Physics 6A

Lab #1: Kinematics, Measurement, and Uncertainty

Introduction:
In this lab, we ask you to design and carry out your own experiments to measure
several everyday quantities: your top running speed, your acceleration as you begin
running from a standing start, or as you try to stop ("decelerate", in everyday language),
and the speed with which you can throw simple projectiles like beanbags. We also
describe several "variants" of each experiment, which you might want to try in addition
to or instead of the main experiments. The more creative you are with your designs, the
better!

In all these cases, we're less interested in the exact numbers you come up with (after
all, your lab TA doesn't actually know your top running speed) than in the thought
process that goes into your experiment. In particular, you will want to pay close
attention to the sources of error or uncertainty in your experiment. When possible, try to
find ways to reduce this uncertainty, to make your results more accurate and precise.
You will never be able to eliminate uncertainty, so you will want to estimate and clearly
explain the remaining source(s) of error in your results.

In Appendix A, at the end of this writeup, we discuss the issue of experimental errors in
greater detail, including a number of informal but quite useful techniques for estimating
the uncertainty in a measurement, and figuring out how it contributes to the uncertainty
of the final result. You should read Appendix A thoroughly before designing your
experiments for this lab.

One thing worth noting right off the bat, though: you'll notice that I tend to use the words
"uncertainty" and "error" interchangeably. This is intentional; I'm trying to guide you
away from a trap that many students fall into. When we speak of experimental error, we
usually do not mean that the experimenter made a mistake or a blunder. Every
measurement is uncertain, and even if the measurements were perfect, the equations
that we are "testing" in a lab are usually a simplified model of how nature really works.
(For example, in part III of this lab, we neglect air resistance while analyzing the flight of
a thrown beanbag. Since air resistance did have at least a slight effect on the flight, our
equations will never be a perfect match to reality.)

So in this and other labs, if we ask you to identify sources of error, don't just focus on
things you feel you might have messed up (although those are worth mentioning too.)
Instead, think of anything which could make your results uncertain, or make reality differ
from a simplified equation that you're using.

And for more discussion of that issue, see Appendix A.

Apparatus:
You'll have access to an assortment of measuring and surveying tools: tape measures,
stopwatches, string, plum bobs, chalk to mark courses, protractors in case you want to
get creative, etc. If you think of any additional tools you'd like to use, contact me
before the lab, and I'll see if I can obtain what you want. If you own any additional
equipment which you think would make an interesting addition to this lab, by all means,
bring it along.

Weather permitting, we'll have lots of space outdoors to set up your experiments. If it's
raining acidic toads or something, we might have to set things up indoors, which will be
more difficult but manageable.

Part I: Measuring a person's top running speed

Your goal here is simple enough: design and carry out an experiment to determine how
fast you can run, in units of meters per second.

That much, everyone should be able to do; you probably already have at least one idea
on how to do it as you read this. A bit more challenging will be the issue of experimental
uncertainty: what are the most important sources of systematic and random error in
your results? How large do you estimate these errors to be? And how can you change
your experimental procedure to make your results as accurate and precise as possible?
(For example, repeating the experiment several times and averaging the results will
help you reduce random uncertainties. But to reduce systematic errors, you may have
to tweak the overall design of your experiment.)

To help you reason your way through these questions, Appendix A includes a discussion
of experimental uncertainty, along with an example drawn specifically from this part of
the lab. It also explains what we mean by "systematic" vs. "random" errors; make sure
you understand that terminology, because we'll use it throughout the course.

When you write up this section of the experiment in your lab journal, be sure to include
at least the following information:

Experimental Procedure:
what did you do, and why? (Explain how the choices you made help you reduce
sources of error.)

Data and Calculations:

all the good number crunching stuff. What were your actual measurements, and how
did you use them to calculate your running speed.

Analysis of Uncertainty:
as we asked above; what are the most important sources of systematic and random
error in your results? How large do you estimate these errors to be? Can you tell if the
error should give you a speed that is faster than your true top speed, slower, or is there
no way to tell? (For systematic errors you will often be able to predict in which direction
they distort your results; for random errors, you usually cannot.) Explain the reasoning
behind your estimates; Appendix A will help.

