Combining measurement methods to get a more accurate EQ - MMM + Gated Nearfield + Klippel Measurements

[johnp98]

I posted this on the HTM-12v2 thread, but thought it was an interesting enough concept that people might want to do it with other speakers and thus I am posting the info here as well.

I had a fun little project today and wanted to add a layer of active EQ to my HTM-12v2 and wanted to compare various EQ possibilities and their results.

Based on the Klippel measurements I made an EQ filter in REW and then did moving mic method (MMM) measurements around my listening position (MLP) and got these results (results are separated to make them easier to see, and unfortunately I was running them with my subs, which are a little hot compared to the speakers.... so just ignore the low end to focus on what the EQ is going after)

Purple = Klippel Measurements
Orange = No EQ on my HTM-12v2 doing MMM at my MLP
Blue = Klippel based EQ on my HTM-12v2 doing MMM at my MLP.

Ok fair enough, but since I have slightly different box dimensions, did constrained layer dampening on the whole inside of the box and on the waveguide itself, and had various layers of different dampening material inside the speaker and because there were concerns with the consistency of drivers in the prior HTM-12v1 I thought I would do some nearfield gated measurements of my own HTM-12v2s to see how the high end compared to the klippel measurements and here is what I got:

Purple = Klippel Measurements
Green = Nearfield measurements at 3ft gated to ~4ms I believe.

So it looks like the high end does line up rather nicely! (Although not perfect, presumably due to some of the above mentioned factors, or just my UMIK-1 and its calibration file not being perfect).

But then I got a random idea that I could use my own nearfield gated measurements to EQ the high end.
I thought that it looked to have reasonable resolution from 1,000hz upward so I made an EQ file for 1,000 - 20,00hz and here are the results:

Purple = Klippel Measurements
Green = Nearfield measurements at 3ft gated to ~4ms I believe.
Blue = Post EQ nearfield measurements at 3ft gated to ~4ms I believe.

Then use the Klippel measurements to make an EQ from ~250hz-1,000hz (trying to bridge down to the schroder frequency of my room) and got this:

Purple = Klippel Measurements
Blue = No EQ - MMM at MLP
Orange = Combo of Klippel and Gated EQs - MMM at MLP

Then I used my raw MMM to make an EQ file from 60-400hz based on the in room response for the left speaker and then again for the right.
Then I spliced all the EQ filters together. Since I had good faith in the resolution of each step / segment I was ok with the sheer amount of filters (24 filters for the right and 28 for the left). This is what I got:

Purple = Klippel Measurements
Blue = No EQ - MMM at MLP
Green = Combo EQ (MMM + Klippel + Gated) - MMM at MLP

Pretty dang happy with these results!

And now to comparing all the EQ possibilities and adding some measurements I had with Dirac and also EQ based completely on MMM at MLP you get this:

Blue = No EQ - MMM at MLP
Purple = Klippel based EQ on my HTM-12v2 doing MMM at my MLP - No EQ of the sub/speaker transition unfortunately.
Red = Global EQ based on MMM at MLP (EQ generated off the above Blue graph)
Blue = Dirac EQ / results - MMM at MLP
Green = Combo of EQs (MMM + Klippel + Gated) - MMM at MLP

Wow, ok that's a lot of measuring for one day!
But as you can see you get better results using this method!

So it looks like splicing together gated nearfield measurements, with Klippel measurements, with in room MMM you can get pretty killer results for your own specific speaker!

If there was no speaker to speaker variation then you probably don't need the gated measurements, and just use the Klippel EQ to the in room low end MMM EQ.

I guess if there was no Klippel measurements then one could take the speaker outside and do much longer gated measurements to try and get down to the Schroeder frequency and then just combine those two methods. Has anyone done this for their system?

And now to leave you with my left and right and L+R measurements with psychoacoustic smoothing to remind myself that the remaining wiggles seen at 1/12 smoothing is actually water under the bridge:

Thanks for slugging through the long post!
And yes it does sound quite wonderful!


