Help with REW RTA measurement using the MMM method

[Ata]

Greetings,

I have recently acquired a UMIK-1 and started learning how to use REW for measurement and correction. First I did the swipe based measurement in a single position, using the excellent guide at

and the measurement curve made a lot of sense, in terms of levels and slopes.

Then I learned about MMM method and RTA and measured the same speakers in the same position, with the following result for L and R channel:

The questions I need help with:

- Why is the level so low? I calibrated the pink noice to 80dbA using REW before the measurements were taken, but I get an average of 43dbA SPL...
- Why is the slope in high frequencies so large, basically -25dbA at 20KHz compared to 1KHz? Subjectively the sound is good and open, and I know about in-room response, and house curves, yet the slope is very extreme in the measurement. Furthermore, using the swipe measurement the speaker measures almost flat 1-20KHz (+- 3db);
- Why do I see SPL output at 20Hz that dwarves my SPL output at 20KHz. These are floorstanders (VAF DC7s) but I know they have little in them under 35Hz. Again, in the swipe measurement I see what I expect, a 24db/octave loss of SPL after 40ish Hz.

Other than this, I am pleased with the EQ result, it sounds much better than without EQ, but the bass is a bit on the low side -- I suspect I need to redo my house curve for this style of measurement (same house curve was perfect for the swipe based measurement):

But herein is another question: Why no EQ above about 1-2 KHz (the rest of the difference is due to the blue line being L+R speaker measurement).

Thanks for your responses!

 

[fluid]

I have recently acquired a UMIK-1 and started learning how to use REW for measurement and correction.

The questions I need help with:

- Why is the level so low? I calibrated the pink noice to 80dbA using REW before the measurements were taken, but I get an average of 43dbA SPL...
- Why is the slope in high frequencies so large, basically -25dbA at 20KHz compared to 1KHz? Subjectively the sound is good and open, and I know about in-room response, and house curves, yet the slope is very extreme in the measurement. Furthermore, using the swipe measurement the speaker measures almost flat 1-20KHz (+- 3db);
- Why do I see SPL output at 20Hz that dwarves my SPL output at 20KHz. These are floorstanders (VAF DC7s) but I know they have little in them under 35Hz. Again, in the swipe measurement I see what I expect, a 24db/octave loss of SPL after 40ish Hz.

Did you load the calibration file for your mic? The UMIK is already calibrated for level when using REW so you should not have to calibrate the input level like you would with an analogue mic. To set the output level did you do something like step 5 in this page?

https://www.minidsp.com/applications/acoustic-measurements/umik-1-setup-with-rew

If you follow that and keep the level of the signal generator and sweep the same then the levels you record should be consistent.

The slope could depend on how you oriented the mic and the calibration file you used. There are two 0 degree and 90 degree.
I prefer to measure with the mic pointing up and using the 90 degree file when doing in room measurements as it is more consistent and easier to reposition the mic correctly.

Because the level is so low you are going to be close to the noise floor of the room, that is why it seems like you have more bass than you do. If you get the measurement up to 80dB you will get a better picture of the natural roll off, although the room itself will support the very low bass increasing output over the anechoic response.

 

[Χ Ξ Σ]

"Pink noise has energy that falls 3 dB with each doubling of frequency. On a spectrum plot it is a line that falls at that 3 dB per octave rate, on an RTA plot it is a horizontal line as the energy in the signal is falling at the same rate as the bins are widening. We perceive pink noise as having a uniform distribution of energy with frequency."
from https://www.roomeqwizard.com/help/help_en-GB/html/spectrum.html

If you want the MMM measurement to resemble the sine sweep without fiddling the settings, the simplest fix is to use white noise instead.

 

[Χ Ξ Σ]

As for the low level, I have noticed that in my MMM results as well. I don’t know the math behind it, but I suppose this is just a result of RTA averaging. I didn’t worry about it because my ear knew that the signal level was way above the noise floor regardless how the software presented it.

 

[fluid]

If you want the MMM measurement to resemble the sine sweep without fiddling the settings, the simplest fix is to use white noise instead.

That isn't what makes the difference. Pink Noise will give a flat spectrum when viewed on a logarithmic scale which is one of the reasons it is used for analysis. Pink noise and a sine sweep will give the same result if the microphone position doesn't move.

When you move the microphone it works as a spatial average, the more you move it the bigger the area that is averaged. Spatial averaging can reduce the amount of room that appears in the measurement, it gets averaged out. Vector averaging multiple sine sweeps gives a similar result.

This paper from Jean Luc Ohl is a good read on MMM

https://www.ohl.to/audio/downloads/MMM-moving-mic-measurement.pdf

The Vector Averaging works much like beamforming, paper attached below.

Here's an example of the difference between a bunch of separate sine sweeps and them being averaged together, the level of the average is lower but by a few to 5 dB and not across the whole frequency range.

The Energy Time Curve shows what has happened

 

[Matias]

I have had the same observation when trying MMM for the first times (low level) which I somewhat fixed by playing pink noise louder, but in my case the treble response was as expected. So I guess it is correct, both mine and yours. And I do EQ all spectrum, never agreed of limiting to just below 1 kHz transition frequency.

 

[thewas]

I have no big deviations larger than a dB when measuring with MMM vs a central single sweep when using var smoothing as recommended for EQ

(have offsetted the right channel ones by -20 dB to be better visible)

larger differences happen when multiple sweeps are averaged depending on the averaging math used.

 

[DJBonoBobo]

As for the lower level, there´s a setting for that:

 

[Presently42]

Here's a link to an asr thread that well described how to use the mmm technique. It covers the extreme slope you're noting (hint: use the correct settings)

 

[Ata]

"Pink noise has energy that falls 3 dB with each doubling of frequency. On a spectrum plot it is a line that falls at that 3 dB per octave rate, on an RTA plot it is a horizontal line as the energy in the signal is falling at the same rate as the bins are widening. We perceive pink noise as having a uniform distribution of energy with frequency."
from https://www.roomeqwizard.com/help/help_en-GB/html/spectrum.html

If you want the MMM measurement to resemble the sine sweep without fiddling the settings, the simplest fix is to use white noise instead.

Bingo! This was it, once I switched over using the RTA rather than spectrum analysis both the slope past 1 KHz and bass response under 40Hz became what they should be:

I could get the levels back where they should be by selecting "Adjust RTA levels" in the RTA Settings dialog.

Thanks a lot for your help!