340 An Electrical Measuring Machine for Engineering Gauges and other Bodies .
By P. E. Shaw , B.A. , D.Sc .
( Communicated by Professor J. H. Pointing , F.R.S. Received December 1 , 1905 , \#151 ; Read January 25 , 1906 .
) S I.\#151 ; PRELIMINARY .
1 .
End standards of length , called gauges , are tested and compared with one another by means of measuring machines , made by many engineering firms in England and abroad .
These machines are identical in principle , and differ little in form from the original one invented by Sir Joseph Whitworth .
Their principle is that the gauge rests against one jaw , fixed , of the machine , whilst the other jaw is moved forward by a micrometer screw until it touches the gauge .
Compression , more or less , of machine and gauge , is required before any indication can be obtained that both jaws are firmly in contact with the gauge .
These machines will in future be spoken of as mechanical-touch machines , as distinguished from the new machines called electric-touch machines .
In practice it has been found desirable to introduce some index of the touching\#151 ; e.g. , a gravity feeler , spirit level , or the raising of liquid in a fine tube .
2 .
The weight of the gauge , often large , should never , as is sometimes dene , be supported by the grip of the two jaws ; for then there is considerable end-thrust of the jaws with consequent longitudinal strain in the machine , in addition to the vertical strain due to the weight of the gauge .
If the whole or part of the weight of the gauge be supported by the jaws , though the latter be parallel to one another initially , they will not be so after being-weighted .
The errors which enter into the measurements made by these machines differ in kind and degree according to the nature of the gauge .
There are three kinds of gauge .
A. Bar Gauges with Plain Parallel Ends.\#151 ; Each jaw has a flat face and each end of the gauge has a flat face .
Each of the four faces has defects in planeness , and in being not strictly parallel to the other three .
If in addition a gravity feeler be used , it introduces extra errors .
To admit so many errors of unknown amount in the measurement of gauges may be permissible in present-day engineering trade practice , but it does not satisfy the Electrical Measuring Machine for Engineering Ganges , etc. 341 demands of exact metrology .
Thus to obtain accuracy , surface contact should be abandoned and point contact used , and these gauges should be measured between two rounded points or spheres , in which case no assumptions are made as to the perfectness or parallelism of the surfaces involved .
For such point contact we require a more delicate means of perceiving contact than the mechanical one ; hence the electric-touch method is employed .
It has been developed by the writer in a series of researches since 1900.* B. Cylindrical Gauges.\#151 ; The flat faces of the jaws touch the cylinder with two line contacts at opposite ends of a diameter of the cylinder .
Non-planeness and non-parallelism of the jaw faces , as also imperfections in the cylinder , introduce errors , though these are less serious than for Class A. Thus , for accuracy , line contact should be superseded by point contact , the measurement to be made between lines or edges on the jaws , the lines being not parallel with the axis of the cylinder .
C. Sphere or Bar with Spherical Ends.\#151 ; The face of each jaw touches the sphere at a point .
Since the jaw faces are imperfect , error can only be avoided by providing that the contact of the two faces with the sphere always occurs at the same places .
Thus , for accuracy , contact may be made , as usual , between flat faces of the jaws , if the surfaces be made true and contact be always at the same points on the flat faces .
From these remarks it appears that , for each kind of gauge , measurement should be made by point contact , the jaws to support no part of the weight of the gauge , and the end thrust on them to be reduced as much as possible .
The method described below fulfils these conditions , and has in addition the advantage of being more sensitive than the old method .
S II.\#151 ; Description .
1 .
The first machine made was derived from a much less sensitive instrument described previously.^ A massive bed and pillar carry a bracket , in which works a vertical micrometer screw with divided head and vernier .
The nut is double , and has a tightening spring between the two parts .
The screw shaft carries a pulley of the same diameter as the divided head ( 6 inches ) .
The top of the screw shaft passes thoroughfare through a plate rigidly screwed to the bracket , the plate serving to steady the axis of the screw .
A heavy pulley stand has a string passing round its three pulleys * See 4 Pliil .
Mag. , ' December , 1900 , and March , 1901 ; 4 Electrician , ' March , 1900 , and March , 1901 ; 4 Roy .
Soc. Proc. , ' 1903 , 1904 and 1905 ; 4 Nature , ' September , 1905 .
t Shaw , 4 Phys. Rev. , ' March , 1903 .
342 Dr. P. E. Shaw .
An Electrical Measuring [ Dec. ] and going to the large pulley on the screw , so that the observer works the pulley handle instead of handling the screw direct .
The advantages of not handling the screw system during a measurement are obvious .
The gauge being on the glass plate and under the screw point , the screw is rotated , and when its top point touches the top point of the gauge an electric circuit is completed and a telephone sounds , as it does whenever the screw point touches or leaves the gauge .
To find whether the gauge is uniform , one point after another on its top surface is brought below the screw point , and the micrometer is read for each place ; the gauge is reversed , the former top face now resting on the glass surface , and readings are taken at various places as before .
Whilst these processes are difficult for a cylinder or sphere , they are inaccurate for bar gauges , since the end of the gauge on which it rests is more or less convex ; the gauge does not rest firm on the glass and errors arise .
Experience with this instrument showed that , while retaining point contact , the electrical touch , and the pulley system of working , we ought to support the gauge in such way as to leave both surfaces free , so that electric contact can be made on each surface .
2 .
The latest form is shown in accompanying figure .
In general appearance it is somewhat like the usual mechanical-touch machines mentioned in S I. There are two headstocks and a table in the centre , all resting on a massive cast-iron bed .
The headstocks each carry micrometer screws and nuts with 1905 .
] Machine for Engineering Gauges and other Bodies .
343 graduated heads , and these measure the gauge , which rests on and is clamped to the table .
