Room/Speaker EQ Approaches: Parametric EQ settings vs Impulse/Convolution

[Wayne A. Pflughaupt]

Option 2 is the worst as you’re only correcting what your mic heard from a single point in space. You should do a sweep measurement as well as MMM RTA of L, R, and L+R. The sweeps will give you insight as to what response deviations are minimum phase or not (as I understand it, these are readily correctable while excess phase is usually not). Then once you have an idea of what you’re dealing with, use an appropriately sloped target response on the MMM measurements to obtain the parametric EQ settings needed for correction.

I find MMM to be a curiosity. If you’ve done it yourself, perhaps you can answer a couple of questions for me?

1. Before EQ, at the various locations where measurements were taken, could you hear a difference with music or other program material compared what was heard at the main listening position (MLP)?

2. After equalizing via MMM, did the results sound better than had been accomplished by EQing based on results for only the MLP, either at the MLP or the other locations where measurements had been taken?

Regards,
Wayne A. Pflughaupt

 

[Blumlein 88]

I find MMM to be a curiosity. If you’ve done it yourself, perhaps you can answer a couple of questions for me?

1. Before EQ, at the various locations where measurements were taken, could you hear a difference with music or other program material compared what was heard at the main listening position (MLP)?

2. After equalizing via MMM, did the results sound better than had been accomplished by EQing based on results for only the MLP, either at the MLP or the other locations where measurements had been taken?

Regards,
Wayne A. Pflughaupt

https://www.audiosciencereview.com/...e-your-in-room-measurements.13540/post-408373

Here is a chart for one spot at the LP and 6 spots around the LP.

I've found if I do a single point measure with pink noise and with a sweep once some smoothing is applied they are identical. I find with MMM you get something very similar, and probably more accurate than a 6 position average using a sweep.

 

[hyperplanar]

I find MMM to be a curiosity. If you’ve done it yourself, perhaps you can answer a couple of questions for me?

Sure!

1. Before EQ, at the various locations where measurements were taken, could you hear a difference with music or other program material compared what was heard at the main listening position (MLP)?

I have to clarify, my setup is very nearfield (~1m equilateral triangle). So for sweep measurements, they were mostly taken with the mic positioned where the center of my head would normally be. The way I did it before MMM was to take around 9 measurements per speaker, moving the mic a bit each time, and averaging them. The MMM measurements were done in a roughly 1.5 foot cube area. That being said, yes, I could hear a difference in music moving my head around a bit, side-to-side or front-to-back as would occur normally. Especially in the lower midrange region, with material such as piano, where some notes tended to sound louder and more resonant.

2. After equalizing via MMM, did the results sound better than had been accomplished by EQing based on results for only the MLP, either at the MLP or the other locations where measurements had been taken?

Yes, absolutely! MMM provided much better results, at least for my use case. I know @QMuse uses MMM for his non-nearfield setup and perhaps could elaborate more, but I found the result of MMM to be much more representative of what I was actually hearing. I was able to EQ some resonances out which were difficult to pin down by doing multiple sweep measurements.

 

[Pio2001]

I find MMM to be a curiosity. If you’ve done it yourself, perhaps you can answer a couple of questions for me?

1. Before EQ, at the various locations where measurements were taken, could you hear a difference with music or other program material compared what was heard at the main listening position (MLP)?

Hi,
In my case, I'm listening 2.1 meters from the speakers, in an untreated room. There are two strong room modes, one at 54 Hz, another at 69 Hz. There is also a strong SBIR caused by the window behind the speakers, with a large dip in the frequency response around 100 Hz, and some peaks around 200 Hz (not exactly the same for the left and right speakers).

I never listened without EQ, as the room modes makes it impossible to enjoy music this way, but after EQ, the sound is mostly the same if I move my head inside the MMM volume (+/- 30 cm around the listening position), except at the very front of the listening zone, where the frequencies of the SBIR are shifted, and the room correction becomes wrong.

2. After equalizing via MMM, did the results sound better than had been accomplished by EQing based on results for only the MLP, either at the MLP or the other locations where measurements had been taken?

I compared some EQ settings made from measurements averaged from 6 positions +/- 30 cm around the main listening position, then from 6 positions again (mostly the same), then from the 12 individual previous measurements (I averaged the two previous averages), then from a MM measurement spanning the same area.