EDIT and Updates:
Here are some other graphs that people were interested in seeing. So I am adding them to the original post.
They are based on a single point measurement at the MLP so one can look at the impulse response, distortion, spectrograph, waterfall etc.

Distortion graphs:
Here are the distortion graphs pre and post EQ (from a point measurement at the MLP).
No EQ:

Post EQ:

So it certainly has increased a bit, but does not look too bad to me. But I don't have lots of experience in this area.

Impulse response:
Measured from a point measurement at the MLP

Left Pre EQ:

Left Post EQ:

Right Pre EQ:

Right Post EQ:

Unfortunately I am not too familiar with how these graphs are posted, so let me know if you want to see them another way or another scale.

The speakers themselves are ~3-4ft from the front wall yet quite close to the side walls (~1ft) but toed in quite aggressively just infront of the MLP to help with time intensity trading that some people do for waveguide speakers. I think because these speakers have a more narrow directivity I thought keeping them that close to the side wall was reasonable (although probably get some SBIR in the low end).

Spectrograph and Waterfall:

Here is the spectrograph and also waterfall plots pre and post (based on a point measurement at the MLP)
Pre EQ:

Post EQ

Pre EQ: I changed the vertical axis to see how long it takes to decay to near steady state.

Post EQ:

 

[amirm]

Very well done and documented! For a while I have been thinking that the correct approach to EQ is to take Klippel measurements and first use EQ to correct the speaker response and then EQ for the room using usual measurements. You took it to the next level by confirming your own sample measurements. This, I am not sure is needed in grand scheme of things and given the amount of work involved. Still, doesn't hurt and can help a bit.

I promoted your thread to home page to get better attention.

 

[johnp98]

Very well done and documented! For a while I have been thinking that the correct approach to EQ is to take Klippel measurements and first use EQ to correct the speaker response and then EQ for the room using usual measurements. You took it to the next level by confirming your own sample measurements. This, I am not sure is needed in grand scheme of things and given the amount of work involved. Still, doesn't hurt and can help a bit.

Yes, I think for the average consumer speaker there would hopefully be so little speaker to speaker variation that doing ones own gated measurements is certainly not needed. I was just paranoid as I have slightly different box dimensions, did constrained layer dampening on the whole inside of the box and on the waveguide itself, and had various layers of different dampening material inside the speaker and mostly because there were concerns with the consistency of drivers in the prior HTM-12v1 and thus I wanted to remove any doubt in my mind.

My general thoughts are:

1) You need Klippel measurements to see if a speaker is any good and has good directivity (and is worth EQing and would not be harmed by it).

2) If you have Klippel measurements then EQ based on that above the Schroder frequency (assuming there is reasonable speaker consistency or you did not tamper with your speaker at all) and then EQ the speaker/room with MMM at the MLP.

3) If you do not have Klippel measurements then you could try and get by with nearfield gated measurements getting you down into the Schroder frequency (so you would have to do it outside with a longer window/gating - to get reasonable mid frequency resolution) and then combine this with MMM at MLP. But I have never done this and I assume you still would always be questioning point #1.

I promoted your thread to home page to get better attention.

Thank you very much!

 

[sweetchaos]

Since I had good faith in the resolution of each step / segment I was ok with the sheer amount of filters (24 filters for the right and 28 for the left).

Did you manually create the filters or let REW do it?
Can you share them? It will be interested to see.
Assuming that someone wanted to recreate this method, can MiniDSP support that many filters or is this strictly for PC setup with Equalizer APO?

Great work, btw.

 

[GWolfman]

Cool stuff!

 

[Lorenzo74]

Did you manually create the filters or let REW do it?
Can you share them? It will be interested to see.
Assuming that someone wanted to recreate this method, can MiniDSP support that many filters or is this strictly for PC setup with Equalizer APO?

Great work, btw.

minidsp (ex. ddrc24) can support a number of filters that can flat out gated measurements.

For the room response it’s more tricky. It mainly depends on the numbers of listening position you want to include in the equalization and the weight you assign to each of them. It’s risky to EQ only one measurement where your “head” is located.

