BUG or STRANGE problem applying correction

[Andrea Tarasconi]

Hello,
I am using REW 5.20.5 to correct my system room response. I have a 2-way stereo active solution with cross-over at 60Hz (18" sub...).
As player I am using JRiver MC28 and Foobar 1.6.10, as cross-over I am using Thuneau Allocator and as convolver the MeldaProduction MConvolverZ.

First I measured the response of the system alone, then I generated the impulse filter and finally I measured again the response of the system with the correction.

What I detected is that the response after the correction (convolver applied) was having big problem over 1kHz, so I simply decided to disable all the EQ filters above 1kHz and so I go finally a decent response.
However, looking at the measured responses, simulations and final results, it is evident that there are too many deviations. I think there is a bug or at least a strange measuring problem.

Index of measurements in the attached files (containing also the embedded EQ filters generated and settings):
* measures 1, 2 and 3 => just 3 repetition of the same full system response at the golden spot (crossover only with phase alignment between sub and sat). NO Convolver.
* measure 4 => time alignment and averaging of the above 3 measurements to be used to generate the room EQ filters.
* measure 5 => import of the above made filter impulse response (just to give it a look)
* measures 6 and 7 => 2 times repeated measure of the full system response at the golden spot with the applied EQ filter / Convolver => To be noted the bad / over-busted response from 1kHz to 7kHz
* measure 8 => import of the manually modified filter impulse response where I removed the EQ filters 6,7,8,9 - All the filters generated above 1kHz (just to give it a look)
* measure 9 => finally obtained good response using the #8 filter for convolver... However, even if it is looking nice and flat, it is not the response I have planned as target in my design (I set a 5dB roll-off at 20kHz that is not there)... This is really a flat-flat response.

Anyone can help me understanding what's going on?
Thanks and regards, Andrea

 

[John Mulcahy]

It looks like you have both speakers playing in your measurements 1-3. If so, that is a bad idea, measure speakers individually. There is also something strange going on, since the measurements show odd distortion artefacts. Perhaps the ASIO buffer setting should be larger.

 

[Andrea Tarasconi]

Hello John, thanks for your prompt reply, but probably I do not have well understood your reply.
I mean:
1) All measurements, not only 1-3, have been done with both L-R speakers concurrently (I also measured each one individually and practically no difference at the golden spot).
2) I never moved the microphone (on a stable tripod), so if the problem would be in the placement, I should have seen it also in the other measurements.
3) The buffer or the ASIO Driver of Behringer is a very "critical" component... I can try to maximize it just for the measurement sessions but...
4) How can I check the "odd distortion artefacts" you mentioned? I am not an expert user of REW... is there an option to see/check that? How you detected it?
I confirm that "listening during the measurements" I heard some problem of distortion but being -25dBFs (at Behringer internal levels) I didn't care about ... even because not being able to understand exactly what was going wrong (which frequencies were having distortion problems).

Again, thanks and regards, Andrea

 

[John Mulcahy]

One problem with playing both speakers at the same time is that the mic is not exactly the same distance from each. That causes comb filtering, but also produces two peaks in the measurement of similar levels which can cause the timing estimate to jump from one to the other. That is visible in the timing difference between the first and second measurements. It is your system and of course you can follow whatever process works for you, but I would personally never use a measurement of both speakers in that way and would expect it to cause problems.

Click on the Distortion graph to see the distortion. I suspect from the previous measurements that there may be dropouts in replay or capture or both, hence the suggestion to use a larger buffer.

 

[Andrea Tarasconi]

Hello John, thank you so much for your kind and clear reply.
Now I understand. I will follow you suggestions on my next measurement session.
Kind regards, Andrea