The results were better starting from the 12-position average and from the MMM.
Averaging 6 positions only doesn't seem to be enough.

Another advantage of the MMM is that it naturally applies a variable smoothing to the measured frequency response.

 

[Blumlein 88]

Just some data. Here is a single sweep, vs pink noise with the microphone moving over a 30 cm corkscrew. I intentionally picked a spot exactly halfway between the front and rear, and the left/right walls. You get that 140 hz hump in that spot. It goes away if you move from exact center.

Psycho-acoustic smoothing. The two curves look about as similar if you do 1/24th smoothing or smoother. Without smoothing the pink noise is variable over about 1/4th the range vs a sweep. Red is the sweep, and green is the pink noise MMM.

Now I didn't bother showing pink noise from the single spot of the sweep. The differences between the sweep and stationary pink noise is nothing or at most the width of the lines on the graph. Pink noise was 32 averages in all cases.

 

[QMuse]

EQ also corrects phase - see Myth #6 at this link.

https://www.ranecommercial.com/legacy/note115.html

Regards,
Wayne A. Pflughaupt

Phase shifts introduced by PEQ filters are small and definitely inaudible. As an example, dotted line shows phase shift of my correction of Sony SS-CS5 speaker.

 

[QMuse]

The main con is that IR convolution is more resource-intensive than the standard parametric EQ IIR filters, and this gets worse the lower you go in the frequency range.

I have seen folks running BruteFIR on RPI3 B+with CPU utlisation of <20%.

You can use convolution as the final touch to linearize the crossover phase shift at a later point—IMO the room correction is waaaay more important.

Indeed it is. Regarding phase correction, what matters beside correcting the passive XO phase shift is to avoid phase cancellation between speakers in the 20-120Hz region as that causes audible dips in FR.

 

[QMuse]

In addition, care should be taken applying different filters to the main L/R speakers. Mismatched filters above ~300 Hz can whack the imaging. Only below that point should mis-matched filters be attempted.

I would say it's 400Hz, rather than 300Hz, but yes, above that you are correcting speaker, not room, so filters should pretty much be identical (assuming both speakers measure similar, as they should). In the 400-20kHz region I usually apply some mild filters to linearize the response around the XO and some very wide filters to get the tonal balance right (see example with Sony EQ filter).

 

[Pio2001]

Phase shifts introduced by PEQ filters are small and definitely inaudible. As an example, dotted line shows phase shift of my correction of Sony SS-CS5 speaker.

View attachment 65719

PEQ filters are minimal phase. Their action on the time domain sounds natural and is not audible as such, unless the change in frequency response is extreme.

But if we try to correct strong room modes using linear phase equalization, instead of minimum phase, the difference is audible as a pre-ringing in low frequencies.
I've confirmed this with a successful ABX test.

So, we could rather say that it's the lack of phase shift in linear phase equalization that is audible.

 

[hyperplanar]

Phase shifts introduced by PEQ filters are small and definitely inaudible. As an example, dotted line shows phase shift of my correction of Sony SS-CS5 speaker.

View attachment 65719

Not so for the steep and deep filters that are usually employed for room mode correction. But conveniently these phase shifts cancel out the phase shifts of the room modes themselves, since they're mostly minimum phase. That's why minimum phase EQ fixes room modes in the time domain as well. Thus there are no upsides to using linear phase EQ for normal room mode correction.

The only time linear phase EQ needs to be used is for excess phase issues, and not in the bass region when possible to avoid potentially audible preringing. Use the "generate minimum phase" function in REW and view the minimum phase and excess phase plots for your measurements—any and all minimum phase deviations are a direct natural consequence of the frequency response, meaning if you perfectly correct the frequency response with minimum phase EQ, the minimum phase line will become perfectly flat as well.

Using a raw in-room measurement of my subwoofer as an example:

It can be seen the phase deviations caused by room modes are minimum phase. Therefore, when the frequency response deviations are corrected for with minimum phase EQ, these minimum phase deviations will also be corrected. The consequence of this is that the time-domain ringing of the modes are cancelled out by the EQ as well.

