UNCERTAINTIES 1.

Understanding uncertainty in the mean In the lab program you will have taken a series of readings of the period of a pendulum
clamp

pin chuck

retort stand length string

brass cylinder

Figure 1. Experimental apparatus for period measurement of a simple pendulum.

To reduce uncertainties the measurement you measured the time for 20 swings instead of the time for 1 swing. Why/how would this reduce the uncertainty? The table of results obtained is as follows:
Table 1

Length Time for 20 swings (0.5) (0.05 secs) cms 10 12.6, 12.4,11.9,12.8,12.7 20 17.5,17.7,17.8,18.0,17.9 30 21.5,21.8,21.2,22.1,22.0 40 25.3,25.5,25.4,25.4,25.2 50 28.1,28.7,28.4,28.3,28.5
TABLE 1

T(secs) (NOTE: 1 swing only) 0.624

T (secs) (also for 1 swing only) 0.009

Determine the mean of these values, the range and the uncertainty in the mean .

The way we write the result of our measurements and calculations is mean uncertainty in the mean (with units, of course) SIGNIFICANT FIGURES quoted in the mean is determined by the value of the uncertainty in the mean. The number of significant figures quoted in the uncertainty in the mean is usually one except in the case where its most significant figure is 1 in which case the number of significant figures quoted is 2. For example, suppose for this experiment you obtained a mean value of 1.6873497 s and an uncertainty in the mean of 0.0356273 s, then the result would be written as 1.69 0.04 s. If the uncertainty was 0.0162918 s then the result would be written as 1.687 0.016 s. The BEST ESTIMATE of the true value of the quantity you are trying to measure is the mean of your values. The spread in measured values is described (rather poorly, but we will fix this later) by the range. However, the spread in the best estimate of the true value (ie the mean) is measured by the uncertainty in the mean. This is because if you were to repeat the above measurements you would obtain a second mean value, this would differ from the mean of the first set of measurements by an amount of the magnitude of the uncertainty in the mean, not the range. Types of Uncertainties (Errors) Absolute (or actual) error - the uncertainty in the measurement could be a limitation of the measuring device and/or the uncertainty obtained from taking a mean of a set of readings. x Fractional (or relative) error the ratio of the absolute error to the measured value, i.e. x

Percentage error the absolute error expressed as a percentage of the measured value, i.e. 100

x x

Combining uncertainties When values which have uncertainties are combined with other values with uncertainties we need to be able to calculate the overall uncertainty produced. There are 2 rules to remember: a) if quantities are added OR subtracted the absolute uncertainties are ADDED b) if quantities are multiplied OR divided the fractional or percentage uncertainties are ADDED. Three points to note: i) if a quantity with an uncertainty is multiplied by a quantity, such as a constant, which has zero uncertainty then the percentage uncertainty remains constant which is equivalent to multiplying both the value AND THE uncertainty by the quantity. ii) if squaring a number the result has twice the percentage uncertainty of the original number. If taking the square root of a number the percentage uncertainty is reduced by a factor of 2. iii) Uncertainties should only be determined to 1 (or at the most 2) significant figures, so you should be able to do the calculations in your head. If one uncertainty dominates an equation, you can disregard the other uncertainties as insignificant as compared to the dominating uncertainty. Otherwise no uncertainty should be disregarded just because it is small it means that you can calculate the quantity to high accuracy. In our SIMPLE PENDULUM example above, the expected relationship between the acceleration due to gravity, g, the period of a pendulum, T, of length, l, is given by:

With the data we have obtained for this , we would normally plot a graph of T vs . However if you wished to use just one of the lengths of string for a calculation of 'g', you would rearrange the equation as:

and calculate the uncertainty in 'g' from the percentage uncertainties:

g l 2 T 100% = 100% + 100% g l T

Examples of Combining Uncertainties: Given at X = 10.00.05, A = 4.50 0.45 and B = 4.00 0.08 determine the value of Y (stated to the correct no. of significant figures and with the accompanying uncertainty) a) Y = X B (ANS: 6 .00.13) 3X b) Y = AB 6(X + A) c) Y = (ANS:21.71.7) B d) Y = A 2 BX (ANS:810180)

X e) Y = A B

1 2

f) Y = ( A 2 B 3 + X 2 ) 2 (ANS: 7.50.6 )

GRAPHS Note: Plotting graphs by computer is excellent IF YOU KNOW WHAT YOU ARE DOING. You will use a computer graphing technique much better if you know the manual operations particularly with regard to finding the line of best fit to a set of data, and hence modelling the data. In the lab we require you to do all graphs by hand. For reports and posters we were happy for students to use the computer but often the usage was poor. In this exercise we suggest you A) plot your own graph, manually, including all plottable error bars (see note b) below). B) optionally replot the same data using computer but make sure you know how to include uncertainties in both the data points and in determining the slope and intercept. 1. Graphing a Set of Data a) graph the following data which has been taken of the photoelectric effect.

