You are on page 1of 39

Investigating the Behaviour of a Nonlinear Pendulum

By James Wright
SCN: 063334731 Name of Centre: Madras College Number of Centre: 5426537

Contents:
Introduction Summary Underlying Physics Experimental Procedure Procedure Experiment 1 Experiment 2 Experiment 3 Results Experiment 1 Sample Calculation Uncertainties Experiment 2 Sample Calculations Uncertainties Experiment 3 Sample Calculation Uncertainties Discussion Analysis of Results Experiment 1 Experiment 2 Experiment 3 Numerical Modelling Uncertainties Conclusion Evaluation of Experimental Set-up Modifications Made During Experimental Work Limitations of Equipment Further Investigation References Acknowledgements 28 29 30 31 36 37 37 37 37 38 39 39 14 15 16 17 19 21 22 25 27 8 10 12 13 3 3

Introduction: Summary: Three practical experiments were carried out: Measuring the time period as a function of the pendulum's length Measuring the time period as a function of the pendulum's amplitude Measuring the instantaneous damping rate as a function of the pendulum's amplitude

The first of the three experiments tests the validity of the well-known relationship between time period and length of pendulum (shown in Eq. 7), which is found to be extremely accurate where the pendulum's amplitude is small and dissipation is negligible. In the second experiment, the effect of large amplitudes on the time period is investigated. Agreement between the observed periods and mathematical analysis in the absence of dissipation is found, showing that time period is a function of amplitude. Experiment 3 investigates dissipation on large amplitude oscillations. The simple textbook treatment gives a damping rate that is independent of amplitude, which is found to be an oversimplification. The "instantaneous damping rate" was measured, after which numerical modelling was used to find an empirical formula for the drag force as a power series in speed. The latter result does not appear to have been published previously. Underlying physics: Newton's second law states that When applied to the pendulum, this gives (Where Fd is the drag force) (Eq. 1) Given a low velocity, Stokes' Law applies[1] and Fd is proportional to velocity, hence (where is constant) (Eq. 2) If the amplitude of the pendulum is kept small the following small angle approximation can be made (Eq. 3) In light of this, Eq. 1 can be re-written as follows

As this is a second order ODE (Ordinary Differential Equation) with constant coefficients, solutions are sought where (Eq. 4) Substituting this into the re-written Eq. 1 gives

As this is a quadratic in , the quadratic formula can be used to solve it

For realistic parameters

So the above equation can be re-written as

This means

and can be split into real and imaginary parts

Putting these values of lambda into (Eq. 4) gives

Using Euler's formula, this can be re-expressed as

This can be simplified, if the following substitutions are used: and and Giving (Eq. 6) Recalling that (Eq. 5)

Yields the familiar pendulum frequency

Therefore (Eq. 7)

If damping is neglected, but larger amplitudes are used, the small angle approximation can not be made, i.e. Newtons second law becomes

This means

and further

After some integration and algebraic manipulation, this can be expressed in the form of a separable first order ODE.
[2]

Where 0 is a constant of integration chosen such that when t = 0, the pendulum is at an extreme, and hence momentarily at rest. Separating the and t dependences gives
[2]

where T is the nonlinear (large amplitude) time period, so changes from 0 to 0 while time increases from 0 to T/4 (a quarter of a cycle). The right hand side can be written in terms of a Complete Elliptic Integral of the first kind, which is denoted by K. Hence
[2]

which gives the time period in terms of the amplitude of oscillation [2] (where T0 is the time period for small 0) (Eq. 8) This can be approximated as a Maclaurin series which is more convenient to use [3] (Eq. 9) The accuracy of this approximation is shown in the following graph:

Time period as a function of amplitude (normalised by the small amplitude period, T0) In this graph, the red line represents the exact solution given by (Eq. 8) and the black line represents the approximation given by (Eq. 9). This shows that at amplitude = 0, error = 0%, at amplitude = /4, error = 0.006% and at amplitude = /2, error = 0.4%.

