You are on page 1of 12
ee eas Pee eee eee eee erate ee ee ee i A TTT SMT TOTTI NE LTT ERRORS OF OBSERVATIONS ANO SIGNIFICANT FIGURES In 1894, Lord Rayleigh, in the course of a series of accurate deterininations of the densities of common gases, observed that although a litre of nitrogen derived from the air weighed 1.2572 gm, an equal volume of nitrogen prepared from its compounds weighed only 1.2506 gm. This apparently small difference is, however, beyond the limit of experimental error involed in the determinations and so some other factor was sought. After a large number of experiments by Ramsay, Rayleigh, and others, it was revealed that this discrepancy was due to the presence of a hitherto unknown element of the air which was called Argon. The estimation of errors of observations thus led to an important discovery. Our senses and judgement may be trusted upto certain limits, beyond which they begin to subject to errors. The study of the law of distribution of errors is of importance because it allows us to form an estimate of the accuracy with which, under given conditions, the measurements can be made. | Absolute accuracy is, of course, unattainable in laboratory measurements but some estimate of the experimental result can usually be formed as to the magnitude of the possible error. An experimental result of unknown reliability is of no importance. Hence an estimate of the accuracy of measurements is very desirable. Inaccuracy of the result may arise from several different causes, 1. Mistakes may arise due to recording of a wrong number -which can be avoided by care on the part of the observer. (ii) 2, Instrumental error arise from imperfections in the work of the instrument maker in constructing and subdividing the scale used by the observer These may be decreased by using more accurate instruments or b calibrating the scales of the instruments used. We shall usually assume that the accuracy of the instrument is such that the instrumental errors are less, than the errors of observations. 3. Systematic error refers to a perturbation which influences all measurements of a particular quantity equally. Systematic errors are apt to arise when some indirect method of arriving at a result is adopted, a direct method being difficult or impossible. For example, the wave length of light cannot be measured directly and a method depending on diffraction or infference is usually employed. 4. Errors of observations arise due to the inherent limitations of the observer's Power of observing and judging. It is necessary to distinguish between the quantity to be determined and the observations that are made in order to compute it. In all cases, measurements are made by observing scale readings, While taking readings, one has to judge whether a specified point is nearer to one or the other graduation mark on the scale with the convention that if it is half-way or more between two consecutive graduation marks, then the one with higher value shall be recorded. In certain experiments, observations may be made either by null method or by deflection method. It is to be remembered that null methods are, in general, more accurate than deflection methods in which the magnitude of a quantity is measured by reading the position of a pointer on a scale. In the null method, the effect of an unknown quantity is balanced against the effect of a standard quantity of the same type. The resulting effect is observed which has to detect the small difference between the two effects. In the case of null method, the accuracy of adjustment to equality with the standard can be increased as may be desired by increasing the senstiveness of the instrument. On the other hand, if the effect of either quantity is measured directly, an instrument is to be used which provides a moderate deflection (say) when subjected to the whole effect or an instrument of relatively low sensitiveness. As a result, an unavoidable error in reading the deflection would give an appreciable effect on the result. Thus null method is preferable over deflection method if accuracy of measurement is considered. As for an example, let us consider the measurement of mass. A spring balance based on the principle of deflection method, produces an extension of the spring due to the weight of the body whose mass is to be determined. Suppose the spring balance produces an extension of 2 cm