• 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

OP
P

ppataki

Major Contributor
Joined
Aug 7, 2019
Messages
1,216
Likes
1,356
Location
Budapest
@ernestcarl
I have set the Live Playback Latency to Minimum
Buffer is set to max in the ASIO driver of my Topping DM7

Here is the latency in DSP Studio:
1660116542061.png



And here is the delay of the filter itself:

1660116610852.png


To be honest I am not too much worried about the fact that I cannot use convolution
I am now going back to using some VST plugins (CraveEQ and CurveEQ) and again what I tend to conclude is that it sounds more 'lively' - there is no difference in the frequency response so I guess what I hear is attributed to differences in the time domain - I am yet to 'prove' that
I have been spending an awful lot of time trying different methods - as soon as I find the one that sounds (and measures) great I will share it
 

ernestcarl

Major Contributor
Joined
Sep 4, 2019
Messages
3,106
Likes
2,313
Location
Canada
I have set the Live Playback Latency to Minimum
Buffer is set to max in the ASIO driver
of my Topping DM7

Here is the latency in DSP Studio:

I would recommend to keep the former close to or exactly at default and the latter set to as low a number just right under the default as long as it plays "Live" without any long delays, stuttering or glitches.

6 seconds can show up during internal playback, but that is not relevant as what is crucial is the delay value shown during "live" playback. As Henrick (from JRiver) says: The second latency value has no meaning unless you are trying to synchronize live playback.




To be honest I am not too much worried about the fact that I cannot use convolution
I am now going back to using some VST plugins (CraveEQ and CurveEQ) and again what I tend to conclude is that it sounds more 'lively' - there is no difference in the frequency response so I guess what I hear is attributed to differences in the time domain - I am yet to 'prove' that
I have been spending an awful lot of time trying different methods - as soon as I find the one that sounds (and measures) great I will share it

No problem. Though, people should know that they can indeed use convolution just fine without causing any perceptible delay. One just needs to utilize it in a more economical way. Say, purely as an experimental test example, I originally shown here.

Though, I did slightly modify the paragraphic EQ values for better off-axis performance around 15 degrees both horizontally and vertically:

1660126542067.png


1660126546946.png



Don't know if it would interest you at all, but I have ordered an FX50 unit that should be arriving today as well. I'm personally curious if the generic FIR EQ correction I modelled out from Amir's "golden" test unit is going to work well with my own.
 
OP
P

ppataki

Major Contributor
Joined
Aug 7, 2019
Messages
1,216
Likes
1,356
Location
Budapest
I would recommend to keep the former close to or exactly at default and the latter set to as low a number just right under the default as long as it plays "Live" without any long delays, stuttering or glitches.

6 seconds can show up during internal playback, but that is not relevant as what is crucial is the delay value shown during "live" playback. As Henrick (from JRiver) says: The second latency value has no meaning unless you are trying to synchronize live playback.
Good point, thank you!
I will take a look at that later today
No problem. Though, people should know that they can indeed use convolution just fine without causing any perceptible delay. One just needs to utilize it in a more economical way. Say, purely as an experimental test example, I originally shown here.
You are right, I shall also experiment with it further - I realized that if I do not touch phase but normal EQ only, then I can use centering = 0 and that will produce zero latency

See the IRs below using exactly the same settings except for the centering value

this is when using '0' (producing close to zero delay)

1660129203950.png


and this is when using 'middle' (producing the above mentioned >300ms delay)

1660129258134.png


These are just REW-RePhase simulations but will perform some measurements too later today

And one more very interesting phenomenon I realized - would appreciate any comments on this one since it is above my head

When I use EQ in REW I can have various strategies: decrease the target curve and use zero boost or I can keep the target high and use boosting EQ
I have realized that this has quite an impact on how the IR will look like!

there are some differences in the FR curves:
1660130145749.png


I wonder if that is the reason for the IRs to be different.....

see the animation showing the IRs in a loop (starting with the one with the red FR above)

IRs.gif



I really wonder if this is only because of the difference in the FR curve, or by something else....
 

Attachments

  • 1660130205748.png
    1660130205748.png
    676.3 KB · Views: 34

OCA

Addicted to Fun and Learning
Forum Donor
Joined
Feb 2, 2020
Messages
649
Likes
473
Location
Germany
The latest REW (currently only early access from the link below) has new features added in trace arithmetic and it's now possible to create convolution filters with inversion without any EQ filters.


