MMM approach and a new calibration app (magic beans)

[PHD]

I wonder what you guys think of this calibration approach?

The app is not yet available for purchase but I was told that it should become available in about a week...

 

[Astudillojr]

I am very curious to this app they used, also because they put the microphone in really interesting positions for measuring, I never measured near the speaker.

 

[GXAlan]

@amirm @Floyd Toole

Can you weigh in on this? It’s a mix of science and handwaving “we made new terms up” explanations.

Basically the idea is
1) Measure speaker nearfield*
2) Measure speaker at MLP
3) Using PEQ to make the measurement at MLP match measurement nearfield will remove the room transfer function, so you get room correction without altering the sound signature of the speaker which you presumably like (which is why you picked it).

* obviously how nearfield you are affects driver integration on the measurements…

Except for YPAO, your curves in Audyssey or Dirac are limited in options.
1) Front. make measurement at MLP of front speakers the target curve for every other speaker. This may not remove the room effect affecting your front speaker.

2) Harman curve, flat, etc. Make MLP of front speakers match an arbitrary curve which may overcorrect if you don’t have a speaker with identical on axis and off axis response and you are going above the transition frequency.

I know YPAO RSC has a mode which corrects some of the room sound while providing no EQ to the speakers.

3) What “Magic Beans” (the app) does is help automate the step of measuring nearfield and measuring at MLP and then figuring out the difference.

With that difference, the app helps you create a target curve such that once you plug it into Audyssey MultEQ-X or Dirac, what ends up happening with the IIR/FIR filters is the equivalent of the PEQ that corrects the response of your MLP to the nearfield measurement.

In my mind, this has the limitation of
A) nearfield measurements are highly sensitive to mic placement (too high vs too low)
B) is dependent on you liking the on-axis response of your speaker, which may or may not be true.

In my experience, some EQ systems like Sony and Yamaha sort of make estimates on what to correct versus what to ignore. Audyssey and Dirac’s estimated target response may not reflect what is happening and they are also making judgments on what to actually apply.

 

[Floyd Toole]

Sorry for the seriously delayed response, I have been engaged in moving from California back to my "home' Ottawa, Canada - winter and all.

The following is a much simplified response. I will soon be working on a 4th edition of my book and this will be covered in great detail in it.

The basic flaw is that a steady-state room curve - which is what is measured at the listening position - is not a “target”, it is a “result”. We know that the shape of such curves is dominated by early reflected sound - off-axis radiation. The only reliable way to ensure neutral communication of sound is to start by understanding the loudspeaker - which is NOT what is revealed in a close-up moving mic measurement. Such a measurement can reveal an approximation to the direct sound/on-axis response, but the off-axis performance remains a mystery, and that is what is mainly responsible for the shape of the steady-state room curve at frequencies above the transition/Schroeder frequency. A room curve is well predicted by the "early-reflections" component of s spinorama ; BUT only above the transition frequency because low frequency performance is dominated by small-room resonances. These must be addressed and, fortunately, at low frequencies steady-state room curves have meaning. With bass accounting for about 1/3 of our overall impressions of sound quality it is clear that one must deal with the upper and lower frequencies differently. If a loudspeaker is well designed, i.e. free from audible resonances, spectrally flat on axis, and smooth off axis, the only adjustments that should be necessary at middle and high frequencies are broadband “tone control” spectral balance tweaking to address variations in program material. Low frequency room mode problems have to be addressed as a separate problem (see Todd Welti papers or my book), and once solved, again only “tone control” adjustments will be necessary for program variations. No fixed “calibration” can be perfect for all program material.

People keep looking for money-making ways to sell “calibrations” and most of them are lacking in some way. This is another one. It has a chance of making a truly bad loudspeaker sound better, but, in my opinion, it has an equal chance of degrading a truly good one. And so it goes . . . Pick the right demo material and the customer will be thrilled.

 

[sweetchaos]

I will soon be working on a 4th edition of my book and this will be covered in great detail in it.

This is exciting!!

Please let us know how to be notified of it when there’s a pre-order or release.