for 25 gm. If the least count of the spring balance be 1 mm, the accuracy in the measurement will be 0.1 cm in 2 cm ie., only 5 per cent. On the other hand, if a physical balance is used, which is based on the null method, the mass will be balanced against standard masses and will be accurately known upto 0.01 gm (supposing 10 mgm to be the smallest weight). Thus the accuracy in the second case is very much higher than in the first case. eee (iy itis obvious that greater the sensitiveness of the method, smaller is the quantity which it can be used to measure. For example, a micrometer microscope and a gerew gauge are sensitive instruments for the measurement of length, but these serot be used 0 measure a large length of, say, 20 cm. For the measurement of Sich large lengths, a less sensitive instrument such as a meter scale should be used. put on the other hand, if a small difference between two lengths is to be measured accurately where one is the accurately known standard length, the micrometer seroscope can be profitably used and thus the measurement may be made very accurate. Different methods of estimating the magnitudes of errors of observations may be employed, the choice depending on the nature of the instrument. In many cases, the quantity can be measured several times and the mean taken. In other cases, circumstances do not permit repetition. (@ Let us consider a case when only a single observation is made eg., the measurement of temperature of a cooling liquid with a thermometer. Suppose the least count of the thermometer is 0.1°C and the observer reads the temperature as 35.5°C. The statement that the temperature is 35.5°C implies that there is reason for supposing- that the temperature really lies between 35.45°C and 35.55°C ie, (@55#.05)°C. The observer therefore states the temperature as 35.5°C with a possible error of 0.1°C or about ‘0.3%. ()) When repeated observations of a quantity are made and mean of these readings is taken, the smaller the difference between the mean and the separate readings observed, the greater will be reliability of the mean as the best possible measured Value of the quantity. The average of the differences between the mean and the separate readings is called the mean deviation. It can be shown that when 10 observations are made, the probability that the actual error is greater than the mean deviation is very small, about 1 in 100, while if 15 observations are made, it is reduced to 1 in 1000. Even if 5 observations are made, the probability.is only 1 in 15. Hence, when a quantity is measured several iimes, the average deviation may be taken as a measure of the possible error. The possible error will depend on other factors besides the calibration of the scale one employs. One important factor is the sensitivity of the instrument which should be taken into account for the assessment of the possible error. In the Meter bridge Experiment, the scale which is used for measuring length of the wire may be calibrated in mm. But the galvanometer may not be very sensitive and as such for a certain range along the wire (say 4 mm) null deflection may be-observed. The Possible error is + 0.2 cm. In sact the position of balance is deduced by finding for what positions of the sliding contact, deflections (first one way and then the other) are just detectable’ and by using these readings a mean position can be found, together with the limits of error. Suppose in a particular case, the mean position of the sliding contact of the wire for null deflection determined in the way stated 7 (iv) above, is found on the scale to be 51.2 cm. But the possible error in this case mentioned above is 0.2. Hence the result of this Measurement may be reported as (51.2 + 1.20.2) cm. This implies that the measured quantity could lie any where between the limits (51.2 - 0.2) cm and (51.2 + 0.2) cm. Hence to increase acuracy in the result, it is necessary to pay attention to the sensitivity of the measuring The consideration of possible errors is Of great importance in deciding what care need be taken in determining the various factors in a complex measurement and what are the best conditions for obtaining an accurate result, observations for the same measurement are Tepeated a number of times and the arithmetic mean is taken, from errors. Let us call n the number of observations ; By 8, B5enna8y the deviation