178 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957 



ment period, and only tubes having control-grid currents due to gas of 

 less than 5 X 10~" ampere were acceptable. This corresponds to a gas 

 pressure of approximately 2 X 10~^ mm of Hg. Very thorough mechani- 

 cal inspections after the 5,000-hour aging were made to insure that there 

 were no observable mechanical deviations that could cause trouble. The 

 history of each group of 28 tubes, from which prospective candidates for 

 the cable were selected, was reviewed to see if any group abnormalities 

 were found. In case a suspected trait was seen, all tubes in the group 

 were ruled out for cable use. 



As an aid in the selection of tubes for the cable, all pertinent data were 

 put on IBM cards. It was then possible to manipulate and present the 

 data in many very helpful ways that would have otherwise been wholly 

 impractical from time and manpower considerations. An over-all total 

 of about half a million bits of information was involved. 



Reliability Prospects 



Questions are frequently raised concerning the probability of tube 

 failures in the system. There are two areas into which failures naturally 

 fall — catastrophic failures and the type of failure caused by cumulative 

 effects such as the decay of thermionic activity, development of primary 

 emission from the control-grid, or the build-up of conductance across 

 mica insulators or glass stems. 



The catastrophic failures might include such items as open connections 

 caused by weld failures or fatigue of materials, short circuits caused by 

 parts of two different electrodes coming into contact or being bridged by 

 conducting foreign particles and gas leaks through the glass or along 

 stem leads. Fortunately these failure rates have been lowered to a point 

 where there are no sound statistical data available in spite of the sub- 

 stantial amount of life testing that has been done. In approximately 

 4,800 tubes made to date, there have been four failures that were not 

 anticipated by the inspections made. All four of these failures were of 

 different types and occurred either at or before 5,000 hours of life. All 

 four were of types more apt to occur during the early hours of aging and 

 handling. 



Of the cumulative types of failure, life testing has indicated no ap- 

 parent problem with either the growth of insulation conductance or 

 primary emission from the grids. As indicated earlier in the paper, therm- 

 ionic life results are such that there is reason to be optimistic that no 

 failures will occur in 20 years. 



