Manual deconvolution of Gaussian peaks

If you have a spectrum that consists of the addition of two Gaussian peaks, you can try to manually perform a deconvolution of this peak with Spectrum Viewer by using DragPlot in combination with LiveCalc. Let's look at the following example, taken from the file DualGauss.svf in the TestSpectra folder. Make sure there are no files loaded in SV, then open the DualGauss.svf file

The displayed spectrum is an addition of two perfect Gaussians. We will try to reconstruct them. First, we will create two Gaussian curves in data set 2 and 3 that we think make up the final plot.

File g Generate signal g Gauss curve

Give for the first curve Center 105, Peak value 80 and Standard Deviation 20. Guess the second curve at 95, 85 and 15. Now add both Gaussian plots to see the result. 


It looks like one of the peaks should be considerably smaller, the sum of 2 and 3 goes through the roof! The slope of plot 2 matches the original very well, so we will try to drag plot 3 in such a way that the sum of 2 and 3 addition (plot 4) will match plot 1 more closely.

Delete plot 4, then start up LiveCalc. LiveCalc is a continuous performed calculation. The calculation takes one or two source plots and puts the result into a new plot, the destination plot.. Every time something changes in one of the source plots (for example when it is being dragged), the destination plot changes with it.

We will now use LiveCalc to do a continuous addition of plots 2 and 3 into plot 4. Additional to that we will use the second calculation to create a difference plot between the addition result and the original plot. Fill in the table like the above picture, make sure "Update during drag" is not checked and click "Apply". This will result in the following screen:

We will now try to drag plot 3 so that the addition (plot 4) will match the original. When the addition result matches the original plot very well, the difference plot (plot 5) should be almost zero: a horizontal line. Therefore it is best to put the difference plot (plot 5) into the right Y-axis.

Now select plot 3 only, and start up Drag Plot. Try to drag plot 3 in such a way, that the residual plot (plot 5) approaches zero. The best way to achieve this is to use the following drag modes:

The purple plot (plot 5) is the residual error in the right Y axis

After dragging you can now see which two plots (2 and 3) approximately make up the original plot.