Why an Electric Toothbrush Is a Timed Control System
An electric toothbrush combines a motor, gears or magnetic drive, moving brush head, timer, pressure sensing, battery and control logic. The head may rotate, oscillate through an angle, vibrate along an axis or combine motions. A label such as “movements per minute” is meaningful only when the counted event is defined.
This is why mathematics matters. Angular displacement is not the same as distance. Frequency is not the same as speed. A two-minute timer is a state machine, not merely a stopwatch. A pressure warning uses thresholds and may include delay or hysteresis. Battery runtime connects power with energy. Measurement becomes useful only when units and event definitions stay consistent.
This article does not give dental advice or claim health outcomes. Follow current guidance from a qualified dental professional and the product’s instructions. Use unplugged diagrams and teacher-supplied data. Do not dismantle a powered or charging toothbrush, immerse chargers, force a stalled head, share brush heads or treat a warning sensor as a calibrated medical instrument.
Quick Reading Routes
- Motion and counting
- Oscillation mathematics
- Timer logic
- Pressure thresholds
- Student investigations
- FAQs
What Exactly Is Being Counted?
An early electric-toothbrush patent describes oscillatory or rotary drive and timing functions. Other designs, such as US6536066B2, describe oscillating heads. Patent documents show disclosed mechanisms; they do not prove that every current product uses the same geometry or performs alike.
Suppose a head completes 7,200 full oscillation cycles per minute. Frequency is 7,200/60=120 cycles per second, or 120 Hz. Period is 1/120≈0.00833 s, or 8.33 ms per full cycle.
If marketing counts each change of direction as one movement, the same mechanical trace could be described as 14,400 directional movements per minute. If it counts one trip from left extreme to right extreme as a stroke, the number may differ again. The mathematics is simple only after the event is defined.
Frequency, angular speed and path length differ
For a head oscillating between −15° and +15°, one full cycle travels 60° in angular path: 30° across and 30° back. At 120 cycles/s, total angular path rate is 7,200 degrees/s, or about 125.7 radians/s of accumulated path.
That number is not the instantaneous angular velocity. A sinusoidal motion slows to zero at each turning point and is fastest near the centre. If angle is θ=A sin(2πft), maximum angular speed is 2πfA, with A in radians.
For A=15°=0.262 rad and f=120 Hz, maximum angular speed is about 2π×120×0.262≈198 rad/s. The mean absolute angular speed over a sinusoidal cycle differs from that maximum.
Bristle-tip distance depends on radius
Arc length is s=rθ. If a bristle point is 8 mm from the rotation axis and the head swings through 30°=0.524 rad from one extreme to the other, its arc length is 0.008×0.524≈0.00419 m, or 4.19 mm.
A point 4 mm from the axis travels half that distance for the same angle. Therefore, “head angle” does not give one bristle-tip speed for the entire brush surface.
Oscillation, Angle, Path and Frequency
Motion can be described as displacement against time. A sinusoid is a useful first model because it separates amplitude, frequency and phase. Real mechanisms may have non-sinusoidal waveforms, compliance, gear play and load-dependent distortion.
Suppose position is sampled every millisecond. Counting zero crossings can estimate frequency, while peak detection estimates amplitude. Noise may introduce false peaks, so an algorithm needs thresholds or a minimum separation between accepted peaks.
Sampling rate constrains what can be measured
A 120 Hz oscillation completes a cycle in 8.33 ms. A sensor sampling at 200 samples/s records only about 1.67 samples per cycle, insufficient to reconstruct shape reliably. At 2,000 samples/s, there are about 16.7 samples per cycle.
The Nyquist condition says a sampling frequency must exceed twice the signal frequency to avoid simple aliasing, but practical waveform measurement often needs much more than two points per cycle. A 240 Hz minimum is a theoretical boundary for 120 Hz, not a guarantee of accurate amplitude or peak timing.
Aliasing can create a false slow motion
If a 120 Hz signal is sampled at 110 Hz, the sampled pattern may appear to vary at 10 Hz. That beat-like alias is not the true mechanical frequency. Stroboscopic videos can show a head seeming to stop or reverse for the same reason.
This is a powerful transfer lesson. Wheel spokes in video, fan blades, sound recording and sensor data all require sampling mathematics.
A Timer Is a State Machine
Many toothbrushes provide a total-session timer and interval cues. A timer patent such as EP2384719A1 describes timing and operating-state logic. The useful mathematical point is sequencing: timers begin, pause, resume, cue and reset according to conditions.
Model four states: OFF, RUNNING, PAUSED and COMPLETE. Pressing start moves OFF to RUNNING. Removing the brush from load might pause in some designs, while another continues. At 30, 60 and 90 seconds the controller may issue a cue; at 120 seconds it may signal completion.
Elapsed time is not always wall-clock time
If the motor runs 28 s, pauses 5 s, runs 32 s, pauses 3 s and then runs 60 s, wall-clock duration is 128 s but active run time is 120 s. A product can define its timer using either basis.
A dataset must state which clock is accumulated. Otherwise, comparing a user’s stopwatch with internal cues can produce an apparent eight-second “error” that is actually a definition difference.
Clock drift accumulates
Suppose an oscillator is fast by 0.15%. Over a nominal 120 s, timing error is 0.0015×120=0.18 s. If cue sound detection has ±0.10 s reaction uncertainty, a single stopwatch test cannot cleanly separate internal drift from human timing.
Repeat automated traces and average cue timestamps, or report a combined interval. More decimal places on the stopwatch do not remove reaction bias.
Pressure Feedback Is a Threshold Problem
Some designs respond to force or pressure-related signals. A sensor-responsive toothbrush patent describes control based on sensed conditions. Such documents illustrate mechanisms, not clinical thresholds or claims for every product.