 

[Wayne A. Pflughaupt]

https://www.audiosciencereview.com/...e-your-in-room-measurements.13540/post-408373

Here is a chart for one spot at the LP and 6 spots around the LP.

I've found if I do a single point measure with pink noise and with a sweep once some smoothing is applied they are identical. I find with MMM you get something very similar, and probably more accurate than a 6 position average using a sweep.

I have to clarify, my setup is very nearfield (~1m equilateral triangle). So for sweep measurements, they were mostly taken with the mic positioned where the center of my head would normally be. The way I did it before MMM was to take around 9 measurements per speaker, moving the mic a bit each time, and averaging them. The MMM measurements were done in a roughly 1.5 foot cube area. That being said, yes, I could hear a difference in music moving my head around a bit, side-to-side or front-to-back as would occur normally. Especially in the lower midrange region, with material such as piano, where some notes tended to sound louder and more resonant.

Hi,
In my case, I'm listening 2.1 meters from the speakers, in an untreated room. There are two strong room modes, one at 54 Hz, another at 69 Hz. There is also a strong SBIR caused by the window behind the speakers, with a large dip in the frequency response around 100 Hz, and some peaks around 200 Hz (not exactly the same for the left and right speakers).

I never listened without EQ, as the room modes makes it impossible to enjoy music this way, but after EQ, the sound is mostly the same if I move my head inside the MMM volume (+/- 30 cm around the listening position), except at the very front of the listening zone, where the frequencies of the SBIR are shifted, and the room correction becomes wrong.

Thanks guys, for taking the time to respond!

I’ll probably get flamed for this, but I admit I’m a skeptic. I’ve been doing the room measurement / EQ thing for 25 years, long before REW. I’ve never had any issues with measuring and EQing from and for the MLP. But if I had a room that gave me the problem of hearing things different just by moving my head, as hyperplanar described, I’d be inclined to give MMM serious consideration!

But for those who don’t have that a room with that problem, I think a lot of people simply get too obsessed with this stuff, paying more attention to getting nice looking graphs instead of what sounds best. I like the way one guy recently said it over at AV NIRVANA: "I’ve had the feeling that my overzealous EQ attempts fixed the graphs but killed the sound."

Regards,
Wayne A. Pflughaupt

 

[Wayne A. Pflughaupt]

Phase shifts introduced by PEQ filters are small and definitely inaudible. As an example, dotted line shows phase shift of my correction of Sony SS-CS5 speaker.

Not so for the steep and deep filters that are usually employed for room mode correction. But conveniently these phase shifts cancel out the phase shifts of the room modes themselves, since they're mostly minimum phase. That's why minimum phase EQ fixes room modes in the time domain as well. Thus there are no upsides to using linear phase EQ for normal room mode correction.

Indeed, the article I linked basically says the same thing both of you did. Here is a relevant excerpt:

Every time the amplitude changes so does the phase shift. In fact, it can be argued that phase shift is the stuff that causes amplitude changes. Amplitude, phase and time are all inextricably mixed by the physics of sound. One does not exist without the others.
An equalizer is a tool. A tool that allows you to correct for a room's anomalies. It must be capable of reproducing the exact opposite response of the one being connected. This requires precise correction at many neighboring points with the associated phase shift to correct for the room's opposing phase shift. It takes phase shift to fix phase shift. Simple as that.

Regards,
Wayne A. Pflughaupt

 

[Wayne A. Pflughaupt]

I would say it's 400Hz, rather than 300Hz...

I can’t disagree with that. I suspect the relevant point is highly dependent on the Schoeder Frequency, which varies from one room to the next.

Regards,
Wayne A. Pflughaupt

 

[Blumlein 88]

Well here is a LP sweep in green, and a fixed pink noise measure in pink at the same location. 1/6th smoothing. Not a lot to pick between them.

There is a greater difference between right and left sweeps 8 inches apart at the LP as shown with green left and dark red right.

Here is your right and left ear at the LP averaged in blue. You have to admit your ears cannot both be in the same place. Pink is the MMM result with the movement restricted to a flat oval 4 inches tall and 8 inches wide approximately. You could say a single point sweep isn't radically different and it isn't. I'd think the combined frequency averaging of pink noise and spatial averaging of the movement is a little bit closer to what is coming out of the speakers than not. Less influence from room reflections.

