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

[marcl]

Dirac 3.9.7 Beta is posted.

The only difference in the release notes is one additional line: Fixed an issue that could cause the app to stall after a failed login attempt.

I would bet they did something else, because what happened with 3.9.2 to 3.9.6?

www.dirac.com/live/faq/helpdesk/dirac-live-beta-releases/

[ttocs]

Feb 25, 2024 10:17:28 GMT -5 marcl said:

Okay, first ... note that I had omitted something from my previous post so I corrected it.  Magic Beans determines a target curve in addition to filter coefficients.  You import the target curve to your system as well as the filter coefficients.  But if you're using Dirac, you can only import the target curve ... just like you would import a Harman curve or whatever.

Now ... the crux of the biscuit ...

I found this video where Joe is doing a demo for some folks, and most of it ends up being a dialogue with Justin Zazzi, Senior Acoustic Research Scientist/Engineer at Harman X.

Justin does what a respectful scientist/engineer would do.  He asks questions, paraphrases, offers analogies ... to try to understand what Joe is saying.  

hmmmmmmm . . . .

Sooooooo, now I get what this is supposed to be doing. I'm not impressed by the "demo". Joe needs to polish up the demo a bit as it kinda reminds me of a tense argument between John Candy and Steve Martin in Planes, Trains, and Automobiles, when Steve Martins blurts out "And by the way, when you're telling these little stories?, here's an idea, have a point!!"

Putting aside their discussion about the usage of some terms and such, the actual demo was unfinished. No results were shown. No explanation of what was really different after the measurement process was completed compared to before, in this particular room with these speakers. Just a lot of mish mash about "targets". Ok, so show the targets, and show a comparison. But instead it just ends like the ending to The Blair Witch Proje . . . . . . . . . . . . . . . . . .

[ttocs]

Feb 25, 2024 10:17:28 GMT -5 marcl said:

Justin does what a respectful scientist/engineer would do.  He asks questions, paraphrases, offers analogies ... to try to understand what Joe is sayin

After watching the video, I didn't see or hear anything from Joe that would make me want to pursue the concept any further.

I heard rambling on about targets, got a sense that it's a difference of near field speaker measurement vs MLP, but then what? (Rhetorical)  How is this better than other methods? Where is the Magic?

Back to Dirac.

[marcl]

Feb 25, 2024 17:04:43 GMT -5 ttocs said:

Feb 25, 2024 10:17:28 GMT -5 marcl said:

Justin does what a respectful scientist/engineer would do.  He asks questions, paraphrases, offers analogies ... to try to understand what Joe is sayin

After watching the video, I didn't see or hear anything from Joe that would make me want to pursue the concept any further.

I heard rambling on about targets, got a sense that it's a difference of near field speaker measurement vs MLP, but then what? (Rhetorical)  How is this better than other methods? Where is the Magic?

Back to Dirac.

He really doesn't have a point, unfortunately.  He came up with a different way to make target curves and filters.  He doesn't have any scientific premise as to why it would be better than other methods.  Other videos (there are several scattered around his and Channa's YouTube channels) show measurements and talk about how to load filters and curves, but he doesn't START with a premise:  what's lacking with current tools and methods?  How does this method solve the problem?

Instead, he basically does this and then he listens and lets Channa listen and they say "wow doesn't this sound great?"  But does it sound better than Dirac or Audyssey?

Reminds me of an audio show in NY 10-12 years ago.  None other than Mark Levinson himself was demoing a software product called Burwen Bobcat, said to make MP3 files sound amazingly better!  People lined up at the hotel room door to go in for the demo.  He'd play a piece of music and say "doesn't that sound great?"  Then he'd play another piece of music and say "amazing, isn't it?" .   Finally someone spoke up and asked "could you turn it off so we can hear the difference?".  Levinson brushed him off saying "uhh, no we don't have time for that".

But I must say ... Joe seems like a nice guy, and sincerely believes he has something valuable.  Oh well ...

[marcl]

Meanwhile .... today I did a Dirac calibration with 3.9.7 Beta.

I had done an alignment of the Aperion tweeters that I added to complement my Magnepan LRS surrounds. I measured the impulse response of each and found the Aperion to be just slightly "faster" than the LRS even though they physically were in line with the LRS tweeter.   But, I wasn't surprised given the difference in size and mass. I would have been surprise if the converse had been the case.  The difference was 100 microseconds, which translates to 1.4 inches. I moved the Aperions 1.4" back, and the impulse responses aligned perfectly.

Did it sound better after the new calibration? I can't say it sounds better, but it sounds awesome. And I know it IS better 

[PaulBe]

Feb 25, 2024 17:23:48 GMT -5 marcl said:

Feb 25, 2024 17:04:43 GMT -5 ttocs said:

After watching the video, I didn't see or hear anything from Joe that would make me want to pursue the concept any further.

I heard rambling on about targets, got a sense that it's a difference of near field speaker measurement vs MLP, but then what? (Rhetorical)  How is this better than other methods? Where is the Magic?

