The TSPN-3 differs from the TSPN-I [3] in having improved stability over a wide temperature range (up to IO0~ while having the same dimensions. This has been attained by using comparatively thick platinum wire and mounts of a new design. The spiral is made of PL-O thermometer-grade platinum, while the holder consists of glass-coated platinum rods (the high quality of the PL-O platinum has meant that chemical-grade glass has been used in the holder). The rods lie parallel around a circle of diameter 3 mm, and the ends are brought together and welded. The number of rods is varied to suit the diameter of the platinum wire and the required sensitivity. The spiral is mounted on these without welding. The gap between the spiral and the rods is about 0.02 mm. Each thermometer is hermetically sealed and has a four-wire connection system. The design completely eliminates any displacement or deformation of the platinum spiral on acceleration perpendicular to the axis, as well as strain in the spiral in response to changing temperatures. The mechanical strength of this thermometer under acceleration along the axis is higher than that of the thermometer described in [4, 5], which has been attained by reducing the length and diameter of the spiral. For example, a TSPN-3 with Ro= i00 ~ has the diameter of the platinum spiral twice that used in the thermometers of [4-6], while the length is reduced by a factor of 3, although the latter thermometers were made from platinum wire of the same diameter. This means that the strength of a TESPN-3 thermometer has been improved by about afactor of 5, while the length has been halved. In addition, the spiral can be attached to the glass-coated rods with lacquer, which makes the thermometer insensitive to vibration. Preliminary tests have shown that the drift in the TSPN-3 thermometers is not more than 3"I0-3~ and that they can be used as first-class standards [I]. The scope for using the TSPN-3 in thermometric standardization was examined by making 13 thermometers from PL-O platinum: eight of them were TSPN-3 and five were TSPN-I. The latter employed platinum wire of diameter 0.07 mm, whereas the former used wires of diameters 0.05, 0.07, and 0.092 mm. The thermometers were annealed at about 500~ before filling with the heat-transfer gas (helium or neon). The nominal values of the resistance Ro at 0~ varied with the design, the spiral diameter, and the platinum wire used; the ratio of R,oo to Ro (where the former is for t = 100~ was in the range 1.392562-1.392574. These different thermometers were compared for stability during tests under identical conditions. First the stability was tested at the triple point of water. The resistance was measured by the method of [7] with an R-348 potentiometer of class 2"10-s. * The sequence was as follows: two cooling--heating cycles over the range from --200 to +I00~ measurement at the triple point, 20 cooling--heating cycles, measurement at the triple point, �9 An error of 2"I0-S% in measuring the above nominal resistances corresponds to 176 and is largely dependent on the stability of the temperature of the potentiometer. Careful adjustment of the potentiometer has allowed the figure to be reduced to 5"i0-~% (176