Foucault's Method in New Settings: September 2014
Foucault's Method in New Settings: September 2014
Foucault's Method in New Settings: September 2014
net/publication/265908688
CITATIONS READS
0 700
2 authors, including:
Gregor Weihs
University of Innsbruck
242 PUBLICATIONS 10,731 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
All content following this page was uploaded by Gregor Weihs on 21 October 2014.
In this paper, we introduce two simple and inexpensive versions of the well-known Foucault
method for measuring the speed of light. In a footprint of just 20 cm by 270 cm with readily available
laboratory items and a webcam, we obtained c = 296720 ± 3000 km/s, and c = 302295 ± 3000 km/s,
respectively, both within less than a per cent of the defined value. The experiment also prepares
students to work with large amounts of data.
data), and not just a few measurements. Notable exam- fits of modern technologies in terms of measurement time,
ples are customer recommendation systems of Amazon, accuracy, cost, or the volume of data.
eBay and similar retailers, or the data streams of the The example that we take is the measurement of the
Large Hadron Collider or the Square Kilometre Array. speed of light, which is one of the fundamental physical
This trend is expected to continue in the foreseeable fu- constants. Strictly speaking, our example is pathological
ture, especially, with the advent of always-online mobile in the sense that the international meter is defined by the
devices capable of continuously collecting and transmit- help of the speed of light and the international standard
ting all kinds of data. of time, and not the other way around. However, first,
till 1983 (i.e., in Foucault’s life), length and time were
However, it also seems that science education does
defined and the speed of light was the derived quantity,
not keep up with the pace of progress in data collection
second, this fact does not reduce the didactic value of the
and analysis capabilities, and that many introductory
experiment itself. We would like to emphasize that while
or even advanced level laboratory experiments are still
the evaluation of the measurements requires some data
conducted with hand-held stopwatches, weights, mercury
processing, this fact does definitely not qualify it as a big
thermometers, and microscopes with engraved scales.
data exercise.
This approach has at least three inherent problems. The
first is that it teaches students how experiments were The paper is organized as follows. In the next two
conducted two centuries ago, but does not tell them how sections, we outline the historical and theoretical back-
to do them now. Second, the amount of data that can ground and derive the expression for c. In Section IV., we
be collected in this way is limited, and inaccurate. This introduce our experimental setup and the critical compo-
also means that statistical evaluation of the results is nents, Section V. contains a detailed discussion of our re-
constrained to a handful of data points. Finally, by the sults, while Section VI. is devoted to a thorough analysis
very nature of the required specialized setups, these ex- of various systematic errors. In the appendix, we present
periments are expensive, and students have to operate a couple of MATLAB (The Mathworks, Inc.) snippets
within the spatial and temporal confines of the labora- that can be used to evaluate measurement data.
tory course. We are convinced that one cannot underesti-
mate the pedagogical benefits of pursuing science on the
kitchen sink: when one can accurately measure some- II. HISTORICAL BACKGROUND
thing relevant (such as, a fundamental constant) with
easily available and cheap everyday items, and without That the speed of light, c, is finite was already conjec-
reference to a dedicated laboratory. It is a very fortunate tured by Galileo in the XVII. century, though, his exper-
coincidence that in our times, everyday items are digital imental apparatus at the time prevented him from giving
gadgets capable of measuring all kinds of physical quan- even an order-of-magnitude estimate for the value. Since
tities, e.g., distance, temperature, acceleration, magnetic then, various methods have been developed.
fields, light intensity, frequency, time etc., and that the It was first Huygens, who, based on the astronomi-
demand for high quality in user experience makes it pos- cal measurements of Rømer in 1676 on the entry into
sible to deliver unprecedented accuracy. and exit from eclipses of Jupiter’s moons, could provide
At the same time, we also recognize the pedagogical a lower bound of about 200 000 km/s. The same mea-
value of discussing how experiments were conducted in surements, repeated with higher accuracy by Delambre
the past and that it would be an irreparable loss not in 1809, yielded 304 000 km/s, astonishingly close to the
to show what could be achieved with devices that we true value. Another astronomical method, the aberra-
would now consider rudimentary. What we would like tion of light, was discovered by Bradley in 1729, with the
2
While it is not absolutely necessary, by passing the light plies that the physical size of the setup can considerably
through the single-mode fibre patch cable (Thorlabs P1- be reduced. We would like to point out that the use of
630A-FC-1), we begin with a perfect Gaussian beam. It a webcam in the context of speed of light measurements
is worth noting that the fault locator can be replaced by was discussed in an interferometric setting in [4].
