A system for comparing thermocouples was developed that eliminates the need for stable furnace temperatures and ice point references. This simplified system can be used without sacrificing accuracy if all units under test are the same type and have similar construction to the reference unit.

This system was used at Hewlett-Packard to calibrate platinum vs. platinum-rhodium type R thermocouples used to calibrate silicon diffusion furnaces. The range of calibration was from 600 to 1100 degrees Celsius.

By using a reference standard with identical construction and materials to the thermocouples being calibrated, all time constants will be the same. Because all units in the test have the same time constant, accurate measurements can be made in a slowly moving furnace. A reference standard and several units under test are placed in the calibration furnace. The furnace is heated to a temperature slightly above the highest calibration point, and then turned off and allowed to slowly cool. At HP the furnace used was a well-insulated fluidized bed that took about 6 hours to cool from 1150 to 600 degrees Celsius.

test point, a set of measurements is taken. Each unit under test is compared several times to the reference standard.

Ice point cold junction references are not used because it is not necessary to know the absolute temperature, closer than a few degrees. The approximate cold junction correction required for the furnace temperature is made by measuring the temperature at the DVM terminals. When doing the calculations for the difference measurements the cold junction corrections cancel.

This system was tested by making repeated measurements with the thermocouples in the same furnace wells. Agreement was about 0.3 deg. at 1100 deg. C. and 0.1 deg. at 600 deg. C. When the reference unit was interchanged with a test unit the agreement was about 0.5 deg. at 1100 deg. C. and 0.2 at 600 deg. C.

Similar results were obtained using the conventional procedure with stabilized furnace temperatures and ice points. Thus no additional errors are introduced using the simplified procedure.

Readings are taken at calibration points as the furnace is cooling down, usually every 50 degrees. The furnace temperature does not need to be stable or known to closer than a few degrees because only difference measurements are being made. The only requirements are that the all the units in the furnace are at the same temperature and that all units have the same time constant.

A low thermal scanner is used to measure differences between each unit under test and the reference standard. A short circuit is placed across one of the scanner inputs enabling the reference unit to monitor the furnace temperature directly. When the temperature falls to 2 or 3 degrees above the


Page 13

Home Page Contents (Page1) Previous Section Next Section