If velocity is increased, the simple drag force equation (Eq. 2) is no longer valid, as the velocity becomes high, Fd takes the form: (Where Cd is the drag coefficient, v is the velocity of the bob, is the density of the air, and A is the cross sectional area of the bob) [1] However, Cd is a function of the velocity, this makes Fd very complicated, but it should be possible to express it as a Maclaurin series in velocity: [4] (Eq. 10) where a1 and a2 are constants which are empirically derived later on in the numerical modelling section, to match the results of Experiment 3. Note that the quadratic term is defined in a manner such that Fd always acts in the opposite direction to the velocity, instead of using v2, v|v| must be used. An example of an oscillatory decay is shown below. The red curve shows the amplitude envelope of the oscillation 0(t), which clearly decays in time.

The decay rate is defined as follows: (Eq. 11) This means that depends on 0 in general. However for the simple Fd as defined in (Eq. 2), which is proportional to velocity, the decay rate is independent of 0 as shown in (Eq. 5). For realistic damping, as defined in (Eq. 10), depends on 0, and may be thought of as an instantaneous decay rate (0) for a particular amplitude. If 0 is large, velocity will be too; this implies an enhanced Fd and 0 is expected to decay at a greater rate. Experiment 3 focuses on measuring (0) and modelling is used to determine the damping coefficients a1 and a2 in (Eq. 10) that describe the observations.

Experimental procedure
For a pendulum, a light inextensible string was connected to a ball-bearing 2cm in diameter (measured using a Vernier calliper). The bob was chosen as it had a high ratio of mass to cross sectional area, resulting in a relatively low damping rate. Light inextensible string

Sellotape

Ball bearing

This was suspended from a set-up consisting of a clamp-stand, a boss-head, a clamp and a Gclamp which clamped two thin pieces of wood together to provide a definite, even suspension point from which the length of the pendulum could be measured. This was then secured to a desk with two additional G-clamps.

Boss-head

Suspension point G-clamps Clamp-stand

A board (110x120cm) was reinforced with a wooden framework in order to make sure it was flat and graph paper was carefully glued to it to use for measurements of vertical and horizontal displacements of the pendulum required in experiments 2 and 3.

Board

To film the pendulum, a camera capable of filming 59.94 frames per second with a resolution of 1920x1080 pixels in progressive scan HD was set up with a tripod. To make sure the board was perfectly horizontal it was placed on two lab-jacks which were adjusted until an accurate spiritlevel showed it to be.

Spirit level

Experiment 1 Experiment 1 involved measuring the time period of a pendulum as a function of its length in the small amplitude limit. Small amplitudes were chosen to allow the use of the approximation shown in (Eq. 2). This approximation was required to provide a simple analytical solution as opposed to the more complex solution used in Light inextensible string Experiment 2, which required the use Pen mark of special functions. Consequently the set-up of this experiment was simpler than experiments 2 and 3 in that no equipment was required to measure the amplitude of the pendulum. The string was held taut across a metre stick and pen marks were added on at ten centimetre intervals from 10cm to 1m. The pendulum was then secured from the suspension point and pulled down until a mark was visible; this was repeated for each marking giving ten readings.

G-clamp

Suspension point Light inextensible string

To process this data, a spreadsheet was used. The frame numbers of two maxima were inputted into the spreadsheet along with the number of cycles that occurred between them. This was done using QuickTime, as shown in the following screenshot:

10

Frame number

A minimum of five cycles between two maxima were used in order to minimise uncertainty in time period. This was then converted into a time period. This time period was then squared and plotted against pendulum length. As an aside, the accuracy of this can be tested by using the gradient of the graph to give a value of gravity and comparing it to the accepted value. All of this is discussed at greater length in the Analysis of Results section.

11

Experiment 2 Experiment 2 investigates the relationship between period of oscillation and amplitude for a range of amplitudes between 0 and 90. As decay is being neglected, the length of pendulum was chosen to be 35.8cm (a relatively short length) to ensure small velocities and hence low drag force. This allows several oscillations to be filmed (to allow an accurate determination of the period) with a relatively small change in amplitude. To compensate for parallax errors, a metre-stick was checked to be completely straight and then set up horizontally in the plane of the pendulum for amplitudes < 45, and vertically for those > 45 to ensure minimal reading errors (this is discussed in detail in the 'uncertainties' section).

