Vous êtes sur la page 1sur 8

The Pendulum Wave

(an essay accompanying the video


“Animated Pendulum Wave”)
by Samuel Peters

The pendulum wave was first conceived by the great Austrian physicist Ernst Mach (1869).1 Only two
other (very different) versions on Youtube are also automatic,2 and none of the other versions use
mirrors.

My longest pendulum completes 51 oscillations in 60 seconds.3 Each successively shorter pendulum


completes one more oscillation during that 60-second time period.

What we observe in the pendulum wave is very different depending on whether it is observed from
the side or overhead.4 From the side we see the pendula cycling through ten discrete stages (before
starting the cycle again). These stages can be described (broadly) as follows:

1) 1 column

2) 1 transverse wave

3) 4 columns

4) 3 columns

5) 2 transverse waves

6) 2 columns

7) 2 transverse waves

8) 3 columns

9) 4 columns

10) 1 transverse wave

11) 1 column5

As one can see, the pattern proceeds to the middle of the cycle and then reverses itself (like a movie
run backwards). Even the order of CW and CCW are reversed. The pace roughly goes from slow to fast
to less fast to slow (at the middle), and then it reverses. The four columns last the least time, the three
last the next least, and the two last the longest.

When viewed from above the sequence is very different:

1) 1 row

2) 1 long transverse wave which gets shorter


3) 2 transverse waves

4) 2 rows

5) 2 transverse waves

6) 1 short transverse wave that gets longer

7) 1 row6

Here, as with the side view, we start and end with one row (here rows instead of columns), we transition
from one transverse wave to two transverse waves (and then reverse that order after passing through
the middle), and in the middle of each we have two rows. In the top view, however, we are looking
along an axis that is perpendicular to the plane that the bobs seem to be traveling in,7 and this makes
the motion seem very two-dimensional.

But what is most vividly different between the two views is the appearance in the side view of the 4
and then three swirling columns. This is the grandest part of the illusion because the motion it suggests
is the most distinctly three-dimensional. Because in the side view we are looking from a vantage point
that is both from the side and from above, we there have a real sense of depth.

However, we cannot say that those swirling columns are objectively observable because before being
able to see them one must mentally presuppose that the balls are vertically connected. If, instead, one
chooses to insist on seeing the balls as moving in independent horizontal paths (or, more likely, if one’s
mind will not allow itself to think of the balls as being vertically connected), one will not see the vortices.
Observing these effects, then, is not always guaranteed. From this standpoint our visual observation of
the pendula cannot (properly speaking) be considered empirical data. On the other hand, since it seems
that one will always view this motion as being of either the one type or the other, our observations can
be thought of a semiempirical.

There are several important lessons we can take from our consideration of the visual appearance of
the pendulum wave. First, what we’ve seen vividly illustrates the fact that different points of view can
yield astonishingly different observations. Secondly, our observation effectively exemplifies the fact
that taking a vantage point perpendicular to the plane of motion reduces dimensionality from three to
two. Finally, we are reminded of how overriding is the impact of one’s mental predisposition to what it
is that one will observe.

Moving beyond visual observation, the viewer of the pendulum wave is compelled to ask himself,
“How can a simple series of pendula create such astonishing complexity?” The answer lies in the fact
that the lengths of the pendula were chosen so as to create integer differences in the number of
oscillations during the chosen time period of the sequence. These different pendulum lengths must be
positioned sequentially and also must be spaced apart at a regular distance.

As a result of the largest pendulum having 51 oscillations per minute, and then the next longest
having 52 (and so on), the frequency of adjacent pendula are roughly 0.016 Hz apart.8 That corresponds
to a six degree9 per second advancement of the position of each shorter pendulum compared to the
longer pendulum it sits next to. This creates the illusion of smooth motion. (Film, after all, maintains its
illusion of smooth motion with twice as much advancement each second.)10 We inevitably identify the
thing that seems to move as a wave. But of course there is no energy that actually passes between
pendula, as would be the case with a real wave.11

What the eye sees is a pattern that definitely exists in real time, but which actually is only observable
in a particular time frame. If we were to slow the rate of progression down so that the difference in
period was much greater (which could be done by making the strings much longer) the appearance of
motion would be so slow that our eyes would not catch it.12