from the arithmetic mean ; then the mean error of a single observation 2 The probable error amounts to5 of this. final result. It is thus important to consider ow the error in the final result can be computed from the errors in the observations (@) Possible error of a sum : Let z= xy . where x and y are two measured quantities. Let x Tange between xy + 8x and %o — 8x, y range between Yo + 8y, and yo— 5, then the uncertainty in z can be obtained from % * b= (x9 # 8x) + (yy + by) and the maximum value of 82 is given by choosing similar signs throughout ie, Bz = Bx + By, As the uncertainty in the sum is the sum of the individual uncertainties ; this can be expressed in terms of relative uncertainties as f & be+ = by wu(2) z xty Perec () (b) Possible error of the difference of two variables : ES ey: +(3) Let As explained above 8z will be obtained from 2 + 5z = (xp + x) - (Yo + Sy). Here maximum value of 8z will be given by choosing negative sign for 8y so that B= bx + By. The relative uncertainty will be given by = ep ey (A) tay This shows that if (x—y) is small, the relative uncertainty will be very large. The measuring instrument should therefore have very small least count. The other way to avoid it is to measure the difference directly rather than obtain it by subtraction between two measured quantities. The results (2) and (4) can easily be obtained by taking logarithm of the expressions (1) and (3), differenciating and finally choosing the signs of the errors occurred in the various measurements so that the total error is maximum. To establish the above, let us again consider zex-y Taking logarithm to both the sides log z = log (x—y). Differentiating, = Zz x-y Maximum value of &z will be given by choosing negative sign for Sy, so that be _ berby Zz x-y (©) Possible error of the product of two or more variables : Let z=xy. Taking logarithm, we get log z = log x + log y Differentiating (z/z) = (6x / x) + (Byly). « A ane the maximum value for 8z/z is given by choosing similar signs for both (@) Possible error of the quotient of two variables : Let z=x/ly log z = log x - log y ~_ ss (vi) Differentiating, we get, (82/2) = (8x/x) - (Sy/y). In this case, maximum value of (82/2) is given by choosing positive sign for 8 and negetive sign for 8y so that (82/z) = (8x/x) + (8y/y). More generally if 2 xy" ey . log z = a log x, + b log y; - c log 2 —d log yp .., & 85, z 4 n Ye Maximum value of 8) will be given by choosing positive values of &; and 8, and negative values of St and By respectively. Thus, Differentiating ay 4 iy at z AO Pai FW. For given possible errors in x, ¥ and z ..., the effect of the possible errors in the Tesult is greater for higher powers of x yorz.. In other words, those quantities mu: st be measured more accurately which are (©) Some other cases of interest : Suppose x, %y yi and y2 are m easured quantities and another quantity z depends on them as shown by the equation Z™ XY + xp. . The possible error &z in z can be found as shown below in terms Of possible errors in x’s and ys 82 = 8 (xy) + 8 (xyp), = (bys + y18x) + Edy + y2bxp). () If z=sin x (8z/z) = cot x &x. (8) Similarly, if z= log x (82/2) = dx/(x log x) (©) Also if z=e (82/2) = & dy. When a measurement is made, one should simplest way of expressing, it is by writing the significant figures. Greater the accuracy in th number of significant figures. To take an exam) estimate its reliability and the result using the proper number of © measurement, the larger is the ple let us suppose that the results rere es aera eeeeee (wii) ts are reported as (i) 3.1 cm (if) 4.20 cm (ii) 5.314 cm using 2, of some aaa respectively In these cases results have been reported upto Sand 4 Sto which best possible estimates could be made. For example, in (i) the the digits pPestimate of last digit which is in the first decimal place of a cm, is 1 te Pothing is known about the digits in the second, third etc. decimal places. In whe experimenter is reasonably certain that the digit in the second decimal place ao similarly in (ii) the digits upto three decimal places could be estimated. ‘The errors in measurement result from natural and instrumental limitations and are quite unrelated to the number of significant figures. It should be borne in mind that writing a number to a given number of significant figures does not imply that the last figure given is known to an accuracy