The ageing behaviour of metal film resistors was studied using high-re
solution in-situ resistance measurements. A temperature storage experi
ment was performed at five different temperature levels. At each condi
tion, the ageing behaviour of 128 resistors was monitored. It turned o
ut that the failure behaviour of the samples is rather complicated. On
e remarkable fact is that measurements coming from two different produ
ction lots led to completely different results. One production lot dis
played monomodal failure behaviour, while for the other lot bimodality
was observed. Serious errors in the calculation of the FTTF (first ti
me to failure) can be expected when this fact is not properly taken in
to consideration. This paper also discusses some important aspects of
the analysis of failure time data, such as (i) the choice of the under
lying failure distribution, (ii) the number of samples to be tested an
d (iii) the failure criterion to be used. (C) 1998 John Wiley & Sons,
Ltd.