Simply measuring pendulum arm lengths, however, isn’t at all sufficient for constructing a working
pendulum wave. There are slight irregularities introduced by the various physical details of your
apparatus whose effects likely add together to overwhelm the precision you seek. In my setup, for
instance, I had to make great efforts to normalize the interface between the top of the pendulum arm
thread and the hole it passed through (to get to tuning knobs at the top of the support board). My issue
seemed to be that there was a tiny extra length to the pendulum arm when the thread was vertical
compared to when it was at an angle (that there was wiggle room at the bottom of the 1/16 inch hole).
I addressed this by wedging the top part of the pendulum-arm thread to the entry of the hole with a
segment of 16-gauge wire, with the wire turned sideways so that the smooth edge was at the point of
contact.13

But these various tiny physical adjustments had to be paired with normalizing the period of each
pendulum with an electronic timer (Arbor Scientific Timer and Photogate) to within 1/1000 of a second.
It took an average of 30 minutes of tuning (making slight adjustments with the tuning knob) to achieve
that level of precision with each pendulum. And as you can see from the video, in spite of having done
all of this, there is still a slight but noticeable discrepancy that appears in the pattern by the end of one
60-second sequence (so that the balls do not exactly swing in perfect unison as they begin their second
cycle).

Interestingly, the string lengths of the pendula do not increase linearly. This fact is dramatically made
clear when the stationary strings are viewed in profile. The strings make a slight concave-downward arc
(where the curve becomes less steep as the string length decreases) rather than a straight diagonal line.
If it were string length that regularly changed rather than period, the illusion of smooth motion would
not exist. The illusion works by having the string length determined by the number of oscillations per
minute, instead of the other way around.

Mathematically this relationship is encapsulated by the small-angle approximation formula14 for the
period of a pendulum: T =2π sqrt (L/ g), where T is the period of the pendulum, L is the string length and
g is the acceleration due to gravity. Here, clearly, T is proportional to the square root of L and not to L
itself. Since square roots do not decrease as fast as their arguments15, decreases in period occur more
slowly than decreases in string length. So to keep decreases in period regular, the decreases in string
length must exhibit the concave-downward arc.16

Another practical issue I had was that I found that if I lined up the axis of the hinges of the “refresh”
board with the axis of the (stationary) pendulum bobs, the displacement of the bobs didn’t look equal
(though they actually were). The longest pendula didn’t seem to have hardly any swing compared to the
shortest ones, and this was visually unsatisfying. So I angled the refresh board so that the longest bobs
were actually displaced more. This fixed the visual disparity. Happily, this also had the more important
effect of making the release angles roughly the same. That made the correction factor for the small-
angle approximation formula the same for each of my pendula, and thus made that source of error
largely disappear.17

What the untrained observer probably finds most impressive about the pendulum wave, however, is
what psychologists call “a violation of expectation.”18 We expect there to be chaos when 15 differently-
sized and independently-moving objects are placed side by side. When instead we see an elaborate and
continuous pattern, the first-time observer is stunned. There is no obvious cause for the astonishing
orderliness of what we see.19

Regarding “Aliasing”

The various patterns observed in the pendulum wave have been described as being examples of
“aliasing.”20 “Aliasing” is also known as “the wagon wheel effect,” since in films moving wagon wheels
often appear to turn backward. Aliasing occurs when you sample a source at an insufficient rate.21 In
the wagon-wheel example, the frequency of the sample is the frame-per-second rate of film, whereas
the frequency of the source is the actual rate at which the spokes of a moving wagon wheel change
places.

At first thought, it seems that the pendulum wave is fundamentally unlike the case of the filmed
wagon wheel because in the case of the pendula there is no physical intermediary between the original
event and its representation. There’s no active sampling going on where adjusting the rate of that
sampling could remove illusory motion. The frequency that is at variance in the pendulum wave is the
very real frequency of adjacent pendula.

