96 Lecture 5 
5.4. PERFORMANCE 
5.4.1. Stability 
5.4.1.1. Supply-Voltage and Ambient-Temperature Fluctuations 
The transistor velocimeters built at NBS exhibit changes in output prf of 
from 5 to 11 ppm for a 1% change in supply voltage. 
The effect of ambient temperature changes on the circuit itself is rather 
small, amounting toa total change in prfof about | part in 35,000 over a tempera - 
ture interval 0 to 42°C. The thermal expansion of the sound path is more serious. 
Most of the transistor velocimeters have a base plate of stainless steel for 
which the mean thermal expansion is 16.4- 10-2 PC. If the standard path length 
is taken as that at 20°C, thenthe error at 0 or 40°C is about 1 part in 3000, unless 
corrections for temperature are made. Most users of the instrument require 
temperature measurements in any case, and their practice is to correct the 
velocimeter readings for temperature. The base plate can be made of invar, as 
it was for the dozen or so vacuum-tube type of instruments made before the 
present instrument was developed. This procedure effectively solves the thermal - 
expansion problem, but makes it necessary to take steps to prevent the serious 
corrosion which occurs wherever the invar touches stainless steel and also at 
tool marks and scratches in the invar. 
5.4.1.2. Frequency Stability 
The term "frequency stability" here denotes the degree to which the prf of 
the velocimeter remains constant if the speed of sound in the liquid remains 
constant. In water, the speed of sound is a maximum at about 74°C [2]. At this 
temperature a variation of 1°C changes the speed of sound by only 13 ppm, and 
a variation of 0.1°C by only 0.1 ppm. The stability is therefore conveniently 
measured in water at 74° + 0.1°C; the prf is obtained by counting for ten sec- 
onds three times a minute. Such a test was made for 6.3 hr on a transistor 
velocimeter. The data scattered about a line representing a drift rate of 4.7 
ppm/hr with a standard deviation of 0.47 count in about 75,500, or 6.2 ppm. 
Vacuum-tube type velocimeters [9], equipped with an invar path and solidly 
backed transducers, are better than this. A 10.5-hr run on such an instrument 
showed a drift of only 0.3 ppm/hr witha standard deviation of 0.72 count in about 
153,000, or 5 ppm. In both cases the standard deviation includes the effect of the 
+1-count inherent counter error, and possibly some errors due to accidental 
counts. Figure 5.7 shows the sound head of such a velocimeter. Note the singly 
bent path and the hard rubber reflector. Figure 5.8 shows the construction of the 
transducer mounts. 
5.4.2, Calibration 
The standard liquid is outgassed distilled water; we use as standard values 
of c those we measured by a pulse technique in 1957 [10]. The water, rapidly stir- 
red, is held at each of several temperatures, measured with a platinum resist- 
ance thermometer, while the prf, f, is measured by counting pulses for 10 sec. 
Corresponding values of c and fare fitted by least squares to Eq. (1), which we 
repeat for convenience 
e 
eer ea (a) 
f c 
