REW Beta Release Bug report: RTA with files, incorrect display when sample count == FFT size

KSTR

Member
Thread Starter
Joined
Jun 7, 2019
Messages
12
[Issue is fixed]

When using the RTA with files (via drag&drop) it I need to make the file one sample longer than the FFT size otherwise no display (overlap=0) or wrong display (overlap=50% or more).
So with eg a 65536 point FFT the file needs to contain at least 65537 samples.

Tested with V5.20 beta 11, using a sine on an exact bin freq, no windowing, where it is easy to see if the spectrum is right.
 
Last edited:

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
2,032
Not how I'd anticipated it being used, but I've fixed that for the next beta. :)
 

KSTR

Member
Thread Starter
Joined
Jun 7, 2019
Messages
12
Great, thanks for that. My use case is block-averaged data (to bring down the noise and uncorrelated stuff) with the block length typically being a FFT size.
 

KSTR

Member
Thread Starter
Joined
Jun 7, 2019
Messages
12
Another small glitch I seem to have found:
When the sample rate of the input file doesn't match the global settings (in "preferences") the distortion meter doesn't seem to work ("distortion N/A").
Normally not a problem with typical sample rates (just set the global rate accordingly) but it is when using a higher rate like 384kHz for the file input and that frequency is not available from the list, working offline. Also, I sometimes set the SR of a file to the FFT size so that "frequency" values actually are bin numbers.
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
2,032
I fixed a few small bugs in the RTA file processing over the last week or so, including some that affect the final values for the distortion display. Best checking again when beta 12 is released. In the meantime if you have any test files I can try I'll give them a go, either attach them here or email them to me at feedback@roomeqwizard.com
 

KSTR

Member
Thread Starter
Joined
Jun 7, 2019
Messages
12
I did a more thorough test now, and it looks like distortion meter does always work with any sample rate (standard or arbitrary) when the "show distortion" button is clicked *before* loading the file. If I load the file first then click it, it always shows "N/A", regardless of sample rate. I always used a full restart of REW to have a clean base first.
Then I did check turning the button off after one succesful load of a file (with button engaged before), load another file, or the same file (doesn't matter), turn the button on again and voila, distortion display works.

So it has nothing to with the sample rate, rather it seems to be a small issue with initializing the distortion meter.

Testfiles (1M FFT with a H2 component, padded with the additional needed sample at the end) attached anyway, same content but with 48000Hz and 56789Hz sample rates in the headers. It's actually a .7z (much better compression than .zip) but I had to rename it to .zip to attach it, can be opened/extracted with 7zip with no problem, though not with Windows built-in zip tool.
 

Attachments

Last edited:

KSTR

Member
Thread Starter
Joined
Jun 7, 2019
Messages
12
Ah, but yet another glitch: when I load a file into freshly started REW with arbitrary SR the frequencies displayed are correct. But when I then load a file with a different SR, the frequency scale isn't updated accordingly. Even when I unload the first file before loading the other. Switching the global SR prior to loading a file seems to fix it.
 
Last edited:

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
2,032
Ah, but yet another glitch: when I load a file into freshly started REW with arbitrary SR the frequencies displayed are correct. But when I then load a file with a different SR, the frequency scale isn't updated accordingly.
Fixed that a couple of days ago. It also affects switching to live data of a different sample rate.
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
2,032
Here is the result of processing those in beta 12, let me know if anything looks incorrect.

48000.jpg

56789.jpg
 

KSTR

Member
Thread Starter
Joined
Jun 7, 2019
Messages
12
Yep John, this looks perfect, and identical to what I get when tinkering with Beta 11. Case closed, also for the original issue.
Big Thanks, Klaus

PS.: In closing, I did a small experiment with the distortion analyzer which demontrates how well you implemented this and how much effort you have put in an extremely compact and precise screen output full of accurate and relevant information. Absolutely top level!
Shown is signal from previous post, but with sample rate set this time to 1048576 to match FFT (frequency==bin# and binwidth=1Hz) and an agressor frequency added which is 60dB louder than the harmonic and only 3 bins apart.
The analyser shows a perfect and clean display for this critical signal, the HD detector has a catch range of no more than +-2 bins (but no less either, well-chosen), and the agressor perfectly reflected as noise component only. Talking -200dB noise floor. I'm impressed...
close_spuriae_rejection.gif
 
Last edited:
Top Bottom