Genelec GLM Review (Room EQ & Setup)

[dominikz]

??? If you put the mic on a tripod, how are you getting all that variability?

This was of course written assuming the microphone was not in the exact same spot between two single-position sweep measurement takes. If you never move the mic, the single-position sweep measurement should of course match well between subsequent takes.
If you can't replicate the exact position between takes, I find MMM results in better repeatability.
Sorry if the content of my post was offending in some way?

 

[ernestcarl]

The context is not what it seems. Correction here is only in bass. There is no need for spatial averaging if that is all you are doing. Yes, you have to worry about non-minimum-phase nulls but you can do that without any spatial measurements.

If you want to do correction in higher frequencies, then yes, you have to have a strategy for smoothing as slight changes in microphone location can make a big impact on measured results. Comb filtering for example is a big problem in higher frequencies which smoothing helps to resolve.

In bass frequencies, modal peaks can have resolution of less than one Hertz. Smoothing can produce incorrect results for filter programming for example.

As I noted, the problem with multipoint measurement is repeatability. You absolutely want to listen to experiment here. To the extent you can't repeat a multi-point measurement because you can't identically move a microphone to different spot in 3-D space, then I rather use single point measurements.

Now, if you have multiple seating locations, then that is an entirely different game and you do need to measure more than one seat to build a model of what to correct. Even here, spreading the measurements across multiple seats equally can reduce the preference in the outcome. From Sean and crew's own AES paper:

The Subjective and Objective Evaluation of Room Correction Products
Sean E. Olive1, John Jackson2, Allan Devantier3, David Hunt4 and Sean M. Hess5


View attachment 152409

Any kind of averaging gives you suboptimal results for the corner cases. Here, if you all you care about is a single location, then averaging just reduces the resolution there, not increase it.

Bottom line: keep bass optimization separate from rest of the frequency response. They are completely different animals as Sean and Dr. Toole would tell you.

Looking at your before and after optimization graph again, the correction does stop right around ~300Hz... so I guess in this limited case something like MMM for extra verification is not exactly needed.

 

[mastachio]

@amirm I know you bought and reviewed the LX504 I own the LX701. What did you think about MCACC?

 

[richard12511]

This seems not correct, but I may be missing something.

GLM could always integrate multiple subs, including time of flight and level to main listening position.
Even in a mixed system where some monitors have their own dedicated sub or adaptive W371A SAM™ Woofer.

For multiple subs outputting the same signal, just calibrate as usual and edit the 'Acoustic settings' of each subwoofer to reduce the ‘Level Compensation’ value. Section 10.4 in the current GLM4 System Operating Manual even has a table if you don't know by how many db.

A top tier multi sub DSP solution needs to consider the combined bass response as part of overall equation. Audiolense, MSO, DLBC, etc. do this; GLM does not. It’s been confirmed by Genelec, but it’s also something they’re currently working on for a future update, so it should be here eventually.

Not saying you can’t get good bass with a multi sub GLM solution . I’m sure it can sound pretty good with the individual calibration that it does. It just won’t be as perfect as what you’d get with multiple Rythmiks and DLBC.

I used 7370s with GLM for a few weeks and it pretty good. One thing I did to get around GLMs multisub limitation is to go in afterwards and apply a 20-100Hz proper full system convolution file via Roon(Jriver supports it to). I just measured the combined post GLM response and used that to gen the file. It was a really nice improvement to my ears. Definitely worth trying if your using GLM for multisub.

 

[Dimifoot]

@amirm We are all here to learn

As I have shown in multiple posts on ASR, it is not just at one listening point either

You could try the Moving Microphone Method when evaluating room eq software

From my experience, MMM can also provide better results (to some degree) precisely because it filters the response.

Users who have experienced room correction using products like Dirac or Acourate (almost a decade old now) know of their benefits of measuring over multiple points vs the single mic point correction

To be honest, I am a little disappointed with the less than scientific approach of room EQ. Repeatability is a cornerstone of the scientific method and It is well known that single spot microphone measurements are not reliable. Even though 9 spot measurements are not that repeatable, they are far superior to single spot measurements.
The suggestion for a before and after Moving Microphone Measurement seems to me a valid one. It has been show to be very repeatable for a FR. This article bij Jean-Luc Ohl has scientific data for the MMM approach.

