| META TOPICPARENT | name="NmrGroups" |
|---|
Contents
Processing Bruker digitally filtered data Sep 2005
There is 1 message totalling 121 lines in this issue.
Topics of the day:
1. Processing Bruker digitally filtered data
----------------------------------------------------------------------
Date: Fri, 2 Sep 2005 09:45:10 +0100
From: Jeremy Craven <c.j.craven@SHEFFIELD.AC.UK>
Subject: Re: Processing Bruker digitally filtered data
This is something I would like to understand better too.
>
>Practically, the group delay has to be dealt with in order to give
>sensible data. There are two schools of though as to how to deal with
>it which should in the ideal case be equivalent. Key to these methods
>is the understanding that a 1st order phase correction in the frequency
>domain is equivalent to a time shift in the time domain. Hence, to get
>rid of the group delay you should either be able to left shift the data
>the correct number of points (N) and then FT, or FT and then apply the
>appropriate 1st order phase correction (360*N degrees).
>
>
>
Actually there is a subtle but important difference between whether you
do a left shift and discard the points that fall off the left edge (and
zero the right hand few points), or instead do a circular shift. A
circular shift of the time domain gives a mathematically identical
result to applying a first order phase correction (from the definition
of the discrete Fourier Transform). A first order phase correction means
that a different phase correction is applied to every POINT of the FID.
(((In raw mathematical terms, phase correction means replacing the real
and imaginary parts of the spectrum by the combinations of the original
real and imaginary parts appropriately multiplied by sin and cos of the
phase correction angle))).
The (non circular) left shift gives a subtly different effect in that
(at least for simple exponentially decaying signals) approximates well
to applying a different phase correction to each resonance LINE (ie take
each component resonance lineshape in turn; apply a zero order phase
correction to each one (but by a different phase angle for each line),
and then add up the result).
These two things are not quite equivalent, and (especially if there are
large signals like water around) mean that attempting to correct a
timing error by applying a first order phase correction can give a
wiggly baseline.
I talked about this in detail in a paper once (The action of time-domain
convolution filters for solvent suppression, Craven CJ, Waltho JP,
JOURNAL OF MAGNETIC RESONANCE SERIES B, 1995, Vol.106, No.1, pp.40-46)
which has some diagrams that might make this more clear.
>Having experimented with left shifting data (either explicitly or using
>AZARA's AVANCE command) and with applying the appropriate 1st order
>phase correction after FFT, I always find better results (flatter
>baselines with "frowns" only at the very edge) with the phase
>correction method. My personal explanation for this is that the group
>delay points must encode something about the shape of the
>baseline/filter function which is critical.
>
>
>
>The only implication of using the phasing method rather than left
>shifting is that manipulations of the FID become more complex. Window
>functions are usually trivially affected, but convolution difference to
>suppress solvent signals is impossible on the unshifted FID. The ideal
>way to deal with this would be to apply window functions etc with a -ve
>time component through the group delay, but this is practically
>difficult and depends absolutely on knowing the sampling regime (sweep
>width). It is simple enough to process the data as normal with the 1st
>order phase correction and then use an inverse FT-convolution
>difference-FT stage to apply the solvent suppression.
>
>
>
We also do it this way, but I have never felt quite happy about it. I
have never felt quite sure about when to apply the window function for
optimal results (ie before the dft, or before the plain ft)
The Bruker software can do convolution difference(BCMOD=qfil or qpolI
think ), so it must do something here. Also there is the CONVDTA
(CNVDATA?) command which converts "digitally filtered data" to "normal".
Does this just do the dft-ift thing too?
I know Igor Barsukov once told me some useful things about all this but
I have forgotten them. I will try to get him to contribute something.
>PPS it would also be nice to have the ability to define an arbitrary
>pivot point for baseline corrections in AZARA as they are often at
>least pseudo-symmetrical about the centre of the spectrum rather than
>the LHS.
>
>
>
I agree.
>PPPS should I have cross-posted to ANSIG-AZARA?
>
>
I don't know, but it is a useful sort of topic for theis general
discussion list as it goes beyond just AZARA.
Jeremy
--
*********************************************************************************
Dr C. Jeremy Craven
Department of Molecular Biology and Biotechnology
University of Sheffield,
Firth Court, Western Bank
S10 2TN Sheffield UK
e-mail: c.j.craven@shef.ac.uk http://www.shef.ac.uk/uni/projects/nmr/CJC/CJC.html
Phone: x24323
From outside Sheffield: 0114 222 4323
From outside UK: +44 114 2224323
Fax: 0114 272 2800
*********************************************************************************
------------------------------
End of NMRGEN Digest - 26 Aug 2005 to 2 Sep 2005 (#2005-2)
**********************************************************
| 