Metre-stick used to find position of pendulum Metre-stick used to find position of pendulum Lab jacks

Initially the pendulum was released by hand, but this would often result in the pendulum going into a plane that wasn't parallel to the board or in-line with the metre-stick, however using a clothes peg to release the pendulum provided a clean swing that stayed in-line. Lighting had to be adjusted as the shadow from the ball due to natural light obscured the markings on the metre-stick; this was achieved by using a lamp to illuminate the previously shaded area. After a large range of amplitudes had been filmed the footage was then analysed on a computer, from which data was extracted. The frame numbers of two maxima were noted, along with the displacements given the left and right edges of the ball-bearing when the maxima occurred, this is shown in the following screenshot:

12

Metre-stick used to find position of pendulum

The recording was paused at a maximum and the grid lines were used to find the readings on the ruler for the bob's position. The readings that corresponded to the left and right edges of the bob were noted. The data was then entered into a spreadsheet. Experiment 3 The procedure of Experiment 3 was very similar to that of Experiment 2 in that the same equipment and set-up was used, and the same measurements were taken from both. However Experiment 3 investigates damping rate, this means a large velocity of bob and therefore a large length of pendulum is required, so a length of 0.944m was chosen. Due to the fact that the string was longer, the arc length of the pendulum's swing covered a larger region of the board. For the sake of minimising parallax error, it was important to keep the camera was facing normal to the pendulum's maximum, to maintain this, the camera had to be moved frequently.

13

Experiment 1

( = Uncertainty)

14

Results

Sample Calculation:
Taking row 3, columns A - E are entered manually, column F is the time period calculated using the following formula:

(frame number) is calculated as column D - column C (1691 - 611) giving 1080 frames. The number of frames per second is constant at 59.94 fps, and the number of cycles is taken from column E, giving 20 cycles.

This result was squared in column H. Column G calculated the error in time period as

The number of cycles is once again taken from column E, this gives

Column I calculated the error in time period squared as follows

This gave

15

Uncertainties
Due to the fact that a camera was used to film the motion of the pendulum, there is no random uncertainty in identifying the duration of the oscillation as there would be if, for example, an operator used a stopwatch. To minimise uncertainty in time period, two frame numbers between a large number of oscillations were noted and the average time period was found. This is possible due to the fact small amplitudes are used (as previously discussed). The camera used takes one frame every 1/59.94s, if two consecutive frames at t = 0 s and t = 1/59.94 s are recorded, the greatest inaccuracy possible would be when a maximum of amplitude occurs directly between the two frames. In this situation either frame would be equally valid and the uncertainty would be 1/199.88 s. As the time period depends on the two frames selected as being closest to two separate maxima, each having a maximum uncertainty of 1/199.88 s, the formula used to calculate the uncertainty in time period is as follows

Therefore

The uncertainty in T2 was calculated as follows

Therefore

The reading uncertainty in length of pendulum was taken to be 1mm and a measurement uncertainty was approximated to be 2mm. It may be noted that the line on the graph (see 'Analysis of Results' section) does not go directly through the origin, it goes instead through (0,0.0087). This suggests that for a time period of 0s, the length of the pendulum must be -2.16mm. The reading uncertainty combined with the measurement uncertainty accounts for this:

Additionally, this offset could be attributed to the inaccuracy of the equation, as using the exact solution would shift all points to the right by a small amount. Another less likely cause would be a small systematic error in the metre stick used to measure the length of the pendulum. 16

(Continued on next page)

17

Experiment 2

18

Experiment 2 (Contd.)

