Processing digitally filtered data (RD)
An extra processing step is required before FT for digitally filtered data (Bruker data from the Avance series of spectrometers such as DRX and ARX, Tecmag NT-NMR data and some JEOL data). This is necessary because the data have been digitally filtered and "decimated" prior to saving the FID. The initial points of the FID are zero. A circular left shift must be performed before the FT. The number of points to shift is calculated from parameters found in the file header (the Decimation Number and the DSP firmware version found in the Bruker file header), but the user can change the number of points. (In a circular left shift, points are removed from the left end of the data and added to the right end.)
** Note: NUTS attempts to perform the RD operation automatically when FT is executed. NUTS will use its calculated value for number of points to rotate. The value can be changed by typing RD. There are 2 ways to avoid having NUTS do automatically apply RD: Type RD and enter 0 for points to rotate, or use the command CT (complex FT) or BT (Bruker FT) instead of FT. The choice of CT or BT depends on whether your data was acquired using simultaneous or sequential acquisition.
This FID is typical of digitally filtered data. Note the odd appearance of the beginning of the FID.
The FID above has been expanded to show the first 400 points, making it easier to see the beginning of the FID.
If an FT is performed on the above FID, a seriously distorted spectrum results. The "wiggles" can be removed with very large linear phase correction (many 100s of degrees) which is not very convenient.
The first 70 points at the beginning of the FID need to be removed, which is done as a circular left shift, so that the points removed from the beginning of the FID are added onto the end. This is performed with the NUTS command RD.
When RD is typed, this screen is displayed. NUTS determines the number of points to shift (Number of Rotations) from the Decimation Number and Firmware Version, which are imported with the data.
If NUTS has been able to determine a default value from parameters in the data header, a non-zero value will be displayed in the Number of Rotations box. Usually the user should just click on OK. If NUTS has not been able to determine the correct number, this value will be zero, and the user must input the appropriate number. This can be determined by examining the beginning of the FID.
The user can experiment with different values for Number of Rotations. After FT and phasing, look at the value for linear phase correction (TP command). When the correct value is used, the linear correction should be fairly small (less than 360 degrees).
The resulting FID now has its maximum at the beginning of the FID. Note the non-zero points at the end of the FID.
FT of the FID after RD gives a reasonable looking spectrum. In some cases, as here, there is some roll-off at the ends of the spectrum. The normal BC command (to remove DC offset and tilt) does not work properly in this case. To compensate, either expand the spectrum so that the curved ends are not displayed and apply baseline correction, or use the BR command to ignore the ends of the spectrum while applying DC and tilt correction.
Any apodization should be applied before RD is executed. Below is a comparison of spectra resulting from exponential multiplication applied after (top) RD and before (bottom) RD. Note that the tailing off of the baseline is worse when the apodization is applied after RD. Retaining the non-zero points at the end of the FID is important for minimizing baseline distortion.
If the acquisition time was short enough that the FID has not decayed to zero, then doing an RD can result in a discontinuity, and the resulting spectrum will be distorted. (See description of zero-filling for an illustration.) Applying a window function before the RD operation, to bring the end of the FID smoothly to zero, will avoid this.
For Bruker data, NUTS determines how many points should be shifted by examining the parameters DSPFVS and DECIM found in the acqus file. Allowed DSPFVS values are 10, 11 and 12. DECIM can have any of several values. Milo Westler and coworkers at the University of Wisconsin generated a look-up table for the correct number of points to rotate, based on these 2 parameters. If either of these values is zero, NUTS concludes this is not digitally filtered data, and so zero points should be rotated.
However, sometimes the information found in the data header is incorrect or insufficient for NUTS to determine the correct number of points, so NUTS has an alternative approach. The DF command examines the beginning of the FID to determine where the "real" start of the data should be, and how may zeroed points precede it. It executes an FT, then attempts to undo the large linear phase correction by applying linear phase of -360 degrees for every zeroed point. This will not work properly for all data, so users are advised to experiment.
RD can be used for other purposes than just to correct digitally filtered data. See example.
Last updated: 3/12/04.