AREA under curve approximation

[jitender]

Which algorithm is used to approximate the area under the curve (peak area) in various proprietary chromatographic data acquistion and manipulation softwares?
Rectangle Rule/Trapezoid rule/Simpson rule/Gauss-Legendre or any other.

What is the criteria of defining small time interval in rectangle/trapezoidal method. Is it defined by the frequency of data acquistion by detector or some interpolation is done.


May I have some good reference discussion about this?

[DR]

PE & Waters both have white papers on their web sites that discuss this (I'm pretty sure).

Most of the differences (and discussion) will be about how they determine if/when there's a peak to integrate. Once that's done, I think most systems just sum rectangles, but the papers talk about that too.

[amaryl]

Check even LCGC articles they are rich source of chromatography.

Regards

Amaryl

[MikeD]

I doubt that the actual peak start/end algorithms are ever published. They are proprietary - but I could be wrong.

I have written my own for applications involving peaks with extreme asymmetry lasting up to 2 hours. My peak start algorithm worked most of the time but it was not as easy as it first appeared. Peak stop never worked very well. They were based on threshold above s/n ratio for a set of 6 area slices incrementing one at a time. Between peak start and peak end it's just adding rectangles. I am sure the commercial algorithms are much cleverer than my amateur efforts.

[unmgvar]

No software company will give the exact algorith used, but they will explain how it mathematically works.

in basic it always evolves around the change in the angle between two baseline points. i know that Chromeleon uses a different apporoach and that Empower is capable of doing it not by checking from the start or end of the peak but also from the apex. don't know if they use it already but they have been talking about it.

[tom jupille]

jitender, to answer your original question, if you have a sufficiently large number of data points, there is only a trivial difference between the results of the different algorithms you cite. I believe that they essentially add the voltage values.