Motion Lab · Guide · glidepathscience.org
GuideMotion Lab
Contents
This Guide explains the ideas and equations behind every Walkthrough step, Concept Challenge question and Predict round, in the order a textbook would take them. Part 1 prepares the Intro track and Part 2 the Advanced track; each takes about fifteen minutes to read. Part 1 follows one run of its own, drawn in its figures, and every section ends with a worked example that is not one of the sim's own problems.
Part 1 · The essentials
Math you’ll need: algebra, slopes of lines, areas of rectangles and triangles; calculus only in “With calculus” boxes.
Position, the x–t graph and the motion diagram
Kinematics begins with a plain question: where is the object, and when? Along a straight line, choose a point as the origin and one direction as positive. The object's position x is its signed distance from the origin, in meters. Neither choice is physical: moving the origin shifts every position by the same amount, and flipping the direction changes every sign, but the motion is the same. The change in position from an earlier moment to a later one is the displacement,
$\Delta x = x_2 - x_1$
where $x_1$ is the earlier position and $x_2$ the later one. The symbol Δ, "change in", always means later minus earlier.
A position–time graph, or x–t graph, plots x against the time t. It is the most often misread picture in kinematics, because it looks like a path. A cart on a level track gives an x–t line that climbs while the cart moves in the +x direction and falls while it comes back; the vertical axis is distance along the track, not height. Reading the graph as a hill is the first mistake to unlearn. What the graph adds to a photograph is the timing.
A motion diagram puts that timing back on the track. Mark the cart's position at equal intervals of time, as a strobe photograph would. Every gap took the same time, so wide gaps mean fast motion and crowded dots mean slow motion.
The rest of Part 1 follows the run in the figure. The cart starts from rest at x = 0, speeds up for 2.0 s, cruises for 2.0 s, then brakes steadily. It stops for an instant at t = 6.0 s, 5.60 m out, comes back, and ends at x = 3.50 m at t = 8.5 s. Its dots are 0.70 m apart during the cruise and only 0.09 m apart just before the turn: the farthest dot is not the fastest, since at the far end the cart is momentarily at rest.
Dot spacing says roughly how fast. To put a number on it, we need the slope of the x–t graph.
Velocity is the slope of position
How fast the position changes, and in which direction, is the velocity. Over an interval Δt, the average velocity is the displacement divided by the time. On the x–t graph it is the slope of the straight line, the secant, joining the points at the two ends of the interval. Shrink the interval around one instant and the secant becomes the tangent there; its slope is the instantaneous velocity v:
$v_{avg} = \Delta x/\Delta t$ v = slope of the tangent to x–t
Velocity carries a sign: a rising x–t line means motion in the +x direction, a falling one motion toward −x. Its size, |v|, is the speed. A straight stretch of x–t has one slope, so the velocity is constant there; a stretch that curves upward (concave up) has a velocity that increases.
The classic slip is to read height as speed. A line high on the graph belongs to a cart far from the origin; a steep line belongs to a fast one. A flat line, however high, is a cart at rest.
A curved x–t graph means a changing velocity. How quickly it changes is the next quantity.
Acceleration is the slope of velocity
The acceleration measures how quickly the velocity changes, and it follows from velocity exactly as velocity follows from position, one graph further down. The average acceleration is the slope of a secant on the v–t graph, and the instantaneous acceleration a is the slope of the tangent:
$a_{avg} = \Delta v/\Delta t$ a = slope of the tangent to v–t
Its unit is m/s². A straight stretch of v–t means constant acceleration; a flat stretch means a = 0, constant velocity at whatever value it has.
A motion diagram can show acceleration too. Draw the velocity as an arrow at each dot; the change from one arrow to the next is Δv. While a is constant, $\Delta v = a\Delta t$, and in general Δv is the acceleration accumulated over the interval, so the Δv arrows point the way the acceleration points. When Δv points along the velocity the cart speeds up; when it points against it, the cart slows down. Two habits cause trouble: calling a large velocity a large acceleration, and forgetting that Δv is final minus initial, so a velocity that ends lower than it began has a negative Δv.
So the sign of a, on its own, does not say whether the cart speeds up. The next section says what does.