Back to Dirac.

He really doesn't have a point, unfortunately.  He came up with a different way to make target curves and filters.  He doesn't have any scientific premise as to why it would be better than other methods.  Other videos (there are several scattered around his and Channa's YouTube channels) show measurements and talk about how to load filters and curves, but he doesn't START with a premise:  what's lacking with current tools and methods?  How does this method solve the problem?

Instead, he basically does this and then he listens and lets Channa listen and they say "wow doesn't this sound great?"  But does it sound better than Dirac or Audyssey?

Reminds me of an audio show in NY 10-12 years ago.  None other than Mark Levinson himself was demoing a software product called Burwen Bobcat, said to make MP3 files sound amazingly better!  People lined up at the hotel room door to go in for the demo.  He'd play a piece of music and say "doesn't that sound great?"  Then he'd play another piece of music and say "amazing, isn't it?" .   Finally someone spoke up and asked "could you turn it off so we can hear the difference?".  Levinson brushed him off saying "uhh, no we don't have time for that".

But I must say ... Joe seems like a nice guy, and sincerely believes he has something valuable.  Oh well ...

Joe needs to spend some time learning enough till he realizes he doesn't know very much.  A real education in any area should be a humbling experience.  He is focusing on the magic to make beans rather than do the real fundamental work.  But, he does seem like a nice guy.

I think " ...Justin Zazzi, Senior Acoustic Research Scientist/Engineer at Harman X.", was at Joe's place just to see if Joe stumbled onto something useful for further real research. Once in a while, a blind squirrel gets a nut.

[KeithL]

Full disclosure... I only watched some of the video...

However, that said, I can see one area in which he might have a point of sorts... and that is the idea of "the moving measurements".
I also suspect that this idea may have come up before...
(Although, that said, while I think he might have a point, I'm not so sure about his solution to address it.)

To pick a single point... at 10 kHz a single wavelength is about one inch.
This means that, if you have a 10 kHz tone, playing from two speakers, and you try to measure it...
If the sound from both speakers is in-phase at one point, it will be out-of-phase an inch away, and so on...
You actually have a "three dimensional comb filter effect" between the two sources.
And, even if you only have a single active source, the contributions from room reverberance must be considered to constitute a whole bunch of sources as well.
And, of course, the measurement element of most real-world measurement microphones is a significant fraction of an inch across.
The result is that it's going to be difficult to get an accurate measurement at all... and especially by only using a few sample locations.
(And, in the human world, we seem not to notice that comb filter effect much, which suggests that our ears and our brains somehow compensate for it.)
I think the idea is that, by moving the microphone around in a pattern, you are effectively "getting an average of an infinite number of measurement points spread over a specific area".
And, presumably, that this is somehow a better representation of what we hear than simply averaging a few measurements at fixed points. 

We also need to remember that, strictly speaking, any and all discussion of impulse and frequency only applies at a single point in 3d space.
So any possible correction that you can calculate or apply will only be perfectly true at one point in 3d space... and slightly less true everywhere else.
And what sort of approximation you apply to get it to work over any larger finite area is some of that "secret sauce" that various room correction software uses.
(That's the part where you're trading between a really good solution for one specific spot and a slightly less good solution that works over a larger area.)

I think his idea is that the "moving measurement" is sort of an ad-hoc way to "create the secret sauce"...
(Move the reference point around over an area to create a solution that will work over an area.)
As I said, I'm not at all convinced that his idea is valid, or will work well, but I can see at least the possibility that it might be worthwhile.

Feb 25, 2024 9:17:20 GMT -5 marcl said:

lrobertson   lcseminole  ttocs I think it was just you guys commenting on this. Tag others if I forgot somebody.

Magic Beans

Okay I looked into this but honestly I'm not going to go back 18 months and watch every video that Joe and Channa made to try to piece together what Joe is trying to accomplish and what (if any) scientific basis he has for his approach to Magic Beans (because I'll bet my magic beans he doesn't).  But I think I have enough info to make some comments:

What do you do with Magic Beans?

1. Take near-field measurements (within a range similar to speaker baffle width), and MLP measurements for each speaker using pink noise, and moving the mic around in a pattern at the measurement position.  No multiple listening position measurements, just MLP.
2. Magic Beans uses this data to (somehow) come up with a response for each speaker.
3. You pick a target curve, and Magic Beans generates the coefficients for PEQ filters to correct the response, and they can be imported into various processors (Emotiva included).  Similar to how using REW EQ automatically generates filters and gives you coefficients that can be imported.
4. You import the PEQ coefficients to your processor.  
5. If you have the Audyssey version that they reference running in your processor, you can import the data into Audyssey   Otherwise you use the PEQ function of your processor.
6. You also import the target curve from Magic Beans.  You can import this target curve to any system, including Dirac.

What DON'T you do with Magic Beans?