 

[joentell]

Basically the idea is
1) Measure speaker nearfield*
2) Measure speaker at MLP
3) Using PEQ to make the measurement at MLP match measurement nearfield will remove the room transfer function, so you get room correction without altering the sound signature of the speaker which you presumably like (which is why you picked it).

Point #3 is a misrepresentation of the method. It is incorrect that we attempt to remove the room response to attempt to retain the speaker’s sound signature. For the most part, the room will do what it’s going to do and not much can be done about it. But at the same time, that fact is why we can use the room to guide us as to how we can make adjustments to the near-field response. The end result of our method should be called “speaker correction” more so than “room correction.”

@Floyd Toole I texted you the other day so we can set time aside some time so I can share with you some of my findings about near-field MMM and how at various distances, it compares with the listening window response, and the early reflections measurements taken with the Klippel NFS. I know you’ve been busy moving, so we’ll try to schedule when you’ve settled in.

For others, I am a student of Dr. Toole’s work, so I would not disagree with his comments about the importance of the steady state response for bass specifically, that EQ should be different above and below the transition region, that a well-designed speaker will not need much adjustment in the midrange and treble region. Those are some of the principles this calibration method is designed to take into consideration.

He’s also correct that we are hoping to eventually make money from this app. This app has been in development for over a year with over 200 iterations and $0 earned. So, yeah, if the method is correct, I believe we deserve to make some money. We’ve earned the right.

I will explain more in a future video. I appreciate ASR, but this is not my platform of choice to discuss or promote my product. I will do my best to respond to questions, but out of respect for @amirm, I prefer not to promote anything here and perhaps just responding to questions about my app can be seen as a form of promotion. If you have questions, you can comment on my videos, and we will have a public Discord specific to the app. If you decide to post here or another forum, then I may feel the need to clarify if there’s a misunderstanding as to how the app works. I think it’s my responsibility to defend my product. And if there’s something I have wrong, I’m open to listening to that also. This app is still under development. I’m all ears at this point.

 

[GXAlan]

Point #3 is a misrepresentation of the method. It is incorrect that we attempt to remove the room response to attempt to retain the speaker’s sound signature. For the most part, the room will do what it’s going to do and not much can be done about it. But at the same time, that fact is why we can use the room to guide us as to how we can make adjustments to the near-field response. The end result of our method should be called “speaker correction” more so than “room correction.”

Thanks, is it incorrect that I said you are attempting to retain the sound signature and it may be a mathematical side effect of what you are doing but not an intentional decision? Does your software use the nearfield but have some sort of empirically derived “weighting” so it’s not a one to one computation?

One really cool thing to try, and I unfortunately have sold my 4319 or else I would have lent it to you, is to take one of the current JBL speakers like the L100 or L83 or L52 which have L-pads to adjust the voltage seen by the different drivers.

This wouldn’t change any directivity or room effects, but would let you see how the software changes the end result of the target curve.

a well-designed speaker will not need much adjustment in the midrange and treble region. Those are some of the principles this calibration method is designed to take into consideration.

How does the software work with bad speakers? Maybe Dr. Toole’s statement is correct. That it is dependent on good speakers to start with, but that’s still an awesome tool if someone with nice Revel speakers or similarly high performance preference score speakers can utilize Magic Beans to get the most of those speakers in various residential environments and it lets you correct slightly above the transition frequency or figure out where the transition frequency should be for your room, etc.

He’s also correct that we are hoping to eventually make money from this app. This app has been in development for over a year with over 200 iterations and $0 earned. So, yeah, if the method is correct, I believe we deserve to make some money. We’ve earned the right.

+1000. I also think you are not marketing this when you are trying to explain the technical part and you aren’t hiding the fact that you are the designer of the app. If this works even in some rooms well, you deserved to profit from taking on all of the initial risks. If you have filed patents, those aren’t free either.

I will do my best to respond to questions, but out of respect for @amirm, I prefer not to promote anything here and perhaps just responding to questions about my app can be seen as a form of promotion. If you have questions, you can comment on my videos, and we will have a public Discord specific to the app. If you decide to post here or another forum, then I may feel the need to clarify if there’s a misunderstanding as to how the app works. I think it’s my responsibility to defend my product. And if there’s something I have wrong, I’m open to listening to that also. This app is still under development. I’m all ears at this point.

