• 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!

Dirac Live 3 vs. Manual Effort with my MCLA

ppataki

Major Contributor
Joined
Aug 7, 2019
Messages
1,216
Likes
1,355
Location
Budapest
I have kind of posted this already in my MCLA thread but I wanted to devote a separate thread for additional discussion

The purpose of this post is merely to share my personal experience about using Dirac Live 3.3.0 for room correction versus correcting the room 'manually' using a different approach that I came up with

Below I will describe both methods and provide measurement results (at listening position)

I have a Murphy's Corner Line Array system (see above link for details) which is a 1-way speaker hence my findings might not apply to multi-way systems

Method 1:
Using Dirac Live 3.3.0 and Dirac Processor 1.5.12 on my PC
9-point measurement using a 'cube' with a diameter of 1 meter

Method 2:
1-point measurement with REW at 75dB for each channel
Exporting the measurement into a text file
In Excel subtracting the measured frequency response values from the value 75 so I get a series of points that will fit on a 75dB target curve
Upload the resulting text file into Voxengo CurveEQ per channel (make sure to turn Minimum Phase mode ON)
Note: CurveEQ also supports multichannel, not just stereo

1651481353831.png


This is actually performing the correction

Now the resolution of CurveEQ below 150-200Hz is low so I need to correct that further by performing these steps:
Measure the CurveEQ results again with REW
Using REW EQ with 1/12 smoothing with a target curve of 75dB I am creating correction filters below 150Hz per channel (only cutting!)
Note: I could also fill in the dips with EQ but in my room that will cause highly elevated ringing so I don't do that

1651481639058.png


Then I upload these into Izotope Ozone 9 Equalizer per channel

1651481726680.png



Now let's see the measurements!
I am applying a fully flat target curve in Dirac so I can compare apples to apples

First let's see how MCLA frequency curve looks like without any corrections (right channel)

1651481917313.png


Pretty disgusting, isn't it? :D

Now let's see the same channel corrected using Method 1 vs Method 2

1651482014348.png


Blue is Dirac, Orange is CurveEQ and Ozone EQ

We can see that Dirac is unable to raise the highs to an adequate level and its flatness overall the whole frequency range is somewhat questionable
Regarding the lows there is a 2dB dip btw 25-40Hz with Method 2 but then there is a dip with Dirac btw 55-70Hz so I would call that kind of a tie

Now let's see the distortion figures:

1651482272109.png


No significant difference

Let's see the phase:

1651482385752.png


No significant difference

Step response:

1651482432701.png


I would say that Method 2 has less ringing vs Dirac but only minor difference

Group Delay:

1651482511901.png


No big difference

Wavelet diagrams:

Dirac
1651482577058.png


There is a 2-3ms energy delay starting at 43Hz

Method 2
1651482647586.png


The energy delay starts only at 32Hz

So that's all, any comments/questions are welcome

A few notes:
- I have tried creating filters in REW for the full range and exporting it into a wav impulse file but that has always sounded 'dead' to me. Using the above method provides a much better, 'live' sound, at least to my taste
- I have tried many different EQs but always ended up with Ozone, for some reason I personally like how it 'sounds'
- It would be nice to increase the resolution of CurveEQ below 200Hz - I have had a conversation about that with its author and he said he was thinking about it
- As stated at the beginning, this whole thing might not apply to multi-way systems since Dirac can accurately time align the signal and optimize the step response curve while that is not part of my method and shall provide inferior results from that perspective compared to Dirac

Thank you
 
Last edited:

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,106
Likes
2,313
Location
Canada
Upload the resulting text file into Voxengo CurveEQ per channel (make sure to turn Minimum Phase mode ON)

Interesting... so does this mean that you are essentially not applying any additional manual phase correction -- besides whatever little Dirac might be doing -- and the low group delay is just the natural phase response of your sealed MCLA system?

BTW, have you tried to verify the effectiveness of the EQ with MMM?
 
OP
P

ppataki

Major Contributor
Joined
Aug 7, 2019
Messages
1,216
Likes
1,355
Location
Budapest
That is correct @ernestcarl
In Method 1 I am just using Dirac (whatever 'magic' it applies)
In Method 2 I am just using CurveEQ (frequency values only, no phase information) and Ozone Equalizer as described above

Would be interesting if anybody could try the same with a multiway system to see what happens...