1. You don't use the filters with Dirac in any way.  There is no way to import filter coefficients if you're using Dirac.
2. This means, you can't use Dirac for impulse response correction and time alignment ... and Magic Beans for filter coefficients.
3. You have to do time alignment some other way, such as manually entering distances or using the automated tool in Audyssey or in some processors.

What is the point of using Magic Beans?

Well ... uh ...  I mean ... there is no scientific precedent or logic to using near-field measurements for room correction.  Taking measurements at multiple listening positions and applying some algorithm to optimize response makes sense.  But taking measurements at ONE listening position and combining it somehow with near-field measurements makes no sense.  If a speaker had an inherent flaw that revealed itself in a near-field measurement, you could argue that it could be corrected with EQ, and in fact you could even legitimately correct a dip in the bass response if there was one.  But it really doesn't matter because it's the listening position response that matters.

So ... the only point of using Magic Beans would be if - for whatever personal reason - someone does not LIKE what Dirac does, or does not LIKE what Audyssey does ... then this would be a way to play with an alternate app until they get what they want. 

p.s.  In at least one video Joe talks about impulse response and time alignment and trying to get similar results to Dirac or whatever using REW.  Note the following:

1. Time alignment and impulse response correction are two different things.  Dirac does both.  There is no way to do impulse response correction without some sophisticated algorithm and filter implementation similar to Dirac.  Maybe Trinnov does it, but I know of no other way.
2. Yes, you can determine time alignment using the impulse response measurements in REW.  These can be converted to distance and used to manually enter distances to a processor.  But if you use Dirac, you can't enter distances manually with Emotiva processors.
3. In one video Channa talks about how, after doing a Dirac calibration, he used the time alignment files in Spatial Audio Toolkit and (amazed look on his face) Dirac had aligned everything so perfectly that he could not hear the slightest discrepancy in the alignment.

p.p.s. Further evidence that the vast majority of people - even those making YouTube videos - ignore the impulse response correction that Dirac does.

[richb]

Feb 26, 2024 12:59:08 GMT -5 KeithL said:

Full disclosure... I only watched some of the video...

However, that said, I can see one area in which he might have a point of sorts... and that is the idea of "the moving measurements".
I also suspect that this idea may have come up before...
(Although, that said, while I think he might have a point, I'm not so sure about his solution to address it.)

To pick a single point... at 10 kHz a single wavelength is about one inch.
This means that, if you have a 10 kHz tone, playing from two speakers, and you try to measure it...
If the sound from both speakers is in-phase at one point, it will be out-of-phase an inch away, and so on...
You actually have a "three dimensional comb filter effect" between the two sources.
And, even if you only have a single active source, the contributions from room reverberance must be considered to constitute a whole bunch of sources as well.
And, of course, the measurement element of most real-world measurement microphones is a significant fraction of an inch across.
The result is that it's going to be difficult to get an accurate measurement at all... and especially by only using a few sample locations.
(And, in the human world, we seem not to notice that comb filter effect much, which suggests that our ears and our brains somehow compensate for it.)
I think the idea is that, by moving the microphone around in a pattern, you are effectively "getting an average of an infinite number of measurement points spread over a specific area".
And, presumably, that this is somehow a better representation of what we hear than simply averaging a few measurements at fixed points. 

We also need to remember that, strictly speaking, any and all discussion of impulse and frequency only applies at a single point in 3d space.
So any possible correction that you can calculate or apply will only be perfectly true at one point in 3d space... and slightly less true everywhere else.
And what sort of approximation you apply to get it to work over any larger finite area is some of that "secret sauce" that various room correction software uses.
(That's the part where you're trading between a really good solution for one specific spot and a slightly less good solution that works over a larger area.)

I think his idea is that the "moving measurement" is sort of an ad-hoc way to "create the secret sauce"...
(Move the reference point around over an area to create a solution that will work over an area.)
As I said, I'm not at all convinced that his idea is valid, or will work well, but I can see at least the possibility that it might be worthwhile.

Feb 25, 2024 9:17:20 GMT -5 marcl said:

lrobertson   lcseminole  ttocs I think it was just you guys commenting on this. Tag others if I forgot somebody.

Magic Beans

Okay I looked into this but honestly I'm not going to go back 18 months and watch every video that Joe and Channa made to try to piece together what Joe is trying to accomplish and what (if any) scientific basis he has for his approach to Magic Beans (because I'll bet my magic beans he doesn't).  But I think I have enough info to make some comments:

What do you do with Magic Beans?

1. Take near-field measurements (within a range similar to speaker baffle width), and MLP measurements for each speaker using pink noise, and moving the mic around in a pattern at the measurement position.  No multiple listening position measurements, just MLP.
2. Magic Beans uses this data to (somehow) come up with a response for each speaker.
3. You pick a target curve, and Magic Beans generates the coefficients for PEQ filters to correct the response, and they can be imported into various processors (Emotiva included).  Similar to how using REW EQ automatically generates filters and gives you coefficients that can be imported.
4. You import the PEQ coefficients to your processor.  
5. If you have the Audyssey version that they reference running in your processor, you can import the data into Audyssey   Otherwise you use the PEQ function of your processor.
6. You also import the target curve from Magic Beans.  You can import this target curve to any system, including Dirac.