I believe most of the folks who are into room EQ already know that MMM and averaging of multiple measurements done from a single points around LP provides practically the same spatially averaged response curve hence I used the term "spatial averaging" instead of MMM as it covers both techniques.

You have to use multiple point measurements and averaging, or (even better and easier and quicker) the moving microphone method (but you have to practice for that).

And "officially":

A single measurement? No multiple microphone measurements with spatial averaging?That doesn't seem like a good idea.

There is no shame in learning

 

[amirm]

@amirm I know you bought and reviewed the LX504 I own the LX701. What did you think about MCACC?

I never tested it formally. Informally I tested my old Pioneer and found the correction very primitive and slight.

 

[amirm]

There is no shame in learning

There isn't. So I hope everyone by now understands that room EQ can operate in two domains so you need to understand the context of discussion:

1. Bass frequencies below transition.

2. Frequencies above.

Strategies for these domains is radically different. Even the implementation is different often as evidenced by phrases like "mixed phase" in EQ lingo.

#2 provides many challenges as wavelengths get small enough relative to our body dimensions where no microphone measurement represents how we hear. As such, all such corrections must be, let me repeat, must be confirmed with listening tests. Lots of different schemes have been developed to deal with this but ultimately you are better off getting an excellent speaker and just overlaying a target curve.

For #1, as I have explained, wavelengths are long and your correction as such can be made without worrying about psychoacoustics. Here, filter resolution becomes very important and hence my caution against smoothing over the response.

I am pretty sure Sean's comment assumed full-range correction and for multiple seats. That is not GLM's focus or capability so yes, it lacks that. His comment does not mean that you must have mandatory features like this for just bass correction.

Most of the time just 2 to 3 filters make a massive difference in bass response and you can just call it done then. GLM automates that which is nice but you can do it manually. And again, I suggest not averaging (MMM or otherwise) so that you can determine the peak frequency more accurately.

Note that there are much more advanced approaches that come into play when you have multiple subs and multiple speakers that you want to optimize for many seats. This can be a nightmare by had and few systems do as good a job as JBL Arcos. See my write ups:

https://www.audiosciencereview.com/forum/index.php?threads/subwoofer-low-frequency-optimization.15/

https://www.audiosciencereview.com/...s/computer-optimization-of-room-acoustics.12/

So yes, as Sean says, GLM is not a sophisticated system.

At some point I will do a tutorial for all of this but for now, please don't blindly follow the crowd with terms like "MMM" and such.

 

[Dimifoot]

There isn't. So I hope everyone by now understands that room EQ can operate in two domains so you need to understand the context of discussion:

1. Bass frequencies below transition.

2. Frequencies above.

Strategies for these domains is radically different. Even the implementation is different often as evidenced by phrases like "mixed phase" in EQ lingo.

#2 provides many challenges as wavelengths get small enough relative to our body dimensions where no microphone measurement represents how we hear. As such, all such corrections must be, let me repeat, must be confirmed with listening tests. Lots of different schemes have been developed to deal with this but ultimately you are better off getting an excellent speaker and just overlaying a target curve.

For #1, as I have explained, wavelengths are long and your correction as such can be made without worrying about psychoacoustics. Here, filter resolution becomes very important and hence my caution against smoothing over the response.

I am pretty sure Sean's comment assumed full-range correction and for multiple seats. That is not GLM's focus or capability so yes, it lacks that. His comment does not mean that you must have mandatory features like this for just bass correction.

Most of the time just 2 to 3 filters make a massive difference in bass response and you can just call it done then. GLM automates that which is nice but you can do it manually. And again, I suggest not averaging (MMM or otherwise) so that you can determine the peak frequency more accurately.

Note that there are much more advanced approaches that come into play when you have multiple subs and multiple speakers that you want to optimize for many seats. This can be a nightmare by had and few systems do as good a job as JBL Arcos. See my write ups:

https://www.audiosciencereview.com/forum/index.php?threads/subwoofer-low-frequency-optimization.15/

https://www.audiosciencereview.com/...s/computer-optimization-of-room-acoustics.12/

So yes, as Sean says, GLM is not a sophisticated system.