Sample Calculation:
Columns A - K are entered manually, time period and time period are calculated using the same method used in Experiment 1. Column N calculated the horizontal position of the bob's centre at its first maximum using the following formula:

Taking row 4, this gives a horizontal displacement from the equilibrium position of

Column P calculates the position of the bob's centre at the second maximum using the same formula, this gives

Columns O and P calculate the uncertainty in both of these centres. The uncertainty depends on the reading uncertainty in the length of the pendulum and that of the offset, as these are both read from a metre-stick with 1mm intervals, the uncertainty in both quantities is taken to be 1mm, meaning that the uncertainty in both centres is given by

Columns R and T calculate the angular amplitude of the pendulum at a certain maximum, this is calculated as

For Column R this gives the amplitude of the first centre to be

For column T this gives the second maxima to be

Columns S and U calculate the uncertainty in amplitude, this is given in terms of the reading uncertainty in length and horizontal displacement

19

For Column S, this gives

For Column U it gives

Column V calculates the average amplitude over the observed interval as follows

This gives

Column X calculates the uncertainty in the average amplitude, this is done as follows

Giving

Finally column Y gives the analytical time period using (Eq. 9) where T0 is defined by (Eq. 7).

20

Uncertainties
The uncertainty in amplitude for <45 was derived as shown: (where l is the length of the pendulum and h is its horizontal displacement) The error in is related to the uncertainties in l and h through the sine function and its derivative [5]

For >45, tan() is replaced with cot(), and horizontal displacement is replaced with vertical displacement. Interestingly, the position of the bob at a maximum was determined by finding the position of the bob on the metre stick, and subtracting the offset from it. Both measurements have a reading uncertainty of 1mm in them. The textbook treatment of combining uncertainties on a linear scale states that the resultant uncertainty in the bobs displacement is 2mm (approx. 1.4mm). However logic dictates that if the reading of offset was read to be 1mm less than its true value, and the position off the bob on the metre stick 1mm greater than its true value - a case which is possible within the given uncertainties, the reading of the bobs displacement would in fact be 2mm off its true value. This contradicts the textbook treatment of uncertainties and would increase the total uncertainty in amplitude. The same principle applies to the uncertainty in measurement of time period, where the uncertainty in each measurement is one half of a frame time (1/119.88s) and the frame number of one maximum is subtracted from another.

21

22

Experiment 3

Experiment 3 (Contd.)

23

( = 'change in', i.e. initial value final value)

24

Experiment 3 (Contd.)

Sample Calculation:
Columns A - V are identical to those in Experiment 2. As dissipation is no longer negligible, the amplitude changes significantly over the course of several cycles. This means that there is an additional error in determining the true average amplitude on top of the uncertainty described in Experiment 2. Column X calculates the uncertainty in amplitude due to its value changing during the observation as follows

For row 20, this means

Column Y combines this with the reading uncertainty to give an overall uncertainty of

Giving

(Note that for smaller amplitudes, their is a greater difference between and ) Column Z calculates the change in amplitude between the two maxima as shown Giving Column AA calculates the error in as

Giving

Column AB calculates the rate of change of amplitude

Giving 25

Column AC calculates damping time as

Giving

Column AD calculates damping rate as

Giving

Column AE calculates the uncertainty in damping rate as

Giving

26

Uncertainties The uncertainty in amplitude due to reading uncertainties was calculated using a similar method as demonstrated in Experiment 2, however as becomes larger, it must be acknowledged that there is a random uncertainty between the true average amplitude and the mean amplitude, this can be ignored in Experiment 2 as is relatively small. The maximum value for this was calculated as /2, as stated in the sample calculation for Experiment 3. These two uncertainties in were calculated using the following formula:

Recalling that means

Recalling that

means

therefore

27

Discussion
Analysis of results:
Experiment 1 A graph was plotted of length of pendulum against time period of pendulum squared. A line of best fit was then inserted which gave values of the gradient and y-intercept of the straight line:

The error bars here were initially included, but were too small to be visible without magnification. The values of these errors can be found in the table of results for this experiment. To test the accuracy of this experiment and the applicability of (Eq. 8), the equation can be rearranged to find a value of the acceleration due to gravity