an inexpensive laser pointer. Our second setup, Setup 2, is shown in Fig. 3. The light
of the fault locator is focused by a lens of focal length
SM
400 mm onto F P , from where it reaches the spherical
mirror SM with a focal length of 2 m. The spherical
Detector
mirror is 4060 mm away from F P , and is tilted slightly,
so that the light is reflected off the rotating mirror RM ,
RM BS located 730 mm away from SM , and finally F M , located
FP
TS FC
3260 mm away from RM . The lengths in the setup are
L 2 400 mm L1 75 mm
Camera
chosen in such a way that the light is focused on the
TS flat end reflector, F M , although, as will be discussed
in Section VI., small longitudinal misalignments do not
Figure 2: Experimental setup Setup 1. RM , F M , SM are influence the results in any significant way. As in the first
the rotating, folding, and back reflector mirrors, respectively, setup, folding mirrors were used between SM , and RM ,
L1 , L2 are lenses of focal length 75 and 400 mm, respectively, and between RM , and F M .
F P is the focal point of L1 , F C is the fibre collimator, BS is
the beamsplitter, and T S are translation stages. Dimensions FM RM
are given in the text.
set by adjusting the clock signal that is fed into the PLL from the focal spot, F P . With the tip-tilt control knobs
loop. Fig. 4 also indicates the connections of the mirror of the mirror holder, SM has now to be aligned so that
assembly: P W R (pin 1) is the power line, whose po- the light is reflected back to the laser. At this point,
tential can be anything between +18, and +36 V, GN D the reflected beam should be focused on F P . Finally,
(pin 2) is ground, EN AB (pin 3) is the active-low motor the camera has to be placed in the diverted focus of the
enable pin (this should be tied to ground), while CLK back-reflected beam. Great care has to be taken to make
(pin 5) is the clock line, which takes TTL pulses with fre- sure that the camera’s plane is as perpendicular to the
quencies between around 300, and 6000 Hz. Pin 4 is an laser beam as possible: failure to do so will results in a
output connected to the magnetic field transducer, and systematic error, which leads to higher speeds of light.
can be used for monitoring the rotation. For a thorough discussion on this, see Section VI.
The advantages of the mirror assembly are that first,
the mirror is monolithic, therefore, it is safe to operate:
no pieces can break off at high speeds. Second, the con- V. EXPERIMENTAL RESULTS
trol electronics makes it possible to adjust the speed by
setting the frequency of the clock signal from a simple As can be inferred from Eq.(2), in order to determine
function generator, and that there is a well-defined lin- the speed of light, one has to measure d1,2,3 , the angular
ear relationship between the rotation speed and the clock frequency ω, and the displacement ∆s. The measure-
frequency. ment can be done in the same way in both setups, and
the steps are as follows. First, one has to determine the
rotation speed as a function of the clock frequency. Next,
the pixel size of the camera has to be measured. This step
amounts to calibrating a ruler. Then the displacement of
the image on the camera has to be measured at various
clock frequencies (this step involves fitting to the camera
images), and by using the pixel size, this displacement
has to be converted to physical units. Finally, the slope
of the displacement-frequency relationship has to be de-
termined, and inserted in Eq.(2).
The rotation speed can be deduced from the time
traces of the photodiode, either by simply measuring the
time difference between an integer number of maxima, or
recording the potential values, taking the Fourier trans-
form, and identifying the strongest frequency component.