If you think holistic this is similar to the approach DIRAC Live try with the mixed phase filters.

by Multiple measurements around listening position and crosscorrelation it is possible to identify early reflection and the “gated” portion of the frequency responce maybe down around 1kHz.

then the room is tamed by a filter that address main response variations with rule of best compromise.
(@Flak may help here).

I do believe what has been very well presented in this thread has been also executed in active speaker like D&D, Kii Three, Neumann at least for the “gated” part of the signal.

Many thanks @johnp98, one question for you:
I would be interested to see the step response (first 10ms) of your Left and Right channels to see how early reflections affect it.

My Best

 

[dominikz]

I posted this on the HTM-12v2 thread, but thought it was an interesting enough concept that people might want to do it with other speakers and thus I am posting the info here as well.

I had a fun little project today and wanted to add a layer of active EQ to my HTM-12v2 and wanted to compare various EQ possibilities and their results.

Based on the Klippel measurements I made an EQ filter in REW and then did moving mic method (MMM) measurements around my listening position (MLP) and got these results (results are separated to make them easier to see, and unfortunately I was running them with my subs, which are a little hot compared to the speakers.... so just ignore the low end to focus on what the EQ is going after)

Purple = Klippel Measurements
Orange = No EQ on my HTM-12v2 doing MMM at my MLP
Blue = Klippel based EQ on my HTM-12v2 doing MMM at my MLP.

Ok fair enough, but since I have slightly different box dimensions, did constrained layer dampening on the whole inside of the box and on the waveguide itself, and had various layers of different dampening material inside the speaker and because there were concerns with the consistency of drivers in the prior HTM-12v1 I thought I would do some nearfield gated measurements of my own HTM-12v2s to see how the high end compared to the klippel measurements and here is what I got:

Purple = Klippel Measurements
Green = Nearfield measurements at 3ft gated to ~4ms I believe.

So it looks like the high end does line up rather nicely! (Although not perfect, presumably due to some of the above mentioned factors, or just my UMIK-1 and its calibration file not being perfect).

But then I got a random idea that I could use my own nearfield gated measurements to EQ the high end.
I thought that it looked to have reasonable resolution from 1,000hz upward so I made an EQ file for 1,000 - 20,00hz and here are the results:

Purple = Klippel Measurements
Green = Nearfield measurements at 3ft gated to ~4ms I believe.
Blue = Post EQ nearfield measurements at 3ft gated to ~4ms I believe.

Then use the Klippel measurements to make an EQ from ~250hz-1,000hz (trying to bridge down to the schroder frequency of my room) and got this:

Purple = Klippel Measurements
Blue = No EQ - MMM at MLP
Orange = Combo of Klippel and Gated EQs - MMM at MLP

Then I used my raw MMM to make an EQ file from 60-400hz based on the in room response for the left speaker and then again for the right.
Then I spliced all the EQ filters together. Since I had good faith in the resolution of each step / segment I was ok with the sheer amount of filters (24 filters for the right and 28 for the left). This is what I got:

Purple = Klippel Measurements
Blue = No EQ - MMM at MLP
Green = Combo EQ (MMM + Klippel + Gated) - MMM at MLP

Pretty dang happy with these results!

And now to comparing all the EQ possibilities and adding some measurements I had with Dirac and also EQ based completely on MMM at MLP you get this:

Blue = No EQ - MMM at MLP
Purple = Klippel based EQ on my HTM-12v2 doing MMM at my MLP - No EQ of the sub/speaker transition unfortunately.
Red = Global EQ based on MMM at MLP (EQ generated off the above Blue graph)
Blue = Dirac EQ / results - MMM at MLP
Green = Combo of EQs (MMM + Klippel + Gated) - MMM at MLP

Wow, ok that's a lot of measuring for one day!
But as you can see you get better results using this method!

So it looks like splicing together gated nearfield measurements, with Klippel measurements, with in room MMM you can get pretty killer results for your own specific speaker!

If there was no speaker to speaker variation then you probably don't need the gated measurements, and just use the Klippel EQ to the in room low end MMM EQ.