Variants for Part I:

If you have time, measure your speed using more than one method, and compare the
results. For example, if you figure out how to change your experiment to eliminate
some form of systematic error, you could run both the "good" and the "bad" version of
the experiment, and see how much difference it makes.

You might also want to bring along a simple vehicle of some sort (bicycle, skateboard,
roller skates...) and measure your speed on that contraption in addition to or instead of
your running speed.

Part II: Measuring a person's acceleration.

Again, the goal is simple. You have already measured your maximum speed; now
design and carry out an experiment to measure your acceleration as you start from rest
and accelerate toward higher speeds.

This may be a more challenging experiment than the last one; your results will almost
surely be more uncertain than before. For example, you might find yourself trying to
use a stopwatch to measure fairly short time intervals, which is never easy. That's all
right; even if all you can do is to get a good ballpark estimate of your acceleration, that's
infinitely better than no estimate at all.

A few ideas to get you started:

Note: the following indented section is meant as ideas to help you design your experiment.
Anyone who tries to treat it as a step-by-step set of instructions for what to do in lab will be
summarily keelhauled! Or at least they would, if we had a boat...

You might have to assume that the person's acceleration is constant (at least
for the first few moments of their journey; clearly they can't keep accelerating
at a constant rate forever, or their maximum speed would be infinite!) This
assumption may not be quite true, even for a brief interval of time...but if you
accept this source of error (systematic or random?) it opens up some possible
ideas for experiments.

In particular, if you assume the acceleration is constant, you could try using
any of the three Equations of Kinematics at Constant Acceleration, which we
discussed in lecture on friday. This suggests at least three possible

v = v0 + at (or v = at , if your starting velocity v0 is zero.)

Can you time how long it takes them to reach the maximum velocity you
measured in Part I? (This might rely on the person judging for themselves
 when they've hit maximum velocity, and calling out to the timer...probably a
significant source of error right there!)

v2 = v02 + 2ax (or v2 = 2ax , if v0 is zero.)

So if you try to use this equation, you'll need to measure the distance the
person travels before they reach top speed, rather than the time it takes.
Which one seems easier to measure accurately?

x = v0 t + 1/2 at2 (or x = 1/2 at2, if v0 is zero.)

If you use this equation, you'll need to measure the time it takes you to travel
a given distance. Might seem a lot like what you did in part I...except this
time, the distances will have to be much shorter. (Why?) One advantage
here is that you can try measuring your time for a several different distances,
and see if you get consistent values for the acceleration a. A disadvantage is
that, once again, you'll have to try to measure some very short time intervals.

Once you've designed your experiment, go ahead and measure your acceleration. As
you write up your experiences in your lab journal, include all the same types of
information we suggested in Part I. It's very likely that your sources of error, both
systematic and random, will be much larger than in Part I.

Variants for Part II:

As in part I, you might try this experiment with vehicles of various sorts, instead of just a
running human. Also, you might devise a way to measure your acceleration as you slow
down from top speed and come to a stop. If you've studied a fair amount of physics or
mathematics in the past (and therefore already know some of the concepts we'll
develop later in the class), you might even be able to figure out how to measure your
acceleration as you run in a circle, changing direction rather than speeding up or
slowing down.

Part III: Measuring the launch speed of a thrown object.

Grab a beanbag or other convenient projectile (we'll have an assortment on hand). Your
goal this time is to devise and carry out an experiment in which you throw the projectile
as hard as you can, and measure its velocity as it leaves your hand.

Once again, there are myriad ways to do this experiment. A couple of thoughts which
might get you started on your own ideas: (Again, these are food for thought, not a step-
by-step set of instructions.)

If you assume that air resistance has negligible effect on the object's flight, then its
horizontal velocity should remain constant throughout the flight. You could just
 stand a known distance from a wall, throw the object at the wall, and have a partner
time how long it's in the air. The advantage here is simplicity. The disadvantage is
that you're trying to time a fairly short flight, which will probably introduce significant
uncertainty (random or systematic?). I suppose you could make it a very long
flight...but in that case, it might become less and less accurate to neglect air
resistance, which introduces a different source of error. (Is this one random or
systematic?)

