Page images
[blocks in formation]
[merged small][merged small][merged small][ocr errors][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small][merged small]

K thru M

8 24-25

28 38

2.06 1.67

0.009 0.008

0.10 0.09

0.30 0.27

O Measurements made with laser No. 109
A Measurements made with laser No. 184
• Measurements made with iodine stabilized

[ocr errors]

in 10


А в




[ocr errors]
[ocr errors]


[ocr errors]



[blocks in formation]
[blocks in formation]


[blocks in formation]

Figure 7.

Control charts on the 10 inch steel and 10 inch glass-ceramic

control standards,

[ocr errors]


0.129 INCH

[merged small][ocr errors][merged small][ocr errors]

2 4 6 8 10 12 14 16 18 20 22 24

26 TIME IN MONTHS Control charts on the 0.129 and 0.150 inch control standards

showing secular changes.

Figure 8.

[blocks in formation]

Figure 9. Correlation chart, length vs vapor pressure.

although both instruments gave the same reading with the interferometer door open, a difference developed once the door was closed. Hygroscopic

, material in the housing is the probable explanation.

The hygrometer sensing element was then mounted inside the housing (a process change) and control block measurements were resumed for period B. A marked improvement in precision was obvious on the control charts (fig. 7) but no significant change in mean values occurred.

Period B measurements, no longer masked by imprecise humidity measurements, indicated the steel block values to be less precise than the glass ceramic values. Temperature was the only parameter likely to affect steel more than glass-ceramic, therefore, it was possible that temperature measurement variability was causing the difference in precision. This theory was tested by attempting to improve the temperature measuring sys

New thermocouples were assembled using higher grade copper and constantan wire and the lead length was reduced by moving the reference tempperature enclosure from a position several feet away to a position in contact with the interferometer housing. The leads do not pass through open air, but go directly into the housing with a consequent improvement in thermal environment for the thermocouple system. Finally, one measuring junction was attached to the stem adjacent to the resistance coil of the SPRT. This allows verification of temperature equality between reference

thermal emfs. The new temperature system was then recalibrated by the method previously described.

Control standard measurements were resumed and results plotted on the control charts, periods C and E. Points taken in period D were omitted from this subgroup for cause.* It seemed apparent from the control chart that no improvement in precision resulted from rebuilding the temperature system, in fact precision was degraded slightly.

In rebuilding the temperature system, the nanovoltmeter was moved close to the interferometer from a location several feet away. Even though circuitry is solid state, the small amount of heat generated by this instrument could introduce perturbations effecting precision. Evidence such as a faster rise in interferometer air temperature when power was supplied to the nanovoltmeter plus the metrological axiom against heat sources being near measurement areas led to moving the meter.

Again the effect of this change was visible on the control chart, period F, with a statistically significant improvement in precision for the 10 inch steel control standard, the one most influenced by temperature. Temperature system rebuilding had indeed improved precision. A small systematic error probably not exceeding 0.001°C may be indicated by the change in mean value of the 10 inch steel błock but the evidence for this conclusion is marginal.**

Measurement precision on the 10 inch blocks (o=0.18 and 0.21 uin.) now approached that of the 0.150 inch control block (o=0.10 uin.) suggesting that length dependent random errors were very small. Until factors causing this variability are discovered it is unlikely that process precision can be improved.

Another round, G, was started with no process changes. Only four measurements were made before laser No. 109 broke down, requiring a new plasma tube. Unfortunately this occurred just before a planned laser calibration which would have allowed evaluation of wavelength, the last source of systematic error remaining to be studied. Judgment of long term stability of this laser must be based mainly on three calibrations: one at the beginning of this study and two about 18 months before. Vacuum wavelength variations of approximately 1 part in 108 were observed. Extrapolation of this evidence is not advisable although th control block data seems to reinforce the picture of a well behaved laser.

At this point measurement precision (30) of the process was about 5

*During this period, the aneroid barometer was carried back and forth between two laboratory rooms. A periodic barometer check indicated a calibration change, probably from mechanical shock. Pinpointing the exact date of the calibration shift was impossible so all values were omitted. **Further tests of temperature measurement effectiveness are given in the Appendix, item 4.

« PreviousContinue »