• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required as is 20 years of participation in forums (not all true). There are daily 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

OP
P

ppataki

Addicted to Fun and Learning
Joined
Aug 7, 2019
Messages
716
Likes
741
Location
Budapest
I think it would be better to create a new thread; @Serkan would you be fine with that please?
Thank you
 

Serkan

Member
Joined
Feb 2, 2020
Messages
63
Likes
32
Yeah, sure m8. I am not very familiar with this forum's rules to be honest...

and sorry for late response, I only just saw this message...
 
Last edited:
OP
P

ppataki

Addicted to Fun and Learning
Joined
Aug 7, 2019
Messages
716
Likes
741
Location
Budapest
Hi @Serkan

We can do two things: either you start a new thread and ask the community for comments or I am also happy to do that for you; please let me know

Yesterday night I was playing around with the new methodology and performed measurements too

See my findings below:

The end result is remarkably good!!!

1660286266043.png


Especially if you take into account that my system is a 1-way MCLA with extremely bad frequency response without EQ
Actually I used some preEQ-ing in Jriver applying shelf filters to first compensate the lows and the highs and then used the new correction method

Before (with shelves!) and after for the left channel:

1660286459273.png


When listening to it it is by far the best sounding result that I have ever had with any convolution type of corrections!
I am now debating with myself to switch to this method vs sticking with the CraveEQ/CurveEQ method......
The advantage of this one is that it is really fast to produce the results (once you are familiar with the steps)

However I also found some weird things:

- if I simulate the convolution (A*B) in REW everything seems to be fine but in reality when I listen to it (and measure) the left channel gets 12dB lower in volume....have no idea why. I compensate that in Jriver so no big issue
- the overall volume is really low, >35dB has to be compensated (again not a real issue but weird)
- I tried using Jriver's convolution and also tried ConvologyXT; the latter is a bit less precise but has almost zero latency! and has a nice normalize feature that auto-compensates for the >35dB loss

Please see the measurements here:

I would love to hear any comments, especially on those weird things (probably I did something incorrectly during the process)
Thank you
 

Serkan

Member
Joined
Feb 2, 2020
Messages
63
Likes
32
I'm glad you liked the results. Some db loss is natural as we're cutting a lot of peaks however, 12 db sounds way too much. I'll have a look at your measurements when I can.

Also please go ahead and move this thread for me.
 
Last edited:

Serkan

Member
Joined
Feb 2, 2020
Messages
63
Likes
32
Hi @Serkan

We can do two things: either you start a new thread and ask the community for comments or I am also happy to do that for you; please let me know

Yesterday night I was playing around with the new methodology and performed measurements too

See my findings below:

The end result is remarkably good!!!

View attachment 223722

Especially if you take into account that my system is a 1-way MCLA with extremely bad frequency response without EQ
Actually I used some preEQ-ing in Jriver applying shelf filters to first compensate the lows and the highs and then used the new correction method

Before (with shelves!) and after for the left channel:

View attachment 223723

When listening to it it is by far the best sounding result that I have ever had with any convolution type of corrections!
I am now debating with myself to switch to this method vs sticking with the CraveEQ/CurveEQ method......
The advantage of this one is that it is really fast to produce the results (once you are familiar with the steps)

However I also found some weird things:

- if I simulate the convolution (A*B) in REW everything seems to be fine but in reality when I listen to it (and measure) the left channel gets 12dB lower in volume....have no idea why. I compensate that in Jriver so no big issue
- the overall volume is really low, >35dB has to be compensated (again not a real issue but weird)
- I tried using Jriver's convolution and also tried ConvologyXT; the latter is a bit less precise but has almost zero latency! and has a nice normalize feature that auto-compensates for the >35dB loss

Please see the measurements here:

I would love to hear any comments, especially on those weird things (probably I did something incorrectly during the process)
Thank you
You have forgotten to use regularisation of 8% during 1/A operation which kept the division peaks in the sub bass area and in return REW had to dim down the SPL of the correction.
 

Serkan

Member
Joined
Feb 2, 2020
Messages
63
Likes
32
I might have over corrected a bit but try the attached convolution filters when you can...

Some reflection at around 22.5 cm away from your mid-woofers and around 4 cm away from your tweeters is effecting the phase responses quite badly and the effect is stronger on the left speaker.

Btw, you can use the shelving, HP, LP and crossover filters of REW (generic EQ have them all), "generate measurement from filters" and convolve that (AxB) with your correction and use that one directly. There's no need to separately use Jriver for that.
 

Attachments

  • ppataki.zip
    663.1 KB · Views: 13
Last edited:

fluid

Addicted to Fun and Learning
Joined
Apr 19, 2021
Messages
594
Likes
1,012
Some reflection at around 22.5 cm away from your mid-woofers and around 4 cm away from your tweeters is effecting the phase responses quite badly and the effect is stronger on the left speaker.
This is a full range line array, there are no separate tweeters and woofers, their measurements are quite different to more regular style speakers.
 

Serkan

Member
Joined
Feb 2, 2020
Messages
63
Likes
32
This is a full range line array, there are no separate tweeters and woofers, their measurements are quite different to more regular style speakers.
I see, excuse my ignorance. I guess it makes a dynamic difference to have many smaller drivers instead of a large one? You still should check for these reflection points though because left and right are quite different to each other.
 
OP
P

ppataki

Addicted to Fun and Learning
Joined
Aug 7, 2019
Messages
716
Likes
741
Location
Budapest
Thank you @Serkan for your comments, appreciate it
I played around with this yesterday and I can confirm that I will stick to this methodology; I totally love the sound it produces; it is very close to what I managed to achieve with CurveEQ (that is also based on inverting the FR) and CraveEQ (to fix the low-end)

I guess in order to further play around this this method I can adjust the target curve + the regularization setting at the 1/A operation (change the 8% to see what happens), right?

I will now open the new thread that we discussed above with @fluid
Also I will open another one for the convolution topic since I found quite a lot of convolution plugins that produce by far less delay vs Jriver's own convolution engine
(I will first perform some measurements and I will share those in that new thread)
FYI @ernestcarl
 
Last edited:
Top Bottom