Moving mic vs. averaged vs. single point measurements - which is better?

[Keith_W]

AFAIK there are three ways we can measure room response for loudspeakers:

- Maximum Length Sequence (MLS), Exponential Sine Sweep (ESS) or single point sweep: The mic is set up at the listening position, and a single sweep is taken.
- Averaged: The mic is set up at various points around the listening position, multiple sweeps are taken, and the responses are averaged.
- Moving mic measurement (MMM): The mic is set up on a boom, and pink noise is played. The mic is then physically moved around the listening position whilst recording the result.

I have always done single point sweeps. This is as recommended by Uli Bruegemann, author of Acourate*. Then I came across this article advocating use of MMM which made the bold claim that there is no correlation between measured response and listening quality for a MLS. It also includes a quote from Floyd Toole who said "spatial averaging of speaker measurements is critical - single point measurements are erroneous and meaningless". This made me think that I should probably be doing MMM measurements. After all, repeated sweeps of my listening room have shown poor reproducibility due to minute differences in microphone placement.

Resources I have found online suggest that averaged sweeps are not done because of the time involved setting up the microphone and running multiple sine sweeps. The favoured approach is MMM.

So my question: should I (and most of us) be changing the way we measure in-room response? If MMM is so superior and so much faster, why aren't more of us doing it? Are there any disadvantages that I have overlooked?

* I tried looking for a source to suppose this claim but I was unable to. I might find it later, but for now you'll have to take my word for it

 

[staticV3]

One advantage of sweep measurements is that you can simultaneously capture frequency response, impulse response, phase response, etc.
However, if all you care about is in-room frequency response, then MMM is better.

With MMM, you also have to be mindful of wind noise and handling noise.

 

[Deleted member 48726]

When I measure with REW I use the moving mic method at and around the listening pos. I don't get the ruler every time I want to listen to music so I want to know what the response is in the _general area_ of the MLP and correct EQ to that.

I've since gone over to using Dirac Live instead and their middle-of-the-road procedure is 13 measurements taken around the MLP. How that data is used is proprietary though.

I tend to agree that a single measurement is close to being meaningless.
You also have discovered it yourself; that the repeatability of single measurements are close to impossible. That "speaks volumes", doesn't it?

 

[ernestcarl]

Nobody needs to use a single method. I use everything. Single point sweep, multi-point, MMM, and quasi-anechoic. If I have the spinorama or family of curves from third party sources -- or even self-made -- I'd use those as well as just another reference.

 

[Speedskater]

My thoughts:
a] the more you equalize transient response, the poorer sustain response becomes.
b] the more you equalize a single listening position response, the poorer room response becomes.

 

[Keith_W]

Yes, I get that you need to do a single point sweep with the microphone centred to obtain information about timing and phase. I do that anyway. What I want to know is the best measurement method for room correction.

I think the first idea that I have ditched is that measurement of a single point in the listening position is inadequate. So thank you for all your responses.

I have never done an MMM before. To me it sounds as if you are literally waving a magic wand and your result appears in much less time than it would take to do a multi-point averaged sweep. I have read that you need to be careful of wind noise and transmitted noise, and make sure you don't wave it around so fast that you introduce doppler shifts into your measurement.

It seems as if multi-point averaged would give me a more consistent and reproducible result. Especially if I was anal about it and construct a jig (basically a board with holes drilled in it) to help position the microphone consistently every time. Of course I would remove the board after positioning the microphone to stop reflections from contaminating the measurement. What I want to know is whether I need to go to these lengths to get these measurements.

 

[3ll3d00d]

If you use the standard approach in acourate (using the mic alignment tool), do you find the resulting filter is materially different each time you measure?

 

[Rja4000]

Dirac uses multiple positions sweep averaged

 

[Keith_W]

If you use the standard approach in acourate (using the mic alignment tool), do you find the resulting filter is materially different each time you measure?

Yes. Even if I don't move the microphone and do two sweeps back to back, there are little variations of 0.5-1dB here and there. The differences are more dramatic if I move the microphone and then replace it and use the mic alignment tool again. I don't really care about tiny differences here and there, I am more concerned about low Q peaks and dips that may be spurious, which I shouldn't be correcting. I normally use a FDW of 12/8, what do you use?

