Vector Avg reducing high frequency SPL

[6speed]

I need help understanding vector averaging or fixing my workflow. What I did was to

1. Take several measurements at various points on my couch.
2. Do not apply windowing or smoothing to individual measurements.
3. Use overlays to align impulse response of individual measurements. This worked beautifully, and I had only one outlier I threw away rather than align manually.
4. Click vector average.
5. Apply variable smoothing to all measurements.
6. Note that vector average (yellow dotted line) plotted a realistic curve of best fit at lower frequencies, but reduced the SPL at higher frequencies.

Questions:

1. Is this an expected result due to ... what? Phase differences between measurements or variations in path length that do not affect longer wavelengths?
2. Did I do something wrong?

 

[John Mulcahy]

Is this an expected result due to ... what? Phase differences between measurements or variations in path length that do not affect longer wavelengths?

Yes, not unusual, for the reason you mentioned. If you look at the phase overlay graph you will probably see the differences.

Did I do something wrong?

Not really, but it is questionable to apply a vector average to measurements made at different positions. For EQ purposes a magnitude average may make more sense, but arguably it doesn't particularly matter since EQ above a few hundred Hz is generally not a good idea anyway.

 

[6speed]

OK, the RMS average produced the SPL curve I was expecting. I had followed a tutorial I had found for attempting to average out enough reflections to make a more accurate room curve, but I never achieved the clean looking step response in the tutorial...which I cannot include a link to because the content filter here thinks it is spam-like. The RMS curve is sufficient for seeing a truer room curve, but the point of cleaning up timing information seems to be to use excess group delay to identify points in the frequency response that should not be EQ'ed because they are not minimum phase. Apparently, I have no points below 100Hz with excess group delay, and no points above 100Hz where the excess group delay is common to any two points on the couch...so the exercise is a waste of time and the vector average is not useful.

 

[6speed]

How about this idea. Attach a piece of string to the center of the bottom of a speaker (or the top by the tweeter), and tie the other end to the mic. Now move the mic around the LP taking measurements at different heights. Should the time of flight be close enough to identical to allow for a reliable vector average? This would be for the purpose of confirm reflections were removed via a cleaned up impulse.

 

[teyhyrh4r]

How about this idea. Attach a piece of string to the center of the bottom of a speaker (or the top by the tweeter), and tie the other end to the mic. Now move the mic around the LP taking measurements at different heights. Should the time of flight be close enough to identical to allow for a reliable vector average? This would be for the purpose of confirm reflections were removed via a cleaned up impulse.

you ever tried this?

 

[6speed]

No, I never tried it. I ended up doing an RTA with the moving mic technique and was able to confirm that all the problems I have 100 - 500Hz are reflections, but it's not like I can do anything about them.