The expected equation relating frequency and the maximum kinetic of the emitted photoelectrons is: KE max = hf W where h is Plancks constant and W is the work function. You should plot the data to obtain a value of h and W to the maximum accuracy which is possible from the data. Note: a) Slope and Intercept These are obtained from the line of best fit where the equation fitting the data is of the form y=mx+b b) Error Bars are the individual uncertainties in the experimental points. These should be plotted IF THEY ARE SIGNIFICANT on the scale of plot chosen. If they are too small to be plotted they should nevertheless be recorded so that they can be included in your uncertainty calculations. Sometimes the individual uncertainties seem to be larger than would be expected from the good fit of the data. Priority in fitting uncertainty lines should be given to the actual data points as compared to the error bars. If the data points give a good fit (i.e. little twist in the LBF) then the uncertainty should be small c) Uncertainties in Slopes upper and lower Just as you have an uncertainty in taking individual readings, you have uncertainties in plotting the line of best fit or LBF. These uncertainty lines indicate the highest and lowest slopes that are possible from fitting a line of best fit. You need to "estimate" these limits. The uncertainty in the slope is the greater deviation from the LBF of the two limit lines. The slope should then be quoted WITH ITS UNCERTAINTY AND WITH ITS UNITS. d) Linearising Data The most analytical treatment we are able to make of data is if it can be represented by a linear curve fitting process. Curves can be very informative as to how , e.g. pressure varies with volume but quantitavely they are hard to interpret. The following set of data is taken of a quantity of gas kept at room temperature (300K) but the pressure and volume are measured at various intervals of increasing pressure.

You should

i) ii)

iii) iv)

plot the informative graph of pressure vs volume. According to the ideal gas equation PV = constant * T, you would not expect a straight line for a plot of P vs V. Do a second plot, with pressure on the vertical axis and a "function of V" on the x-axis, which should produce a line fit graph. Uncertainties will have to be recalculated for the new axis variable (using rules for manipulating uncertainties done in the first assignment). If they are large enough to be plotted they should be done on this graph, to assist in plotting the line of best fit and determining its uncertainty. From your graphing determine the relationship between p and V at constant temperature.

PART OF THE PRACTICAL EXAM will include a question in the style of : A Physical Modelling Paper in 2004. A set of springs is holding up a flat table top, which tends to start vibrating when a certain amount of weight is placed upon it. Physics students decide to investigate the properties of the springs to see whether they can predict the weight to be avoided. Knowing something about Hookes Law, i.e. F = Cx, where C is a constant,the students decide to measure the compression of the springs when a load is placed on the table. Their results are tabulated below. Reading x 0.05 cm m 0.2 (kg) 1 0.10 0.2 2 0.27 0.5 3 0.33 0.9 4 0.45 1.4 5 0.60 1.3 6 0.69 2.2 7 1.2 3.4 In this table, x is the vertical displacement of the table top, and m is the load placed on top of the table. i. [1] Draw a sketch of the experimental arrangement the students may have used ii. [2] Using one of the x values and the corresponding m value from your table, estimate the value of C (DO NOT estimate the uncertainty in this value). iii. [1] Remembering their practical classes the students decide to use a graphical method to find C, rather than a simple calculation as in part (ii)What extra information can they obtain in this way? iv. [4] On a sheet of graph paper plot m versus x (with error bars). Do the results support the students understanding of Hookes Law? v. [4] Using the graph, estimate the stiffness factor of the spring C, including any information you can obtain about the uncertainty.in the value. vi. [1] Summarize the findings of the experiment. Note: graph.

Hand drawn graphs are needed, because that is what you will be required to do in the exam Make sure you know which variable is on which axis Make sure you plot uncertainty lines if there is sufficient uncertainty to plot Calculate your slope FROM POINTS ON THE GRAPH LINE, not from individual data points Your conclusions should relate qualitatively and quantitatively to the data and the results you obtain from your