I don't see where any of those are flattening the curve artificially in some extreme manner. I wouldn't be afraid of using the MMM result if I'm doing PEQ instead of impulse convolution.

 

[Pio2001]

But if I had a room that gave me the problem of hearing things different just by moving my head, as hyperplanar described, I’d be inclined to give MMM serious consideration!

For me (and also for a professional that I know), the purpose of MMM is not to increase the accuracy of the measurement, but to gain a lot of time.
The microphone doesn't have to stand by itself, it can be held by hand (as long as we also move the arm relative to the body so as to cancel the reflections of the sound on the body), and it takes only one run to get the averaged curve instead of 12 (which is roughly the equivalent of an MMM run).

But of course, sweep measurements are still needed if we want to see the decay, the group delay and the excess phase.

I also like to perform a multipoint measurement and look at the individual curves in order to get an idea of the variations according to the position, like this (6 sweeps taken at the edge of a 50x60 cm rectangle) :

But once I've done this once, I made all subsequent measurements in MMM. The one to check the result after eq, the one to check the effect of moving the computer's screen etc. It's way faster than multipoint... and maybe a bit more accurate.

 

[QMuse]

But if I had a room that gave me the problem of hearing things different just by moving my head, as hyperplanar described, I’d be inclined to give MMM serious consideration!

But for those who don’t have that a room with that problem, I think a lot of people simply get too obsessed with this stuff, paying more attention to getting nice looking graphs instead of what sounds best. I like the way one guy recently said it over at AV NIRVANA: "I’ve had the feeling that my overzealous EQ attempts fixed the graphs but killed the sound."

Regards,
Wayne A. Pflughaupt

Unless your room is very "calm" with not many reflections you will always have that problem, and the only thing you can do to avoid it is to base your EQ on a spatially averaged measurement. The point of MMM is it is giving you just that, and you can do it quickly. I usualy take MMM with 60+ samples and I don't even want to think how much time it iwll take me to do 60 sweeps from different positions. Once I fix the frequency response I do a sweep to correct phase and that's it. You can do vector average of a bucnh of sweeps to average phase response but I foudn that unnecesary as phase is ot changing so much with positions as frequency response is.

 

[Wayne A. Pflughaupt]

It all sounds interesting for sure, but I’d be more inclined to jump on board if a few people could say they compared the audible results of EQ with both single MLP and MMM, and they felt the latter sounded better. I’ve been doing this long enough to know that a graph looking better than another doesn’t necessarily mean the former actually sounds better.

Regards,
Wayne A. Pflughaupt

 

[Wayne A. Pflughaupt]

On a separate question, does the current REW permit MMM? I'm using a really old version.

Regards,
Wayne A. Pflughaupt

 

[Blumlein 88]

On a separate question, does the current REW permit MMM? I'm using a really old version.

Regards,
Wayne A. Pflughaupt

What you do is open REW. Open RTA and set it to read 1/48th octave instead of spectrum (spectrum is just FFT).

Pull up the Signal Generator and choose pink noise or PN for periodic pink noise. PN is a little better more stable.

In settings for the RTA you can have say 32 averages or forever averaging.

Let it play the pink noise while you move the microphone and record into the RTA. When satisfied you stop the RTA recording and choose Save. It will show up as a measurement in the normal window.

 

[Pio2001]

It all sounds interesting for sure, but I’d be more inclined to jump on board if a few people could say they compared the audible results of EQ with both single MLP and MMM, and they felt the latter sounded better. I’ve been doing this long enough to know that a graph looking better than another doesn’t necessarily mean the former actually sounds better.

In my modest experience, the type of measurement has no effect on the result.

The weight of individual locations may change the result : are the individual measurements taken around a square? A rectangle ? A circle ? Is there an additional measurement in the center ? Two of them ? Will you move the microphone exactly at ear's level ? +/- 5 cm above and below ? +/- 10 cm ?
What smoothing will be applied to the curves ?
Are we measuring both speakers together, or measuring them one by one ?

All these questions may cause minor variations in the results, while, in an ideal situation, the choice between multipoint or MMM should cause no difference at all.