Subject: noise and SD
From: Stefano
Date: Jun 11, 2010



Hello,
I am trying to use the rh routine to calculate the relaxation parameters. The problem I have is how Sparky estimates the noise and therefore the SD. If I select one peak and I calculate the relaxation parameter, the SD changes randomly upon repeating the calculation over and over. The changes are quite large, even by 4-5 time. What is the best way to estimate the SD then? I understand that the SD depends on the noise, and that the noise is calculated as the mediam of a certain number - to be set - of randomly sampled absolute data heights.
What is the best number of points to use?
Do these randomly selected points possibly include regions of the spectrum where the peaks are, or the peaks are excluded?
What is the best protocol to use to avoid the overestimate of the SD?
Stefano