24 db/oct and 80 with 12 db/oct) in place. Correct?

I just expect an L and an R measurement of whatever configuration you think is performing the best.

Perferably with no EQ above 100Hz.

 

[Geertidow]

Perfect, than I understand it, thanks for helping out. For the measurments I will apply no EQ whatsoever for optimal assessment. Thanks.

 

[Geertidow]

No EQ above 111 Hz, description is in the comments of the file.
Thanks for helping out!
The file is accessible through: https://we.tl/t-jo6kcYCysE
Because of the size (15 MB)