What DON'T you do with Magic Beans?

1. You don't use the filters with Dirac in any way.  There is no way to import filter coefficients if you're using Dirac.
2. This means, you can't use Dirac for impulse response correction and time alignment ... and Magic Beans for filter coefficients.
3. You have to do time alignment some other way, such as manually entering distances or using the automated tool in Audyssey or in some processors.

What is the point of using Magic Beans?

Well ... uh ...  I mean ... there is no scientific precedent or logic to using near-field measurements for room correction.  Taking measurements at multiple listening positions and applying some algorithm to optimize response makes sense.  But taking measurements at ONE listening position and combining it somehow with near-field measurements makes no sense.  If a speaker had an inherent flaw that revealed itself in a near-field measurement, you could argue that it could be corrected with EQ, and in fact you could even legitimately correct a dip in the bass response if there was one.  But it really doesn't matter because it's the listening position response that matters.

So ... the only point of using Magic Beans would be if - for whatever personal reason - someone does not LIKE what Dirac does, or does not LIKE what Audyssey does ... then this would be a way to play with an alternate app until they get what they want. 

p.s.  In at least one video Joe talks about impulse response and time alignment and trying to get similar results to Dirac or whatever using REW.  Note the following:

1. Time alignment and impulse response correction are two different things.  Dirac does both.  There is no way to do impulse response correction without some sophisticated algorithm and filter implementation similar to Dirac.  Maybe Trinnov does it, but I know of no other way.
2. Yes, you can determine time alignment using the impulse response measurements in REW.  These can be converted to distance and used to manually enter distances to a processor.  But if you use Dirac, you can't enter distances manually with Emotiva processors.
3. In one video Channa talks about how, after doing a Dirac calibration, he used the time alignment files in Spatial Audio Toolkit and (amazed look on his face) Dirac had aligned everything so perfectly that he could not hear the slightest discrepancy in the alignment.

p.p.s. Further evidence that the vast majority of people - even those making YouTube videos - ignore the impulse response correction that Dirac does.

I am not sure who has more magic beans the moving mic or Dirac marketing.
I could be thick, but given the length of the sound waves at higher frequency, seems to defy physics.

I remember, when Panasonic Plasma was marketing with Infinite Black, then Infinite Black Pro.
It makes you wonder if these people failed math, or marketing is not tethered to reality or verasity.

At best, Dirac could avoid phase correct when the measured points show different phase at measured positions.

For me, this never passed the smell test at higher frequencies.
End users cannot see the filter corrections applied.
Many users simple apply the correction listen in an uncontrolled environment, so the opportunity of for conformation bias is real.
So, for most this looks scientific but is ultimately, subjective.
That is fine by me, BTW.

There are those that independently measure the after results and that is better.
I am not sure if folks measure impulse at same positions, at higher frequencies that seems daunting.

- Rich

[marcl]

Dirac Version 3.9.7 is now Official


www.dirac.com/live/faq/helpdesk/software-changelog/


Dirac Live 3.9.7

2024-02-28



Features

Faster filter export on most devices

Users experiencing recurring imprecise measurements on Windows PCs now have the possibility to opt-in for using the microphones in Exclusive Mode by adding a Windows Environment Variable:

Click here for instructions how to activate exclusive mode.

Fixes

Fixed the known problem where some NAD and other devices didn’t populate the subwoofer filter unless using version 3.4.4 or older.

Fixed the known problem where some NAD devices got substantially louder sound with Dirac ON than with Dirac OFF.

Fixed a problem when a general and misleading error message was displayed when trying to enable Dirac filters while being in measurement mode.

Fixed a problem that blocked the app window from being moved while being busy with background operations.

Fixed a UI glitch that exposed Windows-style title-bar buttons also in MacOS systems.

Fixed an issue that could cause the app to stall after a failed login attempt.


Known problems

Filter export to devices from Onkyo/Integra/Pioneer take longer time than usual. Please be patient and wait for it – it will eventually complete.

Exclusive mode measurements appear with lower level in the Measurement UI. But they are useful anyway – no need to crank up the volume to unsafe levels.

Older Windows 10 PCs may have problems with the app not displaying properly.

Workaround

Press the Windows button or the search field, type “env” and press Enter

Click the “Environment Variables…” button

Click the “New…” button under “User variables for <user>“

Enter the following:

Variable name: QSG_RHI_PREFER_SOFTWARE_RENDERER

Variable value: 1

Click “OK”

Restart the Dirac Live application


Download


Windows: artifacts.connect.dirac.com/public/accord/release/win64/diraclive-v3.9.7-setup.exe


MacOS: artifacts.connect.dirac.com/public/accord/release/macos/diraclive-v3.9.7-setup.zip

[ttocs]

Mar 4, 2024 17:39:43 GMT -5 marcl said:

Dirac Version 3.9.7 is now Official
Dirac Live 3.9.7
2024-02-28

Known problems
Exclusive mode measurements appear with lower level in the Measurement UI. But they are useful anyway – no need to crank up the volume to unsafe levels.

