Subject: Spectrum noise estimation
From: Tom Goddard
Date: Mar 6, 2009

Nikolaos Sgourakis wrote:

Dear Tom,
...

Can sparky give an estimate of the background noise of a spectrum
(perhaps as the rms of the background signal) ? I have noticed the rh
command for the processing of relaxation data produces errors but it
is unclear in the manual how the original noise in the spectrum is
estimated. I am trying to measure noise in hnNOE spectra, so this
information would be very helpful.

Thanks again,
Nik

Hi Nik,

The relaxation fitting error estimate is a bit weird in that it does not
use a noise level estimated directly from the spectrum. Instead it
infers the noise of the spectrum by looking at the deviation of the fit
exponential curve from the spectrum values use in the fit. Thats a
pretty unreliable estimate since few spectrum data points are used -- so
if the exponential fit is coincidentally very good it will say the error
in relaxation constant is low. Basically it is just measuring the
goodness of fit in terms of the relaxation constant.

http://www.cgl.ucsf.edu/home/sparky/manual/extensions.html#RelaxFit

Sparky can estimate noise in a spectrum in two ways although there is
not a way to use this in the relaxation fit tool. The spectrum settings
dialog (st) can randomly sample the whole spectrum and report a median
value:

http://www.cgl.ucsf.edu/home/sparky/manual/views.html#Noise

Or menu entry Extensions / Spectrum / Spectrum Region RMSD (rm) can
report the mean and standard deviation of spectrum data values for a box
you drag:

http://www.cgl.ucsf.edu/home/sparky/manual/extensions.html#RegionRMSD

Tom