 

[ernestcarl]

... if I was anal about it and construct a jig (basically a board with holes drilled in it) to help position the microphone consistently every time. Of course I would remove the board after positioning the microphone to stop reflections from contaminating the measurement. What I want to know is whether I need to go to these lengths to get these measurements.

Honestly, that sounds unnecessary. Just accept the fact that there's always going to be small differences. Some people who do this for a living seem to favor the "fixed" multi-mic method where they at least have four microphones positioned throughout the listening space.

 

[LTig]

For simple room EQ (like finding EQ settings for a parametric EQ, below Schroder) the MMM method is fast and effective.

 

[ebslo]

TLDR: It doesn't matter (for me anyway, YMMV).

I equalize from single point, but take MMM as sanity check and for final verification. I break it down like this:

• above Schroder: Doesn't matter, should be using anechoic anyway.
• Transition: It's a mess and mostly not minimum phase, so I'm only using very low-Q filters and ignoring the details. For the general trend, MMM and single point agree will enough, so it also doesn't matter.
• Modal (<100Hz or so in my room): The single point and MMM agree almost exactly (presumably because the wavelength is long relative to the area covered by MMM), so it also doesn't matter. I use the single point because it contains the phase info which I can use to see, and sometimes even correct, how the channels sum, and also to check/correct sub alignment.

 

[ebslo]

Maybe I should clarify, it depends how you're doing your EQ.

• If you measure all channels together and apply the same EQ to all channels: use MMM because you don't need any phase info.
• If you measure and EQ each channel individually, use single point. The phase info is necessary to simulate and correct L/R/Sub summation because it will change depending on differences between the EQ on each channel.

 

[3ll3d00d]

Yes. Even if I don't move the microphone and do two sweeps back to back, there are little variations of 0.5-1dB here and there.

that sounds odd, there should be no variation at all if you don't move the mic. For example, you should be able to measure, generate filters, measure with those filters in place and then see measured match the test convolution exactly.

I normally use a FDW of 12/8, what do you use?

I linearise using quasi anechoic data + use MSO to combine multiple subs and then use acourate to bring it all together. For this, I use 9/2 for magnitude and 5/5 for phase atm (though I can't say I can tell the difference between different 2/2 and 5/5).

 

[theREALdotnet]

You can also do the non-moving microphone method, i.e. mic at the listening position while playing ping periodic noise and displaying the RTA on a prominently placed screen. This is very useful while carrying subwoofers around the room, or finding good speaker positions.

REW is very versatile.

 

[Blumlein 88]

Was going to mention that pings or impulses can be used.

I don't know if I posted results of averaged multiple positions vs MMM. On FR they are very close. I don't know how Dirac uses phase in this case. Just listening I didn't hear any sizable difference.

I previously used Tact gear which used multiple pings. Results were sometimes surprising. I think low frequency noise interferes because of the low SNR with pings. Later Tact gear used a full range ping and a filtered low frequency ping to improve that situation.

 

[Keith_W]

TLDR: It doesn't matter (for me anyway, YMMV).

I equalize from single point, but take MMM as sanity check and for final verification. I break it down like this:

• above Schroder: Doesn't matter, should be using anechoic anyway.
• Transition: It's a mess and mostly not minimum phase, so I'm only using very low-Q filters and ignoring the details. For the general trend, MMM and single point agree will enough, so it also doesn't matter.
• Modal (<100Hz or so in my room): The single point and MMM agree almost exactly (presumably because the wavelength is long relative to the area covered by MMM), so it also doesn't matter. I use the single point because it contains the phase info which I can use to see, and sometimes even correct, how the channels sum, and also to check/correct sub alignment.

Maybe I should clarify, it depends how you're doing your EQ.

• If you measure all channels together and apply the same EQ to all channels: use MMM because you don't need any phase info.
• If you measure and EQ each channel individually, use single point. The phase info is necessary to simulate and correct L/R/Sub summation because it will change depending on differences between the EQ on each channel.