However as (Where m is the gradient of the graph) the formula can be simplified to As m = 4.0212 this gives g to be 9.81757ms -2. An online tool[6] was used to calculate the true value of acceleration due to gravity in St. Andrews; this gave a value of 9.82121ms-2, meaning that the calculated value given by this experiment has a percentage error of 0.04%. This shows that the equation (Eq. 8) gives a very accurate approximation for time period when very small (<10) amplitudes are used. 28

Experiment 2 The idea of using a large number of oscillations between two maxima was implemented as a means of reducing the uncertainty in time period. However it may be noted that the validity of using multiple cycles depends on amplitude: as amplitude increases, damping rate increases. This means that at large amplitudes, only a small number of oscillations between each reading is allowed otherwise the apparent average amplitude will not be the true average amplitude. This can be ignored in Experiment 1 as the time period is independent of amplitude for small displacements. However as the amplitude of the pendulum increases, fewer cycles must be taken between two maxima to prevent the uncertainty in amplitude growing. This explains the increase in the size of Yerror bars with amplitude in the following graph:

(Red line = solution given by (Eq. 9), Blue line = readings <45, Yellow line = Readings >45) The uncertainty in amplitude was too small to be visible without magnification. This shows that the values given by the experiment match up with the solution given by (Eq. 9) within the given errors.

29

Experiment 3 As in Experiment 2, a constant length of pendulum was used, the same quantities were recorded for both and they used the same method of obtaining amplitudes and time periods. Once these had been found, damping rate was calculated using (Eq. 11) and plotted against amplitude.

Here the straight blue line is a line of best fit between the points. To complement the experimental results, it is possible to provide numerical solutions of the nonlinear pendulum with realistic dissipation. It is important to realise that the analytical solutions in the 'underlying physics' section make approximations which limit their applicability. Indeed Experiments 1 and 2 were carefully set up to be described by the analytical formulae, but Experiment 3 requires a realistic drag force, and this means no analytical solution can be found. However, solutions can be found numerically to the equations with realistic drag by using software such as Maple. The following two pages give a print-out of a Maple program and the results it generated. The Maple program is a 'numerical experiment' with a virtual pendulum that can be used to study the amplitude decay between two successive maxima, and hence give the decay rate for the parameters chosen. In the Maple program plain black 'Roman' text corresponds to comments that have been inserted to explain what the software is being asked to do; lines beginning with > are lines of Maple program; the latter may be followed by the result Maple has calculated, such as a plot or blue italic text giving the value of a quantity just calculated.

30

31

32

(Eq. 10) shows that for small velocities, the v term dominates, whilst for large velocities the v|v| term will dominate. Hence the Maple program can be run for a very small amplitude of 0.001 rad which will give a small velocity and allow a1 to be adjusted to match the intercept of the straight line on the damping rate axis of 0.0022 s-1. Once a1 has been found, a2 can be determined by requiring the damping rate at 1.4 rad to be 0.0226 s-1 (according to the straight line fit of the data). Note that having adjusted the value of a2, the intercept of the damping rate axis will have undergone a slight shift and require a1 to be recalculated. This was repeated to obtain the new value for a2 and rapidly converged to consistent values. Once a1 and a2 had been determined, the Maple program was run several times for different initial amplitudes and the decay rate calculated according to (Eq. 11). The results are shown in the table below.

These results were plotted and are visible as the red points in the graph below. A line of best fit (red) was plotted and gave the same equation as the previous blue line.

33

Although the damping rate given by Maple was adjusted with a1 and a2 so that it agreed with the observations for = 1.4 rad and the intercept on the y-axis, this does not guarantee agreement for intermediate angles. The fact that the agreement is so good for all amplitudes indicates that the actual drag force is well represented by the two terms in the Maclaurin series. If there had been a discrepancy between theory and observation, it would suggest that higher order terms in the Maclaurin series were required. Evidently this is not the case. The two terms on the right hand side of (Eq. 10) are of comparable size when

for the coefficients determined by Maple. Indeed the plot below shows that the linear and quadratic terms are equal at v = 0.4ms-1.