I have been getting consistently good results despite equalization on the entire frequency band (way above Schroeder's) and also uploaded a video tutorial (along with a link to a comprehensive text guide) explaining the method:

 
OP
P

ppataki

Major Contributor
Joined
Aug 7, 2019
Messages
1,216
Likes
1,356
Location
Budapest
The latest REW (currently only early access from the link below) has new features added in trace arithmetic and it's now possible to create convolution filters with inversion without any EQ filters.


I have been getting consistently good results despite equalization on the entire frequency band (way above Schroeder's) and also uploaded a video tutorial (along with a link to a comprehensive text guide) explaining the method:


Thank you @Serkan, this is marvelous!!! I totally love it!! I have been waiting for this feature for years......
I will play around with it as soon as I have some time and will report my findings
 
  • Like
Reactions: OCA
OP
P

ppataki

Major Contributor
Joined
Aug 7, 2019
Messages
1,216
Likes
1,356
Location
Budapest
One quick question @Serkan: you manually do the IR time alignment in the video
Is there a reason for doing that instead of just simply using the Align IR Start button?
Thank you
 

fluid

Addicted to Fun and Learning
Joined
Apr 19, 2021
Messages
691
Likes
1,196
I will play around with it as soon as I have some time and will report my findings
The features in the insider version are a big leap forward for REW, the important item that is missing as far as I can see is the ability to vary the frequency dependent window length with frequency. This makes a big difference in programs that have it. The walkthrough and demonstration in the video is great, particularly for those without any experience. There were quite a lot of descriptions or choices where I found myself disagreeing, I don't want to drag your thread into a negative direction by listing them, but it felt wrong to avoid saying anything, so I suppose I would just caution to not necessarily copy everything seen or spoken about verbatim.
 

fluid

Addicted to Fun and Learning
Joined
Apr 19, 2021
Messages
691
Likes
1,196
Maybe @fluid we shall devote a new thread for that discussion?
Would be really interesting
Maybe, but it could be hard avoid sounding negative towards Serkan (which is not my intention) unless the discussion was more general. Anyone who goes out of their way to provide useful information should be supported.
 
OP
P

ppataki

Major Contributor
Joined
Aug 7, 2019
Messages
1,216
Likes
1,356
Location
Budapest
Hi @Serkan
I started playing around with this method (only simulations in REW for the time being; I will have time to actually listen to how it sounds later tonight)

One question: why shall we set the Target to 3.6 when doing the 1/A operation? Just trying to understand what I am doing

Then with regards to the IR alignment again: I noticed that in the end the MP version will have the IR start at 0 independently if I set that up manually to be at 0 at the beginning of the process

This is the MP correction IR for my left channel when not applying IR offset at the beginning:

1660204357744.png


and here is the MP correction IR when applying IR correction manually at the beginning of the process:

1660204397980.png


They are the same, no matter if I applied IR correction manually or not

Need to add that in the non-MP correction filters the IRs are indeed different but that difference disappears when I generate the MP versions so I am not sure if the manual alignment is necessary if we are going to use the MP versions at the end of the day

Can you please comment?
Thank you
 
  • Like
Reactions: OCA

OCA

Addicted to Fun and Learning
Forum Donor
Joined
Feb 2, 2020
Messages
649
Likes
473
Location
Germany
One quick question @Serkan: you manually do the IR time alignment in the video
Is there a reason for doing that instead of just simply using the Align IR Start button?
Thank you

One quick question @Serkan: you manually do the IR time alignment in the video
Is there a reason for doing that instead of just simply using the Align IR Start button?
Thank you
IR start can get it wrong sometimes but most importantly you will not see the exact fix REW is applying to be able to adjust your speakers accordingly and IR start doesn't necessarily move the IR peaks to time 0 which is handy for the vector calculations afterwards.
 

OCA

Addicted to Fun and Learning
Forum Donor
Joined
Feb 2, 2020
Messages
649
Likes
473
Location
Germany
Maybe, but it could be hard avoid sounding negative towards Serkan (which is not my intention) unless the discussion was more general. Anyone who goes out of their way to provide useful information should be supported.
Guys, I am fully open to all sorts of comments and criticism. At the end of the day, this is an attempt to get the most out of a REW version which is still not even beta and I am not 100% certain of some of the steps either. I tried to share the steps which gave me the best results.
 

Hayabusa

Addicted to Fun and Learning
Joined
Oct 12, 2019
Messages
787
Likes
519
Location
Abu Dhabi
Method 1:
9-point measurement using a 'cube' with a diameter of 1 meter

Method 2:
1-point measurement with REW at 75dB for each channel

You wanted to compare apples to apples..
So why then 9 measurements in a cube for method 1?
Could this explain the differences in FR you see?
 

