589 The Measurement of End-Standards oj Length .
By P. E. Shaw , B.A. , D.Sc .
( Communicated by Prof. Pointing , F.R.S. Received November 18 , 1910 , \#151 ; Read January 12 , 1911 .
) During the last few years there have been improvements in this branch of metrology .
In 1905 , as pointed out by the writer , * the best gauges were greatly defective as to their end faces and the machines for their measurement were faulty in principle .
Since then Jolianson has , by a secret process , produced gauges incomparably superior to anything previously on the market , with the result that other gauge-makers have been stimulated to improve their standards .
The ordinary bar gauge has flat ends .
This paper deals with this form only .
The ideal gauge would satisfy two conditions : ( 1 ) The ends to he absolutely plane and parallel , though they need not be truly normal to the length of the bar ; ( 2 ) the distance between the planes to be exactly the length for which the gauge stands .
In this paper only condition ( 1 ) is considered .
Testing work too extensive for full publication has been performed , but the following brief account shows some improvements effected both in gauges and in measuring machines .
The method of measuring is fully described in the above paper .
In brief , the machine consists of a massive bed carrying two measuring head-stocks and a table for carrying and adjusting the gauge .
The headstocks each carry a horizontal micrometer screw , working parallel to the bed , and a nut with a graduated head .
These measure the bar , which rests on and is clamped to the table .
In taking a measurement the left screw , say , is brought into electric contact with the gauge , then the right screw is brought into electric contact with the gauge , and wiien the current passes through the gauge from one measuring point to the other , the two divided heads are read .
The difference in the readings is a measure of the distance through the gauge from one measuring point to the other .
In addition to the usual precautions against temperature changes provision is made to avoid shaking by arranging that the graduated heads are not actually touched by hand , but are worked by a hand pulley and string .
There are two novelties in this method of measuring : ( 1 ) the measurement is made electrically between point and point instead of between rough plane * See ' Roy .
Soc. Proc. , ' A , 1906 , vol. 77 .
Dr. P. E. Shaw .
[ Nov. 18 , and rough plane ; ( 2 ) the gauge can be moved , after a measurement in one position , into another parallel position at a known distance away and a new measurement taken .
Thus the two gauge faces are simultaneously explored and errors become manifest .
Improvements Gauges .
Suppose the face of the gauge is circular .
Take 13 places equi-spaced as shown on the face .
The first measure is taken from the centre of the face along a line normal to the face to the other face .
The second measure is taken from point 2 along a parallel line , and so on for 13 measurements .
11 3 fi 2 13 Since the gauge face generally slopes away from the centre in all directions , the centre is a place corresponding to maximum length through the gauge , and we should expect to find the greatest irregularities for the larger numbers , 10 , 11 , 12 , 13 , approaching the edge of the face .
Such , in general , is the case .
The accompanying curves ( fig. 1 ) show at a glance that the measures taken are very unequal for curves A , B , C , D , and less so for E , F , G- .
One curve is produced from the measurements as above explained ; the gauge is then reversed , and the twin curve produced .
This is the procedure for curves A , B , C , E , F. The small vertical divisions on the paper imply 0'5 yu , ( = 1/ 2000 mm. ) ; the horizontal divisions are arbitrary , but the whole length of curve represent generally about 6 mm. Thus the irregularities of the pair of gauge faces are exaggerated 400 times .
Let A = the arithmetic mean of the differences for all the readings at one time .
Let 3 = the arithmetic meail of the differences in the readings for any one place .
Thus A and 3 give respectively measures of the accuracy of the gauge and of the machine .
During the experimental work temperature changes may effect the length of the gauge .
This would increase 3 and , to a less extent , A. 1910 .
] The Measurement of End-Standards of Length .
591 } 3-inch commercial .
) 6-inch commercial .
) 9-inch commercial .
150 mm. commercial .
150 mm. M.A. } M.A. M.A. Fig. 1 .
If both gauges and machine were perfect the curves in fig. 1 would be horizontal straight lines , whereas if the gauge alone were perfect the curves would be sinuous on account of the uncertain action of the machine .
The lower curves in the figure approximate to the conditions of perfection , whereas the upper ones are very sinuous .
A. This represents readings on a 3-inch commercial gauge , measured direct and reverse .
A = 3'6 ytt , 0'4 B. A 6-inch commercial gauge , direct and reverse .
A = 3'2 ju , S = 0*8 ju .
