Anomalies in the -100 to -130db region at different times and settings ? ? ?

[mikelm345]

Hi, when using REW to test my DIY Class A mono block power amps I have noticed some anomolies.

I tested both amps in last few days and at 1.5 watts only 2nd & 3rd Harmonics were present. All other harmonics were below the noise floor in the -125 to -130db region.

However, when testing the same amps, unchanged, today with similar settings I am seeing various harmonics like 7th of 9th ( most undesirable ) at around -100db

I tried many different settings to no effect until I accidently changed the frequency to 1001hz. At this frequency all higher harmonics fell back down into the noise floor.

After trying various frequencies like 500 1k 2k 3k 4k 5k these harmonics like 7 & 9 are once again at a higher level . . .

BUT when I tried changing the frequencies by one digit either up or down i.e. 1999hz or 2001hz, the higher harmonics consistantly fell back into the noise floor.

Logic would suggest that this is some kind of a digit anomaly. I cannot imagine my amp's HD signatures are changing so much with a 1 Hz difference.

So my question is this: Are the rounded frequencies, 1k 2k etc somehow engineered to be more accurate or . . . .
Could there be some kind of digital anomaly being created with these exact frequencies which is not present with the slightly offset frequencies ?

I would really like to know which measurements are most accurate

BTW I did try 5001 5002 5003 5004 5005Hz and at all these values the higher harmonics were consistantly within in the noise floor.

Many thanks

mike mears

 

[John Mulcahy]

A few questions:
Which REW version?
Which audio driver? (ASIO, Java EXCL, Java)
What bit depth is the interface running at? (bottom left of REW main window)
What sample rate?
What dither depth has been set on the generator?

 

[mikelm345]

Which REW version? 5.20.13

Which audio driver? (ASIO, Java EXCL, Java) Java ( Soundcard is a USB external - Asus Xonar U7 )

What bit depth is the interface running at? (bottom left of REW main window) 24 bit in 16 bit out

What sample rate? Tried 48 & 96kb/s but this did not appear impact this issue

What dither depth has been set on the generator? dither option is not checked

I confess that my knowledge of the digital universe is limited

Many thanks for your support

 

[John Mulcahy]

Digital signals need to be dithered to prevent quantisation causing artefacts in the spectrum. The dither should be set to the bit depth of the signal, so a 16-bit output should use 16-bit dither.

The worst signals for generating artefacts are those with a small number of repeating samples. A 1 kHz signal at 48 kHz sample rate has only 48 unique sample values, a small change in frequency produces a signal with many more different sample values which introduces a degree of self-dithering, but dither is the best solution.

With the Java drivers on Windows choose device names that start with EXCL, that will use a WASAPI Exclusive driver and provide the full bit depth the interface supports.

 

[mikelm345]

OK, did that, and now all seems well again.

Thanks again