550 



In test with service weapons, the best results were obtained using a 

 torque wrench which slipped after a predetermined torque was reached. 



In visual practice the damaged diaphragms were measured after 

 removal from the gaviges. A series of diaphragms were meastired in the 

 gauges before and after the shot and the difference compared with the 

 usual measTirements . The measurements in the gaviges averaged .003 in- 

 greater than out of the gaviges. This was less than 1^ of the ordinary 

 damage and either method was considered adequate as long as it was used 

 consistently. 



The thickness of the diaphreigm in the region which remained un- 

 damaged, that is, protected by the clamping plate of the gauge, was 

 measured before and after the shot and found to remain the same vithin 

 .001 in. However, by marking diaphragms with parsillel lines about l/2 

 and 1-1/2 in. from the edge, it was shown that there is movement toward 

 the central or damaged portion of the diaphragm. The amount of slippage 

 was roughly proportional to the damage and varied from 0.20 mm to 0.60 mm 

 for a damage change of 0.32 to O.kS in. 



(v) Summary . If then, for a tjpical experiment, we assume the 

 following errors, we can calculate deviations for a single shot using 

 k gauges. 



Assumed Errors ji Deviation in Damage 



t2 in. Gauge-to-charge distance 



(variation due to charge t3-5i> 



displacement only) 



+6 in. Charge dlsplacaaent 



(vertical) t 2% 



+15° Charge tilt t 1^ 



+ 5° Gauge orientation t 1^ 



Then if there were no other variables we should have the following 

 precision: 



Standard deviation of a single observation t h.3i> 



Probable error of a single observation t 3 '2^ 



These results are to be compared to the standard deviation and 

 probable error as determined from the distribution curves plotted in 

 Figures 17 and l8. To obtain these curves we have used 212 damaged 

 diaphragms. These were damaged by 2 to J+ lb. charges all at W in. 



In Figure 17 the actual deviations in hundredths of an inch from 

 the mean for each series is plotted against frequency of occurrence. 

 Figure l8 shows the percentage deviation frcan the mean for each series. 

 The absolute deviation is nearly constant independent of amount of 

 damage, so Figure 17 is considered more useful than Figure l8. 