of + 1, eg., suppose a distance is measured as 1523 metre. This does not necessarily mean (1.523 + .001) metre the statement simply means that the best estimate of the last digit is 3 and nothing at all is known about the following digit. If the possible error in an experiment is known to be + .006, the final result is conventionally written by stating the possible error together with the result e.g., 1.523 + .006 metre (suppose). Very large numbers having many digits in decimal places can be written in a more compact way as a product of (i) a number between 1 and 10 and (i) a number which is a power of 10. For example 3370 = 3.37 x 103 0.156 = 1.56 x 1071 753,000 = 7.53 x 105. The numerical factor represents the significant figures and the power of 10 is called the order of magnitude. The use of significant figures sometimes invite difficulties which may necessitate a detailed study of the theory of errors. However, it should be remembered that the idea of significant figures helps us avoid misleading numbers and unnecessary calculations. When the values of a physical quantity is calculated by substituting different observed data in a formula, the students often find it difficult to decide the number of significant figures upto which the final result is to be reported. The protedure to be adopted is broadly as indicated below : (i) To substitute in the formula the different observed data upto correct significant figures in each case. (ii) To calculate the final value using a logarithm table and to express it as a product of a number between 1 to 10 and a number which is a suitable power of 10. (ii) To estimate the maximum error for each item of observation and then to estimate the maximum error in the quantity to be determined experimentally. (2) To use this maximum error for deciding the number of significant figures to be used in reporting the final result. — (wily These steps will become clear from the examples given below : [1] Determination of ‘Mechanical equivalent of heat (F)’ electrically Observations : Current (i) = 21 + 0.1 amp Voltage (E) = 29 + 0.1 volt Mass of calorimeter (m,) = 140.03 + .01 gm Mass of calorimeter + water (1m) = 297.11 + .01 gm Initial temperature (@,) = 214 + 02°C Final temperature (0) = 27.0 + 0.2 °C Time for which current flowed (1) = 780 + 1 sec Temperature of liquid after cooling it through the same time (03) = 25.0 + 0.2°C Specific heat of the calorimeter = 0.1 cal/gm/°C. (given)* Calculations : Mass of water (m) = (mz — mm) = 157.08 gm Radiation correction = (®) ~ 03)/2 = 1°C ¢; Final temperature with radiation correction = @ + (8 — 03)/2 = (27 + 1)°C = 28°C Eit 1 Gsm) 1p +0,~8,)/3 0] afl) : 2.9x2.1- 780 © (157.08 +140.03x0.1) (28.0— 21.4) = 4205 joule/cal Taking logarithm of both sides of (1), log J = log E + log i + log # — log (m + msy) - log [@) + (82 - 03)/2 - 0). Hence Differentiating, ay = SE, Bi, 5 8(m+mis;) 3 [0p +(6,-05)/2-0)] Pe Be egy m+ mys; 8 +(8 ~@3)/2-0; = SE HB Blom +m)-+ 516m _ 8p + (58, ~884)/2—50,] Ear ae; m+ms, © + (8) -03)/2-6, Maximum error §J in f will be given by choosing 88, and negative values of Sm, 8m, and 56. Thus, given by * When we use the values of well-known Phy \ es sical constants, we can neglect the consideration of error in these quantities as the values are known with sufficent accuracy. (ix) 8] BE, BB, Bra +m +51 Sm, | 80, +(60, +805)/2 +60, fig eRe Eire m+ ms, 8 + (0, ~8,)/2-0, etl .01+.01+1x.01 4p 2H(2+.2)/24.2 2.9 21 780 157.18 +140.03 «0.1 28-214 1 1 .021*, 6 = sett 29 21° 780° 71.18 66 = .03448 + .04762 + .001282 + 0001168 + .09091 = -1743 8 = .1743 x 4.205 = 0.7329 % error in J = 17.43%* The statement f = 4.205, implies that the value lies between 4.2045 and 4.2055 ie, the figure in 3rd decimal place is uncertain to the extent of 1. Hence it claime an accuracy of about 1 part in 4000. But the maximum possible error is 17.43% which means an accuracy of 1 part in 6. Thus it is obvious that by reporting the value of J as 4.205, we claim a far higher order of accuracy than attainable with the apparatus used. We should report the value of f using appropriate number of figures consistent with the accuracy obtainable in the experimental arrangement. To decide this, we note that on account of an error of + 0.732, the figure in the first decimal place in the value of f becomes uncertain and hence the figures in the second and third decimal places