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

[Blumlein 88]

Just another data point. Here is a measurement of some LSR 305 monitors. One speaker in this case from my left stereo position at the time. I think 11 feet to the MLP.
A grid of sweeps. 6 sweeps on a grid one meter wide and one meter deep, 3 front and 3 rear centered around where my head would be. The bright red line is the average of them. You can see it gets pretty tight above 1 khz and above 2 khz the different positions are pretty much within a +/- 1 db range. More difference lower in frequency. 1/12th octave smoothing. No correction of the speaker the measuring was with a Umik 1.

Same measurements with 1/3 octave smoothing and no average shown as it really isn't needed to see the result.

 

[Blumlein 88]

Here is another data point. This is Revel F12's red is a sweep taken center of where my ears would be and ten inches (1/4 meter) left and right of that and averaged. Green is an MMM result using pink noise with the mic being moved in a figure 8 pattern 20 inches wide. The two big humps are intentional. I moved the speakers to find something like that. I wanted to know if MMM was better or worse vs sweeps at finding such a thing.

And the same with 1/3 octave smoothing.

 

[thewas]

My experience is similar https://www.audiosciencereview.com/...ement-using-the-mmm-method.24131/#post-815429 , but still prefer using MMM as it doesn't take much time and can be smoother at the high frequencies which I only use as a reference for the level "stitching" of my low frequencies correction as I correct the higher frequencies based on anechoic or windowed listening window measurements.

 

[Balle Clorin]

I agree with Toole, at least for upper frequencies . I compared 27 measurements in A cube around my head an averaged them , then it more resembled my moving mic response, especially above 1k. I do single point also but avoid aggressive correction above 500-1000hz, only 1/3 octave to avoid correcting a peak at one ear that does not exist at the other…
I use Trinnov.

 

[3ll3d00d]

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

No I mean measuring them like https://kimmosaunisto.net/Software/VituixCAD/VituixCAD_Measurement_REW.pdf and linearising based on this data. I designed/built my own speakers though so had this data anyway.

MSO creates filters and delays based on measurements from multiple positions. Acourate has no such capability. I have a couple of seats to equalise for in a single row so equalising at LF based on a single position makes no sense really.

 

[Keith_W]

Before you ask, I haven't started using MMM or MSO yet. As per suggestions in this thread, I am going to use a new strategy for room correction.

Below the Schroder frequency, I will use Multi-Sub Optimizer (MSO). In the transitional zone, I will use MMM (Moving Microphone Measurement). Above the transitional zone, I will use the anechoic response of the speaker.

Since I don't have access to an anechoic chamber, I will use the quasi-anechoic measurement using beamforming method. I laid down a tape measure down the center between the two speakers. I then started at 0cm and took repeated sine wave sweeps every 10cm until I covered +1.5m and -1.5m from the MLP. Then all the responses are averaged. The idea is that direct on-axis sound from the speakers are correlated, and room reflections are uncorrelated. It took me almost 3 hours yesterday to perform 30 sweeps. Each time I moved the microphone I had to make sure it was centered between the two speakers, that the microphone wasn't clipping (obviously volume gets louder the closer you get to the speaker), etc. ... it took me a long time.

Here is a sum of five curves into one. 25 more to go. At this point I am really hating Acourate's limitation of only allowing you to load 6 curves at a time.

Finished summing all 30 sine wave sweeps. The red line is a single measurement which is noticeably noisier than the averaged left channel (in cyan) and right (in black). The response is really jagged because room correction has not yet been applied. There is also no smoothing. Yes, I know there is a channel mismatch between left and right tweeters. Each tweeter is active and has an individual volume pot which is unstepped. I will have to go and fix it later.

Next step is to calculate the Schroder frequency. The formula is Fs = 2000 * sqrt (T30/V).
- Fs = Schroder frequency
- T30 = time for impulse to decay by 30dB. From my measurements, it is 0.31 sec.
- V = Volume of the room in m^3. My room is 7m * 5.8m * 2.7m = 109.62m^3.

Therefore predicted Schroder frequency is 106.36Hz. The end of the transitional zone is 4Fs, or 425.44Hz.

(Pardon me for borrowing @mitchco's labels in his article on room correction posted on another forum).

