REW Beta Release Strange measurements in latest beta

MagnusH

New Member
Thread Starter
Joined
Oct 2, 2017
Messages
7
I have used REW for several years, mostly for room correction. But it seems something is broken in the last versions, 5.20 beta 8:
1. I took a moving-mic RTA measurement, made correction and then measured the correction and got -6dB in the 5khz area and a wide area around it. Needless to say, it did not sound good.
2. There is a very noticeable change in frequency response for the different frequency response outputs (the ones you get when you save impulse response). 44khz sounds normal and 192kHz sound very muffled.
3. Not sure about this, but I can't get sweep measurements/corrections to sound good. Might be related to above.

And here is a feature request: in Roon (which I use as player), Low Shelf and High Shelf has a Q, but not in REW, which makes it tricky to experiment and compare. Could LS/HS get a Q in REW as well (I assume it has one, but fixed at a default value).
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
7,212
The shelf filters are defined in terms of a shelf slope parameter, per the RBJ cookbook. It is 0.5 for LS/HS 6 dB, 1.0 for LS/HS 12 dB and 0.9 for LS/HS.

Can you post your measurement including its correction filters? Have you tried importing the corrections back into REW and looking at their responses?
 

MagnusH

New Member
Thread Starter
Joined
Oct 2, 2017
Messages
7
I found the problem, its the higher sample rates that gets wrong. I did a correction from a measurement, exported if (Export filter impulse to wav) to both 44khz and 192khz. Then I imported them and looked at them

44Khz (sound correct):
upload_2019-4-20_11-32-49.png


192khz (sounds muffled):
upload_2019-4-20_11-33-6.png
 

Attachments

  • upload_2019-4-20_11-31-35.png
    upload_2019-4-20_11-31-35.png
    54.4 KB · Views: 4

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
7,212
Do you have a notch filter just below 20 kHz? Filters close to the sample rate limit will have altered shapes, so the 44 kHz sampled version of a filter that high in frequency will be misshapen and the 192 kHz closer to the shape of the analog equivalent filter. In any case, a deep notch up there is unlikely to be a good thing.
 

MagnusH

New Member
Thread Starter
Joined
Oct 2, 2017
Messages
7
Don't remember and now its gone, but did a similar correction and got a similar result (but not quite as bad). Only auto-correct from REW, and REW put a -9.6dB filter at 19897, see below for a screenshot. So its probably what you said, but with regular filters instead of notch.

As you can see, my measurement shows an increase at a very high frequency, which is what REW tries to correct.

I tried a -10dB HS filter instead at 19khz, but then the 192khz became about +6dB higher all over the frequency (the shape was also a little different but not so bad).

Btw, if I want my correction to look exactly like in the EQ window, which frequency should I output to?
upload_2019-4-20_16-33-28.png
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
7,212
The Generic EQ setting is at 48k. I strongly recommend you do not apply any automatic EQ above about 200 Hz. The higher the frequency the less likely the results will be beneficial. If you want to apply filters for tone control better applying them manually and keeping the corrections small and broad.
 

MagnusH

New Member
Thread Starter
Joined
Oct 2, 2017
Messages
7
Cheers, I usually don't do automatic correction, I like to do it manually and trying to use as few filters as possible, and as you say keep the Q down. Just never noticed this behavior before.

I think I will only generate 48khz impulse response and then use that for all sample rates (I use Roon, and it down/upsamples the filter if needed). Seems that is better than risking filters deviating from the view in the window.
 
Top Bottom