The more information a spectroscopist tries to extract from a spectrum, the more important a flat baseline becomes. Examples of areas where baseline distortion creates unwanted difficulties are:
- Small signals in the presence of larger signals
- Spectral integration
- Difference Spectra
- Water Saturation
- 2D experiments, especially NOESY
In some cases, “baseline distortion” can arise from real NMR signals. Broad signals, which can be wider than the observation sweep width, create “distortion”. Examples of this are probe background, wide-line 2H spectra of solids or 31P surface coil spectra. When these special cases of “real signal” are identified and eliminated from the discussion, there are still many cases where instrumentally induced baseline distortion complicates data interpretation. Examples of sources of these instrumental baseline distortions are:
- Digitizer overload (Clipping)
- Preamp and/or receiver overload
- Audio filters
- Slow recovery of RF stage from overload
- Probe acoustical ringing
These are difficult problems to diagnose and often more than one problem is present at a time. The approach is to isolate each variable as much as possible and repair or replace the offending module or find a work-around solution. An easy one to find and eliminate is clipping of the digitizer. Every spectroscopist knows how to detect and eliminate this one, but let’s use it as an example of a particular type of baseline distortion which arises from “overloading”. Using a simple sample such as Ethyl Benzene while observing protons, when the digitizer is not overloaded, the spectrum has a normal straight baseline. If the signal amplitude is increased to the point where the first few points of the FID are clipped, the signals will appear to sit in a shallow “hole” in the baseline. The more the signal is clipped, the deeper the hole.
Other areas of the spectrometer’s receiver can “clip” or overload besides the digitizer. When there is too much signal for an amplifier or mixer to handle and remain in its linear range, the high amplitude areas of the FID are suppressed relative to the low amplitude areas. This creates an amplifier “clip” which results in a baseline distortion similar to digitizer clipping. Another problem with this type of overload is the creation of “extra” peaks in the spectrum by intermodulation mixing of signals from the probe. For the simple case of two strong signals in the spectrum with this type of overloading, the two signals will mix to yield sum and difference signals as well as the main signals. This can be easy to see with wide sweep widths when the signals are confined to a small region near the center of the spectrum. In other cases, it is more confusing since the signals can “foldback” from outside the spectral window and not appear to be either a sum or a difference. But for this discussion, the main area to notice is the shallow well around the peaks. This type of RF overload occurs most easily when running strong samples such as proteins in water. Very slight RF or digitizer overload has been used to create a very slight “hole” around the residual water signal to reduce the water “tail” in water saturation experiments.
Another type of baseline distortion arises from the distortion of the first few points of an FID. This can come from probe acoustical ringing or, more commonly, filter distortion. If the first few points of the FID are distorted, then during the mathematical processes of Fourier Transform and phase correction, a baseline roll is introduced. When there is one large signal (water) in the spectrum, its small distortion leads to a very curved baseline when examined at the amplitude of most of the interesting spectral peaks.
To minimize this problem, it is best to use a good Butterworth filter set at least 50% wider than the acquired spectrum. A Bessel filter has a better impulse response and produces even less distortion but with it’s more gradual rolloff there is a slight reduction in signal to noise.
Either with the proper audio filter or without, a reduction in the baseline curvature can be obtained by adjusting the time between the end of the pulse and the digitizer trigger (DE in Bruker, GE and Nicolet terms) such that the B (first order) phase correction is near zero. For an FID with a distorted first point, this will reduce the mathematical outcome of phasing to a DC offset of the baseline, which is easier to correct. If the digitizer does sequential, rather than simultaneous, data acquisition for each channel during quadrature detection (some Brukers), the transmitter phase also needs to be adjusted to give an A (zero order) phase correction near zero to achieve a similar result.
Last updated: 01/22/03