• 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!

Verifying Dirac Room Correction with REW

psykano

Member
Joined
Apr 29, 2023
Messages
6
Likes
14
Hey, so I figured I might as well post this in case someone else is looking to see what Dirac is actually doing in terms of room correction by measuring before and after using a UMIK-1 microphone and REW using the Diract Live Room Correction Suite. My setup is two JBL 306P speakers about 2.6 meters from the listening position and about 1.4 meters apart. The speakers are also only 6.5 inches from the back wall and the left speaker is about 12 inches from the side wall (actually, window with curtain).

Here are the measurements for the left speaker in Dirac Live. Yellow is the measured FR and purple is the corrected FR.

Screenshot 2025-01-16 at 8.13.19 PM.png


Here are the actual measurements for the left speaker in REW (using Var smoothing). Red is without Dirac enabled and orange is with Dirac enabled:

Screenshot 2025-01-16 at 8.15.58 PM.png


Here is the right speaker in Dirac Live. Yellow is the measured FR and teal (or light blue) is the corrected FR:

Screenshot 2025-01-16 at 8.18.56 PM.png


Here are the actual measurements for the right speaker in REW (using Var smoothing). Blue is without Dirac enabled and teal (or light blue) is with Dirac enabled:

Screenshot 2025-01-16 at 8.20.35 PM.png


Just to round out this post, here are my speaker's measurements before any Dirac Live corrections:

Screenshot 2025-01-16 at 8.33.16 PM.png


Here are the measurements after Dirac Live corrects for bass from ~35 Hz to 500 Hz:

Screenshot 2025-01-16 at 8.34.21 PM.png


Lastly, here are the measurements when the FR is corrected from ~35 Hz to 20 kHz (I, personally, do not hear a difference between this and the above but that's for another discussion):

Screenshot 2025-01-16 at 8.36.10 PM.png


I hope this post has been helpful! :)

As an aside, the other big thing Dirac has fixed in my setup is the timing and gain for better imaging when sitting in the listening position since the speakers are on one side of the room due to living in an open-concept apartment (also, hence, no subwoofer). Of course, that better imaging is gone if moving more than a few inches from the listening position. :facepalm:
 
The process with Dirac involves taking a series of measurements at and around the listening position. Since we can’t be certain how it processes and weights these measurements, it’s difficult to determine how to perform a proper control measurement for a fully accurate assessment.

However, if you want to give it a try, you could take a series of measurements at the same positions as Dirac Live and average them, or perform a MMM at the main listening position.
 
For sure. I didn't mean to compare against REW MMM. I just wanted to show mostly, as usual, the predicted FR in its room correction software is still not indicative of the actual FR when measured. I also wanted to show that it's probably not actually boosting any dip by 10 dB even though it looks like it is in the predicted FR.

After searching, I couldn't find any posts of post Dirac-enabled measured FR as opposed to all the posts showing the Dirac predicted FR. So I hope this helps someone though I'm sure the software will change in the future anyway and, like you said, without us knowing precisely what those changes are! :D
 
... I just wanted to show mostly, as usual, the predicted FR in its room correction software is still not indicative of the actual FR when measured
It is, but the measurement conditions need to be the same... you can either average the measurements in REW as in Dirac, or you can measure a single point with Dirac and compare the result with REW
 
It is, but the measurement conditions need to be the same... you can either average the measurements in REW as in Dirac, or you can measure a single point with Dirac and compare the result with REW

Exactly. Single point Dirac shows in rew that it is doing exactly what it says it is doing.
 
You show L and R speakers, but not Both.

I usually don't listen to music through just the left or the right.

At my place there's an interesting hole around 48Hz when both speakers play, that doesn't show in the individual traces.

If you don't want to repeat the sweeps, a Vector Average of the left and right gives the same result (with single point listening position measurements)

Red and Blue are left and right
Green is with both playing the sweep
VAR smoothing

1737145132152.png
 
For sure. I didn't mean to compare against REW MMM. I just wanted to show mostly, as usual, the predicted FR in its room correction software is still not indicative of the actual FR when measured. I also wanted to show that it's probably not actually boosting any dip by 10 dB even though it looks like it is in the predicted FR.
This was the point I was trying to make -it’s not a trivial task. You can’t truly demonstrate what Dirac Live does without meticulously taking the same measurements in the exact same positions. Even then, you won’t fully understand how its algorithms work, how it calculates corrections, sets thresholds for addressing dips, or how it blends and separates the left and right channel responses, among other factors.

You might think that taking a single measurement in Dirac Live would be enough to prove its functionality. Well, not exactly. That wouldn’t truly showcase how Dirac Live operates. Since you don’t know how it weights different measurements, you also can’t be sure if it behaves differently with a dataset consisting of just one measurement compared to a full suite of measurements.
 
As I think most people use Dirac in full range mode, it would have been more interesting to see what the program does beyond 500 Hz, in comparison to gated measurements in REW of the actual direct response of the speakers themselves.
 
I provided the full range Dirac-enabled frequency response in the last picture. :)

It's pointless to compare against what Dirac says it's corrected FR should be in the app since it's essentially extremely flat.
 
At my place there's an interesting hole around 48Hz when both speakers play, that doesn't show in the individual traces.
This is where DLART will be able to make the most of its advantages.
 
I provided the full range Dirac-enabled frequency response in the last picture. :)

It's pointless to compare against what Dirac says it's corrected FR should be in the app since it's essentially extremely flat.

Yes, but it would still be interesting to see the Dirac curve against a gated measurement of the direct response of the speakers. In that case, we would be able to see if Dirac does something for the higher frequencies it shouldn't be doing.
 
If you don't want to repeat the sweeps, a Vector Average of the left and right gives the same result (with single point listening position measurements)
Would this not only be theoretically and in anechoic environments perhaps? But even then, speakers could influence each other? In the plot ou posted the green also doesn't look like the vector average of the separete L and R traces, right?
 
Would this not only be theoretically and in anechoic environments perhaps? But even then, speakers could influence each other? In the plot ou posted the green also doesn't look like the vector average of the separete L and R traces, right?

Red - Stereo Sweep with both left and right speakers playing
Black - Vector average of individual left and right traces. Individual L and R traces not shown here.
1/48 smoothing

1737203902209.png


The dip is a phase problem in my room.

Almost no difference. Noise difference at the time of the indiviual and stereo sweeps could account for the tiny differences, I suppose.
 
Last edited:
This is where DLART will be able to make the most of its advantages.

I don't have a space for Dirac in my setup or wallet.

The big dip measurably disappears when music has stereo bass.

I didn't know the dip was there until measured, and I still don't seem to hear it,

Sometimes I play a (mono) bass guitar through the sytem, and don't notice it even then.

So, it's an anomaly that doesn't seem to bother me. But its there.



RTA of some random recording from a long while back. The wide 48Hz dip disappears in the Peak reading (red). Black is whatever was playing at that moment,

1737205929177.png



If the bass is mono in a recording the dip is visible in an RTA.
 
I've had Dirac pro live since 2020. I now use a pro analog EQ but think I will hit bypass & run some Dirac correction filters I have saved & compare as I have some time on my hands. Actually, I think I'll remeasure to verify old filters.
 
Back
Top Bottom