The gauge being clamped on the table , is set true with respect to the axes of the micrometer screws , by adjustments of the table .
The left screw point is brought into electric contact with the gauge ; then the right screw point is brought into electric contact with the gauge , and when the current passes through the gauge from one measuring point to the other , the two divided heads are read .
To turn the graduated heads the screw system is not actually touched by hand , but is worked by a hand pulley and string , the former being attached to the base of the headstock , there being a large pulley on and concentric with the head .
Details are shown in figs. 1 , 2 , and 3 .
The left headstock is a replica of the right headstock ; to save space , only the measuring point of the left headstock is shown .
Fig. 1 .
Gauge d rests on the table f and is clamped to it .
The table top can be brought into any desired plane by two rotations , as follows:\#151 ; 3 .
Rotation about a Vertical Axis.\#151 ; The pillar P carries a bracket b to which is fitted an adjusting screw a. The top of the pillar is turned cylindrical and fitting this cylinder is a collar which carries an arm .
A spring attached to arm and bracket keeps the latter always in contact with the end of the horizontal screw a. Thus the table top and gauge can be rotated either way by a small angle .
4 .
Rotation about a Horizontal Axis across the Red.\#151 ; The bracket b has a hinge on the left side , the top plate / hinging there and being held down to adjusting screw hby a spring attached to the bracket below and to the plate above .
The vertical screw h works in the bracket .
Rotation either way is thus provided .
344 Dr. P. E. Shaw .
An Electrical Measuring [ Dec. 1 , Besides the two rotations , the table has three translatory movements:___ 5 .
Translation Horizontally along the Bed.\#151 ; The table touches the bed at five places by studs let into base D of the table .
The table can be placed anywhere along the bed and clamped .
Plate D is grooved to fit part X of the bed ( fig. 2 ) and touches it by five studs , four on X and one on Y. 6 .
Translation Horizontally across the Bed._ Above plate D is a plate C\ ploughed out from front to back in a V-groove as shown on the left side ( fig. 1 ) ; the slide k which carries column A has three levelling screws , two resting in the V-groove and one on the plane on the right .
A spring is arranged to keep the slide pressing firmly on its feet , so that it has freedom only in a horizontal line across the bed .
7 .
Translation Vertically Up and Down.\#151 ; The column A is hollow and can slide up and down outside a pillar carried by k ; the pillar has a triangular cross-section , and the fit between column and pillar can be adjusted by four screws not shown .
The total weight of column , table and gauge is balanced by a lever with counterpoise weight on it .
The fulcrum of the lever can be set at any place desired on a separate pillar .
None of these counterpoising arrangements are shown in fig. 1 .
In order to raise or lower the gauge , the observer works the lever up or down on its fulcrum , this slide motion being very free .
Both the two latter translations have millimetre scales attached , so that the amount of any translation can be known .
Also , after the table is moved , it and the gauge on it can be brought back accurately to any desired place .
8 .
The five adjustments mentioned are sufficient for setting the gauge true .
As will be seen later the rotations take some time to perform , but they are seldom required ; the translations , being slides , are quickly performed .
These slide movements are quite satisfactory .
Screw movements and then rack and pinion movements were tried and found defective ; the first being slow and the second irregular .
In figs. 1 and 3 is shown a nut A , working in bearings Bh B2 and having the screw S in it .
A steel cone n is screwed into the right end of nut A , and bears against the stop m. There is a helical spring F which presses forward against bearing JE\ and back against the nut forcing the cone n against stop m. The point of n lies truly in the axis of nut A and the front face of m is ground truly plane and by a special device is made accurately normal to the V R Fig. 2 .
1905 .
] Machine for Engineering Gauges and othei ' Bodies .
345 axis of the nut N. By this means when the nut is turned it should have a true rotational motion , without any periodic to-and-fro translation along its axis , such as occurs with the usual coned or flat bearings .
Subsequent calibration of the screw shows how nearly this ideal has been approached .
Fig. 3 .
*M F \gt ; 0 N Fixed to nut N is a graduated wheel Q and a double vernier g is shown attached to the casting R. The wheel is centred true with the nut axis by means of the double vernier in the usual way .
The casting R has a front part T to which an upright plate U is screwed .
This plate carries the bearing bracket V on which runs the yoke q and the pulleys rly r2 ( fig. 3 ) .
9 .
The screw S ' would rotate with the nut if free , but as the yoke q which runs on the bracket V is clamped to the screw , the latter acquires a simple translatory movement along its axis , in or out of the nut , according as the latter rotates right- or left-handedly .
The screw spindle carries an index mark by which the position of the screw in the nut can be seen on the fixed scale S ( fig. 3 ) .
In this micrometer system of rotating nut and translating screw , it is essential for accuracy that the nut have no translation and that the screw have no rotation .
The former condition should be achieved by the cone end method of working ( see para .
8 above ) ; the latter by making yoke and bracket rigid and ensuring by a weight that the former presses the latter with constant force .
Backlash and looseness between micrometer screw and nut are minimised 346 Dr. P. E. Shaw .
An Electrical Measuring [ Dec. 1 , by the pull of weights t\ and t2 ( fig. 3 ) from which pulley strings pass over pulleys Ti and r2 to the yoke ends .
By this means the screw is pulled back into the nut by a steady force .
10 .
The casting R is electrically insulated from the plate W by a mica sheet and by having the screws which bind R and W bushed with ebonite .
The base of plate W is grooved to fit the part X of the bed and presses that part by four studs ; one other stud presses on the flat part Y. Thus the headstock rests firmly on the bed at five points of support and has only one degree of freedom , along the length of the bed .
Fixed on the side of W is shown a bar V on which the hand pulley Z can be pulled forward and clamped , so as to take up slack in the pulley cord .
11 .
The two micrometer screws and nuts are cut and ground with great care on the plan adopted by H. A. Rowland , * and when examined under the microscope the screws appeared highly polished and regular .
