The Official Dirac Live Thread: Start w/1,2,3 🙋‍♂️

[LCSeminole]

For those of you following the progress of Dirac Live Bass Control(DLBC), the following is a post on AVS Forum by "Flavio @ Dirac" on their progress and what is coming next.

--------------------------------------------------------------------------------------------------------

Hi all,

we promised to post today so this is a short update from the Dirac team.

We’re looking at the reported bass control issue, and are trying our best to find a solution that will serve all use cases. The problem seen in some scenarios (and the measurements that refer to it) relate to allpass filters before the summation of the redirected bass content, in the case the channels are correlated. Depending on the expectation on the source content this may, or may not make sense. As a few examples show, stereo content is as good as always correlated in the bass, which is why bass control consider the phase relation of stereo pairs. Multi channel PCM (anecdotally) also seem to lack correlation outside of the front pair. However, use cases such as upmixing stereo content to multi-channel seem to produce significant amount of correlation, where the allpass filters on the redirected bass does no longer work as intended.

There are various options to resolve this, each with its own set of considerations on advantages and drawbacks, and we are currently evaluating how to provide these options to users. Next week we will provide written material on the issue and the different methods considered for implementation. Our hope is to be able to have the first option out in a beta version shortly.

Happy Easter
Flavio

[marcl]

Apr 2, 2021 23:12:42 GMT -5 LCSeminole said:

For those of you following the progress of Dirac Live Bass Control(DLBC), the following is a post on AVS Forum by "Flavio @ Dirac" on their progress and what is coming next.

--------------------------------------------------------------------------------------------------------

Hi all,

we promised to post today so this is a short update from the Dirac team.

We’re looking at the reported bass control issue, and are trying our best to find a solution that will serve all use cases. The problem seen in some scenarios (and the measurements that refer to it) relate to allpass filters before the summation of the redirected bass content, in the case the channels are correlated. Depending on the expectation on the source content this may, or may not make sense. As a few examples show, stereo content is as good as always correlated in the bass, which is why bass control consider the phase relation of stereo pairs. Multi channel PCM (anecdotally) also seem to lack correlation outside of the front pair. However, use cases such as upmixing stereo content to multi-channel seem to produce significant amount of correlation, where the allpass filters on the redirected bass does no longer work as intended.

There are various options to resolve this, each with its own set of considerations on advantages and drawbacks, and we are currently evaluating how to provide these options to users. Next week we will provide written material on the issue and the different methods considered for implementation. Our hope is to be able to have the first option out in a beta version shortly.

Happy Easter
Flavio

This is interesting, albeit essentially undecipherable. 

Maybe if we knew what the term "correlation" means in this context it would make some sense.  But essentially, it works for stereo and multichannel sources on their own, but not for upmixed stereo (maybe not for upmixed multichannel either?).   I wonder has this been true all along, or did they just break something?

Interesting that we have a similar (unrelated) issue with our Loudness where it works for stereo but introduces seriously exaggerated bass for upmixed multichannel.
 

[uwe]

Apr 3, 2021 6:25:48 GMT -5 marcl said:

Apr 2, 2021 23:12:42 GMT -5 LCSeminole said:

For those of you following the progress of Dirac Live Bass Control(DLBC), the following is a post on AVS Forum by "Flavio @ Dirac" on their progress and what is coming next.

--------------------------------------------------------------------------------------------------------

Hi all,

we promised to post today so this is a short update from the Dirac team.

We’re looking at the reported bass control issue, and are trying our best to find a solution that will serve all use cases. The problem seen in some scenarios (and the measurements that refer to it) relate to allpass filters before the summation of the redirected bass content, in the case the channels are correlated. Depending on the expectation on the source content this may, or may not make sense. As a few examples show, stereo content is as good as always correlated in the bass, which is why bass control consider the phase relation of stereo pairs. Multi channel PCM (anecdotally) also seem to lack correlation outside of the front pair. However, use cases such as upmixing stereo content to multi-channel seem to produce significant amount of correlation, where the allpass filters on the redirected bass does no longer work as intended.

There are various options to resolve this, each with its own set of considerations on advantages and drawbacks, and we are currently evaluating how to provide these options to users. Next week we will provide written material on the issue and the different methods considered for implementation. Our hope is to be able to have the first option out in a beta version shortly.

Happy Easter
Flavio

This is interesting, albeit essentially undecipherable. 

