NOESY spectra are particularly susceptible to bad baselines because the information of interest is contained in very small peaks. Baseline correction during processing can improve the results, but care to acquire data with flat baselines to begin with reduces the need for correction. Parameters to adjust are opening the sweep width far enough to avoid distortions due to filter cut-off, and adjustment of the pre-acquisition delay to minimize linear phase correction and baseline curvature.
The 3 spectra below are from the same NOESY data set (3.5 mg of codeine in CDCl3) which had substantial baseline curvature. The spectra were processed identically except for baseline correction.
The spectrum above was processed with no correction of baselines.
A BC was performed after FT in the direct dimension, removing DC offset and linear tilt.
A polynomial baseline correction was performed after FT in the direct dimension gives a slightly better result.
The vertical streaks are at the position of the 2 intense methyl singlets, commonly referred to as "t1-noise" (where t1 refers to the indirect dimension). (The above spectra are viewed with the direct dimension running horizontally.) Viewing individual slices, it is apparent that the noise level in these slices is 3-5x greater than in the rest of the spectrum, and some of the slices have severely curved baselines. The appearance of the spectrum can be improved by selectively dividing each of the offending slices by a constant, combined with careful polynomial correction of specific slices, as needed.
This is shown in more detail here.
Last updated: 01/22/2003
Acorn
NMR Inc.
7670 Las Positas Rd 
Livermore, CA 94551
(925)456-1020 FAX
(925)456-1024