But upon further consideration I realized that the sampling being spoken of is the selection and
positioning of the pendulum frequencies during the actual construction of the pendulum wave (and thus
the authors use the term “spatial aliasing”). There is still a temporal component because the only
frequencies displayed are those which are one integer apart in a given time period. But there is also a
spatial component due to the fact that the pendula are positioned so that their frequencies are
sequential and so that the space between them is recurrent. All of these effects insure that we will see
an illusion of motion.22

In summary, my approach here has been essentially descriptive rather than mathematical. The
mathematical equations that model the motion of the pendulum wave are, of course, the final analytical
word.23 But the visual effects of this aliasing are also worth noting. Though they are the result of
contrivance and delusion, they nonetheless enable a deeper appreciation of some of the pitfalls of
observation.24
1. https://sciencedemonstrations.fas.harvard.edu/presentation/pendulum/pendulum-waves

2. The two other (though very different) automated versions on Youtube are the one at the
University of Arizona (completed in 2013 [ https://www.youtube.com/watch?v=TyTwl-
BPs7A]) and the gigantic (12 m x 10 m x 10 m) version from the Netherlands (completed in
2014 [https://largependulumwave.nl/]).

3. My pendulum lengths are the same as those used by the Harvard version, which is itself a
replica of Mach’s version. These dimensions are also used by Paul Liu, who provides
excellent information about the practicalities of constructing a pendulum wave
(hippomath: blogspot.com/2011/06/making-your-own-pendulum-wave-machine.html).

4. From the front no pattern is observable whatsoever since the axis of motion is parallel to
the line of sight.

5. The following is an expanded version:


1) 1 column (for ½ swing)
2) a slow, long-wavelength transverse wave
3) a fast, short-wavelength transverse wave
4) CW (clockwise) vortex*

* The vortices sometimes last several seconds and sometimes last only a
fraction of a second. Partly because of the stop-action capabilities of
video I am able to discern far more of these in the video than I am in
watching the actual pendulum wave without that intermediary. But
the difference in the number of vortices observed is also likely due to
the effects of “aliasing,” where the frame-per-second rate of video
skews our perception of motion. (More on “aliasing” later.)

5) CCW (counterclockwise) vortex


6) CW vortex
7) CCW vortex
8) CW vortex
9) 4 columns (three groups of 4 vertically attached balls and 1 group of 3 vertically
attached balls)
10) CCW vortex
11) CW vortex
12) CCW vortex
13) CW vortex
14) 3 curving columns (3 groups of 5 vertically attached balls)
15) 3 straight columns
16) CCW vortex
17) 3 curving columns
18) CW vortex
19) CCW vortex
20) CW vortex
21) CCW vortex
22) 2 crisscrossing transverse waves
23) 2 columns ← this is the middle

At this point the process goes in reverse back to the starting point (one column) where
the whole sequence begins again.

6. Here’s an expanded version of the top view:


1) 1 row
2) ¼ wavelength of a transverse wave
3) ½ wavelength
4) 1 wavelength
5) 2 wavelengths
6) 3 wavelengths
7) 3 rows which constantly switch places
8) 2 crisscrossing transverse waves
9) 2 rows ← this is the middle
10) 2 crisscrossing transverse waves
11) 3 rows which constantly switch places
12) 3 wavelengths
13) 2 wavelengths
14) 1 wavelength
15) ½ wavelength
16) ¼ wavelength
17) 1 row

7. Viewed from above, the changes in height are not noticeable.

8. Since the longest pendulum has 51 oscillations in 60 s, that is a frequency of 0.85


oscillations per second. Doing the same for the second-longest pendulum (52/60) yields a
frequency of 0.867 oscillations per second. The frequency difference between them is
0.017 Hz.

9. (0.017 osc./s)(360ᵒ/osc.) = 6.0ᵒ/s

10. The 24 frames per second rate of film equates to 0.04 s per frame. The difference in period
of adjacent pendula is 0.02 s.

11. Also the motion of the waveform is not truly periodic since the wavelength keeps changing
(though the overall sequence does exactly repeat itself).
12 . If the longest string were 100 meters long it would make three oscillations per minute
(instead of our current 51), which means it would have a period of 20 s. In order for the
next-longest pendulum to complete one more oscillation per minute, its period would have
to be 15 s--5 seconds shorter (and it would have a length of 56 m). With such a 5-second
difference in period the illusion of motion would be barely perceptible.

13. Cutting metal wire with wire cutters creates edges that alternate (every 90 degrees)
between tapered and smooth, and only the smooth edge would eliminate the wiggle room.