Turning points and the sign rule
Two common beliefs are false: that an object momentarily at rest has no acceleration, and that a negative acceleration means slowing down. At a turning point the velocity passes through zero and changes sign, and the position reaches a maximum or a minimum. There v = 0, but a is the slope of v–t, and nothing about a zero crossing makes that slope zero. If the acceleration vanished at the turn and stayed zero, the velocity would stay zero and the cart would never come back. A ball thrown straight up is the familiar case: at the top it is at rest for an instant and still accelerating downward at 9.81 m/s².
Whether the speed grows depends on the signs of v and a together. That is the sign rule: an object speeds up when v and a have the same sign and slows down when their signs differ. On a motion diagram it is plain to see: a Δv arrow pointing along the velocity arrow lengthens it, and one pointing against it shortens it. A negative acceleration slows a cart moving toward +x and speeds up one moving toward −x; "negative" names a direction, not a slowdown.
Slopes take us down the list, from x to v to a. Areas take us back up.
Areas: displacement, distance and the change in velocity
The operation that undoes a slope is an area. For a constant velocity the displacement is $v\Delta t$, a rectangle under the v–t graph; any v–t graph can be cut into thin, nearly rectangular strips, so in general the displacement is the area between the v–t graph and the time axis. One graph down, the change in velocity is the area under a–t:
Δx = area under v–t Δv = area under a–t
These are signed areas: area below the axis counts as negative. The distance traveled is different. It counts every piece as positive, adds every leg of the trip, and never decreases, so it equals the size of the displacement only when the object never turns around.
The usual errors are bookkeeping: subtracting the pieces when the question asks for distance, adding them when it asks for displacement, or forgetting the starting velocity. With constant acceleration, every area is a rectangle or a triangle, and the bookkeeping becomes three equations.
Constant acceleration
Many motions keep a nearly constant acceleration for a while, such as a car braking hard or a ball in free fall. Let $x_0$ and $v_0$ be the position and velocity at t = 0. With a constant, the a–t graph is a rectangle of area at, so the velocity is $v_0 + at$. The v–t graph is then a straight line whose area is a rectangle $v_0t$ plus a triangle $\frac{1}{2}(at)t$. Eliminating t between the two gives the third equation:
$v = v_0 + at$ $x = x_0 + v_0t + \frac{1}{2}at^2$ $v^2 = v_0^2 + 2a\Delta x$
Here $\Delta x = x - x_0$. The third equation has no t in it, which makes it the one to use for "how far" rather than "how long". Braking to a stop is the standard case: the final v is zero, so the time to stop is $v_0/{|a|}$ and the stopping distance is
$\Delta x = v_0^2/2{|a|}$
The distance grows as the square of the starting speed: twice the speed needs four times the road. The common error is to multiply $v_0$ by the stopping time, as if the car kept its starting speed; its speed falls steadily, so its average is half of $v_0$, and that error doubles the answer.
Every graph so far has been exact. A real experiment hands us something rougher.
Real data: a detector measures position
An ultrasonic motion detector sends out a short pulse of sound and times the echo; knowing the speed of sound c, it turns the round-trip time $t_{echo}$ into a distance,
$d = c\,t_{echo}/2$
with the 2 because the sound goes out and back. A typical detector does this 10 to 50 times a second, to about a millimeter. What matters is what it does not measure: velocity and acceleration. The software computes them from differences of neighboring positions, and each difference divides a small error in position by a short time. A few millimeters of scatter in x become a visible scatter in v and a much larger one in a. A noisy acceleration graph does not mean the cart is jittering or the detector is broken; it is what taking differences of real data does.
Part 2 · Going further
Math you’ll need: derivatives, integrals, the chain rule and exponentials; standard deviations for the data.
Two constants, and where the pieces join
In the language of calculus, velocity and position are integrals of acceleration, and an integral fixes a function only up to a constant. Integrating twice brings two constants, the starting velocity $v_0$ and the starting position $x_0$:
$v(t) = v_0 + \int a\,dt$ $x(t) = x_0 + \int v\,dt$
with each integral taken from 0 to t. Thinking that a(t) alone fixes the motion is the mistake to avoid here. Two carts with identical acceleration graphs have v–t graphs shifted by the difference in $v_0$, and x–t graphs that differ by $\Delta x_0 + \Delta v_0t$, a straight line.
