Advanced

A statistical analysis of NMR spectrometer noise.

Grage, Halfdan LU and Akke, Mikael LU (2003) In Journal of Magnetic Resonance 162(1). p.176-188
Abstract
Estimation of NMR spectral parameters, using e.g. maximum likelihood methods, is commonly based on the assumption of white complex Gaussian noise in the signal obtained by quadrature detection. Here we present a statistical analysis with the purpose of discussing and testing the validity of this fundamental assumption. Theoretical expressions are derived for the correlation structure of the noise under various conditions, showing that in general the noise in the sampled signal is not strictly white, even if the thermal noise in the receiver steps prior to digitisation can be characterised as white Gaussian noise. It is shown that the noise correlation properties depend on the ratio between the sampling frequency and the filter cut-off... (More)
Estimation of NMR spectral parameters, using e.g. maximum likelihood methods, is commonly based on the assumption of white complex Gaussian noise in the signal obtained by quadrature detection. Here we present a statistical analysis with the purpose of discussing and testing the validity of this fundamental assumption. Theoretical expressions are derived for the correlation structure of the noise under various conditions, showing that in general the noise in the sampled signal is not strictly white, even if the thermal noise in the receiver steps prior to digitisation can be characterised as white Gaussian noise. It is shown that the noise correlation properties depend on the ratio between the sampling frequency and the filter cut-off frequency, as well as the filter characteristics. The theoretical analysis identifies conditions that are expected to yield non-white noise in the sampled signal. Extensive statistical characterisation of experimental noise confirms the theoretical predictions. The statistical methods outlined here are also useful for residual analysis in connection with validation of the model and the parameter estimates. (Less)
Please use this url to cite or link to this publication:
author
organization
publishing date
type
Contribution to journal
publication status
published
subject
in
Journal of Magnetic Resonance
volume
162
issue
1
pages
176 - 188
publisher
Elsevier
external identifiers
  • wos:000183372400015
  • pmid:12762994
  • scopus:0043211081
ISSN
1096-0856
DOI
10.1016/S1090-7807(03)00038-7
language
English
LU publication?
yes
id
4f3b0fcd-c077-4d23-bcec-6382e000388b (old id 128046)
date added to LUP
2007-06-29 10:36:50
date last changed
2018-01-07 09:06:39
@article{4f3b0fcd-c077-4d23-bcec-6382e000388b,
  abstract     = {Estimation of NMR spectral parameters, using e.g. maximum likelihood methods, is commonly based on the assumption of white complex Gaussian noise in the signal obtained by quadrature detection. Here we present a statistical analysis with the purpose of discussing and testing the validity of this fundamental assumption. Theoretical expressions are derived for the correlation structure of the noise under various conditions, showing that in general the noise in the sampled signal is not strictly white, even if the thermal noise in the receiver steps prior to digitisation can be characterised as white Gaussian noise. It is shown that the noise correlation properties depend on the ratio between the sampling frequency and the filter cut-off frequency, as well as the filter characteristics. The theoretical analysis identifies conditions that are expected to yield non-white noise in the sampled signal. Extensive statistical characterisation of experimental noise confirms the theoretical predictions. The statistical methods outlined here are also useful for residual analysis in connection with validation of the model and the parameter estimates.},
  author       = {Grage, Halfdan and Akke, Mikael},
  issn         = {1096-0856},
  language     = {eng},
  number       = {1},
  pages        = {176--188},
  publisher    = {Elsevier},
  series       = {Journal of Magnetic Resonance},
  title        = {A statistical analysis of NMR spectrometer noise.},
  url          = {http://dx.doi.org/10.1016/S1090-7807(03)00038-7},
  volume       = {162},
  year         = {2003},
}