This one is very interesting! This is what is seen during Measurement UI when using a USB-C mic with a macOS computer. So now they're saying, "it's ok, really, it's ok, just use the measurement that isn't being heard by the mic."

Well, if one considers that the relationship between sweep volume and room noise, this could be ok. But the mic isn't being used as optimally as it would if the sweep was heard as loudly by the mic in the mic's sweet spot of SPL sensitivity.

Thanks as always Marc for sharing the latest Dirac!

[webmst007]

New Stereonet. interview with Dirac:

www.stereonet.com/au/features/inside-track-dr-mathias-johansson-dirac

Simple intro.

[marcl]

Matthew Poes discusses Dirac Target Curves

[klinemj]

Where, in this 225 page thread, can I find the latest, greatest tips on how to run DIRAC? I recall there being a lot of great tips on how to run it most effectively, but I'm lost as to where to find it. And, I need to run it again.

Thanks!

Mark

[ttocs]

Mar 14, 2024 18:37:13 GMT -5 klinemj said:

Where, in this 225 page thread, can I find the latest, greatest tips on how to run DIRAC? I recall there being a lot of great tips on how to run it most effectively, but I'm lost as to where to find it. And, I need to run it again.
Thanks!

Mark

Hi Mark! Try this just a couple pages back. It's an interesting thing to try.
emotivalounge.proboards.com/post/1121762/thread

But if you mean what settings to use, that will be dependent on some things like which mic you're using and which OS. The UMIK mics with USB-C connection on them can make things more difficult if using a Mac, but on Win they seem to be easier to contend with.

This is what I do:
Mic Gain at 100%,
Master Output very low at first,
Play the pink noise for a few seconds on every channel but don't make any adjustments yet.
After the pink noise has been played on every channel and stopped,
Look for the channel with weakest output and adjust the Master Output until the Level is at your desired setting.
Then match all the other Speaker Channels to that setting.
Then set the Subwoofer(s) to a setting that is +6dB higher than the Speaker channels, this will cause Dirac to play the subwoofer sweeps at the same level as the speakers in Measurement.
Done with Volume Calibration, go to next step.

After measuring all the mic locations you want to use, SAVE THE PROJECT before going to filter design. This will give you options for things you might want to do later, like removing some mic locations.

Something I wasn't aware of until recently is how irrelevant the "depicted" mic locations are in the graphic in Measurement. Other than the one in the MLP, the first measurement which has the most importance because it will include Impulse data, you could pick any of the dots you want in which to store any particular measurement from a mic location. I used to think that the mic locations should be close to what the graphic depicted, but no, it doesn't matter. They are simply a guide, they are just place holders for data from each measurement.

Not every system can use the same exact settings. But, if I use my UMIK-1/USB-A mic with -18dB sensitivity (miniDSP Factory default setting), I set my speakers to -24dB and my subs to -18dB in Volume Calibration.

This oughta be a good start. 

[klinemj]

Mar 14, 2024 19:30:19 GMT -5 ttocs said:

Mar 14, 2024 18:37:13 GMT -5 klinemj said:

Where, in this 225 page thread, can I find the latest, greatest tips on how to run DIRAC? I recall there being a lot of great tips on how to run it most effectively, but I'm lost as to where to find it. And, I need to run it again.
Thanks!

Mark

Hi Mark! Try this just a couple pages back. It's an interesting thing to try.
emotivalounge.proboards.com/post/1121762/thread

But if you mean what settings to use, that will be dependent on some things like which mic you're using and which OS. The UMIK mics with USB-C connection on them can make things more difficult if using a Mac, but on Win they seem to be easier to contend with.

This is what I do:
Mic Gain at 100%,
Master Output very low at first,
Play the pink noise for a few seconds on every channel but don't make any adjustments yet.
After the pink noise has been played on every channel and stopped,
Look for the channel with weakest output and adjust the Master Output until the Level is at your desired setting.
Then match all the other Speaker Channels to that setting.
Then set the Subwoofer(s) to a setting that is +6dB higher than the Speaker channels, this will cause Dirac to play the subwoofer sweeps at the same level as the speakers in Measurement.
Done with Volume Calibration, go to next step.

After measuring all the mic locations you want to use, SAVE THE PROJECT before going to filter design. This will give you options for things you might want to do later, like removing some mic locations.

Something I wasn't aware of until recently is how irrelevant the "depicted" mic locations are in the graphic in Measurement. Other than the one in the MLP, the first measurement which has the most importance because it will include Impulse data, you could pick any of the dots you want in which to store any particular measurement from a mic location. I used to think that the mic locations should be close to what the graphic depicted, but no, it doesn't matter. They are simply a guide, they are just place holders for data from each measurement.

Not every system can use the same exact settings. But, if I use my UMIK-1/USB-A mic with -18dB sensitivity (miniDSP Factory default setting), I set my speakers to -24dB and my subs to -18dB in Volume Calibration.

