38 REPORT— iPnO. 



increase the accuracy of reading, a good plane mirror was substituted for 

 the original concave mirror, and observations were taken with a telescope 

 and scale at a distance of about 3 metres. The definition of the imago 

 was such as to permit of reading with accuracy to a fifth of a millimetre. 

 Owing to the gradual change or ' drift ' of zero, due to imperfect elasticity 

 of the suspension, which is always a serious source of error in galvano- 

 meters of this type, it was found to be impossible to obtain sufficiently 

 consistent observations by the deflection method. To ininimise this 

 source of error the potentiometer balance-method was adopted, and care 

 was taken not to subject the suspension to excessive torsion. To increase 

 the sensitiveness, the iron-copper thermo-junction was replaced by junctions 

 of iron and gei'man silver (30 microvolts per degree), and iron and con- 

 stantan (52 microvolts). The wires employed for this purpose were very 

 fine — about 0"2 millimetre — to minimise the cooling of the junctions by 

 conduction, and their thermo-electric powers were determined by a special 

 series of observations made on the particular pieces employed. Witli 

 these improvements it was optically possible to observe a difference of 

 temperature of a thousandth of a degree with certainty, as it corresponded 

 to a deflection of about a quarter of a millimetre with the iron and 

 constantan couple. 



Thermo-electric Sources of Error. — In observing small differences of 

 temperature with a thermo-couple, assuming that drift of the galvano- 

 meter zero is avoided by employing the balance method, the most trouble- 

 some residual errors arise from accidental thermal effects due to small 

 differences of temperature in other parts of the electric circuit, and in 

 particular at the junctions of the bridge-wire, and at the point of contact 

 of the slider. Tt is usual to employ german-silver or platinoid or platinum 

 silver as the material for the bridge-wire to secure a low temperature- 

 coefficient and high specific resistance. Unfortunately these materials 

 give large thermal effects when joined to copper. The alloy known as 

 manganin is greatly to be preferred to platinoid or constantan in this 

 respect, but its surface is more liable to tarnish. The superiority of the 

 bolometric method (platinum resistance) over the thermo-couple for accu- 

 rate measurement of small differences of temperature depends chiefly on 

 the relative ease with which these accidental thermal effects may be 

 eliminated. In the present instance they were found to be so trouble- 

 some that it was eventually decided to make the bridge-wire and the 

 whole of the circuit, with the exception of the couple itself, of pure 

 copper. By adopting this method the accidental disturbances were 

 reduced to a small fraction of a microvolt, without taking any special 

 precautions to secure uniformity of temperature throughout the various 

 parts of the measuring apparatus. The cold junctions of the thermo- 

 couple were contained in a copper plug screwing into the copper cube, 

 and were assumed to be at the same temperature as the walls of the cube. 

 In order to secure this, and to minimise changes of temperature of the 

 copper cube, it was found necessary to wrap the cube and the projecting 

 plug in a considerable thickness of cotton-wool, even when exposed to 

 feeble sources of radiation. The layer of felt surrounding the cube formed 

 no protection for the copper plug containing the cold junction, and proved 

 quite inadequate to prevent rapid changes of temperature when exposed 

 to strong sources. 



Constant Source of Radiation. — The necessity of a constant source of 

 radiation for comparative pieasurements and tests was recognised at a 



