Strange discrepancy between REW FFT and Audition FFT

[staticV3]

When generating sine tones in REW, exporting as WAV, then analyzing in both Audition and REW, the results are identical:


However, when I generate and export the tone in Audition, then run the FFT with both programs, it looks like this:


Where do these spurious tones come from?

Here's the 1k sine generated in Audition: https://drive.google.com/file/d/1URGuLmOSRjpnrjTbqPlzg8jjiU2eZxq7/view?usp=share_link
(@JohnPM )

 

[KSTR]

@staticV3 , looks like something is going terrible wrong here. Both Adobe Audition (3.0 in my case) and REW generate 100% clean sine and show them as clean sine in the FFT in all cases including export and import, assuming reasonable settings.
EDIT: The file you provided seems to have been corrupted during export, I can re-create the plot you're seeing in REW. REW's FFT is flawless, so it's the file.

 

[staticV3]

Both Adobe Audition (3.0 in my case) and REW generate 100% clean sine and show them as clean sine in the FFT in all cases

Thank you for checking it on your side.
Would you mind creating a 100% clean, 24bit sine wave in both REW and Audition and posting them here?

 

[KSTR]

Will post when I'm on my Windoze machine again.
Audition running in WINE under Linux has some issues (cannot change any parameters for the FFT, drop-down lists close immediately all by themselves :-(

 

[pma]

Would you mind creating a 100% clean, 24bit sine wave in both REW and Audition and posting them here?

Below is the link to a "100% clean, 24bit sine wave" in AA:

https://pmacura.cz/AA_997_24.zip

 

[staticV3]

@pma now on a 2012 iMac instead of my PC, just in case that makes a difference.

REW FFT (64bit processing on):

Mine has less noise and a "tighter" fundamental, but both have the same spurious tones as above.

Audition FFT:

Same as in REW, just without the spurious tones.

So unless I'm mistaken, I'm back where I started. I doubt the files are corrupted, I doubt both my PC and iMac are broken somehow.
Looks like something's wrong with REW's 64bit processing of WAVs.

Edit:
Here's REW without 64bit FFT:

both ugly

 

[JohnPM]

Sorry about that, the code that converts audio files is a bit long in the tooth and hadn't kept up with the other conversion code. Here's how it should look, fixed for the next build.

 

[staticV3]

Thank you, John!

 

[KSTR]

Thank you for checking it on your side.
Would you mind creating a 100% clean, 24bit sine wave in both REW and Audition and posting them here?

Ah, I see the issue now. 24bit export from AA is not perfect (I get the same results with @pma's file, with those little spuriae). I tend to use 32bit float format for exports/imports. 32bit fixed point PCM is also fine with AA, even without dithering.
FFT in AA is fine as long as you haven't saved the file in 24bit because it is operating on the original 32bit float data.
(I promised plots but alas, just right in time for NYE the power brick of my windoze laptop went up in smoke last night, have to wait until a replacement supply is ready).

REW does a better job in creating a dithered 24bit output, 100% clean.

FWIW, sox also doesn't cut it when dithering a 32bit float or 32biut fixed point file to 24bits, same spuriae in the spectrum.

 

[staticV3]

Ah, I see the issue now. 24bit export from AA is not perfect (I get the same results with @pma's file, with those little spuriae)

See JohnPM's response above. The 24bit files from AA are fine, AA's FFT is fine and shows the truth, but REW's FFT is currently bugged, introducing those tones that are not in the file.

FFT in AA is fine as long as you haven't saved the file in 24bit because it is operating on the original 24bit float data.

For me on my two PCs, AA's FFT is fine even if I export the files in 24bit, reimport them into AA, then run the FFT on them.
Not sure what your problem is there. Could you elaborate?

 

[KSTR]

See JohnPM's response above. The 24bit files from AA are fine, AA's FFT is fine and shows the truth, but REW's FFT is currently bugged.

Don't think so, as REW is fine when importing a 24bit dithered file it generated before, so the file exporter/importer appears to be OK.
I'm on REW V5.20.14ea15.

 

[staticV3]

Don't think so

I'm on REW V5.20.14ea15.

That's a beta version of REW.

JohnPM (the developer of REW) has already confirmed here that the stable build of REW is currently bugged, resulting in those spurious tones, so not sure why we're still arguing.

 

[KSTR]

^ We're not arguing. I just had assumed everyone is on the latest Beta...

 

[JohnPM]

Ah, I see the issue now. 24bit export from AA is not perfect (I get the same results with @pma's file, with those little spuriae).

The files are fine, the conversion artefacts are side effects of different choices of normalisation factor (I'd speculate) exposing a need to apply float dither to the converted data from imported files. The code that does that isn't released yet, should get it onto the early access thread next week some time.