the origin. Distances measured vertically are termed ordinates, and those horizontally, abscissæ.

(3) Carefully note which set of readings have to be plotted on the ordinates and which on the abscissæ, and then choose the scales of the axes OP and OQ such that they are as long as possible and include the maximum values to be plotted. Also, if possible, arrange such that one of the smallest divisions represents a simple whole number of one digit. For example, if 33 was the largest number to be plotted and the side of the squared paper contained 100 divisions, let 1 division represent 0·5 only, whence 66 will give the 33; this is far more convenient a scale for future reference in obtaining intermediate values than 1 division representing 0.33 (i. e. 99 to give the 33 approx.). While it is a great advantage for the numerical length of the axes to be as large as possible, so as to enable the curve to be drawn larger and more accurately, the length should be decided by considerations of future reference to it for intermediate values as just mentioned.

(4) The axes must be numbered every 10th division, and under no circumstances with the numbers obtained from experiment. (5) Write along each axis the nature of the quantity plotted on it. (6) Each point must be plotted by finding the point of intersection of the axes representing the two corresponding quantities under consideration at the moment and a distinctive mark there made.

(7) When all the points are plotted, a mean curve, as shown by the full line Fig. 1, must be drawn through as many points as will allow of a uniform line being drawn.

Some points are always sure to lie on either side of this mean line and denote experimental errors. The object of the curve is to correct for these.

(8) In some tests, as for example in “characteristic" determinations with direct current generators, it often happens that curves cross one another and lie close together. In such cases they must be drawn thin and a different notation for the respective sets of points used, such as that represented in Fig. 2.

All confusion will thus be avoided.

00000

FIG. 2.

(2) Calibration and Standardization of
Electrical Measuring Instruments.

General Remarks.-This subject is perhaps one of the most important in connection with electrical testing, and we shall therefore devote some considerable attention to it. It will at once be obvious to any one, without any consideration, that a measuring instrument which is reading incorrectly, or one that has been calibrated so long ago that its present readings may not be true, is a useless instrument, or even worse than this, as one is unconsciously liable to take its reading as correct. The importance of correct reading and accurately calibrated measuring instruments cannot be over-estimated, for on their being so hangs the whole crux of further testing, the results of which would otherwise be quite worthless. This the author would emphasize most strongly, for it is unfortunately his experience, and no doubt that of many more like him, that the average experimentalist is only too ready to take the scale reading of any instrument as correct without in the least troubling himself as to whether it actually is so or not. This no doubt arises from the little extra trouble required to calibrate such instruments prior to starting some particular test.

Measuring instruments may change their constants and develop errors in their scale readings either from continued use, abuse, or in transit from one place to another, some of course being much more susceptible to alteration than others. Hence in all cases where it is desired to obtain accurate results and do good work, the instruments should be re-calibrated and restandardized frequently, and a calibration curve drawn whenever possible with the date of the test inserted. At the very least six determinations should be made, wherever possible, but preferably ten or twelve, as it is not possible to draw a reliable calibration curve on less than six points. In all cases it is of the utmost importance to see that the connecting wires or cables do not magnetically affect the instruments, for it must be carefully remembered that a wire carrying a current, no matter whether it is straight or otherwise, acts as a magnet.

Such inductive effects will be minimized by running or twisting the "lead" and "return together, when the two equal and opposite magnetic effects neutralize. An ordinary flexible twinlead is non-magnetic externally, but it possesses a very small electrostatic capacity.

Instruments are usually calibrated by comparing their readings with those of very accurately calibrated standard instruments. Simultaneous readings must be taken on both to avoid errors due to variation in between. Ammeters are always connected in series and voltmeters are always connected in parallel with their standards.

In the calibration of voltmeters, the employment of keys in any of the branched or parallel circuits containing voltmeters to be calibrated is usually a source of inconvenience and should be avoided, for a key which places, say, a voltmeter of 1500 Ohms resistance in parallel with a similar instrument already reading, will cause this reading to decrease owing to the alteration of the P.D. at the terminals due to inserting such a low resistance meter, and the consequent reduction in the terminal combined resistance.

(3) Calibration of an Ammeter by comparison with a Standard D'Arsonval Ammeter. Introduction.-When a standard current measurer, such as a Kelvin standard balance or a potentiometer set, is not available for comparing the ammeter to be tested with, the following method of calibration may conveniently be employed. It consists in using a good reflecting D'Arsonval galvanometer in conjunction with a low resistance composed of platinoid or other suitable material having a small temperature co-efficient of resistance, which should preferably be known. The resistance of the D'Arsonval galvanometer may conveniently be something like 2000 to 4000 times that of the low resistance to which it is shunted. instrument, its scale, and the resistance should be permanently fixed and standardized carefully by means of a copper or silver voltameter. Then if the current which produces a full scale deflection, with a certain known resistance in series with the galvanometer, is accurately known, the current producing any

The

other deflection with the same resistances will be very approximately in direct proportion and therefore at once known. Some slight corrections might be necessary for great accuracy when subsequently using these particular constants, due to alteration of resistance through change of temperature and to the deviation of the D'Arsonval readings from the direct proportional law, for which correction see Appendix, p. 306.

Apparatus.-Secondary battery B capable of giving the

maximum current required; reflecting D'Arsonval galvanometer G; switch S; key K; carbon rheostat R (p. 393); resistance box (); ammeter A to be calibrated; low resistance PP either of the form shown (p. 401), or simply a sheet of the metal.

N.B. It is assumed that the galvanometer and low resistance in combination has been carefully standardized previously and now constitutes the standard D'Arsonval ammeter.

Observations.—(1) Connect up as in Fig. 3, and adjust the pointer of A to zero and the spot of light from G to the left-hand end of the scale used as a temporary or false zero in this

test.

(2) Insert the proper resistance in () as given from the constants of standardization for the maximum current to be measured and corrected for the temperature of the room at the time of the test.

(3) With R large, close S and adjust the current through the ammeter to be calibrated to about th of the maximum scale reading by means of R. Then note simultaneously its reading A and the deflection D on G when K is pressed.

(4) Repeat 3 for about ten different readings on A rising by about equal increments to the maximum with no decreasings of

current.

(5) Repeat 3 and 4 for a similar descending set of the same readings on G, noting the corresponding ones on A, avoiding all increasings of current, and tabulate your results as follows

(6) Plot curves having values of true current (a) as ordinates and A as abscissæ.

Inferences. Enumerate any sources of error in ammeters generally. What can you infer from your experimental results? Why should the current be so carefully increased only in 4 above and decreased only in 5 above?

(4) Calibration of an Ammeter by comparison with a Kelvin Composite Balance used as a Centi-ampere Meter.

Introduction. The following is a convenient and ready means of calibrating any ammeter reading up to 1 ampere, employing a Kelvin composite balance used in the manner mentioned, as a standard for comparison. A complete description of the construction and manipulation of the instrument will be found on p. 358, to which a reference should be made and the constants obtained therefrom.

Apparatus. Kelvin composite balance K.B. (p. 358); ammeter A to be tested; switch S; adjustable resistance R (p. 399); source of current C at a P.D. of from 40 to 60 volts.

K.B.

(2) Turn the switch in front of the balance to "volts" so as to place the fixed and movable fine wire coils in series with each