resistance change per degree is less. Sippican hand-tailors each probe lead resistance match 

 to less than .03°C error at 25°C. 



• Wire Insulation Leakage 



Generally , (although not always) produces up-scale error, since a leak in the 

 "A" lead will have more effect than a similar leak in "B"; will have more noticeable effect 

 at cold .temperatures where thermistor resistance is higher and is more susceptible to high 

 shunting resistance. Normal wire insulation leakage contributes less than .03°C error 

 at-l.l°C. 



• Thermistor Offset 



May be either up or down scale and results in a temperature reading that is always 

 more or less than actual by about the same amount. 



• Thermistor Slope 



Little or no error around 25°C, increasing in either up or down scale direction as 

 end scale temperatures are approached. Typical sum of offset and slope errors is .1°C max. 

 over the 0°C to 30°C span. 



• Thermistor Insulation Leakage 



Always up scale and more noticeable at cold temperatures where thermistor 

 resistance is high. 



Because Sippican normally in-line production tests each XBT for an accuracy of 

 ±.1°C at room temperature, most of the field failures result from wire and thermistor 

 insulation leakage which will tend to bias failures heavily in an up-scale direction. 



With the aforementioned error sources in mind, the following discussion relates to 

 the 'Summary of Results" in Mr. Anderson's report. 



(1) Two out of eleven 460-m systems malfunctioned to produce erroneous but 

 visually acceptable temperature errors after making accurate drops and yielded 36.5% of 

 total data that was outside stated system error. 



Comment: 



We suspect leakage resistance of launcher and launcher cable was too low. This 

 is normally result of salt water penetration of cable or launcher through physical damage 

 to the launcher or cable. It is visually detectable by installing an A2A test canister in 

 launcher and testing as described on page 1 1 of Volume I of Mr. Anderson's report. It may 

 be compensated, if not excessive, by recalibrating zero and full scale. It is definitely 

 established by the recommended monthly Reference Standard Leakage Test B-1 of 

 Table 5-4, page 5-4 of R-603, Instruction Manual for XBT System or by testing with an A4 

 XBT test box. 



Effects of low-leakage resistance do tend to bias readings upscale by a greater 

 amount at colder temperatures. 



(2) Only 80.1% of 460 m system drops would have been visually acceptable. 



Comment: 



Normal acceptance rate is 90% or better for 460 m probes that are two years 

 old or less. Older probes will fail at a higher rate due to aging of wire and thermistor 

 insulation. If a customer experiences an abnormally high probe failure rate, Sippican will 



136 



