Spatial averaging of measurements for room EQ

[Daverz]

I'm using MMM (moving microphone method) described here. In REW you do it as RTA measurement with pink noise.

Here is an example how it looks with my left speaker:

This was taken from the point at LP where app my left ear is. Red line is a sine sweep and green line is RTA pink noise. As you can see they are practically identical.




This is RTA where I moved mike in a horizontal circle with a diameter of 20cm (so around the area where my head would be):




This is RTA where I moved mike in a horizontal circle with diameter of 1m (so around the area where my sofa is):





This is a single point sweep taken app 5cm behind the first sweep (shown on first pic). As you can see it differs from first sweep quite a bit.



This is sofa spatial sweep vs left ear sweep, for comparison:




With average of 9 sweeps I was getting pretty close measurement to MMM RTA but I consider MMM RTA to better represent spatial average as it has 80+ samples instead of 9. It is also easier and faster for me to make it.

Thanks for this description. Could you further describe your RTA settings and pink noise generator settings?

What are the 80+ samples? Does this refer to FFT block cycles? How long are the cycles?

In the paper they say

"So with long time scanning, FFT analysis and rms averaging, the method combines temporal and spatial averaging. But note that the averaging has to be done on the whole scan duration, not a gliding average, with some constraints on calculation."

I take that to mean no overlap of FFT blocks and no windowing, which makes sense as you don't want measurements "backing up" over eachother, and "Forever" averaging.

In that case, it seems to me you'd get better resolution by simply doing one continuous recording and then post-processing with multi-taper methods to reduce bin leakage. Then import the frequency response. But perhaps leakage is not very important in this method.

 

[Krunok]

These are my RTA settings:

Pink noise generator settings:

I have shown RTA pink noise vs sine sweep in my first chart so you can see they are practically identical. To take that RTA measurement I recorded pink noise with RTA for app 15 sec (same as sweep duration) so app 30 samples were taken with the mic fixed at the same position where sweep was taken.

When doing measurements I move my mic at a steady slow speed (app 5cm/sec, not more) and I keep it pointed at the speaker. I'm using Daytonaudio iMM-6.

 

[Krunok]

What are the 80+ samples? Does this refer to FFT block cycles? How long are the cycles?

When I said "samples" I was referring to what REW is calling "averages". REW takes app2 samples (averages) per second.

 

[Krunok]

Regarding measurement in time domain I'm using sine sweeps. As my speakers are of such design that one of the woofers is on the front panel above the tweeter while the other is on top of the speaker firing upward. Both are working up to 1800Hz. That means a lot of reflections are generated so I have to take sine sweep close to the speaker (20-50cm) in order to get meaningfull phase and step response.

 

[ernestcarl]

I'm using MMM (moving microphone method) described here. In REW you do it as RTA measurement with pink noise.

Here is an example how it looks with my left speaker:

This was taken from the point at LP where app my left ear is. Red line is a sine sweep and green line is RTA pink noise. As you can see they are practically identical.

This is RTA where I moved mike in a horizontal circle with a diameter of 20cm (so around the area where my head would be):

This is RTA where I moved mike in a horizontal circle with diameter of 1m (so around the area where my sofa is):

This is a single point sweep taken app 5cm behind the first sweep (shown on first pic). As you can see it differs from first sweep quite a bit.

This is sofa spatial sweep vs left ear sweep, for comparison:

With average of 9 sweeps I was getting pretty close measurement to MMM RTA but I consider MMM RTA to better represent spatial average as it has 80+ samples instead of 9. It is also easier and faster for me to make it.

I've recently watched a video of this method of spatial averaging: Room Calibration at Echo Bar Studios

I was quite impressed. Not being that good with REW, I had no idea it could be done so easily before.

I guess the simpler method I used needs to be updated/supplemented.

A few peaking PEQs (at 88Hz & 250Hz) were added from my previous -- already minimalist enough -- EQ settings. Reduced the bass volume by a few dBs as well -- I just prefer it linear for critical listening and seems to even lower distortion in the sub bass as well:

13 measurements in front left speaker

13 in front right

Left & Right spatially averaged

I am quite leery of adding any more EQ above 300Hz. Though Mr. Sprinkle seems to think that it is fine up to around 700Hz.

*Hmmmn... Doing some preliminary tests with REW's RTA, I find the FR slope with the pink noise generator looks much, much steeper. I guess that shouldn't really matter?

 

[Xombul]

Is there any material risk of damaging the tweeters as mentionned in REW?

 

[QMuse]

Is there any material risk of damaging the tweeters as mentionned in REW?

Only when doing stepped sine distortion measurements at high levels. Measurement sweeps and RTA pink noise MMM measurements are best done at your "normal" listening levels.

 

[Xombul]

Ok thank you