I think it’s nice to have a forum like here where you can talk and discuss but maybe all you have to do is get the tag “Industry” for your account and then it’s OK? I defer the Amir and the mods.

I think it’s one thing to try to use ASR to just get YouTube monetization and another thing when you are selling a product to be willing to share your thought process and rationale with interested users.

I wrote:

“In my mind, this has the limitation of
A) nearfield measurements are highly sensitive to mic placement (too high vs too low)
B) is dependent on you liking the on-axis response of your speaker, which may or may not be true.”

Maybe A, is figured out with your instructions and how the app works.

Maybe B, isn’t as dependent on the on axis response since you are moving the nearfield a lot, but you would say that you cannot make chicken soup out of you know what, and the higher the preference score (smooth on axis and off axis) the better this tool works too and if you had a completely broken speaker where you need to aggressively EQ it above the transition frequency, this tool isn’t as effective.

I guess the name Magic Beans doesn’t work well at ASR even though the idea of using more points of data to make decisions sounds reasonable to me, and even the decision of where to measure and how to do the moving mic takes a lot of know how.

 

[joentell]

The name Magic Beans is derived from a joke we made that Channa didn't really care how it worked, but just knew it worked for him. I ran with the silly name. Hopefully, the results will speak for themselves. We know there are many in this industry that make magical claims and don't deliver. My aim is to make realistic claims and provide realistic measurable improvements while keeping the lighthearted name.

We don't apply weighting in the traditional sense, but instead allow more correction in the lower frequencies than the higher ones as we don't want to send too much power to the tweeter. We also apply different smoothing above and below the transition region since a user can use multiple subs to target a more idealized curve. There has been a lot of trial and error testing with a handful of people doing private beta testing with me. That's helpful because we get to test with different speakers in different rooms. I've hand-picked people with a decent understanding of DSP and psychoacoustics. I'm slowly expanding the private beta testing with other "experts" to see where they think there's room for improvement.

To simplify the goal of the method, we use the MLP and NF measurements to try to get the NF response flat taking the unavoidable room interactions into account.

I've held an event where someone who was working at Harman brought his JBL 4305's and 4329's. Because those speakers measure so well already, the method determined there wasn't much to "fix" in the mid and upper frequencies. The adjustments only needed to be made to account for the bass interaction with the room.

Anyone can try this method on their own using REW. The purpose of the app is to simplify the process.

Yes, the mic placement during the NF MMM measurement is crucial and will vary based on the driver layout and baffle width. We are still in the process of perfecting the recommended technique for end-users. As you can imagine a measurement of a coaxial speaker, two-way bookshelf, MTM, floor stander, floor to ceiling line array, electrostat, and in-walls/in-ceiling speaker will require varied approaches or an overall approach that will derive the proper response. This is very difficult and just requires a lot of testing and comparison with the anechoic response taken with a Klippel NFS for example.

 

[JPA]

These guys don't understand what a target curve is, which makes me question how much they really understand the fundamentals of room responses and calibrations.

 

[joentell]

These guys don't understand what a target curve is, which makes me question how much they really understand the fundamentals of room responses and calibrations.

Please explain in more detail.

If I were to say my ideal target would be whatever the response would be in-room of an ideal speaker that measures flat anechoically with a smooth/linear off-axis response but accounting for the affect of the bass in-room, is that not concrete enough? I don't believe there is a static fixed MLP target that applies to all speakers in all rooms. The target is to make the speaker more ideal taking into account physical limitations. How that speaker interacts with the room is room, speaker, speaker placement, and MLP dependent.

 

[GXAlan]

The name Magic Beans is derived from a joke we made that Channa didn't really care how it worked, but just knew it worked for him. I ran with the silly name.

I don’t mind the silly name, I just think it was tough for this crowd when some of the technical details are missing.