No MMM yet to verify
 
Last edited:

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,106
Likes
2,313
Location
Canada
Would be interesting if anybody could try the same with a multiway system to see what happens...

You mean regular two- or three-ways etc. and maybe plus sub(s)?

It is very likely that "finished" commercial multi-way boxes would not need as much extensive EQ above the transition zone -- few people are ever going to need to apply boosting EQ of up to +20dB up there... :eek: Hmmn... most people are probably shelving down rather than up, too.


I have a Murphy's Corner Line Array system (see above link for details) which is a 1-way speaker hence my findings might not apply to multi-way systems

At least in here, only a handful of people so far have shared that they have the same/similar setup.


You might want to try at least applying some amount of windowing in the HF considering your technique uses only a single point measurement. I think @fluid mentioned before he used 20 cycles for a line array in the very high frequencies (set in a reflective room environment).
 
OP
P

ppataki

Major Contributor
Joined
Aug 7, 2019
Messages
1,216
Likes
1,355
Location
Budapest
most people are probably shelving down rather than up, too.
Yes, I totally realize that ;)
What I meant is to try this method (vs other, more 'conventional' techniques) on other systems too to see what results it will bring

20 cycles for a line array in the very high frequencies (set in a reflective room environment).

Yes, I am planning to do that, I will share the results here once done
I know already that my FR is considerably different if I apply FDW to the measurement in REW so it would be interesting to use that for the basis of EQ-ing and compare the results also by listening to it
 
Last edited:

phoenixdogfan

Major Contributor
Forum Donor
Joined
Nov 6, 2018
Messages
3,294
Likes
5,070
Location
Nashville
I already have Dirac. Bought it in 2015, version 1.1 for my PC. Made a huge difference then, and I upgraded for free to DL 3 which I use as a VST plug-in for J River. I may eventually use the separate stand alone together with APO Equalizer for my two channel set up.

I also paid the upgrade for multi channel. I use that to correct my 5.1 use case.

Dirac is just so easy to use, and delivers such good results, it's worth it to me. They've upgraded, and to date I've only had to pay $125 for the multichannel, every other upgrade was free.

Moreover, they will soon incorporate their own version of MSO, called Dirac Live Bass Control which will be expensive, but, I think, worth the money as well. Should go live in 4th quarter of this year.

And they have still another upgrade in the pipeline scheduled for release in 2023 which allows the individual speakers to intereact and correct each other, hopefully that will be available for Pc's and workstations as well.
 

fluid

Addicted to Fun and Learning
Joined
Apr 19, 2021
Messages
691
Likes
1,196
You might want to try at least applying some amount of windowing in the HF considering your technique uses only a single point measurement. I think @fluid mentioned before he used 20 cycles for a line array in the very high frequencies (set in a reflective room environment).
I use 15 cycles for basic PEQ to bring things under control then DRC_FIR runs with variable window lengths in my settings. These windows are shorter. The high frequencies have a long window to encompass the length of the speaker. A conventional multiway should use shorter windows here too.

This is an extract from @Wesayso as my settings are very heavily influenced by his

R - Band: 0, 20.0 Hz, width: 13230, Cycles: 6.00 R - Band: 1, 25.2 Hz, width: 9822, Cycles: 5.61 R - Band: 2, 31.8 Hz, width: 7417, Cycles: 5.35 R - Band: 3, 40.0 Hz, width: 5670, Cycles: 5.14 R - Band: 4, 50.4 Hz, width: 4374, Cycles: 5.00 R - Band: 5, 63.5 Hz, width: 3398, Cycles: 4.89 R - Band: 6, 80.0 Hz, width: 2653, Cycles: 4.81 R - Band: 7, 100.8 Hz, width: 2081, Cycles: 4.76 R - Band: 8, 127.1 Hz, width: 1637, Cycles: 4.71 R - Band: 9, 160.1 Hz, width: 1292, Cycles: 4.69 R - Band: 10, 201.7 Hz, width: 1022, Cycles: 4.67 R - Band: 11, 254.0 Hz, width: 811, Cycles: 4.67 R - Band: 12, 320.4 Hz, width: 644, Cycles: 4.68 R - Band: 13, 403.7 Hz, width: 513, Cycles: 4.70 R - Band: 14, 508.2 Hz, width: 410, Cycles: 4.72 R - Band: 15, 641.0 Hz, width: 328, Cycles: 4.77 R - Band: 16, 809.1 Hz, width: 263, Cycles: 4.83 R - Band: 17, 1019.7 Hz, width: 212, Cycles: 4.90 R - Band: 18, 1281.9 Hz, width: 172, Cycles: 5.00 R - Band: 19, 1614.6 Hz, width: 140, Cycles: 5.13 R - Band: 20, 2047.0 Hz, width: 114, Cycles: 5.29 R - Band: 21, 2578.9 Hz, width: 94, Cycles: 5.50 R - Band: 22, 3256.4 Hz, width: 78, Cycles: 5.75 R - Band: 23, 4140.9 Hz, width: 65, Cycles: 6.10 R - Band: 24, 5235.5 Hz, width: 55, Cycles: 6.53 R - Band: 25, 6640.3 Hz, width: 47, Cycles: 7.08 R - Band: 26, 8314.2 Hz, width: 41, Cycles: 7.73 R - Band: 27, 10525.4 Hz, width: 36, Cycles: 8.59 R - Band: 28, 13371.4 Hz, width: 32, Cycles: 9.70 R - Band: 29, 16774.4 Hz, width: 29, Cycles: 11.03 F - Band: 30, 22500.0 Hz, width: 26, Cycles: 13.27
 