34

The change in height required to achieve this speed can be estimated using the principle of conservation of energy

For a pendulum of length 944mm this change in height is achieved for an amplitude of 0.13 rad (7.56), and is indicated by the dashed vertical line in the previous graph. To the left of this line the a1v term is dominant whilst to the right the a2v|v| term dominates. Interestingly (as discussed in 'underlying physics') if Fd is proportional to v, then the decay rate becomes independent of amplitude and there is a suggestion that this can be seen in the small amplitude region of the graph. If the pendulum (length 944mm) is released from 90 conservation of energy suggests its speed will reach 4.30 ms-1 and the graph below indicates that Stokes' law no longer applies.

It is interesting to compare these findings with textbook descriptions of the drag force[1] which are generally characterised in terms of the Reynolds number. Reynolds number is defined as follows: (where L is a characteristic linear dimension eg. radius and is the kinematic viscosity of air, which is 15.11 x 10-6 m2s-1 [7] When Re < 1, Stokes' Law applies i.e. the drag force is proportional to velocity. Taking L = 0.01m indicates the velocity must be less than 1.5 x 10-3ms-1. Whereas when 103 < Re < 105, the drag force is proportional to v2. The definition above suggests that v must exceed 1.5ms-1 for the v2 term to dominate. This is in good agreement with the Maple result, which showed the two terms were comparable when v = 0.4ms-1, hence for greater velocities the v2 term will dominate. 35

Uncertainties The numerical results from Maple will have an error associated with them. It is possible to check the accuracy of these solutions by generating them for situations in which an analytical solution exists with which to compare. For small amplitudes, using (Eq. 3) and no dissipation, simple harmonic motion solutions are found of the form (t) = 0cos(t). If = 2 rad s-1, the period will be exactly 1.0s, whereas the Maple program gave T = 1.00000001s. For large amplitudes, in the absence of dissipation, Maple may be compared with the Elliptic Integral solution (Eq. 8). For = /2 rad, the numerical solution given by Maple finds the time period to be 1.18034070s, which agrees with the elliptic integral solution to 1 x 10-7s. Also, the amplitude should remain undamped and after one cycle had changed by 1 x 10-6 rad. This indicates that the solutions given by Maple are sufficiently accurate to neglect their errors when comparing with the errors in the measured quantities.

36

Conclusion
Experiment 1 showed that (Eq. 7) is valid when the amplitude of the pendulum is very small and its speed is low enough for dissipation to be ignored. Experiment 2 showed that the time period of a pendulum could be plotted as a function of its amplitude accurately in the form of the Elliptic Integral solution or a Maclaurin series expansion. Experiment 3 shows that there is a linear relationship between damping rate and amplitude. The numerical modelling shows that drag force can be expressed in terms of velocity in a Maclaurin series expansion requiring only terms up to the quadratic.

Evaluation of Experimental Set Up


The experiments were set up in a manner that minimised error, for example, the error in time measurements was made as small as possible by using a high frame rate setting on the camera. Setsquares were used to get the camera to face normal to the plane of motion to minimise parallax error. The set up from which the pendulum was hung was checked to not wobble when the pendulum underwent oscillation, G-clamps were used to clamp the clamp-stand to the desk to ensure no movement. The graph paper attached to the board was checked to be completely in-line by placing a spirit-level along the horizontal grid lines and using a plumb-line to check the vertical lines. The lab jacks underneath the board were altered until the grid lines matched the plumb-line and spirit-level. If either the board or the metre-sticks were in close proximity to the pendulum, it could influence the flow of air around the bob, giving unreliable results, to counteract this, the board was placed a suitable distance behind the plane in which the pendulum oscillated.