Given a high enough number of samples, the two methods
deliver the same results. In Fig. 5, we show the measured
rotation speed as a function of the clock frequency, with
a typical time trace of the detector signal on an oscillo-
scope, and its Fourier transform. The period can clearly
be resolved from either the signal, or its Fourier spec-
trum. Note that at high clock rates, the rotation speed
saturates. For this reason, we excluded the last 3 points
from the linear fit, from which we deduced the relation-
Figure 4: Rotating printer mirror, side view (top), and top ship frot (Hz) = (0.167 ± 0.00054) · fclk (Hz) − 0.649 (Hz).
view (bottom). The 30-pin integrated circuit contains the The error of the fit is approximately 0.3%. Given the
motor driver (BD6792FM from Rohm Semiconductors) with precision (in the ppm range) of frequency standards used
the built-in PLL. Control pins are labeled in blue. in modern pulse generators, and the stability of phase-
locked loops used in laser printers, we ascribe the error to
Initial alignment of the setup is performed when the our way of determining the frequency from the Fourier
mirror is stopped (the enable line is high). First, all transform of the time trace. Also note that, since the
mirrors are placed to their respective positions, and F C rotating mirror has 8 facets, the actual rotation speed is
is aligned such that the collimated laser beam can travel only 1/8 of what the detector signal indicates.
to the end mirror, SM . Then L1 is inserted in such a It is worth pointing out that, given the order of mag-
way that the diverging laser light still reaches both RM , nitude of the rotation speed, in the absence of an os-
and SM . After this, L2 is inserted in the path, and is cilloscope, these frequencies can easily be measured by
moved along the optical axis till the size of the light spot means of a smart phone. All one has to do is to convert
reaches its minimum on SM . When this is achieved, the the electrical signal of the photodiode to sound by ampli-
beamsplitter, BS, has to be placed on the left hand side fying it, and connecting it to a speaker, and then record
of the focal point of L1 , at a distance of about 5-7 cm the sound through the microphone. There are countless
5
Position P(µm)
1200 1500
frot(Hz) =0.167 ·fclk(Hz)−0.649(Hz)
1000 1.0 1000
Rotation frequency (Hz)
Power (a.u.)
0.8
800 0.6
0.4 500
0.2
0.0
600 0 1 2 3 4 5
0
Frequency (kHz)
Time (ms) 200 400 600 800 1000
400 0 5 10 15 20
Position p(pixel)
1.0
Signal (V)
200 0.5
Figure 6: Measurement of the pixel size. The statistical errors
0.0 in both the dependent and independent variables are too small
0
0 1000 2000 3000 4000 5000 6000 to be visible. The parameters of the linear fit are indicated
Clock frequency (Hz) in the figure.
1.5 417 Hz
50 Hz
1.0
183 Hz
Position (pixel)
0.5
400 Hz
500 550 600 650 700
Position (pixel)
0
100 751 Hz
Figure 8: Typical profiles taken in Setup 1 at a rotation fre- 200
300
quency of 183 (solid red circle), 417 (empty green triangle), 400
and 885 Hz (solid blue square), respectively. The domain of 500
600
Gaussian fits is indicated by the shaded gray regions, while 700
the solid black lines are the fits.
0 200 400 600 800 1000
Position (pixel)
ter the CCD pixels were converted using the fit from Figure 10: Camera images in Setup 2 at frequencies 50, 400,
Fig. 6. The linear fit to these data yields a slope of and 751 Hz.
(0.130 ± 0.00047) µm/Hz. Given that, with the nomen-
clature of Eq. (2), d1 = 425 ± 1 mm, d2 = 1630 ± 1 mm,
and d3 = 4830 ± 1 mm, and taking all above-mentioned This change in the width can also be seen in Fig. 11,
error sources into account, we calculate a speed of light where we plot the vertically integrated camera images
of c = (2.97 ± 0.03) · 108 m/s. This is within 1% of the for 17 rotation frequencies as indicated. However, despite
defined value of 2.99792458 · 108 m/s, and overall, the the broadening of the profiles, the displacement is clearly
statistical errors are within 1%. visible as the frequency changes.