Maybe if we knew what the term "correlation" means in this context it would make some sense.  But essentially, it works for stereo and multichannel sources on their own, but not for upmixed stereo (maybe not for upmixed multichannel either?).   I wonder has this been true all along, or did they just break something?

Interesting that we have a similar (unrelated) issue with our Loudness where it works for stereo but introduces seriously exaggerated bass for upmixed multichannel.
 

Presumably, the term "correlation" was somewhat inaccurately used in place of "coherence" here.
"Correlation" describes only a connection in the statistical mean, "coherence" the interference ability, i.e., a causal relationship to the same cause (source). With a high degree of coherence, constructive or destructive interference or comb filter effects occur with different paths from the source. This can be the case when individual reflections interfere with the direct sound or with each other or with a channel mix.

Coherence is based on fixed phase relationships between the signal paths. However, if the phase difference is 90 deg, the superposition does not lead to any level change, i.e. neither to constructive nor to destructive interference. It is common practice in studio technology to use a frequency-independent 90 deg phase network to avoid interference when mixing mono from stereo. Better the fixed phase relationship, however, decorrelated by delay and reverb.

Happy Easter
Uwe

[jbrunwa]

New XMC-2 owner here. I haven’t received the Dirac kit yet, but I have to say I’m a little freaked out reading about people blowing their speakers out. It’s also a little daunting to go through 160 pages on this thread and having difficulty understanding what to do from the documents and the forums. So I’m not sure how to proceed when I do get the kit. Am I better off using REW and trying to set PEQ? I do have a umik-1. TIA

[marcl]

Apr 3, 2021 15:34:06 GMT -5 jbrunwa said:

New XMC-2 owner here. I haven’t received the Dirac kit yet, but I have to say I’m a little freaked out reading about people blowing their speakers out. It’s also a little daunting to go through 160 pages on this thread and having difficulty understanding what to do from the documents and the forums. So I’m not sure how to proceed when I do get the kit. Am I better off using REW and trying to set PEQ? I do have a umik-1. TIA

Read the first couple posts on Page 1.  That's enough.  When you set volume levels, no need to set them very loud.  Dirac will tell you if it's too quiet.

[LCSeminole]

Apr 3, 2021 15:34:06 GMT -5 jbrunwa said:

New XMC-2 owner here. I haven’t received the Dirac kit yet, but I have to say I’m a little freaked out reading about people blowing their speakers out. It’s also a little daunting to go through 160 pages on this thread and having difficulty understanding what to do from the documents and the forums. So I’m not sure how to proceed when I do get the kit. Am I better off using REW and trying to set PEQ? I do have a umik-1. TIA

As marcl said in the post before mine, make sure to thoroughly read the first 3 posts of this thread. The Dirac software will tell you if it's too low, so starting off the volume too high to where it blows one's speakers just isn't common sense. I use the UMIK-1 as well, just make sure to use the 90deg narrow band response calibration file.

[jbrunwa]

Out of curiosity, why is an outboard computer required in addition to the PC?

[klinemj]

Apr 5, 2021 18:14:19 GMT -5 jbrunwa said:

Out of curiosity, why is an outboard computer required in addition to the PC?

If you are referring to the "EDNIB" (the Pi-based box), Dirac didn't develop the code to have its filters be compatible with the chips the Emotiva G3P's use (which no other pre-pro that uses DIRAC has). Emotiva needed a solution that took the filters DIRAC generated and translate them into a usable filter by Emotiva's chips. The easiest solution was to have the separate Pi device in the chain.

It's no big deal and works fine.

Mark

[fpv]

Got new lcr speakers and updated my dirac to 3.15 the latest one when i do the dirac calibration my right front speaker getting lower volume and the imaging is gone, tried calibrating with my stereo preset, preset 2 the it's the. Same thing the stereo imaging is gone any advice.
one option i see is to roll back to 3.09 worked great before any suggestion

[foggy1956]

Apr 7, 2021 17:59:28 GMT -5 fpv said:

They are both in phase.
I still have my previous dirac profiles when i switch to them the imaging is perfect but the new update has a tighter low frq response but the LR imaging isn't that good at all nothing changed except my lcr speakers

Dirac is fickle, I've had it ruin my soundstage. But, my current measurement gives me a soundstage that I prefer over reference stereo, no idea why.

[jbrunwa]

I’m pleased to say after my first try with XMC-2 Dirac live, although I didn’t measure all the positions, and I wasn’t even very precise on the mike locations, the exported filter has the speakers sounding great.