The calculated prediction matches very nicely with the measurement. Ideally the modal region should be fixed by bass traps or Helmholtz resonators tuned to predicted room modes and then refined by DSP. Since I don't have adequate bass traps, I am stuck with using DSP only for the time being to smooth out bass response.

Next step is to figure out how to do two things I have never done before - use MSO and MMM. Homework for tonight is to get MSO up and running.

 

[3ll3d00d]

Since I don't have access to an anechoic chamber, I will use the quasi-anechoic measurement using beamforming method.

Painful but full marks for effort

I would run it through macro 1 and compare to the same on your aggregate measure

At this point I am really hating Acourate's limitation of only allowing you to load 6 curves at a time.

Next up, and given you intend to use MSO, enjoying the limitations in acourate's iir filter generator

 

[3ll3d00d]

Did you get mso going? It occurred to me that how to integrate that into acourate might not be obvious so I can post to explain how if required.

 

[Keith_W]

Right now I am still stuck at Step 1. My problem is that it needs a timing reference, and on my system, REW outputs a start chirp but not an end chirp. It probably has something to do with the fact that I need to send REW's sweep through my convolver.

To import it into Acourate ... my plan was to export the files from MSO into EqualizerAPO text format. Then manually enter all the values into Acourate's IIR filter. I have to do them one at a time, maximum of 5 in one go. Then sum them all together and normalize volume. Is there an easier way than this? If you have a better way, I am all ears!

 

[ernestcarl]

Right now I am still stuck at Step 1. My problem is that it needs a timing reference, and on my system, REW outputs a start chirp but not an end chirp. It probably has something to do with the fact that I need to send REW's sweep through my convolver.

To import it into Acourate ... my plan was to export the files from MSO into EqualizerAPO text format. Then manually enter all the values into Acourate's IIR filter. I have to do them one at a time, maximum of 5 in one go. Then sum them all together and normalize volume. Is there an easier way than this? If you have a better way, I am all ears!

Maybe try using a recorded sweep with the reference timing instead? That’s what I do while playing said files in JRiver.

 

[ahofer]

I did both and found very little difference between averaged point measurements and MMM. I came to the conclusion that MMM is sufficient for generating EQ.

Edit: I'm not trying to integrate subwoofers, btw.

 

[Keith_W]

Maybe try using a recorded sweep with the reference timing instead? That’s what I do while playing said files in JRiver.

I tried it tonight (incidentally, only thought of it a few hours before you posted your reply). It worked - thank you for your suggestion

 

[3ll3d00d]

Then sum them all together and normalize volume.

Why would you normalise volume?

I was more thinking of combining that into your sub xo when playing to multiple subs concurrently

 

[Berwhale]

I'm lazy; the 27 measurement positions (9 measurements in 3 vertical planes) required for IKMultimedia's ARC3 system drove me to MMM where I have stayed with my desktop setup

I recently bought a new AVR (Denon X1700h) and got hammered with what I think was another 11 measurement positions

 

[Keith_W]

Why would you normalise volume?

Because I have to level match the sub channel to the rest of the system. Maybe a poor choice of words on my behalf.

I was more thinking of combining that into your sub xo when playing to multiple subs concurrently

I was thinking I would do that, too. Another possibility would be to convolve the result into the room response.

 

[3ll3d00d]

Because I have to level match the sub channel to the rest of the system

MSO exists to design filters for a specific channel source so to use MSO on acourate, you apply the filters via the xo. Once this is done, everything else is as normal. You just have to be v careful to preserve any suggested delays as you change those filters.

 

[Keith_W]

MSO exists to design filters for a specific channel source so to use MSO on acourate, you apply the filters via the xo. Once this is done, everything else is as normal. You just have to be v careful to preserve any suggested delays as you change those filters.

One thing that confused me about the output from MSO is that it recommended an overall sub delay correction, rather than delays for each individual band. Does this sound right to you?

 

[3ll3d00d]

One thing that confused me about the output from MSO is that it recommended an overall sub delay correction, rather than delays for each individual band. Does this sound right to you?

It will vary the values for everything you let it change so if you let it do that, it will surely do it.

 

[RosalieTheDog]

This is probably a ridiculous question, but is there any way to start and stop the RTA measurement with a delay? I can't quite reach the mouse from the MLP. A longer cable is obviously another solution, but perhaps I'm missing some setting.

 

[Ron Texas]

I use MMM. It's an efficient way to get an average of many measurements.