Runs built from pieces of constant acceleration raise a second question: what happens where two pieces meet? The position must be continuous, since a jump in x would need an infinite velocity, and so must the velocity, since a jump in v would need an infinite acceleration. The acceleration may jump in the model: that puts a corner in the v–t graph, not a gap, and needs only an infinite rate of change of acceleration. It is an idealization; a real cart's acceleration takes some milliseconds to change.
Real accelerations cannot jump, so the next step is to let them change gradually.
When the acceleration changes: ramps and jerk
The rate of change of acceleration is the jerk, j, in m/s³; passengers feel it as a lurch. Bars of constant acceleration have zero jerk inside each bar and infinite jerk at each join; ramps, which change the acceleration in straight lines, keep it finite, and their v–t graphs have no corners. The simplest ramp is an acceleration that falls steadily, $a = a_0 - bt$, where $a_0$ is the starting acceleration and b, in m/s³, its rate of fall; its jerk is −b. The change in velocity is still the area under a–t, and integrating once more gives the position:
$j = da/dt$ $a = a_0 - bt$ $\Delta v = a_0t - \frac{1}{2}bt^2$
$x = x_0 + v_0t + \frac{1}{2}a_0t^2 - \frac{1}{6}bt^3$
The tempting mistake is to use Part 1's constant-acceleration equations with a set to its starting value. They overestimate everything, because the acceleration shrinks the whole time.
An acceleration can also depend on velocity or on position instead of time.
When a depends on velocity
A cart coasting against a resistance that grows with its speed, such as a magnet's eddy-current brake, slows more strongly the faster it goes. The simplest model makes the acceleration proportional to the velocity, $a = {-v/\tau}$, where the time constant τ, in seconds, sets how quickly the motion dies away. The equation $dv/dt = {-v/\tau}$ is solved by an exponential, and integrating it gives the position:
$v = v_0e^{-t/\tau}$ $x = x_0 + v_0\tau(1 - e^{-t/\tau})$ $x \to x_0 + v_0\tau$
The cart never stops, since an exponential never reaches zero, yet it never passes $x_0 + v_0\tau$: the remaining distance shrinks as fast as the speed. After one time constant the speed is down to 1/e, 37 %, and the cart has covered 63 % of its total distance; after three, 95 %. Treating the starting acceleration as constant would stop the cart at t = τ, halfway to the true limit.
When the acceleration is given at each place rather than each moment, time is the wrong variable.
When a depends on position
On a rough patch or a spring, the acceleration is set by where the object is. The chain rule removes time: since $a = dv/dt$ and $v = dx/dt$,
$a = v\,dv/dx$ $\Delta(\frac{1}{2}v^2) = \int a\,dx$
The second form, the first integrated over position, says that the change in half the square of the speed is the area under the a–x graph. For a constant acceleration over a distance Δx that area is $a\Delta x$, which gives back $v^2 = v_0^2 + 2a\Delta x$ with no time in sight. The classic slips are adding $a\Delta x$ to a speed, as in $v_{in} + a\Delta x$, which adds m²/s² to m/s, and dropping the 2.
So far every motion was known exactly. The rest of Part 2 turns to motion measured from data.
From readings to derivatives
A detector gives positions $x_{\mathit{k}}$ at times $t_{\mathit{k}}$ a step Δt apart, so derivatives must be estimated from differences. Write $x_+$ and $x_-$ for the readings one step after and before time t. The forward difference uses the reading at t and the next one; the central difference uses the two on either side:
forward: $v \approx (x_+ - x)/\Delta t$ central: $v \approx (x_+ - x_-)/2\Delta t$
Expanding the position about t shows how good each is. With a constant, $x_\pm = x \pm v\Delta t + \frac{1}{2}a\Delta t^2$. The forward difference comes out as $v + a\Delta t/2$, off by $a\Delta t/2$, because it is really the velocity at the midpoint of the step. In the central difference the $a\Delta t^2$ terms cancel, and it is exact for any parabola. A forward difference plotted at t rather than at the midpoint runs early by half a step.
Exact readings are a luxury. With real ones, a difference does something worse than shift the answer.