The calibration , described below , shows that there is a small periodic movement at each rotation of the nut , probably due to its bearings being excentric with respect to its axis .
But this movement is perpendicular to the line of measurement and the resultant errors in the micrometry are of a lower order and probably negligible .
Fig. 4 .
The screw-threads are about 2 inches and the nuts about 4 inches long , so that the screws do not leave the massive nuts at any point in the run .
* See article on " Screw , " 6 Ency .
Brit. ' 1905 .
] Machine for Engineering Gauges and other Bodies .
347 Steadiness in temperature of the screw is thus obtained .
The screw diameter is 1 cm .
, the pitch J mm. , the graduated head has 500 divisions , and the vernier reads tenths ; so one vernier division corresponds to 1/ 10000 mm. ( or 0*1 ( i ) in the micrometry .
The bed is 5 feet long and weighs 200 lbs. ; it rests on three studs , two 1 foot apart under one neutral line and one under the other neutral line .
A voltaic circuit consisting of a cell Ch a resistance box Bh switch x , and a telephone Th are joined to binding screw li on the bed and to another binding screw l2 on each headstock .
The switch is put to right or left according as one wishes to make contact between gauge and left screw or between gauge and right screw .
A better arrangement of circuit is to use two mercury throw-over switches A , B , with the circuit as shown ( fig. 4 ) .
On the right is the gauge ( ?
, with the measuring ends B and L of the micrometers shown .
To make contacts , use connections as shown in the table below :\#151 ; Contact .
Switches .
L , G r , a L , a and G , R L , R 1 and 3 2 and 3 1 and 4 1 and 4 The micrometer microscope Mi rigidly mounted on the right headstock ( fig. 2 ) is used for reading the standard invar scale 8 % .
By this means , as in the Pratt and Whitney measuring machine , the end standards may be compared with line standards of length .
S III.\#151 ; Materials Used .
The bed is of cast iron .
Most of each headstock is of cast iron in one piece ( see illustration , p. 342 ) , but the front plates and brackets and the base plates of headstocks and table are of wrought iron .
Brass is used for the nut bearings and for most of the table .
The micrometer screws are of silver steel , and the micrometer nuts of bell-metal , which is very hard .
The part of the screw spindle projecting from the front of the nut is of first grade invar .
All caps and fittings on the screw ends are also of invar .
Since invar cannot be ground very true , the faces of the caps are of thin steel .
The idea is that these projecting parts , being of small dimensions and necessarily uncovered , are more liable to temperature changes .
The 348 Dr. P. E. Shaw .
An Electrical Measuring [ Dec. 1 , consequent errors in the micrometry are small , since the expansibility of invar is so minute .
The terminal points of the screw spindle are beads of iridio-platinum , which , being hard and non-oxidisable , is the best substance for electric-contact work .
The beads are continuous , with a short piece of iridio-platinum wire , which was hammered into a hole drilled in the end of the invar .
S TV.\#151 ; Adjustments and Tests .
Before the machine is used for actual measurement many adjustments have to be made to ensure accurate working .
In this preliminary work defects in design or make can be detected and rectified .
1 .
Testing the Planeness of the Bed and Setting one Screw Axis Parallel to the Bed.\#151 ; There are three surfaces on the bed on which rest the studs of the uprights .
Two surfaces form an inverted V , and the other surface is horizontal ( fig. 2 ) .
The edge formed by the first two surfaces should be parallel to some third straight line in the third surface .
Place a card on the invar end of the left screw , remove the right headstock , and place a microscope with axis roughly parallel to the bed .
Set the cross-hairs of the microscope on some very small object on the card ; run the headstock along the bed about 1 inch , and work the screw back till the small object is again in focus .
If the image does not come exactly to the cross-hairs , the studs at the base of the headstock are screwed in or out till this condition is obtained .
These studs fit specially tightly in the headstock , and are adjusted by a special spanner .
Eepeat the above double operation in various parts of the bed ; if satisfaction is obtained the right micrometer screw axis is parallel to the ridge formed by the A surfaces , and the three bed surfaces are sufficiently nearly plane .
2 .
Setting the Two Screw Axes Collinear and Arranging that the Bead Ends Meet at their Extreme Points.\#151 ; First set the right screw axis parallel to the bed as above ; next run one headstock along the bed until the bead ends meet .
Set a microscope up horizontal with its axis perpendicular to the bed and focus it on the contact of the bead ends .
Arrange the electric circuit so that the telephone sounds when the beads touch .
If , when the beads are seen to meet in the microscope , they are also heard to meet by the telephone , the adjustment is complete .
The studs at the base of the headstock are screwed in or out till the best result is obtained .
The microscope must next be placed with axis vertical so as to look down 1905 .
] Machine for Engineering Gauges and other Bodies .
349 on the contact from above and the adjustment repeated .
In performing this adjustment the two studs on the same side of the A must be screwed in equally so that the screw axis is moved parallel to itself across the bed ; if this be not done , the former adjustment will be destroyed .
Fig. 5 .
Under the most favourable conditions it is found that the telephone sounds slightly after the beads appear to meet , the distance moved on by the screw end being about 0*5 / jl .
It was found by putting an end-cap on one side ( fig. 5 ) , and viewing with a microscope again , that the same lag , 0*5 / jl , occurs as before .
Hence this amount of movement is required to complete electric touch after the surfaces appear to be in contact .
Unless the bead ends are set carefully , as above , a constant + error will occur in measuring the actual gauge length by a line standard .
3 .
Testing for Backlash of Micrometer Screw.\#151 ; A steady gravity pull is arranged to keep the screw back in the nut .
On trial , backlash is found to be small , and need not be considered ( except in the calibration work ) , for readings are taken at both ends when in contact with the gauge , and the end pressures suffice to eliminate looseness .
