Thanks. I do think it's the periodic noise. Will have to try the 2 channel measurement in ARTA.
However, I did manage to get some very good measurements this morning using other software. White noise was used, but it could have been pink or brown.
The measurements were very repeatable, even tho the mic movements were somewhat random thru a space of about 1 square meter at the listening chair. This did lead to a rather different room EQ - simpler than previously used. The moving mic measurement did show a broad midrange dip that was audible before, but hard to measure.
The technique seems quick and very consistent. At first listen an EQ done on the moving average sounds more natural and well balanced than previous attempts at EQ. Well worth the effort.
However, I did manage to get some very good measurements this morning using other software. White noise was used, but it could have been pink or brown.
The measurements were very repeatable, even tho the mic movements were somewhat random thru a space of about 1 square meter at the listening chair. This did lead to a rather different room EQ - simpler than previously used. The moving mic measurement did show a broad midrange dip that was audible before, but hard to measure.
The technique seems quick and very consistent. At first listen an EQ done on the moving average sounds more natural and well balanced than previous attempts at EQ. Well worth the effort.
That was a real surprise when I first used this technique : repeatability ..... despite the random movements !
And as I explained in my presentation note, MMM is the only method I found that permits quickly to Eq with the same sound (timbre) for left and right (or L/C/R).
Yes, surprising! The inconsistency of my mic movements worried me at first, until it became obvious that they weren't affecting the average. Sure makes that measurements faster and easier. 
I used a 30 sec average. Longer didn't seem to help any. Didn't try shorter.
I used a 30 sec average. Longer didn't seem to help any. Didn't try shorter.
Just a point of clarification. The paper states "Schroeder*4 and Geddes*5 have calculated with different methods". To my knowledge our methods are identical. At least that was what I intended - unless I made a mistake!
I have been using this method for several years. I also was attracted by the speed, repeatability and ease of this method. I get the same response as averaging many different sweeps around the same listening window. It is also repeatable within about ±1 dB 20-20k Hz without regard to the initial placement of the mic. I'm using REW 1/48 octave mode and "forever" averaging. I typically end up with 160-260 averages depending on how fast I move the mic (moderately slow to moderate fast). The response is no different within this range.
The resulting HF (800-20k Hz) listening position response is very similar to a gated HF response at a closer mic position given an adjustment for falloff due to mic distance. My bass range (20-180Hz) is also similar to an individual sweep as the vertical window I use is small (~ 36"W x 18"H) relative to the wavelength so there is not much difference due to mic location anyway.
The largest difference for me is in the mid freq range, maybe from 180-800Hz, where the normal variability of a single measurements with slightly different mic positions is greatly reduced. It avoids running and average multiple sweeps at different mic positions.
I have had good success setting EQ to my favored house curve. I just adjusted my house curve to compliment my favored collection of program material. Other program material would necessitate a different curve however. There is no consistency in source programming. 😡
My question/concern is related to what is being measured. An average over an area around the listening position using RTA methods would maybe an EQ to the power response? I read that power response is not a good way to EQ!
Is this effectively an EQ to the power response?
The resulting HF (800-20k Hz) listening position response is very similar to a gated HF response at a closer mic position given an adjustment for falloff due to mic distance. My bass range (20-180Hz) is also similar to an individual sweep as the vertical window I use is small (~ 36"W x 18"H) relative to the wavelength so there is not much difference due to mic location anyway.
The largest difference for me is in the mid freq range, maybe from 180-800Hz, where the normal variability of a single measurements with slightly different mic positions is greatly reduced. It avoids running and average multiple sweeps at different mic positions.
I have had good success setting EQ to my favored house curve. I just adjusted my house curve to compliment my favored collection of program material. Other program material would necessitate a different curve however. There is no consistency in source programming. 😡
My question/concern is related to what is being measured. An average over an area around the listening position using RTA methods would maybe an EQ to the power response? I read that power response is not a good way to EQ!