This oughta be a good start. 

Thanks! That's what I needed! FYI, I've got a UMIK with Windows. It's been easy in the past, but I knew there were differing recos on mic gain and master output for a while. I just could not remember them as it's been a while!

Mark

[klinemj]

AudioHTIT

Just a quick but HUGE thank you for putting this thread together and keeping all the latest info on the first post!

It's been a while since I needed to run DIRAC, but I'm getting all new amps AND a new PC that will need to have DIRAC (and Roon) installed on it.

I've forgotten so much about how to even download DIRAC, let alone run it! Your post is immensely helpful.

So...THANK YOU!

Mark

[KeithL]

I just noticed one thing you said here that is worth mentioning... at least a little bit.
And that is that I believe that there IS a valid - if "indirect" - precedent for using both "full room" and "near field" measurements.
(Note that I'm not saying that Magic Beans is using the information this way.)

When calculating "the perfect room and speaker correction"... you would ideally have a LOT of information.
That information would include things like the dimensions and shape of the room...
As well as the acoustic properties of each individual surface in that room...
As well as the exact properties of each speaker... including directional information like its radiation pattern at each frequency...
Therefore, at least in theory, having more and different information is always going to be a potential advantage.
(And, in this case, I can definitely see how having "a near-field quasi-anechoic profile for each speaker" would be potentially useful information.)

I'm going to throw out a single, but not uncommon, example where this MIGHT be the case.
Let's say you have a simple 5.1 channel system...
And, before any correction, you find that center channel dialog tends to be somewhat muddled and lacking in intelligibility.
And, in perfect hindsight, we now know this is partly because the floor in front of the couch, between the couch and the center speaker, is shiny bare hardwood.

In that situation we might actually be faced with the fact that "electronic room correction" is not going to be able to provide a perfect solution.
However, knowing details, like the vertical dispersion of the center channel speaker, specifically at midrange frequencies, would be helpful in evaluating possible answers.
(The ideal solutions are going to be to add a rug on the floor in front of the couch... or get a new center channel speaker with more limited vertical dispersion at midrange frequencies.)
(And, even beyond that, knowing the acoustic properties of the ceiling, between the center channel speaker and the couch, will help inform us how much adding a rug will probably help the situation.)

This is not a perfect example by any means...
But my point is that most room correction systems, including Dirac Live, take relatively few measurements, and make a lot of inferences and assumptions as a result.
Dirac Live is "figuring out" both the speaker and room response from just one set of combined measurements.
(One of the benefits of room correction systems like Dirac Live, in the eyes of many users, is that they allow you to avoid spending hours taking hundreds of measurements.)
IN THEORY, having more actual measured data, like the near field response of the speakers, COULD enable such a system to do a better job.
(Of course, whether that will turn out to be the case in practice is going to depend on a lot of other factors, including how effectively that extra information is used.)

I would agree, however, that this sort of detailed analysis of exactly how a room correction system does what it does is mostly "of only academic interest".
You DON'T listen to a compelling explanation of theory; you listen to the results; so there is limited value in "trying to decide whose theory is better" rather than simply comparing the results.
It's also important to remember that, in "advanced room correction theory", a major factor is in determining what should be corrected, because it is audible, and what can safely be ignored, because it is not audible.
And sometimes that just isn't obvious until a system sees a lot of real world usage and testing. 
So, in other words, it's all good and nice that they have a cool idea, but all that really counts is the end result.
(And a lot of talk about how it could or should work isn't going to provide us with that.)

Feb 25, 2024 9:17:20 GMT -5 marcl said:

lrobertson   lcseminole  ttocs I think it was just you guys commenting on this. Tag others if I forgot somebody.

Magic Beans
...............................
What is the point of using Magic Beans?

Well ... uh ...  I mean ... there is no scientific precedent or logic to using near-field measurements for room correction.  Taking measurements at multiple listening positions and applying some algorithm to optimize response makes sense.  But taking measurements at ONE listening position and combining it somehow with near-field measurements makes no sense.  If a speaker had an inherent flaw that revealed itself in a near-field measurement, you could argue that it could be corrected with EQ, and in fact you could even legitimately correct a dip in the bass response if there was one.  But it really doesn't matter because it's the listening position response that matters.

So ... the only point of using Magic Beans would be if - for whatever personal reason - someone does not LIKE what Dirac does, or does not LIKE what Audyssey does ... then this would be a way to play with an alternate app until they get what they want. 

p.s.  In at least one video Joe talks about impulse response and time alignment and trying to get similar results to Dirac or whatever using REW.  Note the following:

1. Time alignment and impulse response correction are two different things.  Dirac does both.  There is no way to do impulse response correction without some sophisticated algorithm and filter implementation similar to Dirac.  Maybe Trinnov does it, but I know of no other way.
2. Yes, you can determine time alignment using the impulse response measurements in REW.  These can be converted to distance and used to manually enter distances to a processor.  But if you use Dirac, you can't enter distances manually with Emotiva processors.
3. In one video Channa talks about how, after doing a Dirac calibration, he used the time alignment files in Spatial Audio Toolkit and (amazed look on his face) Dirac had aligned everything so perfectly that he could not hear the slightest discrepancy in the alignment.

p.p.s. Further evidence that the vast majority of people - even those making YouTube videos - ignore the impulse response correction that Dirac does.