Modifications Made During Experimental Work


The first board used to determine the position of the bob used a slightly bent piece of wood that was smaller than the second one, additionally the first board had graph paper sellotaped to it. Using sellotape had the disadvantage that it was nearly impossible to get the paper to lay completely flat across the board. In light of experimenting with this it became apparent that a larger and completely straight board was required, and so was constructed. The graph paper was then pasted to the board very carefully to make sure it was completely flat. In the first attempt of Experiment 1, amplitudes 15 were used, this lessened the validity of the small angle approximation and in doing so gave readings that were slightly inaccurate. The experiment was then repeated for amplitudes 3-4 which gave much more consistent results. For experiments 2 and 3, it was imperative that the pendulum stayed in-line with the ruler in order to get an accurate measurement of its position. Initially it was released by hand which was found to cause the plane of motion to rotate over several cycles. After this several other release mechanisms were tried out, the most successful of which was a clothes-peg. An improvement that could be made would be to use a solenoid release mechanism, as this would remove human influence.

Limitations of Equipment
Throughout the report, the string used for the pendulum had been described as 'light' and 'inextensible', because it's mass is negligible and it is relatively inelastic. However as the ball experiences a centrifugal force that varies with its speed, the string may stretch, resulting in small fluctuations in length of string. The only way to avoid this would be to use an even less extensible string such as steel wire. However this would introduce a greater mass, reducing the validity of determining the centre of the entire system's mass as being the centre of the bob. Additionally, a wire such as this would be likely to resist motion at the suspension point, resulting in energy being lost though bending the wire, thus producing additional dissipation. As previously mentioned, caution was taken to ensure that the board had minimal influence on the flow of air around the bob, 37

it is possible that moving the board back further would have benefited the accuracy of the experiment. Another issue this would create would be the fact that the camera can only focus on a certain range, meaning that either the board or the pendulum/metre-sticks would be out of focus. One solution to this would be to move the camera further back and zoom in, this would increase the depth of field which the camera could keep in focus, this would require a camera with high optical zoom capabilities or a powerful zoom lens. The pendulum used had a tendency to go 'off-line' at large (>80) amplitudes, this limited the range of possible angles that could be accurately measured. Whilst the frame rate of the camera used is significantly higher than that of an average camera, the error in time period could be reduced by using a camera with an even higher frame rate.

Further Investigation
The behaviour of damping rate at small amplitudes would be an interesting topic to investigate, as it should become independent of amplitude (as discussed in the numerical analysis section). The dependence of the drag coefficients a1 and a2 on the bob's mass and cross-sectional area could also be examined, and this could provide useful empirical formulae for spheres of arbitrary size and mass. Driving the pendulum with its natural frequency at small amplitudes will cause the amplitude of oscillation to grow resonantly, which in turn will change its natural frequency (as shown in Experiment 2) resulting in more interesting behaviour than the that exhibited by a linear, driven simple harmonic oscillator. To investigate this, the homogeneous second order ODE used in Maple could have an inhomogeneous driving term added. Another interesting relationship to investigate would be the effect of dissipation on the large amplitude time period. This could be done using a Maple simulation or by carrying out a practical experiment. Currently there is no theory as to what the effect would be.

38

References:
[1]

[2]

[3]

[4]

[5]

[6]

[7]

D.J. Tritton, 'Physical Fluid Dynamics', Oxford University Press Inc., New York, 1994. Pages 12, 32-33, 110-111 A. Belndez et al. 'Exact solution for the nonlinear pendulum', Revista Brasileira de Ensino de Fsica, v. 29, n. 4, p. 645-648, 2007. D.G. Simpson, 'The Nonlinear Pendulum', http://www.pgccphy.net/ref/nonlin-pendulum.pdf Accessed August 2011 M.E. Bacon and D.D. Nguyen, Real-world damping of a physical pendulum, Eur. J. Phys. 26 (2005) 651655, doi:10.1088/0143-0807/26/4/011 http://a-s.clayton.edu/campbell/physics/phys2211/Supplements/err2211.htm Accessed October 2011 http://www.wolframalpha.com/input/?i=geogravity Accessed October 2011 http://www.engineeringtoolbox.com/air-properties-d_156.html Accessed December 2011

Acknowledgements
St Andrews University for use of Maple software and library facilities.

39

You might also like