Alternatively, you could take advantage of a result we mentioned in lecture on

Friday: if you launch a projectile horizontally with velocity v0 from a height h above
the ground (and if, again, we neglect air resistance), then the distance it will travel
before hitting the ground is:

L = v0 (2 h / g)1/2 where g = 9.8 m/s2 (near the Earth's surface)

If you launch the projectile from a significant height above the ground -- the wall
overlooking the courtyard between Thimann Labs and our usual lecture hall makes
a great launch point -- and measure where it lands Note that if you design your
experiment around this equation, you won't need to measure any times -- only
distances, which you may find introduce far less random uncertainty.

Once you've designed your experiments, carry them out and discuss your design and
your results in your lab journal, just like you did in parts I and II.

Variants for Part III:

If you've got time, you might want to try both the ideas suggested above, and see if they
give you reasonably similar results.

You might also try throwing the object straight up into the air; by measuring the time
before it hits the ground, you can figure out how fast it was moving as it left your hand.
(To figure out how to do this, think again about the equations of kinematics at constant
acceleration, using -g as the vertical acceleration.) Does the speed of a projectile that
you throw straight upward the same as the speed of one which you throw horizontally?
(It might not be; you're using the muscles in your arm differently in the two cases. In
fact, you may well be able to predict in advance which throwing motion gives you a
higher launch speed.)

You may want to repeat this experiment with a much lighter projectile: a ping-pong ball,
or even an inflated balloon. I can guarantee you that you will have a much harder time
getting meaningful data about the launch speed; if you are using any of the methods we
hinted at above, you will face huge systematic errors due to air resistance. But it can be
interesting to try the experiment anyway; it drives home the judgment calls you make
when designing any experiment. Sometimes a simplifying assumption (such as
neglecting air resistance) is an excellent idea, other times it is a disaster. If you pay
close attention to sources of uncertainty in your results, you'll be able to tell the
difference.
 Done?
Congratulations; thanks for helping us experiment with a new type of kinematics lab. If
you've still got time left, you may want to go back and try some of the variants listed for
each part of the lab. Or you could work to refine your write-ups of what you've already
done...or invent and carry out some entirely new experiments that seem more
interesting than the ones we suggest here.

Thanks again, and I hope you've had fun.

Appendix A: Experimental Error and Uncertainty

(vital for this lab)

Any time you try to measure what's going on in the physical world, and to match it
against a model of what the laws of nature say should happen, you'll never get a perfect
match.

There are two main reasons for this: your measurements were imperfect (because of
the limitations of instruments and of the people operating them) and your model was
imperfect (because you made simplifying assumptions which were approximately true,
but not exactly.)

So when designing and analyzing an experiment, you always want to be aware of

sources of uncertainty and error. When you report results, you want to say not only "we
calculated the launch velocity to be 43 m/s", but also something like, "we think this
result is accurate to +/- 5%".

There are two types of error that arise in experiments: random and systematic. Often,
you'll deal with them in sharply different ways:

Random Errors:
These errors will be different each time you repeat the experiment. They often arise
from the imperfections in a measuring instrument, or the difficulty in using it perfectly. If
you use a ruler to measure the length of a string, you might be able to read it to better
than +/- 1 mm. If you just eyeball that same string, maybe you can only estimate its
length to within +/- 20 cm.

The reason this error is random should seem clear: when you measure the length of
one string, you might get an answer that's a millimeter or two too long. When you turn
to the next string, maybe you'll measure it a millimeter too short.

If random sources of uncertainty are important to your experiment, then it's a good idea
to make several identical runs and average the results together. This has two benefits:

• Taking multiple trials helps you estimate the random uncertainty.

 If you make the same measurement several times under identical circumstances, then
you can see at a glance how much the result seems to vary. If you feel like applying
some statistics, you can calculate the standard deviation of the measurements. If you
don't want to go into that level of mathematical detail, you can simply say, "it looks like
most of these measurements are clustered within a few millimeters of each other. I'll
call the random error +/- 3mm."