C. A 9-inch commercial gauge .
This shows an advance in accuracy , the machine is much more true than the gauge .
A = 2-6 fu , S = 0'4 D. A comparison between a 150 mm. commercial gauge which has A = 3*5 / jl and a much better one , marked M.A. , for which A = TO / u. ( This gauge and those used for curves E , .
F , G- , were kindly made for the writer by Messrs. Manlove , Alliott , Ltd. , Nottingham , who took much trouble during these experiments to perfect a lapping method of their own .
) E. This is a good gauge marked M.A. A = TO / u,8 = 0-5/ u. Dr. P. E. Shaw .
[ Nov. 18 , F. Another special gauge marked M.A. After taking a set of readings giving one curve , the gauge was re-lapped , measurements repeated , and the second curve produced .
A = 0*6/ 4 , S \#151 ; 0*4 fi .
( 4 .
A mild-steel gauge .
This is the.best curve produced so far .
The commercial gauges mentioned are made by various well-known firms , whose names cannot , of course , be given ; the gauges had been in sound keeping since they came from the makers , and were in a state of good preservation , so that the tests are fair .
From the above experimental curves and figures it is obvious that ( 1 ) it is possible to make far better gauges than those commonly issued as standards ; ( 21 the best gauges so made are so nearly perfect in form that such errors as are found may be due to the imperfect working of the measuring machine and not to the gauge itself .
Improvement in the Measuring Machine .
At this juncture it seemed advisable to improve the machine .
The table carrying the gauge was the chief source of error .
It is very difficult to provide a reliable mechanism for translating the gauge parallel to itself .
Any looseness or rotation occurring during this translation will introduce errors in the measurements .
The table , as described in the former paper , was discarded .
In its place is one involving a different principle .
There is a plane surface plate set perpendicular to the bed .
The plate on which the gauge rests is pressed by springs so as to rest against the surface plate , touching it at three points .
If the surface plate is true and rigid , and the three feet remain always pressed against it , the movements of the gauge resting firmly on .the plate must be strictly parallel .
The surface plate is set perpendicular to the run of the bed by an optical arrangement .
The Jolianson gauges are made in slabs , the cross-section in all cases being 35 mm. x 10 mm. They are used singly or composite ; in the latter case two or more are wrung together with pressure so as to adhere , and form a rigid composite gauge equal to the sum of the component parts .
Whether .single or composite these gauges were known to be very accurate , but it was hoped that with the improved measuring machine as described , any irregularities could be detected .
The following tables are typical of results found:\#151 ; 1910 .
] The Measurement of End-Sta Length .
593 Johan son Gauges .
2-inch single .
r 0-850 inch .
2-inch composite .\lt ; 0 " 450 " [ 0 -700 " A. B. B-A .
A. !
b- B-A .
487 -2 487 -2 0-0 487 -4 487 -3 -0-1 487 -1 487 -1 o-o 487 -2 487 -1 -o-i 487 -2 487 -0 -0-2 487 -1 487-1 -o-o 487 -0 487 -2 0-2 487-1 487 -0 -o-i 487 -0 487 -2 0-2 487 -1 486 -8 -0-3 487-1 487 -2 o-i 487 -2 486 -8 -0-4 487 -0 487 -2 0-2 487 -1 486 -8 -0-3 487 -0 487 -3 0-3 487-0 486 -8 -0-2 A = 0 -2 fx .
A = 0 '3 | A = 0 '5 / a. A = 0 *4 , u. A = 0 " 5 , u. !
A = 0 -4 The first number in Column A gives the result for one place on the gauge as 487'2 / x. This is the difference in readings of the two headstocks .
It is not the actual length of the gauge , which for present purposes is immaterial .
On moving the gauge parallel to itself , the number obtained is 487'1 fi , and so on .
After an interval of time , Column B is obtained and Column B\#151 ; A results .
Thus , consistency in The Columns A and B taken vertically shows accuracy in the gauge , and consistency when they are taken horizontally shows accuracy in the machine .
Inspection of these results shows ( 1 ) that the 2-inch single is ( naturally ) more .consistent than the 2-inch composite , the mean values of A being 025 and 045 / a respectively ; ( 2 ) that the inconsistencies of the machine , as shown by the B \#151 ; A column , are of the same order as the irregularities of the gauges .
So the apparent errors in the gauges may be attributable to the defects in the machine .
In order to see how ordinary gauges stand tests with the new machine , two 2-inch specially good standards were tested and came out as shown .