4 .
Setting the Two Translations of the Table Perpendicular to the Bed.\#151 ; As will be shown later , it is possible to make a cylindrical gauge very true both in straightness of axis and constancy of diameter .
Having tested a cylinder and found it nearly perfect , it can be used as follows for setting the table .
Mount it on the table with its axis horizontal and across the bed .
Two processes follow :\#151 ; ( a ) Set the gauge perpendicular to a nut axis .
( b ) Set the translation of the table parallel with the gauge axis .
Dr. P. E. Shaw .
An Electrical Measuring [ Dec. l ( a ) Mount on the right screw end , E , a cap , having a projecting excentrie point ( fig. 6 ) .
Unclamp the yoke from the screw , so that on rotating the Fig. 6 .
nut , the screw spindle turns with it and the excentric point describes a circle perfectly normal to the nut axis .
Rotate the gauge on a vertical axis by the rotating screw of the table , until the excentric point just touches it , as shown by the telephone , on passing at points a , 1 .
The gauge axis is now perpendicular to the screw axis .
( 6 ) Remove the cap with excentric and clamp the yoke to the screw , care being taken to make the linear scale and divided head scales agree .
Bring up one bead end to make contact with the cylinder , the table and gauge being moved up and down so that the latter makes bare contact in passing .
Read the micrometer head .
Run the table and gauge across the bed to various positions , making contact and taking readings .
The cylinder being true , we can at once see if the translatory movement of the table is straight and perpendicular to the bed .
Set the line of this run parallel to the gauge axis , by adjusting the studs at the base of the table until the micrometer readings agree at the ends .
This throws out adjustment ( a)\#151 ; so ( a ) and ( b ) are repeated alternately till each is completed .
If intermediate readings are different from those at the ends , the V-groove at the base of the table is not straight .
In the measurement of a cylinder later ( see Table III ) the nonstraightness of the groove is shown by irregularity in Column A. In that table two consecutive readings 5 mm. apart differ by 10 ya , at most thus the angular displacement of the gauge is 1/ 500 radian = 7 ' and versin 7 ' = 0*000002 .
Thus the error produced , due to non-straightness of the groove , in measuring this gauge is , at worst , 1/ 500000 x 20 mm. or less than 2/ 5 of a vernier division , which is negligible here ; the error produced would only 1905 .
] Machine for Engineering Gauges and other Bodies .
351 be important for a long gauge .
The groove should be ground until it satisfies the above tests better .
Having adjusted the cross horizontal translation the vertical translation can be done in like manner .
The gauge is put on the table , with axis vertical , the excentric point and the rotating screw being used until the gauge is perpendicular to the screw axis ; the rest of the adjustment is obvious , the three levelling screws being used for adjusting the column .
5 .
Testing for Flexure of the Bed.\#151 ; In finding the actual length of a gauge , one headstock is moved , causing change in the flexure of the bed .
( In ordinary comparative work of testing and comparing gauges , the headstocks are stationary , and no error , due to this cause , arises .
) Put a plane-faced gauge on the table .
Bring up each screw end to touch its faces .
Read one screw head , say the left , and then move the right headstock away on the bed to a definite distance .
Any change in the contact reading of the left end is due to change in flexure of the bed .
From a series of such observations the error for any movement can be allowed for .
In the present apparatus it amounts to 3 / jl in some cases , the bed not being strong enough for the heavy headstocks .
But little error can enter into the micrometry , for the correction to be made in any case is known .
S Y.\#151 ; Calibration of Micrometer Screws .
Each screw is tested in two ways , ( a ) for long run , by comparison with the units on a line standard of length , ( b ) for short run , by interference bands:\#151 ; ( a ) An invar standard metre , made by the Society Genevoise , is read by a microscope on the right headstock ( fig. 2 ) .
The headstock is given small movements along the bed by a jockey screw in the usual way .
On the end of the left micrometer screw is placed a cap with a plane face as in fig. 5 .
Bring the end of the right screw into electric contact with this cap. Set the microscope cross-wires on a line of the scale , and by the iockey screw move the right headstock back 1 mm. of the scale , and run the micrometer screw forward , by rotating the nut , to make contact again .
The error in 1 mm. run of the screw as compared with 1 mm. of the scale is thus found .
The process is reversed and then repeated at any part of the two screws .
It is found that the probable error of one microscope reading is about 0*3 / jl , whereas the probable error in making the electric contact is about 0*05 / jl .
The tables of results need not be given .
The average run per mm. for 7 mm. of the right screw was found to be 1*0003 mm. ; for the left screw the result was 1*0002 mm. The line standard is here taken to be correct .
The above process gives us the average pitch of the screw and also the 352 Dr. P. E. Shaw .
An Electrical Measuring [ Dec. 1 , behaviour of the screw in different places .
In an attempt to make a perfect screw it would be a means of finding whether or not the screw and nut are in accord .
Thus , in this machine , both right and left screws give distinctly larger runs for every second millimetre .
For instance , the right screw has the following values for consecutive mm. : 1*0001 mm. , 1*0018 mm. , 0*9979 mm. , 1*0011 mm. , 0*9991 mm. , 1*0017 mm. By separate measurement it can be shown that this periodic error is not due to the line standard .
Fig. 7 .
A B ( b ) Sodium light is used from a bunsen burner , the heat of which is carried away by a funnel and stove piping .
On each screw end is fixed a cap with plane end , to which is screwed a microscope slide .
On one slide A is a small lens ( fig. 7 ) mounted by Canada balsam .
The sodium light falls on a glass slide at 45 degrees , and is reflected to the microscope as shown in the figure ( which is a plan ) .
The Newton 's rings formed between the lens and slide Bi are caused to open from or close to their centre according as one screw end is moved to or from the other screw end .
