430 THE BELL SYSTEM TECHNICAL JOURNAL, MAY 1952 



a comparison between the present status and that in September, 1949. 

 Estimates of the half-hfe of a statistical family of devices are at best 

 arbitrary and necessarily amount to extrapolations of survival curves 

 assuming that a known survival law will continue to hold.* In Septem- 

 ber, 1949, life tests on type A units had been in effect some 4000 hours. 

 With the assumption of an exponential survival law, it was not possible, 

 on the basis of a 4000 hour test, to estimate the slope sufficiently accu- 

 rately to warrant a half -life estimate in excess of 10,000 hours. These 

 same type A units have now run on life test for approximately 20,000 

 hours. With the more reliable estimate of survival slope now possible, 

 the half-life is now estimated to be somewhat in excess of 70,000 hours. 

 It should be emphasized, however, that these are type A units of more 

 than two years ago made with inferior materials and processes. It is 

 believed that those units under current development, being made with 

 new materials and processes, are superior; but, of course, life tests are 

 only a few thousand hours old. Although these new data are encouraging, 

 it is still too early to extrapolate the data such a long way. 



Temperature Effects 



Transistors like other semiconductor devices are more sensitive to 

 temperature variations than electron tubes. In terms of the linear 

 equivalent circuit elements, the collector impedance, re , and the current 

 gain, a are the most sensitive. Over the range from — 40°C to 80°C the 

 other elements are relatively much less sensitive. For type A transistors 

 these temperature variations in r<; and a are shown in Fig. 22. While 

 these curves are definitely not linear, an average temperature coefficient 

 for fc of about — 1 per cent per degree was estimated for the purpose of 

 easy tabulation and comparison in Fig. 21. 



Thus, for the early type A, Vc fell off to about 20 to 30 per cent of its 

 room temperature value when the temperature was raised to -f 80° C; 

 at the same time a increased from 20 to 30 per cent over the same 

 temperature range. Today, this variation has been reduced by a factor 

 of about four for re in most point-contact types, the variations in the 

 current gain being relatively unchanged. Fig. 23 illustrates the tem- 

 perature dependence of re and a for the M1729 transistor now under 

 development. Again, for purposes of easy comparison in Fig. 21, the 

 actual dependence of Fig. 23 was approximated by a linear variation and 



* Estimates of life, of course, depend upon definitions of "death". For these 

 experiments, the transistors were operated as Class A amplifiers. A transistor is 

 said to have failed when its Class A gain has fallen 3 db or more below its starting 

 value. 



