Abstract :
The term uncertainty is used rather than instability. Stability implies that a device can be calibrated to a known and repeatable small error. As will be shown for the temperatures considered, this definition of stability may not be true. Two configurations of platinum/platinum-10% rhodium thermocouples were developed for the Tory II reactors. The manufacturer checked outputs at 2400°F. Subsequently, we ran tests between 2000 and 2400°F to determine how dependable the initially measured outputs were. The uncertainty turned out to be about ±10°F, over the rather narrow test band. As a result of these first uncertainty tests, we decided to see what effect, if any, different types of junctions have on the uncertainty. This report has to do primarily with results of tests on crimped, twisted, hydrogen-arc-welded, and resistance-welded junctions. A horizontal tube furnace was used from 1500 to 2300°F and a vertical tube furnace from 2300 to 2800°F. Some unexplained things happened when the transfer to the vertical tube furnace was made. Reference grade, 10-mil thermocouple wire with alumina tube insulation was used for most tests. In general, the uncertainty shows up as an error band of fairly constant width. This means that it cannot be expressed as a percent of indicated temperature. The band width tends to become narrower with increasing temperature up to 2300°F or so; 20°F is still a "good" figure for it at 2300°F. Hydrogen-arc welding should be avoided. The importance of an annealing cycle to test temperature is shown by the curves.
