REW misses a peak???

[Anthony]

Last night when running my new REW setup (v4, MobilePre soundcard, and ECM8k mic), I found an interesting problem.

I did the usual measurements and eventually tried to apply filters. I should save the image and post it here, but suffice to say, I only had about five peaks in the 20 to 200 Hz range. One was big at around 60 Hz, the rest were pretty small except one around 80 Hz.

Here's the thing: REW didn't seem to see the hump at 80Hz, even though it was considerably bigger than all but one other peak. It suggested filters, I applied them, and it flattened out a lot, but there was that 80 Hz hump.

Couple of things:
This hump was in the middle of the crossover rolloff, so peak is relative to that . . . however,
All but the 60Hz peak were on that rolloff section as well

I just found it odd that it didn't see that one, but picked out smaller ones to flatten.

Anyone else seen this behavior? Other than a manual filter there, is there any fix or way for it force or "re-examine" the peaks?

Just curious,
A

How do you know there is a peak if REW doesn't recognize it?

[Anthony]

You can see it in the response curve. It identifies several peaks automatically, but for some reason glossed over this one. It's on my home laptop, so I will try to post it tonight for clarification.

[JohnM]

REW is fairly fussy about the shapes of response features it identifies as peaks. This is because it is targeting peaks due to modal resonances, as these are correctible using parametric EQ filters. Filters need to be accurately matched to the resonance they are correcting, otherwise they can do more harm than good. If you email me the .mdat file from your measurement I'll be happy to look at it and see whether REW should have identified a peak but didn't. You can manually add additional filters, REW will optimise them to match the response in the same way as filters REW applies itself to peaks it has identified.

[brucek]

REW didn't seem to see the hump at 80Hz, even though it was considerably bigger than all but one other peak.

This is not unusual at all. You do of course have to ensure your target is at a level that will force REW to see the peaks sometime. Either way, if you don't agree or want to modify the recommended filters, just manually modify the recommended ones and/or add your own until the predicted corrected response looks to your liking... This is all part of the fun of REW. I think the best part is playing with the filters in REW until the curve looks exactly how I want it and then to see that once I enter those filters, that the actual result looks like the REW predicted result..

brucek