In this article, John Dolan suggests that 5X the vendor specification is good for noise and drift when executing a performance qualification. Unfortunately, the article doesn't explain why this number is ok. Does anybody know why?? Is it to take into account external factors such as interference from other systems connected to the same power supply, age of system optics, temperatuer and humidity of the area the detector is located??? Please give me some idea why this value is allowable.

It probably comes from a wealth of experience.

I don't like to use the vendor spec for noise because there is always the pressure to have better specs than your competition. Some vendors spec the noise with a dry cell, some with no flow, etc.

Most of the vendors now have their own OQ/PQ protocols that they will allow you to review. Somewhere in these documents will be a limit for noise and drift. These are generally more useful because they have to pass these limits under more realistic conditions.

Experience is great, but when you have to defend it to the FDA on where these specs came from, that is a different game entirely.

I have always used the limits from the vendor's qualification protocol and have not had any adverse comments from the FDA. If you are rolling your own PQ protocol you could always attach a copy of the vendors protocol and have an easy reference as to where the limits came from.

Conversely, you could come at it from the opposite direction. Take your method that places the most strigent requirements on S/N, give it a little safety margin, and determine the maximum amount of noise the method could tollerate. Call this value your PQ spec.

Either way you could now justify your noise limit. Good Luck