MMM approach and a new calibration app (magic beans)

[Rednaxela]

What I find incredibly puzzling is how a room response transfer function, i.e. a MLP MMM minus a NF MMM, can serve as something to EQ towards. Intuitively, this transfer function feels as being the core problem, while in MB it represents the exact opposite - the core solution. Both below and above Schroeder at that.

Not that I'm disputing it, but it's just very hard to wrap my head around, let alone to theoretically validate.

 

[joentell]

he's good

He's a genius in what he does. I can ask and he makes it happen. But that's not to diminish my contribution. I came up with the method and did all of the real-world testing.

Steve Wozniak couldn't have created Apple by himself.

 

[joentell]

What I find incredibly puzzling is how a room response transfer function, i.e. a MLP MMM minus a NF MMM, can serve as something to EQ towards. Intuitively, this transfer function feels as being the core problem, while in MB it represents the exact opposite - the core solution. Both below and above Schroeder at that.

Not that I'm disputing it, but it's just very hard to wrap my head around, let alone to theoretically validate.

It isn't as intuitive as I wish it were. But if you consider the room's transfer function (smoothed) is the house curve that you will target for the MLP response, you'll find that it flattens the nearfield response above the transition region, and give you the bass response your room is inevitably demanding. If you think about it another way, the main difference between the Main Listening Position response of the speaker and the room transfer function response above the transition region, is the deviation of that speaker's response from flat.

Yet another way to the think of what it does above the transition region is imagine you have a speaker with a perfectly flat on-axis response and smooth directivity. If you look at the MLP response and compare that to the room transfer function (MLP - NF) what you'll find is the MLP measurement above the transition region is exactly the same as the room transfer function, so the corresponding correction would be 0, no correction.

 

[levimax]

It isn't as intuitive as I wish it were. But if you consider the room's transfer function (smoothed) is the house curve that you will target for the MLP response, you'll find that it flattens the nearfield response above the transition region, and give you the bass response your room is inevitably demanding. If you think about it another way, the main difference between the Main Listening Position response of the speaker and the room transfer function response above the transition region, is the deviation of that speaker's response from flat.

Yet another way to the think of what it does above the transition region is imagine you have a speaker with a perfectly flat on-axis response and smooth directivity. If you look at the MLP response and compare that to the room transfer function (MLP - NF) what you'll find is the MLP measurement above the transition region is exactly the same as the room transfer function, so the corresponding correction would be 0, no correction.

I think MMM measurements have been under appreciated and am happy to see you are looking at them closer for practical uses. One question I have is how does the NF MMM vs MLP MMM account for the baffle step? It seems like the MPL has it but the NF would "miss" it? or am I missing something?

 

[joentell]

I think MMM measurements have been under appreciated and am happy to see you are looking at them closer for practical uses. One question I have is how does the NF MMM vs MLP MMM account for the baffle step? It seems like the MPL has it but the NF would "miss" it? or am I missing something?

Using a MMM technique where you move the mic straight back from 0.5 - 1 meter allows you get the effects of the baffle step in most cases. There are speakers that require a measurement further back.

 

[Edthomp]

I am relating to what I think the intent of the method you've codified is for, but I need help in closing the loop.

1) I have a Trinnov Altitude 16 that I currently generate PEQs in REW to the shape of a Harmon-like target curve. Then I manually enter the resultant PEQs in the Trinnov.

2) But as you've pointed out, how do I know the target curve I am using is correct for my room? I don't.

3) If I understand correctly the MB software will produce the PEQs to enter into the Trinnov, but will it also give me the end-state target curve instead that I could enter into the Trinnov and then let the Trinnov optimizer do its acoustic correction to the target curve produced from MB? Meaning, no PEQs are entered into the Trinnov.

4) Given in general that a downward sloping curve seems to be the accepted recommendation for good sound, if we subtract the MLP MMM from the NF MMM - what happens if the speaker isn't voiced for a general downward slope? Does the MB software have a generalized Harmon-target curve that it then massages to produce a Harmon-like target curve for the given room/speaker combination? Or am I misunderstand the MB recommended target curve for this use case?