Thank you for your valuable response. In the past, I have tried "very tight" control, in that I linearized the drivers with nearfield measurements, time aligned the drivers, at the listening position, then used a MLS with high frequency windowing to acquire a high resolution curve which I used to apply overall room correction to the entire frequency range. The subjective impression of these filters was that they sounded too bright and seemed to have a "smearing" quality to the upper midrange / lower treble. I then repeated the measurement, placing the mic as closely as I could get it (even using a plumb bob duct taped to the ceiling) and it looked to me as if the target curve was as I designed it.

I then decided to forgo correcting anything above 400Hz altogether (the Schroder frequency in my large listening room is probably lower than this) and I was stunned by the results. It sounded so much better.

So what I do is EQ each channel individually, and then do an overall correction for all the channels at the listening position. I will try your suggestion - thank you.

I linearise using quasi anechoic data + use MSO to combine multiple subs and then use acourate to bring it all together. For this, I use 9/2 for magnitude and 5/5 for phase atm (though I can't say I can tell the difference between different 2/2 and 5/5).

By quasi-anechoic data are you referring to the beamforming measurement technique described here? Also, what is the advantage of using MSO over Acourate?

 

[MengW]

AFAIK there are three ways we can measure room response for loudspeakers:

- Maximum Length Sequence (MLS), Exponential Sine Sweep (ESS) or single point sweep: The mic is set up at the listening position, and a single sweep is taken.
- Averaged: The mic is set up at various points around the listening position, multiple sweeps are taken, and the responses are averaged.
- Moving mic measurement (MMM): The mic is set up on a boom, and pink noise is played. The mic is then physically moved around the listening position whilst recording the result.

I have always done single point sweeps. This is as recommended by Uli Bruegemann, author of Acourate*. Then I came across this article advocating use of MMM which made the bold claim that there is no correlation between measured response and listening quality for a MLS. It also includes a quote from Floyd Toole who said "spatial averaging of speaker measurements is critical - single point measurements are erroneous and meaningless". This made me think that I should probably be doing MMM measurements. After all, repeated sweeps of my listening room have shown poor reproducibility due to minute differences in microphone placement.

Resources I have found online suggest that averaged sweeps are not done because of the time involved setting up the microphone and running multiple sine sweeps. The favoured approach is MMM.

So my question: should I (and most of us) be changing the way we measure in-room response? If MMM is so superior and so much faster, why aren't more of us doing it? Are there any disadvantages that I have overlooked?

* I tried looking for a source to suppose this claim but I was unable to. I might find it later, but for now you'll have to take my word for it

Spatial stability of the frequency response estimate and the benefit of spatial averaging
https://assets.ctfassets.net/4zjnzn...mate_and_the_Benefit_of_Spatial_Averaging.pdf


Spatial averaging is assumed to produce a better representation of room acoustics than single point measurements, and thereby be more useful as a reliable starting point for system equalization. However, no significant difference was found between the single point measurement and the spatial average for small spatial average displace- ment measurements taken in professional listening rooms. The spatially averaged responses do not deviate significantly from the single point measure- ment at the listening position for small spatial averaging displacements (± 0.1 m).

fyi.

 

[RayDunzl]

I don't wave my head around while critically listening, so I don't wave the mic around while critically measuring.

Seven measures across the sofa:

Red center, Green left, Black right of center, 1/3 smoothing, both speakers speaking.

 

[kemmler3D]

My general thinking on this is that you should measure (and EQ for) the space where you want EQ to be effective. If MLP is one seat then you don't need to average much space. If MLP is the whole room then do the whole room, or decide which parts of the room you're willing to sacrifice.

I did MMM and averaged sweeps for my couch (about 6' wide) over the past couple days. Big subjective improvement in bass and low mids either way. The measurements were pretty similar from both methods, which gives me confidence.

So far I've spent about an hour on the EQing (not in any rush) and I've already got the measured in-room response down to basically +/- 3dB or better (the whole response lies within 5dB) on the couch from 20hz and up... with only 1 sub! (The mains will do 35hz without much complaint, which helps, and I set the sub xover to 100hz for maximum overlap) I feel pretty fortunate so far, but I will probably discover something ugly if I keep measuring and listening long enough