[LCSeminole]

For those of you following the Dirac Live Bass Control issues, Flavio from Dirac just gave an update yesterday on their progress. The link below has an attachment at the end of the post which gets more into the specifics of Dirac's assumptions behind their DLBC algorithm. Also copied below the link is the post if you don't feel like going straight to Flavio's post on AVS Forum.

www.avsforum.com/threads/the-official-dirac-live-thread.3137722/post-60644462

-------------------------------------------------------------------------------------------------------------------------------------


Hi everyone,

we at Dirac appreciate your commitment and attention in regards to our Bass Control algorithm. We have taken this subject very seriously from day one, even if we have not publicized our process in working on solutions. Now, following extensive internal testing and consultation with a number of affected users, we have sufficient data and understanding to offer two solutions, both of which we expect be available within a few weeks publicly as well as free of charge. Please see the attached document for more info on these. However, if you do not feel satisfied with these solutions and would like to return the product anyway, we, as always, offer full refunds for your purchase. Thank you all for your attention and care. Without you, Dirac Live would not be the product it is today. Thanks again and we will keep you updated.


Update on bass control issue

For some use cases it has been determined that an unexpected amount of correlation is present between the source channels which can lead to cancellations in the digital domain. This mainly seems to concern some upmixed two channel cases, but can also be observed if a low frequency signal is panorated front to back in object audio formats.

There are several options to work around this, each with their own set of advantages and concerns. We are currently considering to add two options that allows a user to choose between three different filter design schemes for bass control:

Co-optimize the subwoofers, then apply traditional bass management.

Advantage: Multiple subwoofers will be co-optimized from a spatial variation perspective
Drawback: Main to sub integration will be uncontrolled

Co-optimize the subwoofers, then use fewer degrees of freedom to work with the phase. More specifically, use a single allpass filter for the redirected bass and individual allpass filters for the full range channels

Advantage: Multiple subwoofers will be co-optimized from a spatial variation perspective. Main to sub integration is controlled. Identical to option 3 for stereo use cases, better than option 3 if channels are correlated, worse if they are not.
Drawback: Having non complementary allpass filters can result in undesired phase shifts significantly above the cross over point, resulting in distorted imaging from pairs of speakers, other than the front pair.
Note: The exact impact and frequency of occurrence of the drawback above must be investigated further.

Bass Control as today

Advantage: Works great for many use cases, like stereo playback with multiple subwoofers, traditional channel based PCM audio.
Drawback: Can have issues with upmixed stereo content, and if present in the content, moving low frequency sources.

The plan is to allow users the choice between these three that is used for a particular filter set. This allows a user to have multiple filter setting (assuming the equipment have the filter storage to upport this) that can be switched between depending on content, if desired.

We hope to have option 1 in beta within the next two weeks. For option 2 we are in the process of implementing that still, and will get back to you with an expected availability date once we have more information.

We have also made a document available reasoning around this, in technical terms, for a stereo use case and a channel based multi channel use case. It does not consider the stereo upmixing use case however.

Dirac Research

[domimag]

Hello all

I just finished reading this 160 page topic and thank you to all of the contributors!

It was a real challenge because it's not my mother tongue (French)

For me, I am now satisfied with DIRAC (v.3.0.14)

My system :
- Preset 1 (cinema) - 7.1.4
- Preset 2 (music) - 5.1

Emotiva XMC-2 (v. 2.2)
Pioneer LX500 (cinema)
Sony X800 M2 (music)

speakers (active):
PSI Audio A21 (front)
PSI Audio A17 (center, surround)
JBL 305P MKII (back, height)
SVS SB3000 (sub)

XMC-2, Raspberry and PC are connected to the same ethernet hub, linked to internet

"Microphone gain"
- I avoided a lot of "noise" and "dismiss" detection problems by setting the microphone gain not to rise -30 db but stay under -50 db (as mentioned by some NAD users)

"Master output level"
- I adjust the volume of the subwoofer not with the software "black tab" but directly on the subwoofer. (- 23 db). I found more accurate "bass" doing like that.
- I do the same with my speakers (amplified speakers). This is obviously not possible with passive speakers.

In "direct mode", levels are now perfect between the five speakers.

I have sometimes had, as I have read several times here, problems with distortion with the treble, and even the lack of treble, especially on the surround speakers.
I also had volume level changes depending on the sessions with the same filter

I fix them (hope to do so...) by doing the following:
- a "factory reset"
- I entered, like for the first use, my complete configuration (about 20 minutes)
- I exported to the XMC-2 a cinema filter on preset 1 and music on preset 2 and checked that the two filters suited me perfectly for listening.
- I made a save of this configuration in the internal memory of the XMC-2

Since then, I have added a filter to each preset (in a blank slot) and for 3 weeks, I no longer have any of the issues mentioned.
When my 3 slots are occupied, I will do a "factory reset" again followed by a backup but I will no longer export a new filter by overwriting a previous filter.