 

[joentell]

I am relating to what I think the intent of the method you've codified is for, but I need help in closing the loop.

1) I have a Trinnov Altitude 16 that I currently generate PEQs in REW to the shape of a Harmon-like target curve. Then I manually enter the resultant PEQs in the Trinnov.

2) But as you've pointed out, how do I know the target curve I am using is correct for my room? I don't.

3) If I understand correctly the MB software will produce the PEQs to enter into the Trinnov, but will it also give me the end-state target curve instead that I could enter into the Trinnov and then let the Trinnov optimizer do its acoustic correction to the target curve produced from MB? Meaning, no PEQs are entered into the Trinnov.

4) Given in general that a downward sloping curve seems to be the accepted recommendation for good sound, if we subtract the MLP MMM from the NF MMM - what happens if the speaker isn't voiced for a general downward slope? Does the MB software have a generalized Harmon-target curve that it then massages to produce a Harmon-like target curve for the given room/speaker combination? Or am I misunderstand the MB recommended target curve for this use case?

1. You can apply the custom curves into the Trinnov using PEQ's or by drawing in the curve. We have an export option that aligns with the specific frequency values in the Trinnov. (You can use the arrow keys and pg up/ pg down to move faster.) This is the recommended method since the Trinnov is able to make it's corrections and the target curve is placed last.

2. Correct.

3. Correct.

4. There is no pre-determined target curve except that the nearfield response above the transition region, which we determine using the measurements, is set to a flat NF target accounting for directivity mismatches that might not be wise to boost. Below the transition region, we're looking at the average room response of your ear-level speakers to see what the bass rise looks like and we use that as the target. That will almost always have a downward slope due to room gain, the transition region is usually pretty flat, then the high frequencies typically have a downward slope. There are edge cases where this isn't the case. For example, someone using MB to calibrate nearfield studio monitors might find that there's very little high frequency roll-off.

 

[Rednaxela]

@joentell Thank you for taking the time to explain it to yet another person on the internet.

Yet another way to the think of what it does above the transition region is imagine you have a speaker with a perfectly flat on-axis response and smooth directivity. If you look at the MLP response and compare that to the room transfer function (MLP - NF) what you'll find is the MLP measurement above the transition region is exactly the same as the room transfer function, so the corresponding correction would be 0, no correction.

Right. For this part the penny seems to have dropped. The core of the theory is that deviations from this ideal speaker will show up in the MLP MMM, but not in the room transfer function. Thus, the RTF is a stable reference, informed by

1) the room, and
2) the MLP, the MMM of which we seek a correction for.

Because the RTF represents the response of an ideal speaker in a given room at a given MLP, and won’t change as long as the room and MLP stay the same, one can view the RTF as the ideal response of any speaker in a given room at a given MLP. Above Schroeder and for all intents and purposes.

With this out of the way, what you propose is a way to find this RTF.

Most interesting.



Why the RTF is also valid reference below Schroeder is still a bit of a mystery to me. A nice one for me to chew on today.

 

[Newman]

What value does it provide if you already have spinorama data for your speakers?

 

[joentell]

@joentell Thank you for taking the time to explain it to yet another person on the internet.


Right. For this part the penny seems to have dropped. The core of the theory is that deviations from this ideal speaker will show up in the MLP MMM, but not in the room transfer function. Thus, the RTF is a stable reference, informed by

1) the room, and
2) the MLP, the MMM of which we seek a correction for.

Because the RTF represents the response of an ideal speaker in a given room at a given MLP, and won’t change as long as the room and MLP stay the same, one can view the RTF as the ideal response of any speaker in a given room at a given MLP. Above Schroeder and for all intents and purposes.

With this out of the way, what you propose is a way to find this RTF.

Most interesting.



Why the RTF is also valid reference below Schroeder is still a bit of a mystery to me. A nice one for me to chew on today.

You explained it better than I ever have, so thank YOU!