I guess if there was no Klippel measurements then one could take the speaker outside and do much longer gated measurements to try and get down to the Schroeder frequency and then just combine those two methods. Has anyone done this for their system?

And now to leave you with my left and right and L+R measurements with psychoacoustic smoothing to remind myself that the remaining wiggles seen at 1/12 smoothing is actually water under the bridge:

Thanks for slugging through the long post!
And yes it does sound quite wonderful!

Great work! I've been thinking about trying something similar so it's really great seeing someone go for it and provide such nicely organized and well presented results!
I bet it sounds great!

However, I do have a general (theoretical) concern - I worry that using just the on-axis gated measurement for the HF EQ might not work as well for speakers where there is significant difference between anechoic on-axis and LW responses. Any thoughts on this?

 

[Absolute]

I've done this type of combined EQ¹ a few times (not with Klippel measurements) on my M2-clones, but it never sounded right. I tried to narrow the cause down and ended up concluding that too aggressive EQ in the bass where you have dips due to SBIR will introduce audible ringing.

Spectrogram will confirm this. At one point I tried to replace the EQ for sbir cancellations with a bag of Rockwool insulation and the sound was a lot better for it.

I think we need to combine additional measurements to determine what to EQ and by how much for optimal result, so if OP have time I'd suggest looking at the minimum phase plot and the spectrogram to see the in-room effect of EQ based on MMM below 1 kHz.

If not on vacation I would do this myself.

¹ - Single-point EQ below 100Hz both individual left and right speaker and combined, MMM over that separately, gated near-field measurements vs MMM near-field, EQ based on time domain instead of frequency response below 100 and 200 hz with MMM above etc.

 

[thewas]

Thank you for presenting a very nice approach which I also had been using, just as a general remark to everyone who is doing such EQing, please be careful with narrow negative filters and even more positive ones (narrow and wide). The first ones can make the sound lifeless as we measure the steady state in the modal region in the room but our perception doesn't fully work like that and the second being resonators(!) can make the sound boomy. The problem is that we can't really fix a poor room response by just adapting the input to it. So please always directly compare different filter configurations and even with no filter as the more I deal with DRC the more I find that corrections that look nice at the frequency response often sound worse than ones which don't give nice smooth curves.

 

[bigjacko]

I have used MMM to eq my speaker without any other measurement, it sounded pretty good afterwards. One thing I found is that the high frequency measurement is pretty consistent even coming back next day and do another measurement, but lower than 1k Hz I can't get consistent measurement if I come back after a break. I think maybe I was not very consistent for my position and volume of measurement but I tried my best. Do you have this kind of issue, and specially at some room modes or SBIR?

 

[abdo123]

To be honest I really think you're over compensating, Orange in Figure 1 is very very smooth and the remaining messiness is probably because the MMM range was limited in one of the 3 axes during measurements (Listener boundary interference response, LBIR) or because of Speaker boundary interference response. You should not compensate for these two things unless this is a dedicated listening space and only one person will listen at a time.

My System has an olive preference score of ~8.5 (my subwoofers are sealed and don't extend very low) and I got a fairly smooth response when I did 19 spatially averaged measurements in Dirac.

If you get more than +/- 1 dB error rate above the transition frequency and your speakers are anechoicly flat then you're not really measuring the in-room response, you're only measuring that particular listening position.

If you get more than +/- 5 dB error rate between 100Hz and the transition frequency then you're dealing with Boundary interaction response related issues.

my 2 cents on the fastest way to get a good in-room response:

20Hz to transition frequency: Correct based on what you measure, because you hear what you measure (room is modal).

2KHz to 20Khz: Correct anechoicly only, room is not modal anymore, it is stochastic and has little influence on the sound at this point. What you measure is not exactly what you hear, your ears have different directivity than the microphone, you also have two ears and they're located differently than where the microphone is. And most importantly, your brain is capable and will filter some reflections out, the microphone will not.

transition frequency to 2KHz: correct both anechoicly and based on in-room measurements, make sure you don't over compensate, try to get as many measurements as possible (sensibly).