At some point I will do a tutorial for all of this but for now, please don't blindly follow the crowd with terms like "MMM" and such.

Don't underestimate ASR members
They might have the knowledge you are referring to, and more.

This is a snapshot moment of a calibration process. One speaker+sub, 9 measurements, 30 cm apart (to the left-right-front-back etc)

As you can see there is variation in the response in the Main Position Listening "box", even below the transition frequency.

In order to properly evaluate results of a Room Eq software we obviously have to take that into account, even if we are addressing only the bass.
Its very easy to get wrong conclusions if you don't have proper repeatability.

 

[thorvat]

There isn't. So I hope everyone by now understands that room EQ can operate in two domains so you need to understand the context of discussion:

1. Bass frequencies below transition.

2. Frequencies above.

Strategies for these domains is radically different. Even the implementation is different often as evidenced by phrases like "mixed phase" in EQ lingo.

#2 provides many challenges as wavelengths get small enough relative to our body dimensions where no microphone measurement represents how we hear. As such, all such corrections must be, let me repeat, must be confirmed with listening tests. Lots of different schemes have been developed to deal with this but ultimately you are better off getting an excellent speaker and just overlaying a target curve.

For #1, as I have explained, wavelengths are long and your correction as such can be made without worrying about psychoacoustics. Here, filter resolution becomes very important and hence my caution against smoothing over the response.

I am pretty sure Sean's comment assumed full-range correction and for multiple seats. That is not GLM's focus or capability so yes, it lacks that. His comment does not mean that you must have mandatory features like this for just bass correction.

Most of the time just 2 to 3 filters make a massive difference in bass response and you can just call it done then. GLM automates that which is nice but you can do it manually. And again, I suggest not averaging (MMM or otherwise) so that you can determine the peak frequency more accurately.

Note that there are much more advanced approaches that come into play when you have multiple subs and multiple speakers that you want to optimize for many seats. This can be a nightmare by had and few systems do as good a job as JBL Arcos. See my write ups:

https://www.audiosciencereview.com/forum/index.php?threads/subwoofer-low-frequency-optimization.15/

https://www.audiosciencereview.com/...s/computer-optimization-of-room-acoustics.12/

So yes, as Sean says, GLM is not a sophisticated system.

At some point I will do a tutorial for all of this but for now, please don't blindly follow the crowd with terms like "MMM" and such.

Most differences between spatially averaged measurement and single point measurement can be seen below transition frequency. Moderate differences can be seen in the transition region (say 400-900Hz) and practically no differences can be seen above app 900Hz.

And that is logical: as you are moving mic around different positions in the room you can expect larger differences between measurements in the regions where room affects sound more.

 

[nai]

For the time being, multi-sub integration solutions, such as Trinnov, antimode x4, glm，dirac automated testing will not be very good, it is best to manually optimize

 

[ernestcarl]

As you can see there is variation in the response in the Main Position Listening "box", even below the transition frequency.

Not wanting to reveal the actual tediousness of my own "process", I actually have gigabytes worth of data for individual channel (up to x50 points per channel) measurements within "a box" of varying sizes and across multiple listening seats. It never ain't pretty, which is why I ultimately prefer some kind of simplified averaging like MMM at the end of the day if I'm just focusing on the frequency magnitude.

*all speakers playing in full-range

Ironically in the above example, I am very fortunate to only have a single relatively "well-behaved" subwoofer. Three or fours subs would require something more sophisticated like a multisub optimizer (MSO). I tried variations of different positions for the sub in the rear of my MLP and also got more differences depending on which side the driver was facing (i.e. left,right,front,back). MMM is quite a godsend in that it cuts hours worth of measurements into a mere few minutes -- let's say in just a minute or two for each channel.

 

[quantum_wave]

At some point I will do a tutorial for all of this but for now, please don't blindly follow the crowd with terms like "MMM" and such.

Oh, yes!, please!. We, the public, would very much benefit from this tutorial. Would be great to get it as a Youtube vid.

 

[thorvat]

I am pretty sure Sean's comment assumed full-range correction and for multiple seats. That is not GLM's focus or capability so yes, it lacks that. His comment does not mean that you must have mandatory features like this for just bass correction.

