• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

What is Dirac doing?

Symphoniac

Member
Forum Donor
Joined
Jan 5, 2024
Messages
70
Likes
23
I ran Dirac this morning and got a really weird result where it's dialing in a dip in the 2to 4k region. The reason why this is weird is because I don't have the full range version of Dirac and am therefore limited to 500hz, so I don't understand why it's messing with the FR above that. Needless to say I won't be using this filter, but just curious how this happened. Has anyone seen this behavior before?

GREEN: Raw response
BLUE: Response w/ Dirac filter

old_measurement.jpg
 

Attachments

  • Ref3_current_dirac.jpg
    Ref3_current_dirac.jpg
    65 KB · Views: 190
Last edited:
Assuming I made a mistake previously, I tried running Dirac again this morning and got another puzzling result. This time Dirac is now dialing down a very broad region between 4k and around 14k. Again, this is very puzzling, given that I do not have a full range Dirac license, and the correction 'curtain' was set to 500hz. FYI, this is using the Dirac functionality of the NAD C658. What is going on here???

GREEN: Raw response
RED: Response w/ Dirac filter

new_measurement.jpg
 
Are those single / individual sweeps or an average of more than one?

Any changes in the room or to your personal location during measurements?
 
Are those single / individual sweeps or an average of more than one?

Any changes in the room or to your personal location during measurements?
Each line on the graph is just a single measurement of both L and R channels taken together at the main listening position (which in my experience is the same as taking L and R measurements separately and averaging them together). Mic was not moved between the measurements done with and without Dirac. Only time mic was moved was between the two different graphs since they were taken on different days, but I took pain to place it in the same MLP and height, so even the two raw measurements taken on different days lie 99% on top of each other when I overlay them in REW. Otherwise, no change in any furniture, not even a single pillow moved between the two measurements (with and without Dirac) or even between the two different days. The NAD C658 (the device that runs Dirac) was purchased open box, so I'm wondering if there might be an issue with it. Oh and regarding 'personal location', I make sure the step outside of the room when the measurement is running, whether Dirac or REW.
 
Last edited:
I'd suggest the following test to check the validity of the measurements.
  1. Use just one speaker.
  2. Place the mic and don't move it for the duration of the measurements.
  3. Measure 6 times
    • A with DIRAC OFF
    • B with DIRAC ON
    • C with DIRAC OFF
    • D with DIRAC ON
    • E with DIRAC OFF
    • F with DIRAC ON

You should see A,C,E graph land pretty much on itself. Similarly for B,D,F graph must land on itself.
If they don't then something else is going on.
 
Assuming I made a mistake previously, I tried running Dirac again this morning and got another puzzling result. This time Dirac is now dialing down a very broad region between 4k and around 14k. Again, this is very puzzling, given that I do not have a full range Dirac license, and the correction 'curtain' was set to 500hz. FYI, this is using the Dirac functionality of the NAD C658. What is going on here???

GREEN: Raw response
RED: Response w/ Dirac filter

View attachment 389982
This looks like it's messing with the negative part of the impulse response while trying to do its phase correction.
 
Me neither, that's why I measured with REW. Graphs are from REW, not Dirac.
Is it L+R that was measured in REW now (or did you apply the direct average of L and R measurements)?
Does Dirac currently compensate for a lot of delay on one speaker?
Some delay compensation + measurement position is not centered + average it
Then a kind of roll-off, or a combination of things, might look "like it's happening."

Can you measure L and R separately and show me?
Original L
Original R
Dirac L
Dirac R
 
Is it L+R that was measured in REW now (or did you apply the direct average of L and R measurements)?
Does Dirac currently compensate for a lot of delay on one speaker?
Some delay compensation + measurement position is not centered + average it
Then a kind of roll-off, or a combination of things, might look "like it's happening."

Can you measure L and R separately and show me?
Original L
Original R
Dirac L
Dirac R
Will try to do this this coming weekend; I'm based in US and am slammed with work following the long holiday weekend :(
 
I did just find this on the Dirac website. It refers to a "sample rate" setting in the Dirac software that could be responsible, but I can't seem to find this setting anywhere in my Dirac application (v 3.10.3).

d.png
 
Assuming I made a mistake previously, I tried running Dirac again this morning and got another puzzling result. This time Dirac is now dialing down a very broad region between 4k and around 14k. Again, this is very puzzling, given that I do not have a full range Dirac license, and the correction 'curtain' was set to 500hz. FYI, this is using the Dirac functionality of the NAD C658. What is going on here???

GREEN: Raw response
RED: Response w/ Dirac filter

View attachment 389982

This tallies rather with my own ear-based results when running the sub 500 Hz version of DL in my NAD M33. The top end loses a little of its sparkle when a DL filter is engaged. I've mentioned this repeatedly on forums, but many claim this is impossible with a digital processor.

My conclusion is that the entire frequency range has to endure this processor, even though DL can only adjust sub-500 Hz. This signal complex processor does the signal no favours, though it may flatten the lower frequencies. For this reason I keep my M33 with No Filter engaged and I can enjoy a distinctly more life-like performance, despite the lower bass being perhaps slightly less than perfect. That certainly suits me and others who have listened when I switch between filter and no filter.

Your graphical display seems to support my view that DSP should only be used in active systems where the top end can be sent without DSP to the top drivers while the amp that delivers just the bass can be fitted with a DSP filter to manage solely those lower frequencies delivered to the bass driver.

Better than DSP is to address the cause of the problem. With great speakers and an accurate amplifier, it seems a travesty to mess with this nice clean signal in order to fix (or rather to adjust for) a few room anomalies. These can normally be mitigated to an acceptable degree by physical means - carpets, curtains, furnishings, etc and perhaps if really necessary, some form of artificial room treatment. But the last thing that should be resorted to is messing with a perfect signal being sent to accurate (as far as they ever are) speakers.

Some hostile replies expected, but I'm pretty thick-skinned! Perhaps some will agree that Signal Processing should in principle be avoided wherever possible.
 
Last edited:
Perhaps some will agree that Signal Processing should in principle be avoided wherever possible.
Nope. If that dip in the treble is being caused by Dirac even when it's limited to sub 500Hz, that's either a bug or there's some issue other issue going on. Simply having DSP is not the problem in-and-of itself. I would advise the OP to contact Dirac's support to see what they have to say.

If your suggestion was merely to limit room correction to the lower frequencies where the room dominates and leave excellent performing speakers alone on the higher frequencies where the speaker dominates you'd get more support.
 
Even when Dirac is set to modify just a frequency range, Dirac applies volume and delay variations between channels. That affect all the frequency range. Thus you get variations out of the configured frequency range.
 
I submitted a ticket with Dirac, so will follow up here when I hear back, thanks to all for your thoughts/suggestions.
 
Even when Dirac is set to modify just a frequency range, Dirac applies volume and delay variations between channels. That affect all the frequency range. Thus you get variations out of the configured frequency range.
That makes sense, although if it's causing an unwanted dip in an out-of-band frequency range I would call that an issue that needs addressing. It should apply those corrections differently, or be able to correct the dip.
 
Back
Top Bottom