We don't apply weighting in the traditional sense, but instead allow more correction in the lower frequencies than the higher ones as we don't want to send too much power to the tweeter. We also apply different smoothing above and below the transition region since a user can use multiple subs to target a more idealized curve.

That is key and not something that I appreciated from the videos. This makes a lot more sense besides the risk to speakers but because comb filtering and measurement issues are a bigger issue at higher frequencies which is the whole reason to EQ to anechoic data not just in room data.

If you think about making cookies, it’s butter, flour, sugar, and eggs but people have different recipes. Even if your weighting is empirically determined, that makes Magic Beans a lot more interesting now.

This is very difficult and just requires a lot of testing and comparison with the anechoic response taken with a Klippel NFS for example.

This is also helpful to know since it addresses one of the concerns or areas of skepticism. “We have gotten this to work with these speakers so far….”. That also makes sense that there are differences in nearfield measurements and that Magic Beans accounts for this through empiric testing and trial and error. So there are “compatible speakers” that have been well established to have good correlation.

 

[joentell]

I don’t mind the silly name, I just think it was tough for this crowd when some of the technical details are missing.



That is key and not something that I appreciated from the videos. This makes a lot more sense besides the risk to speakers but because comb filtering and measurement issues are a bigger issue at higher frequencies which is the whole reason to EQ to anechoic data not just in room data.

If you think about making cookies, it’s butter, flour, sugar, and eggs but people have different recipes. Even if your weighting is empirically determined, that makes Magic Beans a lot more interesting now.




This is also helpful to know since it addresses one of the concerns or areas of skepticism. “We have gotten this to work with these speakers so far….”. That also makes sense that there are differences in nearfield measurements and that Magic Beans accounts for this through empiric testing and trial and error. So there are “compatible speakers” that have been well established to have good correlation.

Thanks for taking the time to ask questions. It helps me also.

There's a reason this has been in development for over a year and is still not available to the public. I don't want to release something that isn't correct from a theoretical or practical standpoint. We are still making adjustments as we speak. They are minor, but without distortion and max SPL data on each speaker, we want to err on the side of under correction rather than overcorrection. We know what the ideal response is, but whether the speaker can actually do it is another story. It's laughable, but I've even gone so far as to test the method on the built-in speakers on an Android tablet to see what happens at the extremes.

 

[PHD]

Regardless of what exactly the Magic Beans app does. While @joentell assured me on the YouTube comments that there shouldn't be any issues with the moving mic method, I'm still unconvinced about this approach. Couldn't the fast movement/shaking of the mic introduce measurement noise and errors, due to air friction, vibrations, or, a Doppler effect? MMM might be a good idea for subwoofer calibration, but high frequencies are very directional.

Why not take a few static near-filed measurements and have the app average them? Could you use the MMM with conventional MLP-based calibrations such as Auddyssey or Dirac? I'm tempted to try holding the mic and moving it in circles during Auddyssey measurements and see if I get better results.

 

[joentell]

Regardless of what exactly the Magic Beans app does. While @joentell assured me on the YouTube comments that there shouldn't be any issues with the moving mic method, I'm still unconvinced about this approach. Couldn't the fast movement/shaking of the mic introduce measurement noise and errors, due to air friction, vibrations, or, a Doppler effect? MMM might be a good idea for subwoofer calibration, but high frequencies are very directional.

Why not take a few static near-filed measurements and have the app average them? Could you use the MMM with conventional MLP-based calibrations such as Auddyssey or Dirac? I'm tempted to try holding the mic and moving it in circles during Auddyssey measurements and see if I get better results.

MMM won't work with Audyssey's auto-calibration since they use fast sweeps. MMM requires pink noise. We specifically use 16K periodic pink noise.

 

[GXAlan]

I don't believe there is a static fixed MLP target that applies to all speakers in all rooms. The target is to make the speaker more ideal taking into account physical limitations. How that speaker interacts with the room is room, speaker, speaker placement, and MLP dependent.

I know ASR is not your preferred discussion format, but I would make this conjecture that goes from hand waving to makes sense.

The weakness of Dirac/Audyssey is that they measure in mono. Distance from speaker is known but angle is not.