Is this effectively an EQ to the power response?
Good question. And "what do we hear as the tonal balance?" would be another.
My findings in the midrange are similar to yours. That's the region that shows the most difference from fixed position in my measurements.
My findings in the midrange are similar to yours. That's the region that shows the most difference from fixed position in my measurements.
Okay, I get that.
It seems to work well in my case. I see a reasonable correlation to sweeping the MF+HF at closer mic distance using the appropriate window, so I guess that, at least in my case, this method is just a good. It's nice also because one measurement works for EQ of the whole frequency range. We just have to be ready to adjust the slope of the house curve to fit the situation and our preference.
I do have 2 or 3 peaks/dips in the mid freqs that are room related. They do not show up at the closer mic distance. For that reason I normally only sparingly address them.
It seems to work well in my case. I see a reasonable correlation to sweeping the MF+HF at closer mic distance using the appropriate window, so I guess that, at least in my case, this method is just a good. It's nice also because one measurement works for EQ of the whole frequency range. We just have to be ready to adjust the slope of the house curve to fit the situation and our preference.
I do have 2 or 3 peaks/dips in the mid freqs that are room related. They do not show up at the closer mic distance. For that reason I normally only sparingly address them.
Unless I misunderstood, in your (recommended) book on page 186, you said "an identical result was shown by Schroeder using an entirely different technique than that used here".
Maybe the methods are identical and both based on Green's function of parallelepidic rooms, but the calculation techniques are different, or did I miss something ?
It is... but the measurement result looks generally closer to "first reflection curve" than to "power response" as per the Toole-Olive team terminology.
I did an online calculator that generally gives a fair basis to each one's target curve.
Why not give it a try : http://www.ohl.to/calculators/targetcurve.php
Interesting calculator. Why can't I enter real values? It only allows a limited few values in the cells.
Not that it matters much. I use a fairly typical drop of 1dB/octave with a hinge point at 800Hz. I've tried other hinge points, but nothing as low as 200Hz
Not that it matters much. I use a fairly typical drop of 1dB/octave with a hinge point at 800Hz. I've tried other hinge points, but nothing as low as 200Hz
My bad. Misread your write up. Yes you were right, the techniques were different. Schroeder used a purely statistical argument to derive the stats and I just simulated a room response and then did the stats on that. I had forgotten that - like everything else these days. What I was trying to say is that what we derived were identical. Schroeder did not have a computer when he did his foundational paper so all he could do was a purely statistical approach. His 50's paper still sits in my mind as one of the most fundamental papers in room acoustics. When I first read it, it was only available in German. I later got the AES to publish this paper in English - done by Prof. Schroeder's own hand. My translation was not bad, but he was a master writer as was everything that he did. This AES paper is must reading IMO.
This paper was targeted only on large rooms, but measurements show its validity also for our small listening rooms, above certain frequencies.
When we are confident about a measuring method, the target response is the main remaining point.
We know that the trend is :
- a downward slope between 0.5 and 1dB/oct at mid frequencies
- a more pronounced slope at highest frequencies to account for air HF attenuation
- a low frequency curve more or less similar to the spectral enveloppe of the natural response of the room.
But this may be another (important) topic.
It is both important and another topic. There was a good thread about it a few years ago.
http://www.diyaudio.com/forums/multi-way/172806-flat-not-correct-stereo-system.html
A new thread with some reasoning and research behind the target curves would be a good idea. I'm simply using the old B&K curve.
To my knowledge the AES is the only place that it is in English. The German version was in Acoustica if I remember correctly.
Interesting technique. My only question is why are you using pink or white noise to measure in room? (unless you listen to a lot of pink noise for fun...) I have only ever used pink/white noise to set XO and levels near-field in quasi-anechoic. For in-room integration noise does not work because it is not music and music does not stick around long enough to excite most of the problems that might be seen using Noise.
I am puzzled
I am puzzled