Do you think that recording several times in the same slot could be the source of the problems?

Thank you and my apologizes for my poor english.

[jbrunwa]

I read the info in the first 3 posts and watched videos. I apologize if I missed this anywhere, what settings are important to have correct before starting Dirac other than speaker selection? For example, speaker distance, small or large, crossover frequencies, slope, levels, loudness, DRC, etc? Or does Dirac ignore these? If not, and these need to be set before Dirac, what is the recommendation for selecting crossovers, slopes, etc? One of the links talks about crossovers in Dirac, do I go back and set them in XMC-2 speaker preset, or is Dirac independent? Thanks in advance

[geebo]

Apr 11, 2021 17:54:30 GMT -5 jbrunwa said:

I read the info in the first 3 posts and watched videos. I apologize if I missed this anywhere, what settings are important to have correct before starting Dirac other than speaker selection? For example, speaker distance, small or large, crossover frequencies, slope, levels, loudness, DRC, etc? Or does Dirac ignore these? If not, and these need to be set before Dirac, what is the recommendation for selecting crossovers, slopes, etc? One of the links talks about crossovers in Dirac, do I go back and set them in XMC-2 speaker preset, or is Dirac independent? Thanks in advance

Dirac ignores all of that when taking measurements and creating filters. However, any Level setting you have done in the Preset's settings is added to the Dirac levels afterwards. I usually set my levels to zero then adjust them with an SPL after Dirac. All speakers are usually with .5 or 1.0 dB of each other after Dirac. If you forget to zero your levels it's no big deal, just zero them out afterwards.

[jbrunwa]

Apr 11, 2021 18:19:13 GMT -5 geebo said:

Apr 11, 2021 17:54:30 GMT -5 jbrunwa said:

I read the info in the first 3 posts and watched videos. I apologize if I missed this anywhere, what settings are important to have correct before starting Dirac other than speaker selection? For example, speaker distance, small or large, crossover frequencies, slope, levels, loudness, DRC, etc? Or does Dirac ignore these? If not, and these need to be set before Dirac, what is the recommendation for selecting crossovers, slopes, etc? One of the links talks about crossovers in Dirac, do I go back and set them in XMC-2 speaker preset, or is Dirac independent? Thanks in advance

Dirac ignores all of that when taking measurements and creating filters. However, any Level setting you have done in the Preset's settings is added to the Dirac levels afterwards. I usually set my levels to zero then adjust them with an SPL after Dirac. All speakers are usually with .5 or 1.0 dB of each other after Dirac. If you forget to zero your levels it's no big deal, just zero them out afterwards.

Great, thanks! So it’s just Dirac and done!

[geebo]

Apr 11, 2021 18:56:35 GMT -5 jbrunwa said:

Apr 11, 2021 18:19:13 GMT -5 geebo said:

Dirac ignores all of that when taking measurements and creating filters. However, any Level setting you have done in the Preset's settings is added to the Dirac levels afterwards. I usually set my levels to zero then adjust them with an SPL after Dirac. All speakers are usually with .5 or 1.0 dB of each other after Dirac. If you forget to zero your levels it's no big deal, just zero them out afterwards.

Great, thanks! So it’s just Dirac and done!

Yep, but you might want to check channel levels when finished.

[jbrunwa]

Apr 11, 2021 19:18:54 GMT -5 geebo said:

Apr 11, 2021 18:56:35 GMT -5 jbrunwa said:

Great, thanks! So it’s just Dirac and done!

Yep, but you might want to check channel levels when finished.

Ok, thanks.

[kempf]

Apr 11, 2021 20:00:24 GMT -5 jbrunwa said:

Apr 11, 2021 19:18:54 GMT -5 geebo said:

Yep, but you might want to check channel levels when finished.

Ok, thanks.

Hi!"

After doing Dirac i have to redo channel levels with a sound level meter, there are always one or 2 speakers in my setup (9.1.6) that at totally out level.

[Dreamer]

I have my Dirac equipment setup correctly, networks firing away, everything's ready to go. Except when I go into setup, speakers, Dirac.......I get "Dirac not available"
Why?
What have I missed?