Two 2-inch Ordinary Gauges .
G-auge A. 1 Gauge B. 655 *4 654 -2 655 *0 653 -8 655 *1 654 -5 655 *2 653 -9 655 -2 654 -2 A = 0 -4 A = 0 -7 594 The Measurement of End-Standards of Length .
These gauges are thus less satisfactory than the Johanson gauges of the same length .
But the superiority of the latter is much greater than appears , for the test was made over about three-fourths of the end faces in all cases .
In the ordinary gauge this area is about 20 sq .
mm. , whereas in the Johanson gauge it is 350 sq .
mm. , or 17 times the area .
Any one who has attempted to make accurate plane surfaces in metal will allow that it is a wonderful achievement to make two planes of this size , plane and parallel to within 0*3 micron .
2-inch single Johanson .
2-inch composite Johanson ?
In the accompanying curves ( fig. 2 ) , the small divisions vertically represent 0*5 micron .
The length of the ordinary gauge curves is about one-sixth of that of the Johanson gauges , corresponding to the diameter of the faces .
Until recently the gauge maker has insisted that the measuring machine is more accurate than the gauges it is used to measure ; and that the gauges are themselves more accurate than is required for practical engineering purposes .
But we here see that the best machine is powerless to detect errors in these wonderful Johanson gauges , and an advance in accuracy of the former is demanded .
The writer is now designing a machine which he hopes will give consistent readings to OT micron , under the trying conditions that the gauge shall be moved parallel to itself between the readings .
In recent years the work of Michelson and of Fabry and Perot has placed the measurement of line standards of length on a high level of accuracy .
An even further advance is promised by the use of Grayson 's rulings by Dr. Tutton.* If a much improved measuring machine on scientific principles can be produced , the measurement of end standards should exceed in refinement that now obtaining in the parallel branch of metrology , since the latter suffers from the optical limitation of the microscope .
2-inch ordinary .
Fig. 2 .
* See ' Phil. Trans. , ' A , 1909 , vol. 210 .
On the Absolute Expansion of Mercury .
595 The writer wishes to tender his thanks to Mr. J. M. C. Paton , of the firm of Manlove , Alliott , Ltd. , for much assistance , freely given , in the making of gauges ; also to Dr. E. T. Glazebrook , for giving him every facility in tests made on his machine at the National Physical Laboratory ; also to the Eoyal Society for a grant which defrayed some of the cost of ' this work .
On the Absolute Expansion of Mercury .
By H. L. Callendar , M.A. , LL. D. , E.E.S. , and H. Moss , B.Sc. , A.E.C.S. , Demonstrator of Physics , Imperial College of Science and Technology , S.W. ( Eeceived November 19 , 1910 , \#151 ; Eead January 12 , 1911 .
) ( Abstract .
) Kegnault 's experiments on the absolute expansion of mercury were reduced by Wullner , and subsequently by Brooch .
Their reductions differed by nearly 1 per cent , at 300 ' C. Begnault 's own reductions differed by nearly 1 per cent , from either at 40 ' C. Chappuis ' later determinations by the weight thermometer method , which were not absolute , agreed , fairly well with Wiillner 's reduction of Begnault at low temperatures , but differed from Wullner in the opposite direction to Brooch by more than 2 per cent , when extrapolated to 300 ' C. The object of the present investigation was to repeat Eegnault 's method on a larger scale with modern appliances , and the apparatus was designed to secure an order of accuracy of 1 in 10,000 , or 0'01 ' C. , which , it is believed , has been substantially attained .
The principal modifications made in Eegnault 's apparatus were as follows : \#151 ; ( 1 ) In place of the single pair of hot and cold columns , each 1*5 metres long , employed by Begnault , six pairs of hot and cold columns , each nearly 2 metres long , were connected in series as a multiple manometer , giving nearly eight times the expansion obtainable with Eegnault 's apparatus .
( 2 ) The difference of level to be measured was directly referred to a standard Invar scale divided in millimetres with divisions 4 microns wide , by means of a pair of telescopes with micrometer eyepieces reading to .O'OOl cm .
The difference of level corresponding to the fundamental interval 0 ' to 100 ' C. was 20'5 cm .
, so that the limit of accuracy of reading was 1 in 20,000 of this interval , or 0'005o C. The lengths of the hot and cold columns , each nearly 2 metres , could be read to 0'01 cm .
, giving the same order of accuracy .
( 3 ) The mean temperatures of the hot and cold