Note that even if you do not take multiple measurements, you should still try to
estimate the random errors in your data. This could take the form of a common-sense
order-of-magnitude estimate: "Do I think I measured that time correctly within 0.001
seconds? Naw, no way. Within 0.01 seconds? Still no. Within 0.1 seconds?
Maybe...still sounds a little too precise. Within 1 second? I was way more accurate
than that. Call it +/- 0.2 seconds."

• Taking multiple trials actually reduces the random uncertainty.

(Once you average them together.)

If you repeat a measurement several times and average them together, the random
errors tend to cancel out, and the average moves closer to the "true" value.

Systematic Errors:
A systematic error is one which will affect your results the same way every time you
repeat the experiment. Averaging together many trials doesn't help, because the
systematic error affects all of them.

This type of error often arises from the simplifying assumptions you make in your
calculations. If you assume that air resistance doesn't affect a projectile, while in reality,
air resistance is slowing the projectile by 4% during its flight, then your results will be off
by 4% no matter how many times you repeat the experiment.

Since you can't get rid of systematic errors by taking lots of data, you may want to
redesign your experiment to reduce or eliminate their effect. When that's not practical,
the best you can do is be aware of systematic errors, and estimate how badly they
distort your results. One thing that is nice about systematic errors is that you can often
figure out which direction they modify your resuts. Air resistance, for example, will only
slow a projectile down, never speed it up.

Absolute vs. Percentage Error:

These are two ways of expressing the same thing: "t = 4.0 s +/- 0.2 s" means exactly
the same thing as "t = 4.0 s +/- 5%" (because 0.2 / 4.0 * 100% = 5%). Often, it's more
meaningful to express errors or uncertainties as percentages. After all, an uncertainty
of +/- 3 minutes could be a miniscule error if you're measuring the age of a star, but a
huge error if you're measuring someone's time in a one-mile race. Something like +/-
2% is much easier to interpret.
Having said that, there are times when absolute error is more important. We discussed
this in lecture on Friday...but as a reminder, here are two approximate but very useful
rules of thumb:

• Whenever you are going to add or subtract two numbers, the one with the greater
absolute uncertainty determines the uncertainty of the result. If x = 500 cm +/- 5 cm, and
y = 50 cm +/- 2 cm, then x + y = 550 cm +/- 5 cm.

• Whenever you are going to multiply or divide two numbers, the one with the greater
percent uncertainty determines the percent uncertainty of the result. The above values
of x and y , expressed with percent uncertainties, are: x = 500 cm +/- 1%, and y = 50 cm
+/- 4%. Thus, xy = 25000 cm

2
+/- 4%.

The above rules are not exact -- there is a more exact technique called adding in
quadrature which would give better results -- but they make a good first step toward
dealing with uncertainty in calculations.

An example, tying it all together:

Abe and Beth mark off a twenty-meter long course on the walkway outside Thimann
Labs. Beth waits at the starting line until Abe, who is waiting at the finish line shouts
"Go!" and simultaneously starts his timer. Beth sprints to the finish line, and Abe stops
the timer as she passes the line. They calculate Beth's speed by the obvious method:

v = distance / time.

There are several sources of error in their technique. One is a systematic error: they
want to measure Beth's top speed, but since she started from rest, she was running
slower than her top speed for much of the course. With a little thought, they could find a
way to eliminate this source of systematic error from their experiment (can you figure
out how?)

Their other source of uncertainty is a random one: Abe has estimated that his
measurements of Beth's running time are only accurate to +/- 0.2 seconds. So if Beth's
average time was t = 4.0 seconds, then they've got a 5% uncertainty in her time. The
percent uncertainty in distance is probably much smaller -- it's easy to measure the
distance from the start to the finish line, but hard to time when a person crosses these
lines -- so when they divide distance by time, that 5% uncertainty will carry through to
the velocity calculation.

Again, there are ways they could redesign the experiment and reduce this time
uncertainty to much less than 5%, even if Abe never gets any better at using the
stopwatch -- can you figure out how?