173 



mean curve but since the temperature of the water in the barrel was not suffi- 

 ciently homogeneous, the observed deviations from the mean could not be attri- 

 buted entirely to the CTI, On the basis of these calibrations the CTI thermom- 

 eter was judged to be a more reliable instrument than the conventional bucket 

 type mercury thermometer commonly used for obtaining surface temperatures 

 at sea. 



A second series of calibration data was obtained from field observations. 

 The CTI underwater head was held just awash at the side of the ship and the 

 deck-side units read at a time when the instrument indicated the water to be 

 homogeneous. At the same tinme a sample of water to be used for chlorinity 

 titration was taken immediately adjacent to the underwater head. The tempera- 

 ture at the sampling spot was obtained either with the CTI thermometer assum- 

 ing the calibration to be adequate, as was usually done, or with a calibration 

 "bucket-thermometer. '' Conductance computations were made from these data 

 as in the case of "barrel" calibrations. Thus, a comparison of the chlorinity 

 computed from CTI measurements with a standard consisting of the chlorinity 

 titration was possible. In theory this method of calibration has the advantage 

 of comparing the test instrument with the standard under the same conditions 

 that exist during normal use of the instrument. 



In calibrations of this kind an unavoidable subjective decision is often 

 required. Ideally, the sampling done by the test instrument (the CTI) and the 

 standard (the chlorinity sample collecting bottle) should be in a homogeneous 

 body of water. Unfortunately, this frequently is not the case as it is not uncom- 

 mon to observe small, rapid changes in the conductivity reading and small slow- 

 er changes in the temperature reading of the CTI, The observer, therefore, is 

 frequently required to decide when the fluctuations are sufficiently small to be 

 neglected. 



The observation that the time constant of the conductivity unit is smaller 

 than that of the thermal unit leads to an important property of the instrument. 

 If these constants were equal, the combined input signals when converted to 

 chlorinity would be distorted (the peaks of the fluctuation reduced). With un- 

 equal constants there will be both distortion and phase shift. It is thus an open 

 question whether the chlorinity values from CTI data and from chlorinity titra- 

 tions are sufficiently comparable. They probably are, but as yet there is no 

 completely objective way of answering the question. 



A set of calibration figures is presented in figure 3, in which the CTI 

 conductivity scale readings has been plotted against conductivity computed from 

 chlorinity titration and CTI temperature data using the interpolation formulas of 

 Thomas, Thompson and Utterback (1934). It has been estimated that a Zo band 

 around the mean would be of the order of 0.5 millimho (approximately 0.5Cl°/oo). 

 This amounts to about ten times the variation expected from design considera- 

 tions had the standard been absolute. 



Thus, neither the barrel nor field calibrations as so far discussed provide 

 data from which a satisfactory picture of the behavior of the instrument can be 

 found. 



Since these data were taken, investigations of three aspects of the cali- 

 bration problem have been initiated. They are: (1) static calibration, (2) char- 

 acteristics of the H-type conductivity cell, and (3) precision estimates from 

 paired measures. 



