Question about noise measure

[FrenchFan]

Hi @John Mulcahy

Perhaps you remember, I'm the person who started the discussion
about cross-correlation under ASR, which you implemented in REW.

I'm restarting my distortion measurement in my DIY tool, integrating a
certain number of BINs around each harmonic frequency.

I'm using REW/RTA as a reference and trying to get as close as
possible to the REW measurements.

I have a problem with the "noise" measurement on a test setup, which
also occurs in REW.

Maybe you'll have an answer.

You'll find the test pattern in the first figure, which also contains the
measurements at different sample counts.

The reference noise of the test pattern is -118.6 dBFS.

We notice at 1M samples that the measurement is no longer viable.

I compared my results with those of REW, which behaves the same way.

Figures 2 and 3 show the comparison between my results and those
of REW at 32k and 1M sample counts.

What reassures me is that our noise calculations must be very close.

The only explanation I have is the decrease in the width of a BIN by
a factor of ~1000 between 32k and 1M, thus a decrease in precision in
float, since the calculation ends with a subtraction of two very close
numbers, but I'm not sure that's the explanation.

I think you've probably already seen the problem and have an explanation.

PS: I'm using REW version 5.40 BETA 50.

 

[John Mulcahy]

I hadn't come across that, but a harmonic at the level of the fundamental is perhaps somewhat unusual . There is an easy fix, start the summation of the N+D values above the fundamental at the end of the data and work backwards.

 

[FrenchFan]

I hadn't come across that, but a harmonic at the level of the fundamental is perhaps somewhat unusual . There is an easy fix, start the summation of the N+D values above the fundamental at the end of the data and work backwards.

It's true that I use a lot of patterns to disrupt my script in relation to REW.

This type of signal is simply too rare to be analyzed.

I'm trying to correct it myself.

Thanks for the reply.