OP
P

ppataki

Major Contributor
Joined
Aug 7, 2019
Messages
1,216
Likes
1,355
Location
Budapest
Thank you @fluid
Is there a kind of a step-by-step manual for DRC as I see it is not the most user-friendly application I have ever seen.... :)
I have tried DRC Designer but it produced horrible results no matter what settings I used
 
OP
P

ppataki

Major Contributor
Joined
Aug 7, 2019
Messages
1,216
Likes
1,355
Location
Budapest
@ernestcarl @fluid I tried the FDW method in REW using 15 cycles and using exactly the same method as described above (CurveEQ + Ozone EQ)

This is how the optimized FR looks like comparing normal measurement vs the FDW one:

1651737224358.png


And this is how it looks like when I switch off FDW for the FDW measurement:

1651737300123.png


Unfortunately I guess that this is rather what I hear with my ears since when I listen to the optimized FDW measurement it sounds overly bright with less impactful bass and accentuated lower mid frequencies

At first it sounds more 'detailed' and 'airy' but then I realize that it is just harsh and unnatural

I think I would just stick to the 'normal' measurement without FDW
 

fluid

Addicted to Fun and Learning
Joined
Apr 19, 2021
Messages
691
Likes
1,196
The comparison you have made tells you nothing about the frequency dependent window other than when applied to your EQ method the result is very different. The purple curve is flat to rising so it's not at all surprising that you don't like it.

If you want to post or send me the raw base L R impulse responses I can run them through my processing and send you a convolution filter to test.
 
OP
P

ppataki

Major Contributor
Joined
Aug 7, 2019
Messages
1,216
Likes
1,355
Location
Budapest
Hi @fluid
Please see attached the raw measurements per channel
Thank you for your help, much appreciated!
 

Attachments

  • 20220503 Raw.zip
    2.8 MB · Views: 63

fluid

Addicted to Fun and Learning
Joined
Apr 19, 2021
Messages
691
Likes
1,196
It's been a while since I've done this do I shall have to dust the cobwebs off DRC_FIR :)

The 600 to 1k dip seen in the FDW of the left channel is due to the peak energy being quite late there in comparison.

LFDW.png
 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,106
Likes
2,313
Location
Canada
Fluid would be far more qualified to help you with how to use FDW in an array type situation in-room. My speakers are "finished" boxes in a treated room where most of the heavy-lifting EQ necessary to improve sound is mostly in the bass or lower half of the response -- even the Sceptre S8 with its complex (and somewhat rougher) on- and off-axis small horn diffraction pattern and inherent box cabinet weaknesses has been equalized and "voiced" heavily with DSP internally from the get-go. Most I had to deal with (dips and peaks) higher up were likely, largely inconsequential.

Even FDW of 15 cycles, variable smoothing, and manual PEQs within REW of a single-point measurement could be considered usable enough for a full-range correction. Honestly, this is peanuts compared to something like yours -- DIY line-array in a reflective room. *On hindsight, I'm also deeply familiar with my own room and speakers... so I already know where to draw the line when it comes to DSP equalization in my own limited situation.

