Dear Rainer
(Given the absence of any replies so far, I’ll copy this to the whole list.)
With exponential weighting things are relatively straightforward: in both Topspin and VnmrJ the parameter lb/LB is the width in Hz of the Lorentzian broadening produced by the weighting, so the time-domain weighting function is exp(-π lb t). (Note that there is an error of a factor of 2 in the description of the command em in the Topspin documentation).
(There is a further slight problem that in the Topspin versions I have seen the implementation is not quite correct - because of the way Bruker do digital signal processing, a group delay is introduced that is not accounted for in exponential weighting. This only matters if LB is a significant fraction of the spectral width in Hz, but with high LB values, > SWH/25, peaks break up into wiggles.)
There are two ways in which Gaussian weighting is commonly used: a simple unshifted Gaussian to apodise a truncated FID (e.g. in 2D NMR), and a time-shifted Gaussian (equivalent to an exponential multiplied by an unshifted Gaussian) for Lorentz-Gauss resolution enhancement.
In VnmrJ these are both simple to implement. The parameter lb determines the time constant of the exponential, as before, and and this is multiplied by a Gaussian of time constant gf. Thus the overall weighting function is exp(-π lb t) exp(-t**2/gf**2). To apodise a truncated FID use gf=at/2; for Lorentz-Gauss resolution enhancement, to convert a Lorentzian line of width lw into a Gaussian line into a Gaussian of width gw use lb = –lw, and gf=0.533/gw.
In Topspin things are more complicated, both for apodisation and for resolution enhancement. The weighting function is proportional to exp[-π t/LB – π LB t**2/(2 GB AQ)] (because the first point of the weighting function is not unity, absolute integrals will change depending on the parameters, whereas in VnmrJ they remain unchanged).
The parameter gb (0 < gb < 1) is the fraction of the acquisition time AQ at which the total weighting function reaches a maximum. This means that it is not possible to produce a perfect Gaussian window function for apodisation purposes because this would require gb=0 (this is why Bruker 2D processing tends to use a lot of weird and wonderful window functions like phase sifted sine bell squared (aka cosine squared)). A simple way around this problem is to set LB to a very small negative value (e.g. -0.01) and use GB = -π LB gf**2/AQ**2, where AQ is the acquisition time and gf is the Gaussian time constant required (usually AQ/2).
For Lorentz-Gauss resolution enhancement, to convert a Lorentzian line of width lw to a Gaussian line of width gw set LB to minus lw (not half lw, as in the Topspin documentation) and set GB to 0.4413 lw/(gw**2 AQ).
I tend to use little au programmes to set the Gaussian parameters - let me know if you would like copies.
I hope this helps!
Best wishes
Gareth Morris
On 6 Apr 2017, at 08:47, Wechselberger, Rainer [JRDBE] <RWECHSEL_at_ITS.JNJ.COM> wrote:
> Hi there,
> I’m not sure if this was discussed before, but couldn’t find anything of it…
> Is there anybody who has a recipe to translate Bruker window functions into Varian ones and vice versa?
> We are working on a general method and like to understand how we can translate Bruker lb and gm parameters to the respective Varian parameters and the other way round. This turned out to be not really straight forward. A lot has been written about window functions in general but I thought maybe somebody went through this exercise already and can save us some time? We don’t have any Varian here, so that makes that a bit difficult…
> Thanks a lot!
> Rainer
>
> Rainer Wechselberger
> Janssen Pharmaceutica
> Beerse, Belgium
=============================================
Prof Gareth A Morris FRS
School of Chemistry, University of Manchester
Oxford Road, Manchester M13 9PL, UK
Tel (0) 161 275 4665
g.a.morris_at_manchester.ac.uk
=============================================
Received on Fri Apr 07 2017 - 04:38:07 MST