14. The mathematical derivation of the equation typically used for the period of a pendulum is
dependent on keeping the angle small enough that the sine of the angle (in radians) is
essentially equal to the angle itself. (We can see that 0.175 radians [=10ᵒ]
and sin (0.175 radians) [= .174] are only 0.6% different.) The small-angle
approximation is ultimately made so that the period of the pendulum can be expressed only
in terms of L and g. (The approximation itself is justified by a Taylor series.)

15. Sqrt (16) is 4, but sqrt (4) is 2.

16. Even if we don’t use the small-angle approximation, we are still forced to accept numerous
simplifying assumptions: that the cord is massless and inextensible, that the bob can be
treated as a point mass, that no energy is lost to friction, that the support apparatus
doesn’t move, that g remains constant (to name what may only be the most evident
assumptions).* This list of prerequisite assumptions makes it clear that the world we are
investigating is elemental and elusive, and that mathematical descriptions are always
generalizations. The difference between these types of generalizations and the metaphors
of language** (which, in science, mathematics historically displaced) is the margin of error.
With mathematics we can quantify and systematically reduce the margin of error where
with spoken language that cannot be done with any precision. (Interestingly, the entire
mathematical treatment of simple harmonic motion [SHM] is based on a quasi-metaphorical
substitution: the mathematics of uniform circular motion [UCM] is accepted as a valid
replacement for the description of SHM based on the fact that the profile of an object in
UCM and the appearance of that same object in SHM are indistinguishable. This substitution
has the effect of making one conclude that it is no wonder that Pi turns up everywhere if
you’re allowed to pull off tricks like that.)

*https://en.wikipedia.org/wiki/Pendulum_(mathematics)

** If compared only on the basis of utilitarian longevity, the metaphors of


language are but a frothy eruption of mood and circumstance which dissipate
in time like popping soap bubbles. But of course there are other standards
for comparison.

17. See Caxton’s discussion of this (www.instructables.com/id/Unique-Pendulum-Wave-and-


Release-Mechanism/ ). My angle of release was about 30 ᵒ--far exceeding the limits
imposed by the small-angle approximation.
18. See Christina J. Howard in Attention Perception and Psychophysics (pp. 2087-2095) et al.

19. Our forebears may have reacted to such displays of unaccountable intelligence by
supposing the intrusion of the supernatural. Today’s trained observers respond to such
rustlings in the bushes by systematically flushing out quivering perpetrators like Pi. But
they still wonder how Pi got there.

20. “Pendulum waves: A lesson in aliasing,” by James A. Flaten and Kevin A. Parendo
(American Journal of Physics 69, 778 (2001) (http://dx.doi.org/10.1119/1.1349543).
Curiously, those authors only consider the patterns observed from above.

21. www.physlink.com/education/askexperts/ae490.cfm

22. In developing their claim that the patterns we see are the effect of spatial aliasing, the AJP
authors state that “the pendulum bobs can be observed at all times” and that it is only their
spatial locations that are “discrete.” One wonders, however, whether we can be so
confident that the brain continuously absorbs temporal information. Psychologists suggest
that when we observe any continuous (analogue) information, there is a rhythm with which
we absorb the information (which could be thought of as our “sampling rate”):
“The first step to understanding, brain scientists say, is to realize that there is no
Cinemascope screen in the brain where all the pieces come together. But if there is no
screen, on what physical principle is consciousness organized? A growing number of
scientists say the answer must lie in some form of timing. An image may be
reconstructed from all cells that are firing in a particular rhythm at a particular instant.”
(http://www.nytimes.com/1995/03/21/science/how-the-brain-might-work-a-new-
theory- of-consciousness.html?pagewanted=all)

23. See the AJP article for a full-blown mathematical analysis.

24. When I try to relate all of this to art (as I am sometimes wont to do), I come up with (what
I’ll loosely call) a Nyquist* Theory of Art: Artists look at the continuum of life and sample it
according to an internal rhythm. What they produce is an alias—a pattern that overlooks a
great deal and yet nonetheless faithfully represents aspects of the continuum. The meaning
we find in art is real in the sense that we do actually resonate with its rhythm. But that
meaning also is an illusion in the sense that so much has been overlooked by the sampling
that compelling arguments can be made for entirely different patterns.

*Nyquist is the name of the engineer who proposed that we can only avoid aliasing
if our sampling rate is at least twice the frequency of the signal that it is sampling.

Vous aimerez peut-être aussi