FDW 15 cycles and variable smoothing:

FDW 15 cycles at 2.3m center of couch MLP.png

If I weren't so obsessed, a single HF shelving filter (variable "tone control" depending on source content and listening volume level) is fully adequate to fix whatever is bothering my own ears above the line.
 
Last edited:

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,106
Likes
2,313
Location
Canada
Hi @fluid
Please see attached the raw measurements per channel
Thank you for your help, much appreciated!

BTW, post DRC would be even better if you could find and tame the direct cause of that strong reflection at ~8ms in the mids visible in the wavelet graph.
 
OP
P

ppataki

Major Contributor
Joined
Aug 7, 2019
Messages
1,216
Likes
1,355
Location
Budapest
BTW, post DRC would be even better if you could find and tame the direct cause of that strong reflection at ~8ms in the mids visible in the wavelet graph.

Once I have the DRC convolution files from @fluid I will give them a try and will of course share the measurements + my subjective listening opinion here
 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,106
Likes
2,313
Location
Canada
I got bored... so to pass the time, I fuddled with your measurements some and created "guestimate" EQ corrections. It's likely "warmer" in comparison to the one you originally made via REW -- and I did not bother boosting below 20Hz -- which you can always change if desired.

FDW 15 var smoothing -- unless otherwise indicated
1.png 2.png 3.png 4a.png 4b.png 5.png 6.png 7.png



8.png



Reverted the EQ settings to generic -- my bad :oops: -- so there is no quick way to transfer the presets into rePhase where you simply generate a convolution file for testing. You'd have to manually enter each one and use "proportional" for the EQ type if you want to try it out.


Wavelet spectrogram:
1/6 res, 40 dB scale, normalized
wave 1.png wave 2.png wave 3.png

BTW, I would not worry about the increase in GD in the low bass as it is mostly below one cycle.


---

*quick tip: under the 'filter tasks' tab inside the EQ window, click generate measurement from predicted (but do disable windowing first). It's easier to work this way than to re-remeasure every time you want to make modifications to the response.

** apply a global negative gain in rePhase to avoid clipping from the boosting applied

***mdat split into two files
 

Attachments

  • pptaki line-array EQ predictions pt1.zip
    3.7 MB · Views: 57
  • pptaki line-array EQ predictions pt2.zip
    3.2 MB · Views: 59
Last edited:

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,106
Likes
2,313
Location
Canada
A giant low bass cut that REW's auto EQ created seems to be the culprit of the increased GD. Didn't notice it before.

spl phase 1.png spl phase 2.png

It's "fixed" in this revision...
 

Attachments

  • pptaki line-array EQ predictions - slight infra bass extend.zip
    2.3 MB · Views: 56

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,106
Likes
2,313
Location
Canada
@ppataki

For some weird reason the "speaker EQ" filters seems to not work correctly with rePhase. I mean if you combine all filters and create a single convolution file the HF part appears to be over-boosted. o_O

If you try some of the filters out the results might turn out frustratingly unlike the prediction...

I've transferred filters into rePhase and adjusted the "speaker EQ" filter set again (first EQ bank) closer to "flat" so a shelving filter can easily be applied at any point. A -2dB high-shelf filter at 3kHz in the last active paragraphic gain EQ bank of each saved profile has been enabled by default.

Anyway, I suspect using the A*B trace arithmetic function in REW combined with filters generated in rePhase may be more reliable.

rePhase Window Left.png rePhase Window Right.png 1651841010132.png

Filters are all in there -- one flat, and the other with a -2dB high-shelf filter. Just load each convolution file or a stereo config settings file linking to the files... dunno if you know how to do that part. But, anyway, this should be pretty doable if a PC is being used as even the max tap settings should work fine.

Ideally, above the transition zone, the "speaker EQ" should be derived from spatially averaged measurements, and possibly also verified with MMM from each channel. Lacking more information, the best I can do is simply use the aligned averaged of the left and right channels [FDW 15 cycles] to create a very generalized flat speaker EQ profile above 1kHz.
 

Attachments

  • ppataki rePhase EQ.zip
    231.4 KB · Views: 51
Last edited:
OP
P

ppataki

Major Contributor
Joined
Aug 7, 2019
Messages
1,216
Likes
1,355
Location
Budapest
Thank you @ernestcarl for all this help, really appreciated!
Inspired by your above posts I have started playing around again with REW and RePhase
That method to generate a measurement from the filters is extremely cool and useful!!
I will post my findings here later on
 
Top Bottom