[marcl]

Mar 19, 2024 9:34:28 GMT -5 KeithL said:

I just noticed one thing you said here that is worth mentioning... at least a little bit.
And that is that I believe that there IS a valid - if "indirect" - precedent for using both "full room" and "near field" measurements.
(Note that I'm not saying that Magic Beans is using the information this way.)

When calculating "the perfect room and speaker correction"... you would ideally have a LOT of information.
That information would include things like the dimensions and shape of the room...
As well as the acoustic properties of each individual surface in that room...
As well as the exact properties of each speaker... including directional information like its radiation pattern at each frequency...
Therefore, at least in theory, having more and different information is always going to be a potential advantage.
(And, in this case, I can definitely see how having "a near-field quasi-anechoic profile for each speaker" would be potentially useful information.)

I'm going to throw out a single, but not uncommon, example where this MIGHT be the case.
Let's say you have a simple 5.1 channel system...
And, before any correction, you find that center channel dialog tends to be somewhat muddled and lacking in intelligibility.
And, in perfect hindsight, we now know this is partly because the floor in front of the couch, between the couch and the center speaker, is shiny bare hardwood.

In that situation we might actually be faced with the fact that "electronic room correction" is not going to be able to provide a perfect solution.
However, knowing details, like the vertical dispersion of the center channel speaker, specifically at midrange frequencies, would be helpful in evaluating possible answers.
(The ideal solutions are going to be to add a rug on the floor in front of the couch... or get a new center channel speaker with more limited vertical dispersion at midrange frequencies.)
(And, even beyond that, knowing the acoustic properties of the ceiling, between the center channel speaker and the couch, will help inform us how much adding a rug will probably help the situation.)

This is not a perfect example by any means...
But my point is that most room correction systems, including Dirac Live, take relatively few measurements, and make a lot of inferences and assumptions as a result.
Dirac Live is "figuring out" both the speaker and room response from just one set of combined measurements.
(One of the benefits of room correction systems like Dirac Live, in the eyes of many users, is that they allow you to avoid spending hours taking hundreds of measurements.)
IN THEORY, having more actual measured data, like the near field response of the speakers, COULD enable such a system to do a better job.
(Of course, whether that will turn out to be the case in practice is going to depend on a lot of other factors, including how effectively that extra information is used.)

I would agree, however, that this sort of detailed analysis of exactly how a room correction system does what it does is mostly "of only academic interest".
You DON'T listen to a compelling explanation of theory; you listen to the results; so there is limited value in "trying to decide whose theory is better" rather than simply comparing the results.
It's also important to remember that, in "advanced room correction theory", a major factor is in determining what should be corrected, because it is audible, and what can safely be ignored, because it is not audible.
And sometimes that just isn't obvious until a system sees a lot of real world usage and testing. 
So, in other words, it's all good and nice that they have a cool idea, but all that really counts is the end result.
(And a lot of talk about how it could or should work isn't going to provide us with that.)

Feb 25, 2024 9:17:20 GMT -5 marcl said:

lrobertson   lcseminole  ttocs I think it was just you guys commenting on this. Tag others if I forgot somebody.

Magic Beans
...............................
What is the point of using Magic Beans?

Well ... uh ...  I mean ... there is no scientific precedent or logic to using near-field measurements for room correction.  Taking measurements at multiple listening positions and applying some algorithm to optimize response makes sense.  But taking measurements at ONE listening position and combining it somehow with near-field measurements makes no sense.  If a speaker had an inherent flaw that revealed itself in a near-field measurement, you could argue that it could be corrected with EQ, and in fact you could even legitimately correct a dip in the bass response if there was one.  But it really doesn't matter because it's the listening position response that matters.

So ... the only point of using Magic Beans would be if - for whatever personal reason - someone does not LIKE what Dirac does, or does not LIKE what Audyssey does ... then this would be a way to play with an alternate app until they get what they want. 

p.s.  In at least one video Joe talks about impulse response and time alignment and trying to get similar results to Dirac or whatever using REW.  Note the following:

1. Time alignment and impulse response correction are two different things.  Dirac does both.  There is no way to do impulse response correction without some sophisticated algorithm and filter implementation similar to Dirac.  Maybe Trinnov does it, but I know of no other way.
2. Yes, you can determine time alignment using the impulse response measurements in REW.  These can be converted to distance and used to manually enter distances to a processor.  But if you use Dirac, you can't enter distances manually with Emotiva processors.
3. In one video Channa talks about how, after doing a Dirac calibration, he used the time alignment files in Spatial Audio Toolkit and (amazed look on his face) Dirac had aligned everything so perfectly that he could not hear the slightest discrepancy in the alignment.

p.p.s. Further evidence that the vast majority of people - even those making YouTube videos - ignore the impulse response correction that Dirac does.