1400 640 7
630 6
Integrated intensity (a.u)
851 Hz
Peak position (pixel)
Peak position (µm)
801 Hz
1350 620 5 751 Hz
701 Hz
651 Hz
4 601 Hz
610 551 Hz
501 Hz
1300 3 450 Hz
600 400 Hz
2 350 Hz
300 Hz
∝ 0.1300 ·frot(Hz)
590 250 Hz
1250 1 200 Hz
150 Hz
100 Hz
580 0 50 Hz
0 200 400 600 800 1000 0 200 400 600 800 1000 1200
Rotation speed (Hz) Position (pixel)
Figure 9: Position of the reflected beam as a function of the Figure 11: Vertically integrated camera profiles in Setup 2 as
rotation speed in Setup 1. On the right vertical axis, the a function of the frequency.
same data are shown in units of the CCD pixels. The peak
position can be obtained from Ppeak (µm) = (0.130±0.00047)·
frot (Hz)+1247.6(µm). Error on the data points is not visible. In Fig. 12 we plot the beam displacement as a function
of the rotation speed, similar to Fig.9. The linear fit
We now discuss measurements in Setup 2. Typical to these data yields a slope of (0.899 ± 0.0059) µm/Hz.
camera images at frequencies 50, 400, and 751 Hz, re- Given that d1 = 4060 ± 1 mm, d2 = 730 ± 1 mm, and
spectively are shown in Fig. 10. As opposed to the other d3 = 3260 ± 1 mm, and considering all error sources, we
setup, the laser spot is stretched vertically over the whole calculate a speed of light of c = (3.02 ± 0.03) · 108 m/s.
length of the camera (720 pixels). Also note that as the Our experimental conditions and results are summa-
frequency increases, so does the width of the images. We rized in Table I.
7
650 slightly with respect to the translation stage till all mea-
1200 sured profiles are identical. As with the systematic error
600 discussed above, corrections are quadratic in the angle.
x
VI. SYSTEMATIC ERRORS
d3 P
We have already indicated the magnitude of statistical RM d1 S'
errors: the calibration of the CCD is about 0.2%, the x d3 d2 Δs
V S
rotation frequency’s is about 0.3%, the length measure- d'1
ment’s is less than 0.1%, while the Gaussian fits to the Δs' L
profiles contain an error of about 0.2%. However, in ad- V'
dition to these, there are a number of systematical errors
that one has to consider. Figure 13: The concept of the measurements, with focusing
One we have already pointed out, namely, if the cam- errors. Notation as in Fig. 1. P is the image of the source S.
era is not perpendicular to the laser beam, all displace-
ments will be measured shorter and this will lead to a The magnitude of the correction will depend on two
seemingly larger speed of light. One way of removing parameters of the setup, d2 , d3 , and the inaccuracy in
this error source is to slightly rotate the camera without the focusing, x. Note that for d2 = 0, i.e., when the
moving it, and repeat the measurements multiple times. rotating mirror is next to the imaging element, the first-
The smallest value of c should correspond to the perpen- oder correction is zero. In the first setup d2 /d3 ≈ 1/3,
dicular configuration. However, since this correction is while in the second case, d2 /d3 ≈ 1/4. Therefore, an up-
proportional to the cosine of the angle of deviation from per bound for the correction in Eq.(3) is x/(4d3 ). Given
the normal, errors are of second order. that d3 ≥ 3 m, we incur an error of 1%, if x ≈ 0.12 m. It
Second, if the camera’s plane is not parallel to the axis is reasonable to assume that the focus can be determined
of the translation stage, the pixel size will be inferred in- with 10 cm accuracy, even if the imaging elements have
correctly, and this, again, will lead to a seemingly higher such long focal length. Therefore, we can conclude that
light speed. As mentioned above, a trivial test for this the error related to imperfect focusing is less than 1%.
is the beam profile measured at various positions of the Finally, the lens, the only glass element in the first
translation stage: all other conditions being identical, a setup, has finite width with a refractive index larger than
simple translation should result in identical profiles. If one, and this adds to the total length between the focal
this is not the case, then the camera has to be rotated point and the end mirror. This extra optical length can
8
be measured and added to the path, provided the refrac- We demonstrated that with readily available and inex-
tive index of the glass is known. Of course, the second pensive optics, and a bit of data processing, acceptable
setup does not suffer from this kind of error. accuracy (results within 1% of the defined value) can be
achieved. We also discussed a range of systematic errors,
and pointed out several possible improvements. The ex-
VII. CONCLUSION periment teaches students the historically important Fou-
cault method, and modern data evaluation concepts at
In conclusion, we presented a simple version of the Fou- the same time.
cault method for the measurement of the speed of light.
