551 
=| 
amplifier. Since the length of the intercept caused by imperfections in the amplifier used in these 
experiments is of the order of 0.4 microseconds, i.e. 0.000% milliseconds, it is clearly legitimate 
to neglect the effect of distortion in the amplifier, The length of the intercept caused vy distortion 
in the bar can be estimated and calculations show that the approximately linear portion of the 
(k'é/e., t) curve for bar B, if produced backwards, will cut the t - axis at t = +3 microseconds 
approximately. Since the ordinate of this curve begins to differ from zero at about t = —- 20 
microseconds, the length of the intercept is about 23 microseconds = 0,023 milliseconds, which is of 
the same order as that actually found here in Figure 3. tt should be remembered that the errors in 
the theory on which the calculations are based are likely to be greatest in the neighbourhood of 
t= 0; it is nevertheless clear that the initial curvature of the curve of Figure 3 may be caused by 
dispersion in the pressure bar and that it is difficult to decide from experiments with a bar of this 
length anc diameter, whether the applied pressure dces or does not rise instantaneously from zera to 
a finite value, 
Considering next the portion of the (€, t) curve of Figure 3 lying between t = 0,035 milliseconds 
and t = 0,103 milliseconds, the equation to the straight line of closest fit drawn through the 
experimental points Is found, by Awbery's method", to be 
Go 1. 3556t— 00328) ise ieisiel wulsisie at asisie (S55) a (citliny cms), ta innmiMiiseconds)s 
The curve shiwn in Figure 3 has been drawn using this equation between the limits of t for which it 
applies and, on the scale of this figure, it is worth noticing that, apart from the point at 
t = 0.111 milliseconds, the scatter of the experimental points around the straight line is small. 
From equation (3.5), it follows that, over the linear portion of Figure 3, ¢ = 1.356 x 
10? cm./sec., and hence, from equation (3.4), 
Sy 1.356 x 103 = 2.780 x 10° dynes/sq.cm. 
as) 
1] 
2.05 x 10 
2.780 x 10° dynes/sq.cm. = 18.01 tons/sq.in. 
vo 
" 
When t exceeds 0,143 milliseconds, the (€, t) curve shows a maximum when t = 0,164 milliseconds, 
followed by a minimum when t = 0 225 milliseconds, and a gradual, irregular rise which extends as far 
as t = 0.46 milliseconds; finally, this rise is followed by a series of irregular peaks. 
The maximum and the minimum in the region extending from t = 0.164 to t = 0,225 milliseconds, 
are probably due to dispersion effects in the bar, associated with a decrease in the applied pressure. 
whilst it is true that the effect only becomes prominent with a parallel plate condenser unit if the 
applied pressure undergoes very sudden variation, it must be remembered that the (response frequency) 
curves for a cylindrical condenser unit show that this type of unit magnifies and reverses the signs 
of the high frequency components in a pulse, relative to the low frequency components. It follows 
that dispersion effects will show up more prominently in a record taken with a cylindrical condenser 
unit than in one taken with a parallel plate unit. For these reasons, it is difficult to make any 
quantitative deductions from the (€, t) curve when t exceeds 0.14 milliseconds, 
The periods of the small oscillations superposed on the record when t exceeds 0,225 mill tseconds 
are of the order which one would expect from considerations of the propagation of a pulse consisting 4 
of extensional waves, on group velocity theory. This is shown by the agreement between these 
periods and the periods of the dominant groups which have been calculated, and which are represented 
to scale in Figure 3 by the lengths of the small sine-curves, C, placed at intervals of 0.05 milliseconds, 
As to the series of irregular peaks which occur when t exceeds 0.46 milliseconds, it was once thought 
that they might be due to flexural or transverse vibrations of the pressure bar, caused for example 
by an asymmetry of the charge relative to the pressure end of the bar; the investigatian of the 
propagation of a pulse of flexural vibrations #n a cylindrical bar, given in the appendix, shows 
however, that the astest possible group of this type will not reach the measuring end of the bar 
within the time covered by the record of Figure 3. jt may be that the peaks in question are caused 
DY svccee 
* J, H. Awbery, Proc. Phys. Soc., Vol. Ui, p.384, (1929), 
