Export, Import and Compare

[nc535]

I'm exporting measurements from REW for FIR equalization and getting questionable results. This has led me to import my exported measurements back into REW and compare. Surprisingly, I find that they don't match.
When exporting measurement as text, then importing the export, I find small differences, not really enough to worry about but something I'd like to understand. The imported IR is very worrisome. The picture changes when one applies an FRD to the curves. What causes these differences? Are the differences in the exported files or are they an artifact of the import process?

Here is a picture after export and import. Its taken in room with delayed reflections. The purple is the imported IR, the light green is imported FRD, and is mostly on top of the dark green original measurement. The purple is the issue. If the FIR tool sees that purple curve instead of the green, it isn't likely to know what to do, speaking colloquially

after importing I applied an FDW5 to all 3 curves and then removed it. Now the picture is less frightening but differences remain:

All of this leads me to ask myself what the heck is going on. I wonder if I should be applying a gate to remove the really long delayed echos or apply and remove that FDW5 before exporting to DRC.
I attached zipped mdat curve in case you want to look at it.
Thanks in advance for any help!

 

[John Mulcahy]

You are not comparing like with like. The measurement had 1/48 octave smoothing, a soundcard cal file and a mic cal file whereas the imported IR had no smoothing, a different t=0 position, different window widths, different window types and no cal files. Here they are when they have the same settings, green is the imported IR and red is the measurement, though it can't really be seen as it is identical.

Things to bear in mind:

• Impulse responses do not include calibration files, the cal files need to be loaded
• Measurements being compared should use the same window types, the same window widths and have the same t=0 position and window reference position
• It is a bad idea to have a left window of zero, even if you have placed the t=0 point at the start of the impulse (which was not the case for the imported IR which had t=0 at the peak). Even then it provides no benefit but risks truncating something that should be included in the response

 

[nc535]

That explains it although I don't recall setting any smoothing on the measurement. I have run into the cal file thing before but it keeps slipping my mind.

Assuming I have accurate cal files and for both mic and sound card, so that the response after corrected per cal file is indeed correct, is there a way to export such that the corrected response can be viewed on another machine or in another program that doesn't have those cal files available?

 

[John Mulcahy]

Text exports of the frequency response includes the cal file effects and mdat files of measurements include the cal files data.

 

[Wesayso]

That explains it although I don't recall setting any smoothing on the measurement. I have run into the cal file thing before but it keeps slipping my mind.

Assuming I have accurate cal files and for both mic and sound card, so that the response after corrected per cal file is indeed correct, is there a way to export such that the corrected response can be viewed on another machine or in another program that doesn't have those cal files available?

As soon as you have used a frequency dependent window and go back to normal viewing, the measurement will show with a 1/48 smoothing, you can even see it in the legend below the graphs. If you unsmooth that graph, the only difference would be the CAL files.

 

[nc535]

that explains the 1/48. since I have an FDW applied by default

thanks