[1] G. W. Clark, American Journal of Physics 24, 189 [11] J. A. Deblaquiere, K. C. Harvey, and A. K. Hemann,
(1956). American Journal of Physics 59, 443 (1991).
[2] G. W. Clark, American Journal of Physics 69, 110 [12] K. Aoki and T. Mitsui, American Journal of Physics 76,
(2001). 812 (2008).
[3] T. J. Belich, R. P. Lahm, R. W. Peterson, and C. D. [13] A. Ronzani, F. Maccarrone, and A. D. Lieto, Euro-
Whipple, American Journal of Physics 65, 186 (1996). pean Journal of Physics 29, 957 (2008), URL http:
[4] T. Lahaye, P. Labastie, and R. Mathevet, American Jour- //iopscience.iop.org/0143-0807/29/5/009.
nal of Physics 80, 497 (2012). [14] H. E. Bates, American Journal of Physics 56, 682 (1988).
[5] D. J. D’Orazio, M. J. Pearson, J. T. Schultz, D. Sidor,
M. W. Best, K. M. Goodfellow, R. E. Scholten, and J. D.
White, American Journal of Physics 78, 524 (2010).
[6] L. T. Dillman, American Journal of Physics 32, 567 Appendix A: MATLAB code
(1964).
[7] J. M. Feagin, American Journal of Physics 47, 288 Here we list matlab snippets that can be used for
(1979). the evaluation of images. The usual workflow is to cre-
[8] H. M. Morrison and J. A. Driedger, Physics Education
ate a profile similar to that in Fig. 8 with the function
(1980), URL http://iopscience.iop.org/0031-9120/
15/2/006. create_profile, and pass the output to the function
[9] J. Brody, L. Griffin, and P. Segre, American Journal of fit_profile, which will print the parameters of the best
Physics 78, 650 (2010). Gaussian fit to the console. gauss simply defines the fit
[10] J. Rogers, R. McMillan, R. Pickett, and R. Anderson, function, and it can easily be replaced by other, more
American Journal of Physics 37, 816 (1969). appropriate forms, if necessary.
f u n c t i o n p r o f i l e = c r e a t e _ p r o f i l e ( fn , range1 , r a n g e 2 )
% Returns and d i s p l a y s a p r o f i l e ( a v e r t i c a l l y i n t e g r a t e d image segment )
%
% Input :
%
% f n : t h e f i l e ( camera image ) t o r e a d
% range1 , r a n g e 2 : t h e l i m i t s o f t h e v e r t i c a l i n t a g r a t i o n
im = imread ( f n ) ;
im1 = sum ( im , 3 ) ; % Turns t h e RGB image i n t o g r a y s c a l e
f i g u r e ( 1 ) ; i m a g e s c ( im1 )
p r o f i l e = sum ( im1 ( r a n g e 1 : range2 , : ) ) ;
f i g u r e ( 2 ) ; p l o t ( p r o f i l e , ’ ro ’ ) ;
f u n c t i o n y = g a u s s ( par , xdata )
% Returns a Gaussian f u n c t i o n e v a l u a t e d a t p o i n t s g i v e n i n xdata
%
% y = A∗ exp (−(x−x0 )^2/ sigma ^2)+ o f f s e t
%
% Input :
% par : an a r r a y o f p a r a m e t e r s i n t h e form [ A, x0 , sigma , o f f s e t ]
% A: a m p l i t u d e o f t h e Gaussian
% x0 : c e n t r e o f t h e Gaussian
9
% sigma : s t a n d a r d d e v i a t i o n o f t h e Gaussian
% o f f s e t : o f f s e t o f t h e Gaussian
% xdata : p o i n t s where t h e f u n c t i o n i s t o be e v a l u a t e d