OCA

Addicted to Fun and Learning
Forum Donor
Joined
Feb 2, 2020
Messages
649
Likes
473
Location
Germany
Hi @Serkan
I started playing around with this method (only simulations in REW for the time being; I will have time to actually listen to how it sounds later tonight)

One question: why shall we set the Target to 3.6 when doing the 1/A operation? Just trying to understand what I am doing

Then with regards to the IR alignment again: I noticed that in the end the MP version will have the IR start at 0 independently if I set that up manually to be at 0 at the beginning of the process

This is the MP correction IR for my left channel when not applying IR offset at the beginning:

View attachment 223577

and here is the MP correction IR when applying IR correction manually at the beginning of the process:

View attachment 223578

They are the same, no matter if I applied IR correction manually or not

Need to add that in the non-MP correction filters the IRs are indeed different but that difference disappears when I generate the MP versions so I am not sure if the manual alignment is necessary if we are going to use the MP versions at the end of the day

Can you please comment?
Thank you
The intention is to check L&R distances to LP and correct them and no harm in adding the offset to the measurement while you are at it. It seems to be the same however there will be very small differences and they will be carried over in the IR windows ref time (as -0.000) of the impulses which will eventually create calculation differences especially in later stages. I haven't included excess phase reversion in the tutorial as it's still work in progress but inverted EPs are very sensitive to time inaccuracies for example.
 

OCA

Addicted to Fun and Learning
Forum Donor
Joined
Feb 2, 2020
Messages
649
Likes
473
Location
Germany
Hi @Serkan
I started playing around with this method (only simulations in REW for the time being; I will have time to actually listen to how it sounds later tonight)

One question: why shall we set the Target to 3.6 when doing the 1/A operation? Just trying to understand what I am doing

Then with regards to the IR alignment again: I noticed that in the end the MP version will have the IR start at 0 independently if I set that up manually to be at 0 at the beginning of the process

This is the MP correction IR for my left channel when not applying IR offset at the beginning:

View attachment 223577

and here is the MP correction IR when applying IR correction manually at the beginning of the process:

View attachment 223578

They are the same, no matter if I applied IR correction manually or not

Need to add that in the non-MP correction filters the IRs are indeed different but that difference disappears when I generate the MP versions so I am not sure if the manual alignment is necessary if we are going to use the MP versions at the end of the day

Can you please comment?
Thank you
Most vector calculations are actually done on the fft (fast fourier transformation) of the impulses and 0 db will give division by zero or infinity errors thus they are applied an offset which we need to compensate for. The exact number 3.6 was given to me by John Mulcahy himself :)
 

OCA

Addicted to Fun and Learning
Forum Donor
Joined
Feb 2, 2020
Messages
649
Likes
473
Location
Germany
Very interested in that too, please let us know once available
It's quite ready and literally upgrades the overall sound by a notch but I have not been able to eliminate the pre-echo from high frequencies without using a combination of Octave, SOX and Audacity and still bit of hit and miss even with these tools. Once I come up with a REW solution I will happily share. Happy to discuss if anyone here has managed to do EP inversion successfully.
 

OCA

Addicted to Fun and Learning
Forum Donor
Joined
Feb 2, 2020
Messages
649
Likes
473
Location
Germany
The features in the insider version are a big leap forward for REW, the important item that is missing as far as I can see is the ability to vary the frequency dependent window length with frequency. This makes a big difference in programs that have it. The walkthrough and demonstration in the video is great, particularly for those without any experience. There were quite a lot of descriptions or choices where I found myself disagreeing, I don't want to drag your thread into a negative direction by listing them, but it felt wrong to avoid saying anything, so I suppose I would just caution to not necessarily copy everything seen or spoken about verbatim.
For starters SPL equalization explanation in the video is wrong (it equalizes a total of the octave value selected around the frequency), I have fixed that in the text guide but I cannot edit the published video.
 

fluid

Addicted to Fun and Learning
Joined
Apr 19, 2021
Messages
691
Likes
1,196
For starters SPL equalization explanation in the video is wrong (it equalizes a total of the octave value selected around the frequency), I have fixed that in the text guide but I cannot edit the published video.
That was one :) Most are like that, fairly minor in significance. The description of the Harman curve, targets and use of windowing would be the main things I see quite differently. Even though you're open to criticism (which is admirable) I don't want to say more unless ppataki wants the discussion here.
 
Top Bottom