Gain Instabilities in Photomultipliers: How Accurate are Photon Counting Measurements?

Experiments performed on five commercially available photomultiplier tubes indicate that gain instabilities can be an important source of error in photon counting measurements at the 1% level. It is shown that the error cannot be significantly reduced by standard differential measurement techniques. Analysis of time variations in the pulse height distribution is shown to be a sensitive diagnostic tool for the measurement of gain variations. Using this technique it is found that gain variations occur at counting rates as low as 100 Hz. It is argued that such errors will be present at some level in all tubes. Several calibrating schemes capable of reducing the error to below the 0.1% level are discussed.