Noise grows with every derivative
Call the standard deviation of each reading's random error $\sigma_x$, with independent errors from reading to reading. A difference is a weighted sum of readings, and independent errors add in quadrature: the result scatters by $\sigma_x$ times the square root of the sum of the squared weights, not by the sum of the weights. The central difference has weights ±1/(2Δt); the three-point second difference, $a \approx (x_+ - 2x + x_-)/\Delta t^2$, has weights 1, −2 and 1, divided by $\Delta t^2$:
$\sigma_{\mathit{v}} = \sigma_x/(\sqrt{2}\,\Delta t)$ $\sigma_{\mathit{a}} = \sqrt{6}\,\sigma_x/\Delta t^2$
The $\sqrt{6} = \sqrt{1^2 + 2^2 + 1^2}$ is the quadrature at work; adding the weights as 1 + 2 + 1 = 4 gives the worst case, not the typical scatter. The surprise is the rate: with Δt = 1/f, $\sigma_{\mathit{a}}$ grows as the square of the sampling rate f, so sampling faster makes differences noisier. (The Motion Detector tab's central option applies the central difference twice, $(x_{+2} - 2x + x_{-2})/4\Delta t^2$, with a quarter of the three-point scatter.)
Faster sampling pays off only when many readings are used at once, which is what smoothing and fitting do.
Smoothing and fitting
A moving average replaces each value by the mean of N neighbors, and averaging N independent readings shrinks their scatter by $\sqrt{N}$. A centered average adds no delay but rounds every corner. A trailing average uses only the latest N readings, so it can run live, but its window reaches back $(N - 1)\Delta t$ and its value belongs to the window's middle:
lag = $(N - 1)\Delta t/2$ noise ÷ $\sqrt{N}$
Taking the whole window as the lag doubles it, a frequent error. Where the acceleration is known to be constant, a better tool is to fit a parabola, $x = c_0 + c_1t + c_2t^2$, to all the readings by least squares and take $a = 2c_2$. For N readings spread evenly over a time T, the fit's uncertainty is about
$\sigma_{\mathit{a}} \approx \sigma_x\sqrt{720/N}/T^2$
At a fixed rate N grows with T, so this falls as $T^{-5/2}$: recording 1.5 times as long cuts it by $1.5^{5/2} \approx 2.8$. The price is the model. A fit assumes the acceleration really is constant across its window. A turning point is no trouble, since a steady acceleration carries the cart straight through it, but a window that includes a change in the acceleration, such as a corner between bars or a bounce, mixes two motions, and its small error bar then means nothing.
A fit needs a stretch of steady acceleration. A bounce is the opposite.
A bounce the detector cannot resolve
A ball's bounce lasts a few milliseconds, far less than the time between detector readings. A ball dropped from a height h lands at $\sqrt{2gh}$ and leaves at e times that, where the coefficient of restitution e is the ratio of leaving to landing speed. With up positive, the bounce changes its velocity by
$\Delta v = \sqrt{2gh}\,(1 + e)$ $\int (a + g)\,dt = \Delta v + g\,t_c$
The area under the net acceleration across a contact of duration $t_c$ is Δv. Measured above the free-fall line, −g, the spike's area is the floor's share, $\Delta v + g\,t_c$, nearly the same for so short a contact. A detector reading 20 times a second sees only a velocity that changes between two readings, so its acceleration spike is spread over a sample interval: far lower than the real one and dependent on the rate, but with very nearly the right area. A smoother spike is not a more faithful one.
Noise and lost spikes are random or easy to spot. The last error is neither.
The detector's ruler: the speed of sound
The detector's distances rest on the speed of sound it assumes, and that speed depends on the air's temperature. In dry air at T degrees Celsius, the straight-line form of $331.3\sqrt{1 + T/273.15}$ m/s is
$c \approx 331.3 + 0.606\,T$ reading = $d \times c(T_{cal})/c(T)$
A detector calibrated at $T_{cal}$ converts every echo time with $c(T_{cal})$. In warmer air sound is faster, the echo returns sooner, and every distance reads short; in colder air every distance reads long. Getting that direction backward is the usual slip. Since v and a come from the positions, they are off by the same factor. This is a systematic error: it shifts every reading the same way, so no averaging or fitting removes it. Recalibrating in the room does.