This plan of calibration is not satisfactory if the screw or nut be bent or " drunk , " since then the plate B moves to and from A at each rotation in a periodic way .
But the method indicates any such defects in screw and nut , however small .
Curve I ( fig. 9 ) obtained thus shows that the right screw-nut system acts as if bent .
From the curve we can find the extent of this rhythmic error .
The length AC ( fig. 7 ) is 50 mm. , the to-and-fro movement 1905 .
] Machine for Engineering Gauges and other Bodies .
353 on the curve is 15 / i. The angular movement of the screw in this to-and-fro movement is 15/ 50000 radians = 60 " .
The length of screw spindle from the bead-end to the nut is 100 mm. , so the diameter of the circle described by the screw- point is 30 / jl .
This movement , being perpendicular to the line of measurement , is immaterial .
Fig. 8 .
L To avoid any such to-and-fro movement , and obtain the true screw calibration\#151 ; the plan adopted is to have the interference produced at a place on the axes of the screws as in the elevation ( fig. 8 ) .
The left screw end has a rod of invar K screwed to it ; another invar rod L is screwed to K. On the end of L is screwed a microscope slide .
The right screw-end cap has a small lens mounted on it .
Set up the slide A to reflect sodium light , view with the microscope Mi the Newton 's rings formed between S and the lens surface .
The micrometer scale is read for every 50 rings passing the cross-wires ( the length traversed is thus 14*73 / x , taking \ = 0*589 / jl ) .
The interfering surfaces are never more than 200 half wave-lengths apart\#151 ; so that the rings are quite clear always .
Also , as the surfaces are so near , the number 0*589 / jl may be taken without considering the exact values for the Di and D2 sodium lines .
As soon as 200 bands have been read , the left screw is advanced to bring the interfering surfaces into contact again .
Then proceed as before .
Curve II is the result , in which the periodic error of Curve I has vanished .
Curves I and II refer to position 25\#151 ; 26 mm. on the right screw .
Dr. P. E. Shaw .
An Electrical Measuring [ Dec. 1 , Fig. 9 .
.120 .OOO Curve W Curve 1L Curve IK Microns Curves III and IV refer to 15\#151 ; 16 mm. and 25\#151 ; 26 mm. respectively on the left screw .
The scale of abscissae is an open one , so that plotting in this way is a very severe test of a screw .
Each millimetre requires about 80 readings , each of 50 bands .
Where a 1905 .
] Machine for Engineering Gauges and other Bodies .
355 curve looks irregular , readings are repeated , but in nearly all cases the readings were found correct to 0*05 / / , .
The millimetre of the screw corresponding to Curve II is more regular than any other tested , the length of run of the screw for 50 bands lying in all the 80 cases between the narrow limits 14*72 jjL to 14*90 ya. These calibration readings are very exact , and can almost always be repeated to X/ 20 = 0*03 micron .
Higher accuracy could be attained by special precautions , but gauge work does not call for it .
Some precautions taken in this interference work are :\#151 ; ( 1 ) Leave the apparatus with sodium light going half an hour before taking readings , to allow the bands to become steady .
( 2 ) Interpose several thick wooden screens between burner and machine , allowing only a narrow line of light to fall on the surfaces .
( 3 ) Avoid backlash by working only one way before a reading .
( 4 ) Count rings by twos , thus reducing the eye-fatigue by nearly one-half .
( 5 ) The invar and glass surfaces and two divided circles , are the only parts not covered by thick layers of felt .
( 6 ) The angle of incidence of the light on the interfering surfaces must be 90 ' , within 2 ' , in which case the error ih the readings is less than 1/ 1000 in the value of \ which is immaterial if the results are not used in a cumulative way .
It should be noticed that it is impossible to have error due to a miscount of bands in such an accurate apparatus as this , since one count never differs from the next by more than 0*1 ya , generally much less , whereas error of one band would produce a sudden change of 0*3 ya and would be at once detected .
S YI.\#151 ; Applications and Results .
The method of connecting the electric circuit differs according to the object to be tested .
In S II something has been said on this point .
1 .
Bar Gauges with Flat Ends.\#151 ; These are measured between two points .
The gauge is put on the table and clamped .
It must first be set so as to have one face normal to a screw axis .
If after performing the adjustment described in S II the two translatory movements are found to be straight lines , the gauge can be set by them .
If not , the excentric movement is used for each gauge .
As will be seen from the contour curves following , the flat ends are never true planes , but for the above purpose the gauge is taken to be set when the contact readings on all points on a circle near the edge of the face are identical .
The flat faces are roughly 6 mm. across , the left screw point is brought into electric touch with the gauge at the centre of the left face , marked VOL. LXXVII.-----A .
2 C 356 Dr. P. E. Shaw .
An Electrical Measuring [ Dec. l 1 in column A ( Tables I and II ) .
The right screw point is then brought up to touch the right face , the circuit being now arranged to pass from the left screw through the gauge to the right screw .
The two micrometers are read as in columns B and C. Then the table and the gauge with it is moved 1 mm. to the left , the contact is now made at place 2 in column A. The micrometer head readings are taken and entered in columns B and C as 25 mm. / / / " ; i .
: blank ' / / 25 mm. '\ 6d Set ii blank 25-mm .
Contours .
1905 .
J Mo , chine for Engineering Gauges and other Bodies .
357 before .
The whole face , except near the edge , is thus tested in L3 symmetrical places .
W \Ui Set i W Set ii 150-mm .
Contours .
Itesults follow for two new gauges , made by the best firms , in Tables I , II .
The numbers in columns B , C , and in other tables below , are those read 2 c 2 358 Dr. P. E. Shaw .
An Electrical Measuring [ Dec. 1 , on the micrometer heads , so that they give only comparative values from place to place , and would have to be all changed alike to represent the absolute values of the length of the gauge .