There are three aspects to this: what can be measured, how can the measurements be analyzed ... and what (if anything) can you actually DO about any of the anomalies that are found?  Fair to say you can always measure more than you could ever actually fix, in practical terms.  Dirac recommends a minimum of 5 and up to 17 measurements but all around the listening area; Steinway-Lingdorf RoomPerfect uses measurements from all over the room; Trinnov has their special "3D" mic.  But in the end, what can be done with the results?  It comes down to simple IIR filters that most systems use, or a combination of FIR and IIR ... impulse and phase correction ... and the most sophisticated so far, ARC.

I'm going to stick my neck out and guess that the Bean people are not throwing as much academic chops at the analysis as the other big players.  Just sayin'.

Now, having ACTUAL anechoic nearfield data for each speaker - i.e. not waving a mic close to a speaker and deciding to CALL it nearfield - would allow the system to possibly correct intrinsic resonance or phase issues with the speaker itself independent of room issues.  Then with that result, measuring the room would reveal only room-specific issues, some of which could be corrected.  But I'm talking about real nearfield in an anechoic chamber or with a Klippel.

Dirac uses the multiple measurements to determine what can and can't be corrected so they do no harm.  Maybe the others do that, maybe not ... I think Audyssey would be prone to over-correcting based on some reviews and measurements I've seen.

I sure would like to see actual polar plots of all the speakers I have!

[marcl]

Todd at AVNirvana has a discussion with Mathias Johannson from Dirac about the new RP22 standard, Bass and Multi-Subs.

Pay particular attention at 22:40 where he says with DLBC they do not to amplitude correction, but rather they do phase correction with All-Pass filters .... and 45:00 when he talks about the differences between DLBC and ART.



If you want to get a copy of the CEDIA RP22 Immersive Audio Design Recommended Practice, start here ...

cedia.org/en-us/advocacy/standards-best-practices/immersive-audio-design-excellence/#:~:text=The%20new%20recommended%20practice%20for,middle%2C%20high%22%20cost%20models.

[geebo]

Feb 8, 2024 7:36:13 GMT -5 marcl said:

As it happened I was going to do a Dirac calibration yesterday because I changed my center channel configuration. I was wondering if proximity to the couch and reflections from the leather were causing some funny stuff in the center impulse and frequency response, so I was going to try moving the MLP measurement forward a few inches and pile up a blanket on the headrest. Then I saw this posted on the AVS Dirac Forum:

www.youtube.com/watch?v=QFbO_x77zjQ&t=1s

You don't really have to watch it, I'll explain … the guy goes on for 33 minutes about how he spent a week trying dozens of measurement patterns.  His premise is that in reality nobody in the house cares about imaging and soundstage but him, so he set out to optimize it for his chair.  He has what looks to be a 5.1.4 system with the theater seats against the back wall so surrounds are mounted to the wall either side and a bit above. Atmos rear top position is compromised a bit.

Here's a graphic I made to illustrate. All measurements on the same plane ... just a sixpack with the MLP rear center 8” from the couch (about where your nose would be) and everything 8" apart all on a flat plane at ear level.. 
View Attachment

The measured results were excellent with much better impulse responses and flatter frequency response. Audible improvement was immediately evident, and in some ways astounding! The first transient I heard - metal on metal striking a triangle - was sharper and clearer than I had ever heard it. The 2-Channel image on all my benchmark tracks dropped very solidly to dead center in front of me, and a bit more forward ... less diffuse. All in a good way. And then Atmos imaging ... spectacular, precise ... and I confirmed with Dark Side of the Moon ... the lunatic (INDEED) is in my head!

Measurements.   Dirac filters were done with flat target curves and curtains to the edge of the measured response for each speaker.  I used the Spears & Munsil newest test disc to set levels after the filters were loaded.  I think it's the most accurate and consistent method.  I used Spatial Audio Toolkit files to do the 7.1.4 sweeps with REW measuring results.  Frequency response is with Psychoacoustic smoothing.  I've also included here C80 Music Clarity.  Clarity measurements above 20db are considered good.
View AttachmentView AttachmentView Attachment


p.s.   When setting up Volume Calibration in Dirac I set the mic level to 96% which results in a Master Gain of -22db in order to get sub level at -18db and all other speakers at -24db.  The noise floor in my room at the time of measurement was 44dbC.  These settings resulted in the volume of the measurement sweeps to peak at 88db without clipping.   Setting the mic level a bit lower than 100% causes a higher Master Gain which in turn causes higher volume of measurement sweeps.  In my system lowering the mic level from 100% to 96% increases the peak measurement volume from 75db to 88db.  better signal to noise ratio for the sweeps.

I just ran Dirac using this mic placement and I'm very pleased with the results. And it's faster and easier, too. Thanks for posting.