High frequency low sound pressure level measurement improvement

[Johan.Brans]

Hi all!
I made a frequency plot of a sharp noise at is generated by one of my products. This noise is generated at roughly 5,6kHz. I have the feeling that the fft has difficulties generating a proper spectrum with so little background noise. Is there a way to reduce the thick band of sound pressure levels in the higher frequency region?
The measurement is performed with the following conditions:
MiniDSP UMIK-2
48kHz
mode: spectrum
no smoothing
fft length: 256k
no averaging
window rectangular
max overlap 50%

 

[John Mulcahy]

Looking at the pattern of the noise I'd guess you are on Windows 10 and using the Java driver option. If you install FlexASIO, switch to the ASIO driver option and select the UMIK as the input device in the ASIO control panel launched from the REW soundcard preferences things will clean up a little. The Java driver on Windows is limited to 16 bits and Windows uses a short dither sequence when truncating the word length so the noise floor is raised as a result. It won't help hugely though, best getting the mic closer to the source of the noise to raise its level. If you get it to be the highest peak in the spectrum you could then turn on coherent averaging which will lower the noise floor.