Column D is the sum of columns B and C ; the larger the sum the less is the thickness of the gauge on the line in question .
The small disagreement between the results for the two sets is shown in column A. The gauge is well covered throughout the readings . .
For the 25 mm. gauge , Table I , the errors arising for any one position lie between 0 and 0*4 / / , .
These errors can be attributed mostly to bad polish of the surfaces .
Grooves are visible and small changes in setting the surfaces would bring the measuring point now over a ridge , now over a hollow .
The errors in reading the micrometer can be ignored , for if any observation be repeated before the gauge be moved , the difference in the reading never amounts to more than 0T / a , generally less .
Thermal expansion may produce a small deformation , since about f hour elapses between a reading-in Set i and the corresponding one in Set ii .
The differences in the gauge thickness in different places amount to 3*6 / i. To show the nature of the gauge , contour figures are drawn , one for each set .
These curves are not contours for one surface in the usual way , but represent the joint effect of the two end faces of the gauge .
Table I.\#151 ; 25 mm. Gauge .
A. Set i. Set ii .
a \#151 ; TV \#151 ; T ) B. C. D = B + C. B ' .
or .
D ' = B ' + C ' .
1 116 *5 147-9 264 *4 117 -6 146-9 264 -5 + 0T 2 117 *5 146-3 263 -8 117 -4 146 -7 264-1 4-0 *3 3 117 *3 146-9 264 -2 118 -2 146 -3 264 -5 + 0-3 4 117 *9 147-7 265 -6 118 -7 146-9 265 -6 o-o 5 117 -0 149-9 266 -9 118 -0 148-9 266 -9 o-o 6 117 -0 147 -4 264 -4 117 -5 147-1 264 -6 + 0-2 7 117 -8 147-6 265 -4 118-4 147-3 265 -7 + 0-2 8 166 -6 147-5 264-1 117 -9 146-6 264 -5 + 0*4 9 118 0 147-5 265 -5 118 -1 147-7 265 -8 + 0-3 10 118-0 147-5 265 -5 117 -7 147-9 265 -6 + 0T 11 117 -7 146 -0 263 -7 117 -3 146-6 263 -9 + 0-2 12 118 -9 148 -5 267 *4 119 -5 147-9 267 *4 o-o 13 119 -0 147 -5 266 *5 119 -6 147-1 266 -7 + 0-2 The units are microns ( toVo mm*)* In the 150 mm. gauge differences are slightly more , \#151 ; OTya to \#151 ; 0*5/ 4 ( thermal expansion exercises more influence for long gauges ) than for the 25 mm. gauge .
1905 .
] Machine for Engineering Gauges and other Bodies .
359 Table II.\#151 ; 150 mm. Gauge .
A. Set i. Set ii .
A = D'\#151 ; D. B. C. D = B + C. B ' .
C ' .
D ' = B ' + C ' .
1 395 -0 96 *2 491 *2 395 -0 96-0 491 -0 \#151 ; 0 *2 2 397 *0 93 *8 490*8 396 -0 94-5 490-5 -0*3 3 396 *0 94 *9 490 -9 395 -0 95 -5 490 *5 -0-4 4 398 -7 93 *5 492 -2 398 -0 93 -9 491 -9 -0*3 5 \#151 ; .\#151 ; \#151 ; \#151 ; \#151 ; \#151 ; 6 394 *0 98 *7 492 -7 396 -0 96-6 492 -6 -0T 7 396 0 95 *5 491 *5 396 -0 95 -0 491 -0 -0-5 8 398 *3 96 -7 495 -0 399 -0 95 -5 494 -5 -0-5 9 \#151 ; \#151 ; \#151 ; \#151 ; I \#151 ; \#151 ; 10 396 0 93 *7 489 *7 396 -0 93-4 489 -4 -0-3 11 397 *0 93 *4 490 -4 397 5 92-8 490-3 -0T 12 398 -0 91 *7 489 -7 399 -0 90-5 489 -5 -0*2 13 396 -0 92 *9 488 -9 396 -0 92-7 488 -7 \#151 ; 0*2 The units are microns .
But differences in the gauge thickness from place to place amount to 5*8 [ jl .
The contour figures are given .
The faces are not normal to the length of the gauge , and no readings can be taken at places 5 , 9 ; when in these positions one screw end touches the gauge , the other screw end does not make contact on the other face , but , if continued , would meet the gauge on its side The actual length of this gauge was found to be , at 18'*5 C. , 150*0332 mm. and 150*0325 mm. in two distinct evaluations , taken at place 1 , Table II .
2 .
Cylindrical Gauges.\#151 ; These are measured between lines or edges .
The screw ends have cylinders or edges mounted perpendicular to the screw axes .
A preferable method is to work between the screw points and to move the table carrying the gauge up and down past the screw points , so as just to establish electric touch in passing .
In this way the straight lines mentioned above are virtual lines due to the passage of the measuring points past the gauge .
Before commencing measurement the axis of the gauge is set perpendicular to the screw axes .
Put the gauge on the table , its axis being horizontal and across the length of the bed ; take a contact between the left screw end and one end of the gauge .
Move the gauge by the table till its other end makes contact with the left screw end .
If the two contacts have identical micrometer readings the gauge is set .
The results for a f-inch gauge are shown in Table III .
Headings are made along the gauge at six places from 0*5 cm .
to 3 cm .
from one end .
After one set of six readings the gauge is rotated on its axis to positions 45 ' , 90 ' , 135 ' , 180 ' from the original one , and for each position six readings are taken as before .
360 Dr. P. E. Shaw .
An Electrical Measuring .
[ Dec. 1 , Set i takes about three-quarters of an hour and is completed before Set ii is commenced ; corresponding measurements are in one line .
In no cases do the result of the two sets differ by more than + 0*2 / / , .