Take the same LCR speakers in an anechoic room. They are perfect 360 degree radiators where the response is identical on and off axis.

If you measure at the MLP, you get perfect on axis measurement so you don’t need Magic Beans. Irregularities in FR are real and you can EQ them to your preferred tilt.

Now imagine LCR speakers that are not 360 degree radiators. You have anechoic on-axis data. In one scenario, those LCR speakers are parallel and pointing straight with no toe in. In another scenario the LCR speakers are setup so that the L/R are toed in toward the MLP. How do they measure differently?

In the first scenario, a speaker whose dispersion characteristics were such that you lost x dB every y degrees from 20-20kHz, would mean that you could EQ full range since the difference in volume from speaker rotation represents the entirety of the difference.

But if we look at even the JBL 708P you are still shifting 3-4 dB in that +/- 20 degree range.

Option 1 is to just correct to a few hundred Hz and ignore the rest. Problem is that you lose the ability to EQ above the transition frequency.

Option 2 is to measure the axis to the MLP in anechoic, apply the PEQ at that point. In an anechoic room, that works.

Now we go in room. What you want to do is identify the room effects and identify the sound from the axis at the listening point.

With a standard Dirac/Audyssey, they take averages at multiple spots hoping that the weighting is such that it each measurement is equal and you can take more measurements near a seat you want.

Since you say MMM at different distances correlates with the NFS at different distances, are you saying that Magic Beans takes the nearfield at a few positions and the MLP and then picks a curve to send to Dirac such that things instead of pulling everything down or up to the target curve, it does “50%” correction (or some amount other than maximum) to account for the imperfect in room measurements?

A lot of nuggets of details that you have added in these posts such as MMM at different distances compared to NFS data and “weighting” the correction differently in the bass versus treble and picking how much of a correction to do along the way and what to ignore make Magic Beans seem a lot more scientific, but I am still trying to wrap my head around what you are doing exactly that makes it better than just going with transition frequency only bass room target (if you have an awesome speaker like a Genelec or Meyer Sound) or biting the bullet and doing full 20-20 correction at the MLP (if you have something like a Bose 901
which is inherently dependent on full range EQ even in anechoic conditions). The Bose 901 is a unique case since it does get closer to that consistent FR at larger areas since you are getting a mix of all those reflections smeared in time anyway. Its almost like a moving speaker as opposed to moving mic when you are measuring at a few different positions.

 

[amirm]

Folks, Joe is one of the "good guys." Please assume good intentions as you respond to him.

 

[joentell]

Folks, Joe is one of the "good guys." Please assume good intentions as you respond to him.

Thank you @amirm
I just want to respect this place by not promoting my app here. I understand that by answering questions, it can be seen as such, but it is not my intention.

 

[LIΟN]

so I can share with you some of my findings about near-field MMM and how at various distances, it compares with the listening window response, and the early reflections measurements taken with the Klippel NFS.

Can you just show some little data in the thread as a preview?
Your approach looks very interesting, I'm rooting for you!!

 

[joentell]

Can you just show some little data in the thread as a preview?
Your approach looks very interesting, I'm rooting for you!!

I can make a video about it eventually. Here's a measurement I just found that I took using MMM overlaid over a Klippel NFS one. This is prior to me realizing that UMIK-1 is slightly out of calibration (treble is too hot) from being dropped one too many times. I should be getting a Cross Spectrum Labs UMIK-1 soon.

 

[LIΟN]

I can make a video about it eventually. Here's a measurement I just found that I took using MMM overlaid over a Klippel NFS one. This is prior to me realizing that UMIK-1 is slightly out of calibration (treble is too hot) from being dropped one too many times. I should be getting a Cross Spectrum Labs UMIK-1 soon.

Really cool!! Is it a method of applying a kind of weight (or correction value?) to the measurement of mmm?

I should be getting a Cross Spectrum Labs UMIK-1 soon.

By calibration, and by the characteristics of mmmm, there may not be much difference, but the outcome of microphones such as Earthworks' m23/m30 may also be curious.!