172 



from the cell design and electrode treatment shown to be desirable by laboratory 

 experience to meet the need for adequate flushing of the cell, mechanical stabili- 

 ty, and other factors connected with operation at sea. The effect of these 

 changes on the precision and accuracy of the instrument can be only estimated, 

 so that calibration is imperative. Calibration implies the existence of a stand- 

 ard with characteristics at least as good as the expected characteristics of the 

 instrument. A generally accepted rule of thumb is that the standard should be 

 an order of magnitude better than the test instrument. 



Experience with the CTI has emphasized another requirement of calibra- 

 tion systems that frequently can be or has been neglected. The CTI, like all 

 in situ conductance instruments, was made to function in an environment in 

 which the variable to be measured fluctuates in time. This raises two difficul- 

 ties in the calibration of such an instrument. First, if the time constant of the 

 instrument is large (slow response time) compared with the fluctuations in the 

 environment, a calibration of the instrument made under static conditions will 

 not provide a true indication of the reliability of a measurement made in the 

 fluctuating environment. Second, if the time constants of the standard and of the 

 test instrument are very different, com'parison of the two in a fluctuating sys- 

 tem may not give a reliable measure of the accuracy and precision of the test 

 instrument. 



In the following discussion some of our experiences in attempting to cali- 

 brate the CTI will be presented. The behavior of this particular instrument is 

 given because it is the most familiar to the author. However, many of the fea- 

 tures are believed to be common to other similar instruments, especially to the 

 STD which uses a thermometer unit similar to the CTI, and to the proposed NEL 

 instrument which is reported to use an H-type conductivity cell. 



The first calibrations of the CTI were carried out in a 55 gallon barrel. 

 The sea water in the barrel was both cooled and diluted by the addition of ice. 

 CTI readings, standard thermonneter readings, and samples for chlorinity titra- 

 tion were taken after each addition of ice had melted and vigorous stirring 

 achieved at least a steady thermal state as shown by the instruments. The CTI 

 thermometer readings were then compared with the standard thermometer val- 

 ues for each set of observations, thus providing a calibration of the CTI ther- 

 mometer in terms of the standard. National Bureau of Standards certified mer- 

 cury thermometer. Conductance values were calculated from standard ther- 

 mometer tennperatures and chlorinity titration results, using the Thomas, 

 Thompson and Utterback (1934) interpolation formulas. These computed val- 

 ues were then compared with similar values obtained from CTI temperature 

 readings, CTI conductance readings, and a previously measured cell constant. 

 Thus, the behavior of the CTI was compared with a standard composed of a 

 National Bureau of Standards certified thermometer and the chlorinity titration. 

 This technique was noted at the time to be crude, as thermal equilibrium in the 

 container was difficult to maintain, especially when the ambient temperature 

 was several degrees from the test temperature. It was under these conditions 

 that the difference in time constants between the mercury standard and CTI re- 

 sistance thermometers was noted. 



"Barrel" calibrations of this sort provided a means of setting the range 

 adjustments and making circuit refinements in the instrument. They are in- 

 adequate, however, for precise calibrations when standard deviation of the or- 

 der of hundredths of degrees centigrade and hundredths of parts per thousand 

 chlorinity are to be noted. These calibrations did, nevertheless, indicate that 

 the CTI resistance thermometer was a stable instrument. Successive ther- 

 mometer calibrations made several months apart gave essentially the same 