The gauge is a very good one ; the differences in various places are not more than 0*7 / x , though there is a distinct tapering in every position , the handle end being thickest .
There is a slightly different angle of taper in the different positions .
The 180 ' position results are almost identical with those for the 0 ' position , as they ought to be .
Table III.\#151 ; f-inch Cylindrical Gauge .
Centimetres from end .
Set i. Set ii .
A = D'-D .
A. B. D = A + B. A ' .
B ' .
D ' = A ' + B ' .
0 ' ... 0*5 236 *9 197-8 434 *7 233 *5 201 -2 434 -7 o-o 1*0 242 *0 192 -6 434 *6 238 -0 196 -7 434 -7 + 0T 1*5 241 T 193 -5 434 -6 238 -0 196 *7 434 -7 + 0T 2 0 234 *4 199 -8 434 -2 231 0 203 T 434 -1 -0T 2-5 226 *6 207 *6 434 -2 223 -0 211 *2 434 -2 0 0 3*0 235 *3 199 -0 434 -3 231 -0 204 -2 434 -2 -0*1 45 ' ... 0-5 237 *6 197 -0 434 -6 231 -0 203 -6 434 -6 o-o 1 *0 243 *0 191 -7 434 -7 238 *0 196 -5 434 -5 -0-2 1 *5 241 *0 193 -4 434 -4 236 *5 197 -8 434 -3 -0 1 2-0 234 0 200 *2 434 -2 230 -0 204 -0 434 -0 -0-2 2 5 220 *0 208 -3 434 *3 223 -0 211 *2 434 -2 -0 1 3 0 233 *0 201 *3 434 -3 230 -0 204 T 434-1 -0*2 90 ' ... 0*5 234 *0 200 -6 434 -6 230 0 204 -6 434 -6 o-o 1 *0 240 *0 294 -4 434 -4 236 -0 198 -5 434 -5 + 0T 1-5 240 *0 194 -3 434 *3 237 0 197 *5 434 '5 + 0-2 2 0 233 *5 200 -8 434 -3 230 -0 204 -2 434 -2 -0 1 2*5 225 *0 209 -1 434*1 223 -0 211 -2 434 *2 + 01 3 0 233 *0 201 T 434 T 229 '0 205 *2 434 -2 + 0T 135 ' ... 0*5 234 *0 200 *5 434 *5 231 0 203 *4 434 -4 -0*1 1 *0 240*0 194 -6 434 -6 238 -0 196 -5 434 *5 \#151 ; 0 *1 1*5 240 -0 194 -3 434 -3 236 -0 198 -3 434 *3 0*0 2 0 233 0 201 -0 434 0 229 -0 205 -2 434 -2 + 0*2 2-5 225 -0 209 -2 434 *2 221 0 213 -3 434 -3 + 0 *1 3-0 234 0 200 *3 434 *3 229 -0 205 *2 434 *2 -01 180 ' ... 0*5 235 -0 199 -7 434 -7 225 0 202 -6 434 -6 -0*1 1 0 237 -8 196 -8 434 *6 238 -0 196 -5 434 -5 -01 1 *5 239 -0 195 *4 434 -4 238 *0 196 -4 434 *4 o-o j 2*0 232 -0 202 T 434 T 230 -0 204 -2 434 -2 + 0*1 2*5 224 -0 201 -2 434 -2 222 -0 202*2 434 *2 o-o 3*0 232 0 202-2 434 *2 230 -0 204 -2 434 -2 o-o The units are microns .
Another cylinder tested was a f-inch " gravity feeler " ( see Table IV ) .
1905 .
] Machine for Engineering Gauges and other Bodies .
361 Table IV.\#151 ; 4-inch " Gravity Feeler .
" Millimetres from end .
B. c. D = B + C. B ' .
O ' .
D ' = B ' + C ' .
A = D'-D .
2 125 0 233 -3 358 *3 121 *0 237 *2 358 *2 -01 4 119 0 239-1 358 T 117 -0 241 *1 358 -1 o-o 6 113 9 244-3 358 *2 115 *0 243 *2 358 -2 o-o 8 115 0 243 -2 358 *2 118 *0 240*3 358 -3 + 01 10 122 -0 235 -6 357 *6 124 -0 233 *8 357 *8 + 0-2 | 12 132-0 227 -2 359 *2 132 *0 227 *3 359 -3 + 01 The units are microns .
This cylinder should be specially perfect .
The first 6 mm. are fairly uniform , but after that the variations amount to 1*5 / * .
Still this and the f-inch gauge are much more perfect than the bar gauges with flat ends .
3 .
Spheres.\#151 ; These are measured between planes .
In theory this can be done by glancing contact on the same plan as for cylinders , but in this case it must be done in two directions : ( a ) up-and-down ; ( b ) to-and-fro across the bed .
To do this would be to use a virtual perfect plane containing the two straight lines above ; no errors would occur due to an actual material plane .
But the easier method is to employ end caps with plane ends and to measure between these two planes .
In this case there are two preliminary operations : ( a ) measure the imperfection of the planes by methods already indicated , and , as far as possible , make the planes true ; ( b ) set each plane perpendicular to the same screw axis and therefore parallel to one another .
The latter operation is performed thus : The two translations of the table being adjusted as in S IV , para .
4 , if the translations are not true straight lines the excentric cap is used , but if they are true a cap is put on the left screw point and adjusted by screws till the readings on the cap are the same at all points on a vertical and horizontal diameter .
This is done by clamping a rod with contact point on the table and moving table and rod to-and-fro and up-and-down as usual .
Next a cap is put on the right screw point , the rod is reversed on the table , and the right cap set in the same way .
Now since the cap surfaces are perpendicular to the same axis , they are parallel to one another even if the two screw axes be not quite parallel .