Let me add a comment to this part as well:

I am, on the other hand, pretty sure Dr. Olive (@Sean Olive) very much understands that room EQ is mostly about correcting response below transition frequency and not about full-range correction.

I am also pretty sure that he advocated for using room EQ based on spatially averaged measurements also for listening to the music in a single seat scenario and not only for multi-seat scenario.

I am hoping he will find time to clarify his post by himself as posts from such experts give all of us the oportunity to directly learn from one the best experts in modern audio technology.

 

[thewas]

Don't underestimate ASR members
They might have the knowledge you are referring to, and more.

This is a snapshot moment of a calibration process. One speaker+sub, 9 measurements, 30 cm apart (to the left-right-front-back etc)

View attachment 152453

As you can see there is variation in the response in the Main Position Listening "box", even below the transition frequency.

In order to properly evaluate results of a Room Eq software we obviously have to take that into account, even if we are addressing only the bass.
Its very easy to get wrong conclusions if you don't have proper repeatability.

Still those differences won't get you further in the case of room correction as you still have to decide on which you base your EQ.
It would be more helpful for the case of EQ to create the average of those and compare that to a single measurement in the middle of that 30cm cube and a MMM one in that volume, am quite positive the differences will be rather small as they were also in my case.

 

[thorvat]

Still those differences won't get you further in the case of room correction as you still have to decide on which you base your EQ.
It would be more helpful for the case of EQ to create the average of those and compare that to a single measurement in the middle of that 30cm cube and a MMM one in that volume, am quite positive the differences will be rather small as they were also in my case.

They would certainly be small if you measure in the 30cm cube, but usually you base your single-seat measurements on the larger cube, say 60-100cm.

 

[Dimifoot]

Still those differences won't get you further in the case of room correction as you still have to decide on which you base your EQ.
It would be more helpful for the case of EQ to create the average of those and compare that to a single measurement in the middle of that 30cm cube and a MMM one in that volume, am quite positive the differences will be rather small as they were also in my case.

I am not referring to calibration strategy (that’s a different topic), but to calibration software evaluation/reviews

 

[onion]

I used 7370s with GLM for a few weeks and it pretty good. One thing I did to get around GLMs multisub limitation is to go in afterwards and apply a 20-100Hz proper full system convolution file via Roon(Jriver supports it to). I just measured the combined post GLM response and used that to gen the file. It was a really nice improvement to my ears. Definitely worth trying if your using GLM for multisub.

Are you able to give an idiot's guide on how to do this? I have three GLM subs in my music room setup and would love the bass management to approach the quality in my movie room (4-subs, miniDSP).

 

[amirm]

This is a snapshot moment of a calibration process. One speaker+sub, 9 measurements, 30 cm apart (to the left-right-front-back etc)

30 cm apart? What does this have to do with the discussion we are having? Once more, if you are optimizing for a single seat, you can make a static measurement, find the offending room modes and program your filters (which is what GLM is doing). Moving the mic to other positions which I presume you then averaged, only produces weaker correction but less extreme for other locations. That is what it does. There is no magic that pops out a more correct answer this way.

I really don't understand this fascination with averaging. Averaging is a filter that reduces resolution to aid a human in understanding a larger picture. Here, we are interested in finding precise values to program an EQ for deep room modes. A single measurement gives us that without the complexity of measuring multiple locations, having your body in front of mic as you swing it around in MMM, etc.

Use MMM if you want to know overall tonality of your room. Or high filtering of the response. But for room modes, it is more work to get less.

 

[thewas]

I am not referring to calibration strategy (that’s a different topic), but to calibration software evaluation/reviews

Why should that be different? As said in the modal region it doesn't change much for a single listener calibration whether you use a single point or multiple ones. I personally also usually use MMM as it just takes a minute and is less sensitive to placement for higher frequencies but acknowledge also the disadvantage that it doesn't have phase information.

 

[thewas]

They would certainly be small if you measure in the 30cm cube, but usually you base your single-seat measurements on the larger cube, say 60-100cm.

Not necessarily and depending on how much you will later move your head, for example for monitoring usually small distances are recommended.