The precise RTF is not ideal below the transition region because it might be a bit jagged. Bass is an area where we can be more idealistic, especially if you're using multiple subs. For this we smooth the target response which allows the RTF to inform us as to the shape and rise. We want to use that as a reference for correction within reason. We don't want to boost 10dB if all the room will do is cancel that out anyway. But at least, we can remove some of the unwanted gain caused by a speaker being near a boundary. The goal isn't to fight the room with DSP below the transition region because that's a losing battle. The best way I can think of it is how your car's shocks should smooth out small bumps, but if you ride over a hill, you're just going to have to accept some of that. Some issues just need to be handled with physical means i.e. moving speakers or moving the MLP. If you're in a large null, the best thing to do is to move your MLP. Looking at the RTF will inform you of these issues though.

 

[joentell]

What value does it provide if you already have spinorama data for your speakers?

I answered this in my recent video on my channel in Part 2 of my target curve video.

 

[Rednaxela]

What value does it provide if you already have spinorama data for your speakers?

As I understand it, a near field listening window response captured as proposed by the method under discussion, also includes potential deviations from the ideal introduced by the speaker's specific placement. The result of a correction towards the MB-generated target will include a correction of these deviations too. I believe this is what @joentell refers to here:

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.

Provided this is of value to you, this is something that might be hard to derive from spinorama data alone.

Hope I got it right and hope it helps.

 

[joentell]

As I understand it, a near field listening window response captured as proposed by the method under discussion, also includes potential deviations from the ideal introduced by the speaker's specific placement. The result of a correction towards the MB-generated target will include a correction of these deviations too. I believe this is what @joentell refers to here:



Provided this is of value to you, this is something that might be hard to derive from spinorama data alone.

Hope I got it right and hope it helps.

This is correct. The easiest example is if a speaker is behind an A.T. screen. That will affect the high frequency response, but anechoic data doesn't take that into account. Anechoic measurements will allow a speaker designer to create the best possible speaker, but I'm of the mindset that DSP correction for speakers in a room require in-room measurements for the best results.

 

[Keith_W]

What value does it provide if you already have spinorama data for your speakers?

The Spinorama tells you if the speaker you are considering is well designed or not, and it will give you some idea how it will behave in your room. But once it's in your room, Spinorama data is less relevant. What is far more important is the actual behaviour of the speaker in your room.

Speaker designers aim for design that measures as perfectly as possible under anechoic conditions or with a Klippel. They do not know how it will behave in your room.

We hobbyists have to find a way to make the speakers we purchased work in our rooms, which means restoring the tonal balance of the design which was altered by the room. This is where this app comes in.

 

[Rednaxela]

[...] which means restoring the tonal balance of the design [...]

I'm not really sure this is either the objective or the result of the MB app.

In principle, the app aims at correcting every deviation from an ideal that is captured in the NF MMM. This includes tonality issues within the speaker itself. As I understand it, these tonality issues are not restored, but fixed. At least an attempt is made to do it.

Perhaps in other words, with my own somewhat less-than-ideal speakers, with the MB app method I expect a MLP response that tries to sound like a corrected NF LW response, not necessarily the speaker's actual NF LW response.

Not sure that's also what you're saying?

 

[Newman]

The Spinorama tells you if the speaker you are considering is well designed or not, and it will give you some idea how it will behave in your room. But once it's in your room, Spinorama data is less relevant. What is far more important is the actual behaviour of the speaker in your room.

Speaker designers aim for design that measures as perfectly as possible under anechoic conditions or with a Klippel. They do not know how it will behave in your room.

We hobbyists have to find a way to make the speakers we purchased work in our rooms, which means restoring the tonal balance of the design which was altered by the room. This is where this app comes in.

Thanks, I get the point you and Rednaxela make, but not sure I agree, especially the 'far more important' claim.

The spinorama give the direct sound information. This Toole says listening tests reveal to be the single most important thing to get right above the transition zone. So how can anything be 'far more important' than that?

Also, bass extension and EQ are quite highly rated for importance, maybe 30% of total preference weighting, and this information is available from spinorama (extension) and normal bass EQ methods.

Speaker placement information could be of some value, so that's interesting. But if it suggests adjustments that negatively affect the direct sound quality, then it is IMHO more likely to be risky than helpful.