The actual measurement of the spheres is as follows : Bring up the left cap to produce electric touch with the sphere at the centre of the former .
Bring up the right cap. Bead both micrometer heads .
Proceed in this way for various diameters of the sphere , care being taken that the table does not move during the measurements .
By such means we shall always make 362 Dr , P. E. Shaw .
An Electrical Measuring [ Dec. 1 , contact on the same points on the caps , and small defects in planeness or parallelism will not vitiate the results .
The results for two 1-inch standard steel balls are shown in Table Y. The diameter of each sphere differs in various places by not more than 0*6 g , whereas the mean value for sphere a is 1*0 n greater than that of the other .
Table Y.\#151 ; Comparison of Two Standard 1-inch Balls .
I Position .
a. 0 .
A. B. A + B. A. B. A + B. a 190 *0 252 *7 442 *7 191 -0 252 *5 443 -5 b 189 *0 253 *5 442*5 191 -0 253 -0 444-0 c 189 -0 253 *7 442 -7 191 *0 252 -7 443 -7 d 190 -0 252 4 442 -4 191 -0 253 -4 443 -4 Mean 442 *6 Mean 443 -6 The units are microns .
* 4 .
Bar Gauges with Spherical Ends.\#151 ; As a rule the radius of the spherical ends are half the length of the bar .
The method of plane end caps is used as in para .
3 above .
5 .
Testing the Planeness of Surfaces.\#151 ; In previous sections methods are given for gauging the accuracy of a plane metallic surface .
6 .
Testing Non-Metallic Substances , e.g. , Glass Plates.\#151 ; In some apparatus ( e.g. , echelon gratings ) it is important that the two surfaces of a glass plate should be parallel .
The following is a method for this purpose:\#151 ; * Put an end cap on the left screw end and fasten to it by an ebonite plug ( fig. 10 ) a steel wire having a binding screw b at one end and a small steel sphere at the other , the latter being just clear of the front surface of the cap. Instead of joining the circuit , as in fig. 4 , bring one wire to b and the other to the headstock .
When the screw is advanced c touches the glass first and is then pushed back to touch the end cap , completing the circuit through the telephone as usual .
The same fitting is put on the right screw end .
The rest of the micrometry is obvious .
As Table VI shows , the accuracy is slightly less than in the work on conducting bodies .
* The method of Fabry and Perot ( 'Ann .
de Chimie , ' vol. 25 , p. 98 , 1902 ) is more sensitive than the above method in a test of parallelism , but it does not give a measure of the thickness of the glass plate , as does the above .
1905 .
] Machine for Engineering Gauges and other Bodies .
363 Fig. 10 .
Table VI.\#151 ; Optically worked Glass Plate .
Centimetres from end .
Set i. Set ii .
A = D'-D .
A. B. T\gt ; = A + B. A ' .
B ' .
D ' = A ' + B ' .
0*5 329 *0 100 *8 429 *8 329 *2 100 *8 430 -0 + 0*2 1*0 345 *8 84 *8 430 *6 345 -2 85 '5 430 -7 + 01 1*5 373 T 58 *8 431 -9 373 *5 58 *4 431 *9 o-o 2*0 399 T 33 *2 432 '3 397 *5 35 *6 432 -1 -0*2 2*5 424 *3 8*9 433 *1 421 *8 11 *4 433 -2 + 01 3*0 463 *2 -30 T 433 -1 463 *2 -30 *4 432 *8 -0*3 The units are microns .
This plate is somewhat wedge-shaped along the line of measurement selected ; it is the most perfect plate tested .
SVII.\#151 ; General Remarks on the Methods .
1 .
The contact surfaces are polished with dry rouge and chamois leather , clamped on the table , and left there for half an hour at least to allow temperature to settle .
2 .
There is no spark at " break " of such length or intensity as need be 364 Electrical Measuring Machine for Engineering Gauges , etc. considered , the potential difference of the surfaces can be made very small , say 1/ 20 volt ; the circuit resistance is 1000 ohms .
3 .
It is best to bring up right and left screws two or three times before taking readings , since the table .yields slightly before the screw , but finally stops due to the small back pressure of the other screw .
4 .
When measuring cylinders by a glancing contact , there is a small noise as of make-and-break made when the surfaces are about 3 ya apart .
This ( unexplained ) sound warns the observer , the true loud contact being made at the proper place with certainty .
5 .
The gauges used are made by the best firms .
If bar gauges with flat ends cannot be considerably improved ( see S VI , para .
1 ) they should be rejected unless strict point contacts are used as in this paper .
But it is impossible to use point contact except with a delicate method such as that of electric touch .
6 .
No stress has been laid on the actual length of the gauges tested , though this could be performed with an accuracy at least as great as with usual methods .
But on the comparative work on gauges shown above , stress is laid , as that work is unusually accurate .
The probable error in a single reading with the electric touch is about 0*05 ya. 7 .
In a new machine some improvements in detail should appear:\#151 ; ( a ) the beds should be stronger and the headstocks lighter , to reduce flexure of the bed ; ( b ) the screws could be improved by a longer application to them of Rowland 's cutting and grinding processes ; ( e ) the end caps , etc. , may be replaced by fittings which screw on the end of the screw-spindle , thus forming a more rigid and more easily-fastened system ; ( d ) the interference bands might be from sources of cadmium , mercury , and hydrogen as used by Michelson , Pulfrich , and Tutton respectively ; ( e ) a stiff spring might be put in one or both headstocks , so that , if contact be overrun , no vital part of the micrometer would be strained ; ( / ) the table should be more massive .
The table plays a very important part in the work , and accuracy depends more on its rigidity than one would expect without actual trial .
But the general form of the table with its quick-slide movements seems quite satisfactory .
The author wishes to thank the Royal Society for the grant , which defrayed the greater part of the cost of this machine .