 

[holbob]

Thank you for presenting a very nice approach which I also had been using, just as a general remark to everyone who is doing such EQing, please be careful with narrow negative filters and even more positive ones (narrow and wide). The first ones can make the sound lifeless as we measure the steady state in the modal region in the room but our perception doesn't fully work like that and the second being resonators(!) can make the sound boomy. The problem is that we can't really fix a poor room response by just adapting the input to it. So please always directly compare different filter configurations and even with no filter as the more I deal with DRC the more I find that corrections that look nice at the frequency response often sound worse than ones which don't give nice smooth curves.

What would be the maximum q you would recommend for the negative filters? I tend to stick to 8 which is where where a lot of software tops out at, but is that too high? I can eq my sub to a flat response using 10 manual 8 q filters in my minidsp, but some of the filters are quite large in dB due to a dip in the middle bass. What is the maximum amount of dB recommended in a negative filter?

It would be nice to have a "community approved" thread with basic recommended eq tips. The problem on our forum with eq advice, is that it is readily and kindly dished out with the best of intentions - but it is MASSIVELY contradictory and lacks thorough explanations.

One person will be saying don't eq below schroeder - which they say is 200hz. Another is saying don't eq below schroeder - which they say is 600hz. Another says don't eq below 1000hz. Another is saying eq full range with very low q filters. Now we are saying add eq based on kippel measurements, which are full range often with very high q filters with less than 1db filters. Well surely that wouldn't even by audible making it pointless???

Community approved advice would cut out the time spent in the burrow. I've been in it since 2020 now!

 

[thewas]

What would be the maximum q you would recommend for the negative filters? I tend to stick to 8 which is where where a lot of software tops out at, but is that too high? I can eq my sub to a flat response using 10 manual 8 q filters in my minidsp, but some of the filters are quite large in dB due to a dip in the middle bass. What is the maximum amount of dB recommended in a negative filter?

Depends on the frequency range, in the bass an upper limit of 8 seems reasonable to me, higher frequencies of course less.

It would be nice to have a "community approved" thread with basic recommended eq tips. The problem on our forum with eq advice, is that it is readily and kindly dished out with the best of intentions - but it is MASSIVELY contradictory and lacks thorough explanations.

I can fully understand your frustration, the problem is that the topic is imho not well examined enough so that there is a common agreed state of art approach and possibly depends also on many individual parameters.

Community approved advice would cut out the time spent in the burrow. I've been in it since 2020 now!

If it comforts you, I am reading and experimenting on DRC for more than 10 years now and still am on the search of the optimum path.
In the end I can only recommend everyone to try many different approaches and by direct comparison (not only quick but also long term pleasure and fatigue) find the one that works best for your setup.

 

[Mauro]

Very well done and documented! For a while I have been thinking that the correct approach to EQ is to take Klippel measurements and first use EQ to correct the speaker response and then EQ for the room using usual measurements. You took it to the next level by confirming your own sample measurements. This, I am not sure is needed in grand scheme of things and given the amount of work involved. Still, doesn't hurt and can help a bit.

I promoted your thread to home page to get better attention.

Wouldn't a near field measurement pick just the on-axis response? Why would we might want to calibrate a speaker even above 1kHz based on on-axis response? Not many speakers have a good directivity index, even the HTM-12 has its own problems at 1,5KHz.

And then using the MMM with a directional mic to prove that everything worked fine does not seem a good idea to me.
As far as I understand, above 5KHz the MMM is unreliable: check this video at 8:30 by Charles Sprinkle (Kali Audio, ex JBL engineer).

Maybe @amirm or @Charles Sprinkle can tell us more and point us in the right direction on how to validate our in-room measurements and what's the best method to use above 5KHz...

 

[daftcombo]

No need to EQ so tighly, especially if the speaker directivity isn't perfect.

Under 1 kHz though, as directivity issues are rare, and MMM technique hard to do in-room, I'd EQ following the Klippel measurements any time.

Then only correct room modes.

 

[johnp98]

Great work! I've been thinking about trying something similar so it's really great seeing someone go for it and provide such nicely organized and well presented results!
I bet it sounds great!