become meaningless. So, the figures in the second and third decimal places need not be reported. Thus the value of J in this case can be reported as J = 4.2 + 0.7 Joule/cal. (2) Determination of Young’s modulus (Y) by the bending of a beam See Expt, No. 5). Observations : Length of the beam (I) =9+1cm Breadth of the beam (b) = 785 + 001 cm Depth of the beam (d) = .785 + 001 cm 0505 + 0.0005 cm 80 cm/sec? (given) Depression of the beam for 200 gm (y) Acceleration due to gravity (g) {l'may be noted that if the error in a quantity Is less than 0.1%, its contribution toward the total error is not significant. Hence it may be neglected. , he OF the 17.43% error in the final result, a total of as much as 8.2% were contributed byte in observation of.the current and the voltage and 9.1% of the temperature: The “ould be improved by using more sensitive ammeter, voltmeter and therm ) Calculations : mgl® 200980 x90" Y Gbd3y 40.785 (0.785) 0.0505 = 1.862 x 10!2 dyne/cm? Taking logarihm, log Y = log m* + log g +3 log I - log 4 — log b - 3 log d - log y. Differentiating, vase iii aa ay Maximum error in y will be obtained by choosing negative signs for 5b, 5d and dy. Thus the maximum relative error is given by BY _38, & , 38d | By Y 1 b da y 3x.1 .001 , 3x.001 , .0005 ee CL 40008 90.785 = .785 = .0505 1 1 ax 5 = 300 "785 * 785 * 505 = .003333 + .001274 + 3 = .001274 + 5 x .001980 (From table on pages 395 - 396) = 01843 SY = .01843 = 1.862 x 10!2 x .03428 x 1012 We find that, on account of an error + 0.034, the figure in the second decimal place in the value of Y becomes uncertain and hence the figures in the third and fourth places becomes meaningless. Thus the value of Y should be reported as Y = (1.86 + 0.03) x 10!2 dyne/cm? (3) Determination of refractive index (Wt) and dispersive power («) for the material of a prism [See Expt. No. 25] Observations Angle of the prism (A) = 60° 46’ 20” + 20” x 1/180 Angle of minimum deviation D,, (red) = 48° 2’ 40” + 20” x 1/180 (yellow) = 48°30’ 20” + 20” x 7/180 (green) = 49°27’ 10” + 20” x n/180 *m is the mass of the slotted weights of known ma: . Error in m has been neglected assuming, . it to be negligibly small. yr (xi) Calculations S 1.6077 (red) sin (A +Dm)/2) sin (A/2) 1.6212 (Green) aking logarithm, log p= 10g [sin Ata -2s{sn 4] 2 oe = 1.6121 (yellow) *) _ S{sin(A+Dy)/2) _ 8 {sin (A/2)] uw) sin(A+Dy)/2 sin (A/2) _BA+BDy 605(A+Dy)/2_ 8A cos (A/2) 2 sin(At+D,)/2 2 sin(A/2) Hence the maximum relative error is given by (for red ray) (*| _BA+8D yr oop AtDm , 5A cot (A/2) Trea Differentiating, ( + 2 2 naa (A/2) we cot 54°24'30" +20, cot 30° 23" 10” 2 “180 = 00006443 + .00003245 = .0001068 (®W)req = -0001068 * 1.6077 = 0.0001717 We find that, on account of an error + 0.0001717, the figure in the 4th decimal place in the value of becomes uncertain and hence the figures in the 5 and 6th become meaningless. Thus the value of rea should be reported as trea = (1.6077 + .0002) Similarly, Uyettow = (1.6121 + .0002) and green = (1.6212 + .0002). Again, © = (Hy — Hr) / (Hy -2) = 02203 = 2.203 x 107. Taking logarithm, log © = log (Hy ~ Hs) ~ log (Hty - 1). fo _ Sig Br Bay @ Bg-Hy by-} Hence, the maximum relative error is given Py Bo ig, @ Py-tr by * Be = 0.0655 x 107. Thus, the result should be reported as ® = (2.20 + 0.06) * 107 Differentiating, = 02974. ee | (xii) (4) Determination of Mutual inductance of a coil [See Expt. No. 56]. Observations : First throw of the galvanometer (6;) = 13.9 + .1 cm Eleventh throw of the galvanometer (811) = 8.0 + .1 cm Constant deflection (9) = 3.5 + 1 cm Periodic time of the galvanometer coil (for 10 oscillations) Small resistance used (1) = 59 1/10 sec. (r) = 0.00053 ohm (given) Calculations : Logarithmic decrement = 2.308 105 =2208 0.05514 (log ®; — log 811) ~ 1 0.55141 M= dae a 2 = 2.048 x 103 henry. Calculation of errors : log 2 = log 2: Os 10g [log ® —log 611] 8A _ Slog a = log 611) _ (50; /6; A log 6 ~log 0 10g.9 “+ maximum value of relative error BA _ (861 /8)) + (861/611) _ (.1/13.9)+.1/8.0) x log8;-log®;~ —~—~—~=~«O399SS 01969 2.303 = "5399 % gg 2399 = 0.004528 , Again log M = log T + log r + log 6; — log 2n — log 6 - log (1 + 4/2) SM ST , 86 _ 3 _ 51 +A/2) epee Taeealo Nag aBR tg Maximum relative error is given by SM _5T , 80 , 8 | (6A)/2 Ayal 1, 0922 M T "6 14%/2 “55 “Tp9*35* 1.02757 0.05471. 0.109 = 10-3, 20 * 0.1) « 10% henry, 6M M

You might also like