@Rednexala a corrected NF response is also available from spinorama, and more accurately.

cheers

 

[Rednaxela]

@Rednexala a corrected NF response is also available from spinorama, and more accurately.

How would this work, for instance in the case of @joentell's A.T. screen example?

The easiest example is if a speaker is behind an A.T. screen. That will affect the high frequency response, but anechoic data doesn't take that into account.

 

[Keith_W]

Thanks, I get the point you and Rednaxela make, but not sure I agree, especially the 'far more important' claim.

The spinorama give the direct sound information. This Toole says listening tests reveal to be the single most important thing to get right above the transition zone. So how can anything be 'far more important' than that?

OK, let's take an example. Suppose you bought a perfect speaker. The designer used an anechoic chamber, placed a mic 1m from the speaker, and tuned it until it is perfectly flat. You take it home and put it in your room. You place your mic 1m away and you see a perfect flat response above the transition zone, below this it's a dog's breakfast. You move your mic 2m away. The frequency response starts to fall. Re-measure at 3m, 4m, etc. You will see that as you get further away, the falling frequency response becomes more pronounced. Thus the tonality of the speaker depends on listening distance. Sit too far away, and you will get bass heavy sound with a pronounced downward treble tilt, regardless of what the predicted in-room response says. You don't have to take my word for it. Get your mic out and go check it for yourself. Why bother with the predicted in-room response when you can easily check your actual room response?

The fact is, once loudspeakers leave the sanitized test conditions where they were designed and into the great unwashed public like you and me, we do all sorts of awful things to speakers. Just look at some of the systems posted on ASR. You will see coffee tables in front of speakers, bookshelf speakers on computer tables, bookshelf speakers in the corner of a kitchen, dipole speakers pushed up against the front wall, and so on. So never mind what the Spinorama says. Spinoramas are artificial, what matters is what you measure.

Toole himself says that "broad tone controls" can be applied to the upper frequencies. Granted, the Magic Beans app does not do that, its aim is to analyse and remove the "room transfer function". But still - we need to realise that it's not only freqs below the transition zone that are affected by choice of speaker placement and MLP, but also the upper frequencies.

 

[joentell]

Granted, the Magic Beans app does not do that, its aim is to analyse and remove the "room transfer function".

I agree with most of what you said, but I want to clarify that we can't really remove the "RTF". All we can do is use that as a guidelines to inform us about how to correct the speaker response. All of the so-called "room correction" software should really be called speaker response correction.

 

[Sokel]

OK, let's take an example. Suppose you bought a perfect speaker. The designer used an anechoic chamber, placed a mic 1m from the speaker, and tuned it until it is perfectly flat. You take it home and put it in your room. You place your mic 1m away and you see a perfect flat response above the transition zone, below this it's a dog's breakfast. You move your mic 2m away. The frequency response starts to fall. Re-measure at 3m, 4m, etc. You will see that as you get further away, the falling frequency response becomes more pronounced. Thus the tonality of the speaker depends on listening distance. Sit too far away, and you will get bass heavy sound with a pronounced downward treble tilt, regardless of what the predicted in-room response says. You don't have to take my word for it. Get your mic out and go check it for yourself. Why bother with the predicted in-room response when you can easily check your actual room response?

The fact is, once loudspeakers leave the sanitized test conditions where they were designed and into the great unwashed public like you and me, we do all sorts of awful things to speakers. Just look at some of the systems posted on ASR. You will see coffee tables in front of speakers, bookshelf speakers on computer tables, bookshelf speakers in the corner of a kitchen, dipole speakers pushed up against the front wall, and so on. So never mind what the Spinorama says. Spinoramas are artificial, what matters is what you measure.

Toole himself says that "broad tone controls" can be applied to the upper frequencies. Granted, the Magic Beans app does not do that, its aim is to analyse and remove the "room transfer function". But still - we need to realise that it's not only freqs below the transition zone that are affected by choice of speaker placement and MLP, but also the upper frequencies.

The whole Toole idea is we can hear through the room above transition,just like we hear an instrument playing in a room and we don't feel the need to "correct"' it.
And it makes sense.