However, I do have a general (theoretical) concern - I worry that using just the on-axis gated measurement for the HF EQ might not work as well for speakers where there is significant difference between anechoic on-axis and LW responses. Any thoughts on this?

Thanks for all the thoughts and feedback.

The main reason I wanted to try this out was because the HTM-12v2 have recently been measured at Erin's Audio Corner and they had a reasonable frequency response, but what really stood out was the normalized horizontal polar plot:

Because of the great normalized horizontal polar plot I felt confident that if I could EQ the speaker for the on axis response and then everything direct and radiated would follow suite. Certainly if there was a speaker with very poor directivity and radiation I would be cautious with this approach.

I did not want to EQ the high end based on my listening window, as then I would be EQing the direct and reflected sounds which for the high frequencies is not ideal.
My original measurements even seem to prove this point as one can compare the results of EQing based MMM at the MLP vs this combo approach measured via MMM at the MLP

Blue = No EQ - MMM at MLP
Purple = Klippel based EQ on my HTM-12v2 doing MMM at my MLP - No EQ of the sub/speaker transition unfortunately.
Red = Global EQ based on MMM at MLP (EQ generated off the above Blue graph)
Blue = Dirac EQ / results - MMM at MLP
Green = Combo of EQs (MMM + Klippel + Gated) - MMM at MLP

So comparing Red vs Green it seems that the combo approach does better at the MLP.

 

[johnp98]

To be honest I really think you're over compensating, Orange in Figure 1 is very very smooth and the remaining messiness is probably because the MMM range was limited in one of the 3 axes during measurements (Listener boundary interference response, LBIR) or because of Speaker boundary interference response. You should not compensate for these two things unless this is a dedicated listening space and only one person will listen at a time.

I have a single listening location in a dedicated listening space, so EQing for what I hear at the MLP is ok with me. Sorry occasional guests.
I took the MMM over a reasonable distance ~3ft in all 3 axis, so I think it should give a fair representation of what the speakers are doing in the modal region. Certianly it should show SBIR, but the LBIR should be a little more diffuse due to the MMM measuring distance.

The original orange line in Figure 1 is the HTM-12s with no EQ, so certainly there is room to EQ the transition/modal region for sure.

my 2 cents on the fastest way to get a good in-room response:

20Hz to transition frequency: Correct based on what you measure, because you hear what you measure (room is modal).

2KHz to 20Khz: Correct anechoicly only, room is not modal anymore, it is stochastic and has little influence on the sound at this point. What you measure is not exactly what you hear, your ears have different directivity than the microphone, you also have two ears and they're located differently than where the microphone is. And most importantly, your brain is capable and will filter some reflections out, the microphone will not.

transition frequency to 2KHz: correct both anechoicly and based on in-room measurements, make sure you don't over compensate, try to get as many measurements as possible (sensibly).

I would agree with your approach and I think my measurements and approach followed this.
20Hz to transition frequency = MMM
Transition frequency to 1KHz = Klippel (but one could see if they could get gated measurements to get this low)
1KHz to 20Khz: Nearfield gated measurements

 

[johnp98]

What would be the maximum q you would recommend for the negative filters? I tend to stick to 8 which is where where a lot of software tops out at, but is that too high? I can eq my sub to a flat response using 10 manual 8 q filters in my minidsp, but some of the filters are quite large in dB due to a dip in the middle bass. What is the maximum amount of dB recommended in a negative filter?

I also would be interested in peoples perspective on this. Certainly interested in what people think is the max db for negative and positive filters.
I guess one question I always had, is people do not like using positive filters or 'boosting'.
Certainly there is no point to have positive filters if you are trying to fill a null or a SBIR (as no amount of extra energy will fill that null).
Others mention that there is the risk of over exerting the speaker or amp, but if one adds negative pre-amplification to match the positive boost then that is totally acceptable is it not? Would that not come out equivalent to a negative filter everywhere outside of the boosted area?

It would be nice to have a "community approved" thread with basic recommended eq tips. The problem on our forum with eq advice, is that it is readily and kindly dished out with the best of intentions - but it is MASSIVELY contradictory and lacks thorough explanations.

