When integrating a group of closely eluting peaks in a chromatogram, is it more correct to draw a baseline from valley to valley for each peak or to draw one common baseline, extending from the start of a group of signals to the end of the group where the baseline is flat again, and drop a vertical line down for each peak? I've seen it done both ways, but the results can be quite different - especially when an internal standard is not used.

I believe the correct way to integrate the scenario you describe is to use one baseline for all the peaks and drop vertical lines. It may be necessary to preview a blank injection to be sure that the area of concern has no discernable absorbance there, but i think that the results generated are the best.

There are some interesting articles on this common problem, which have involved mathematically generated gaussian peaks of known size, and evaluating the correct integration method as resolution decreases to below 1. Conclusions however do vary in these studies (mainly dependent on relative peak size), but one important observation is that for poorly separated peaks, height is superior to area.

If you want some references I can dig them out.

Agreed that in general dropping perpendiculars is more accuarate however this supposes that the peaks are gaussian. Skimming can be more accurate if the first peak tails substamtially.

Agreed too that dropping perpendiculars is reliable and fast because Gauss or Exponentially modified Gauss convoluton does not make any significant change in area of unresolved Peaks. As our datasystem UniChrom does(www.nas.minsk.by) - using Gauss forms only CPU consumer but not reliability producer.

We use height in this case also. As long as you integrate the same way for standards and unknowns, wouldn't the peaks work either using the perpendicular drop OR the valley-to-valley method? Providing that the unknowns and standards have the same compounds in them of course.

As I recall, the first peak can be accuratley measured with the common baseline, and a perpendicular line drawn down from the peak. Since the second peak is tailing on the first, how can you be sure that the highth is accurate ( isn't it falsely elevated?) Drawing a baseline from valley to valley for the 2nd peak will result in an angled baseline , then dropping a perpendicular to that baseline, should be more accurate.