Reference
Common mistakes
Part 1
- Reading an x–t graph as a path. A rising line means farther along the track, not uphill.
- Reading height as speed. Speed is the steepness of x–t; a flat line is at rest.
- Mixing up average and instantaneous. The secant gives the average; the tangent the value at an instant.
- Setting a = 0 at a turning point. v passes through zero; a keeps the slope of v–t.
- Reading negative acceleration as slowing down. Compare the signs of v and a.
- Confusing displacement and distance. Area below the axis subtracts from one and adds to the other.
- Using the starting speed for a whole stop. The average speed is half of it: $\Delta x = v_0^2/2{|a|}$.
Part 2
- Using constant-acceleration equations when a varies. Use areas, $v\,dv/dx = a$, or the exponential.
- Adding errors linearly. Independent errors add in quadrature, and faster sampling makes differences noisier.
- Expecting averaging to fix a calibration. A wrong speed of sound shifts every reading the same way.
Equation sheet
| Part 1 | |
|---|---|
| slopes | $v = \Delta x/\Delta t$ (x–t); $a = \Delta v/\Delta t$ (v–t) |
| areas | $\Delta x$ = signed area under v–t; $\Delta v$ = signed area under a–t; distance counts every leg positive |
| sign rule | speeding up when v and a share a sign |
| constant a | $v = v_0 + at$, $x = x_0 + v_0t + \frac{1}{2}at^2$, $v^2 = v_0^2 + 2a\Delta x$ |
| stopping | at the turn v = 0; braking distance $v_0^2/2{|a|}$ |
| catch-up | from rest: $\frac{1}{2}at^2 = v_Bt$ |
| Part 2 | |
|---|---|
| integrals | $v(t) = v_0 + \int a\,dt$, $x(t) = x_0 + \int v\,dt$ |
| ramp | $a = a_0 - bt$: $\Delta v = a_0t - \frac{1}{2}bt^2$ |
| a(v) | $a = {-v/\tau}$: $v = v_0e^{-t/\tau}$, $x \to x_0 + v_0\tau$ |
| a(x) | $\Delta(\frac{1}{2}v^2) = \int a\,dx$, $v^2 = v_0^2 + 2a\Delta x$ |
| noise | $\sigma_{\mathit{v}} = \sigma_x/(\sqrt{2}\,\Delta t)$ (central), $\sigma_{\mathit{a}} = \sqrt{6}\,\sigma_x/\Delta t^2$ (three-point) |
| smoothing | trailing average of N: lag = $(N - 1)\Delta t/2$ |
| fit | $a = 2c_2$, $\sigma_{\mathit{a}} \approx \sigma_x\sqrt{720/N}/T^2$ |
| sound | $c \approx 331.3 + 0.606\,T$ m/s (T in °C); reading = $d \times c(T_{cal})/c(T)$ |
Symbols: x position (m) · t time (s) · v velocity (m/s), |v| speed · a acceleration (m/s²) · $x_0$, $v_0$ values at t = 0 · $v_B$ the passing vehicle's speed · $a_0$, b a ramp's start and rate of fall · τ time constant · Δt time between readings, f its rate · $\sigma_x$, $\sigma_{\mathit{v}}$, $\sigma_{\mathit{a}}$ scatter of x, v, a · N readings over a time T (in the fit row) · c speed of sound · in the sound row, T the air temperature (°C) and $T_{cal}$ the calibration temperature.
Further reading
OpenStax, University Physics Volume 1, chapter 3, "Motion Along a Straight Line" (openstax.org, CC BY-NC-SA 4.0), covers Part 1 at the same level. On students' difficulties with motion graphs: L. C. McDermott, M. L. Rosenquist and E. H. van Zee, Am. J. Phys. 55, 503 (1987), and R. J. Beichner, Am. J. Phys. 62, 750 (1994). On fitting and its uncertainties: J. R. Taylor, An Introduction to Error Analysis, 2nd ed. (1997). The Field Manual lists every source the sim uses.
Motion Lab v1.4 · © 2026 Dr. David Meier · CC BY-NC-SA 4.0 · glidepathscience.org