When evaluating force-degraded samples for peak purity, we used a non-degraded control sample as our standard. Almost every time we try to generate peak purity numbers for the control sample, we get a failing peak purity number, although we are almost always certain there are no co-eluters. Then begins the tedious process of trying to adjust the peak purity parameters so that the control will pass. Once we get parameters that give a passing peak purity value for the control, we use these same parameters for the force-degraded samples. Often it seems the only way to get the control to pass is to set the noise threshold with a user value instead of allowing the software to calculate it. I dislike doing this but it is often the only way to generate results. I have turned to HP representatives for assistance many times with this issue and they always answer some other question instead of the one I am asking. Why can't I rely on the software to calculate the noise threshold and get meaningful results? I have 7 years experience in method development/validation and have seen this happen over and over again with a wide variety of products and mobile phases. Does every one have this problem, or am I missing something?

P.S. I am more confused than ever since they went to the new software.

NOOOOOOO!!! you're not alone. From the bottom of the world (Argentina) we are sharing the problem. The wonderful algorithms of Peak Purity never have worked even, as you are saying, the sample is a reference, very pure (as we tried after a DSC and GC-MS checking) substance. They have never answered our prayers.
Use manual settings
If there are anybody who knows something about this very interesting topic please, people, don't forget to write!!

Yes, I also have faced the same problem many a times.The peak purity of same sample passes in Waters (Millenium software).

I've had this problem as well, on the HP1100. I got the impression that the peak purity calculation was very subjective.

I did finally get the thing to do the job for me, but I was never really satisfied with it. As I recall, some of what it took to get it to behave was to limit the region over which the peak purity was evaluated. The logic of this is that you cannot expect the noisy regions at the front and tail of a peak to calculate as "pure," when compared to the relatively noisless regions at the peak.

As I (dimly) recall, I managed to set the parameters such that no spectrum would be included in the calculation that was from a region of the peak less than 10% (???) of the full peak height. I'm sorry that my memory of this is too dim to instruct you better. If you cannot figure out how to do this, email me privately and I'll try to help more.

By the way, it may be worth considering using other means than this subjective calculation to do peak purity determinations. For one thing, if the spectrum of the contaminent is very similar to that of the analyte, then the peak will be calculated as pure even if highly contaminated.

LC-MS is one such approach, but even chromatography itself can often be used. Try optimizing separation conditions (low flow, heated column, multiwavelength detection) and then varying chromatographic conditions (column temperature, %B, identity of organic component, etc.) to move the peaks around in a controlled manner. You may well find that you move a little minor component out from under the major component in this way.