925-456-1020 info@acornnmr.com

Segmental baseline correction

BLC – Segmental Baseline Correction

Some baseline distortions are not simply smooth curves, and so are not corrected well using a fit to a polynomial. The baseline can often be corrected using a series of polynomial corrections applied to sub-sections of the spectrum. The first step of the process is to decide which parts of the spectrum (or expanded zoom region) contain just noise, and which contain peaks. Next, the spectrum is divided into regions that can be baseline corrected using a polynomial. Each region is fit, and the calculated baseline subtracted. For regions containing peaks, a DC offset and linear tilt correction is applied, based on averaging points at the ends of the adjacent regions.

The process of determining the difference between noise and peaks requires an estimate of the noise level in the particular spectrum. NUTS automatically determines the rms noise of the spectrum. The spectrum (or expanded region) is broken into segments, and a determination made as to whether or not each segment contains any peaks. A segment of the spectrum is determined to contain a peak if its (maximum – minimum) value exceeds a defined multiple of the rms noise. By default, it will use as that multiplier the RM value (same as is used in peak-picking), but the user can specify a different value. By default, the spectrum is divided into 64 segments, but the user can specify how many points each segment should contain.

The command can take the following command line arguments:

pts – sets the number of points for each segment

mult – sets the RMS multiplier used to distinguish peaks from noise



help – displays a message explaining command syntax and allowed arguments

With no command line arguments, the blc command divides the displayed region of the spectrum into segments containing the specified number of points, and examines each segment to determine which contain just noise and which contain peaks. The decision criterion is presence of data points exceeding mult times the RMS noise. A larger number will cause more segments to be identified as noise, including segments containing small peaks. A smaller number will cause fewer segments to be identified as noise, and segments containing only noise may be treated as peaks. The segments determined to be noise are marked with red bars. Individual segments can be selected or deselected by clicking on the segment with the left mouse button. The first and last segments must be selected.

When all appropriate noise segments have been selected, there will be regions consisting of a series of adjacent selected segments, separated by one or more segments that are not selected. When the fit is executed, a polynomial fit is performed on each region composed of adjacent selected regions. Each region consisting of adjacent un-selected segments is corrected using DC and linear tilt. This is also done for single, isolated un-selected regions. In other words, by selecting sections of baseline carefully, the spectrum is divided into regions each of which is amenable to correction using a polynomial.


A – calculate the baseline and subtract it from the spectrum

M – change the RMS mutliplier

P – change the number of points in each segment

S – select all segments

Z – unselect all segments

If this is performed on a zoom region, rather than the entire spectrum, a DC correction is applied to the parts of the spectrum outside the zoom region, to prevent creation of discontinuties.

The example below illustrates use of this feature.


The hump centered around 10 ppm cannot be corrected well using a 5th order polynomial. The results of an attempt are shown below.

Careful examination of the shape of the hump suggests that breaking that part of the spectrum into 2 regions, each of which has lower-order distortion, might work better. In the figure below, a single segment has been de-selected, indicated by the arrow.

The resulting corrected baseline is shown below.

See also: polynomial baseline correction

Last updated: 9/5/07