One person will be saying don't eq below schroeder - which they say is 200hz. Another is saying don't eq below schroeder - which they say is 600hz. Another says don't eq below 1000hz. Another is saying eq full range with very low q filters. Now we are saying add eq based on kippel measurements, which are full range often with very high q filters with less than 1db filters. Well surely that wouldn't even by audible making it pointless???

Community approved advice would cut out the time spent in the burrow. I've been in it since 2020 now!

I agree, but I can take a stab at it and also put my understanding of why

10hz - 200-400hz (Below ones schroeder frequency which is room dependent):
Room is modal, room is dominating the speakers, you are hearing the steady state of the bass, your tower speakers are probably not in the right place to put sufficient energy into this region, you probably need subs and multiples of them if you care about seat to seat consistency.
This area is prime EQ material!
How to EQ seems to be best if you can truly get a sense of what the speakers and subs are doing in the room, so you want a range of measurements over your MLP. So it seems that a wide MMM in all 3 axis over your MLP is the best way to do this.
Our hearing is more sensitive to narrow frequency's in this region and thus you don't want to EQ a very smoothed / low resolution measurements (So want to use Var or 1/24 in REW)
EQ the peaks and if the nulls are narrow then they are likely SBIR and no amount of power and bosting will improve them so dont boost them.

Above schroeder frequency
Well this is where al the controversy comes in and it is maybe easier to talk about the downsides of each approach.
Single point measurement = well this will have a combo of direct and reflected sound and comb filtering etc, so is too messy to EQ

MMM = still is a combo of direct and reflected sound, so if the speaker has poor directivity then you are correcting a combination of the two yet your brain primarily looks for the direct sound and then merges the reflected sound into it. So you really would want the best on axis sound, but have no way of knowing what is what with your measurements.

Nearfield gated measurements = Ok getting closer, now you have the direct sound, but its hard to get a long enough window to have high resolution down to the schroeder frequency. I think if you were to EQ the high end of a speaker this is what I would use if I did not have anechoic measurements.

Anechoic / Klippel = Gold standard, shows the direct and reflected sound, so now you can EQ the direct sound and also know the speaker directivity and radiation and if those will also follow suite.

The resolution and amount you want to EQ is up to you, but certainly our hearing is likely lower resolution than many of the small filters we make in this region.

I 100% agree that the number of filters I made is quite ridiculous, but since I am using a computer and not limited by filters I wanted to do this more as a proof of concept rather than true comparison of A vs B audibility.

 

[johnp98]

Wouldn't a near field measurement pick just the on-axis response? Why would we might want to calibrate a speaker even above 1kHz based on on-axis response? Not many speakers have a good directivity index, even the HTM-12 has its own problems at 1,5KHz.

I thought that the normalized directivity was one of the best I had seen, so good enough to focus on the direct sound for the high end and I thought the best way to measure that is nearfield gated measurements.

And then using the MMM with a directional mic to prove that everything worked fine does not seem a good idea to me.
As far as I understand, above 5KHz the MMM is unreliable: check this video at 8:30 by Charles Sprinkle (Kali Audio, ex JBL engineer)..

I would agree with this, that is why I showed the nearfield gated measurement showing pre and post and how EQ improved it.

Maybe @amirm or @Charles Sprinkle can tell us more and point us in the right direction on how to validate our in-room measurements and what's the best method to use above 5KHz...

Certainly I don't have their knoledge, but I think MMM for the low end and gated or Klippel for the high end seems to make sense to me.

 

[johnp98]

No need to EQ so tighly, especially if the speaker directivity isn't perfect.

Under 1 kHz though, as directivity issues are rare, and MMM technique hard to do in-room, I'd EQ following the Klippel measurements any time.

Then only correct room modes.

I would agree with this.

I guess I wanted to give this a try, more as a proof of concept rather than saying one needs an exorbitant amount of filters to get good sound. Certainly looking at your measurements with psychoacoustic smoothing from time to time can put things in better perspective.