Strictly, pressure is force divided by area, p=F/A. A handle sensor may estimate force or deflection rather than contact pressure distributed over bristles. Calling every warning “pressure” is convenient, but the measured variable and calibration should be checked.
One threshold can chatter
If a warning turns on above 2.0 N and off below the same 2.0 N, noise around the boundary could make it flicker. Hysteresis might turn on above 2.0 N but stay on until force falls below 1.7 N.
For a trace 1.8, 1.95, 2.05, 1.98, 2.02, 1.75, 1.65 N, a single threshold flips four times. The hysteretic rule turns on at 2.05 N and turns off only at 1.65 N, giving one stable episode.
Delay rejects brief transients
A controller might require the signal to exceed a threshold for 100 ms. A 40 ms bump is ignored; a 160 ms interval triggers. Sampling at 100 Hz gives ten samples in 100 ms, so a consecutive-sample rule can implement the delay.
Delay trades responsiveness against false alarms. A longer validation interval rejects more brief spikes but takes longer to warn. There is no universal optimum without the product’s design objectives and evidence.
Calibration needs known inputs
Suppose supplied calibration points are 0.5 N→0.62 V, 1.0 N→1.08 V, 1.5 N→1.55 V and 2.0 N→2.07 V. A linear fit can convert voltage to estimated force, but residuals should be inspected.
Temperature, housing flex, battery voltage and loading direction may change response. A consumer warning indicator is not automatically a calibrated force instrument.
Battery Energy and Runtime
Energy in watt-hours is average power times time. A hypothetical 3.7 V, 0.80 Ah battery has nominal energy 2.96 Wh. If electronics and motor average 1.2 W during brushing, an ideal continuous runtime is 2.47 h.
If each session lasts 2 minutes, the ideal count is 2.47×60/2≈74 sessions. Real count is lower because conversion losses, standby, battery protection, ageing and load affect usable energy.
Session count is a discrete estimate
If usable energy is 2.20 Wh and each session consumes 0.035 Wh, quotient is 62.9. The complete-session count is at most 62 under that model, not 62.9 completed sessions.
Charging energy from the wall can exceed stored battery energy. Comparing charger input with motor output requires consistent boundaries and time intervals.
Worked Example: From Motion Trace to Summary
A supplied optical trace records 240 complete cycles in 2.00 s. Frequency is 120 Hz and rate is 7,200 cycles/min. Peak angles average +14.8° and −15.2°, giving peak-to-peak angle 30.0° and amplitude about 15.0° around a −0.2° centre offset.
At radius 7 mm, extreme-to-extreme arc is 0.007×(30π/180)=0.00367 m, or 3.67 mm. A full out-and-back cycle covers 7.33 mm at that radius. Multiplying by 120 cycles/s gives 0.880 m/s of accumulated path rate.
This is not the fluid or bristle-tip instantaneous speed at every point. Bristles flex and different radii travel different arcs. Report it as a geometric path estimate from the measured angular trace.
If cycle count uncertainty is ±1 and time uncertainty ±0.01 s, lower frequency is 239/2.01≈118.9 Hz and upper is 241/1.99≈121.1 Hz. Reporting exactly 120.000 Hz would be unjustified.
Worked Example: Timer and Pressure Events
A 125 s wall-clock record includes active intervals 0–31, 34–63, 66–96 and 99–125 s. Total active time is 31+29+30+26=116 s. If cues occur at active 30, 60 and 90 s, wall-clock cue times are shifted by pauses.
The first cue appears near 30 s, the second after 63 s wall time, and the third after 96 s. A naive wall-time analysis might call later cues inaccurate. The state-aware analysis explains the offsets.
In the same trace, estimated force exceeds 2.0 N for 70 ms at active second 18 and for 240 ms at active second 72. With a 100 ms validation delay, only the second event triggers. If hysteresis clears at 1.7 N, warning duration continues until that lower threshold is crossed.
This example joins two independent state machines. The pressure warning need not reset the session timer; design documents determine the interaction.
Common Misconceptions
“Movements per minute always means full cycles”
It may count strokes, directional changes or another defined event. Read the definition.
“A 120 Hz head moves at 120 metres per second”
Hertz counts cycles per second. Linear speed also needs amplitude and radius.
“Two minutes on a stopwatch must match active timer time”
Pause and resume rules can make wall-clock and accumulated run time differ.
“Pressure warning means calibrated contact pressure”
The device may sense force, deflection or a proxy. Contact area and calibration matter.
“More motion guarantees better oral health”
Motion metrics alone cannot establish a health outcome. Technique, individual needs and professional guidance remain separate.
How Students Can Learn and Transfer the Mathematics
Primary learners can count intervals and read simple timelines. Secondary learners can use frequency, period, angle conversion, arc length, duty cycle and energy. Older students can study sinusoidal derivatives, sampling, aliasing, calibration regression and finite-state logic.
Parents can ask, “What counts as one movement?”, “Which clock is the timer using?” and “Is this sensor estimating force or pressure?” Those questions turn vague specifications into mathematical definitions.
The same ideas transfer to robotics, wearable sensors, motors, machine monitoring, digital signal processing and human-machine interfaces. Mathematics strengthens technical judgement; it does not replace dental advice or guarantee a career.
A Topic-Specific Mathematics Laboratory
Use paper mechanisms, recorded traces, spreadsheets and unplugged models. Never run a brush against improvised loads, open a sealed handle or expose a charger to water.
Investigation 1: Define the counted event
Draw one sinusoidal cycle and label full cycle, half-cycle, extreme-to-extreme stroke and direction change. Convert 7,200 cycles/min into each alternative count.
Write a specification sentence that removes ambiguity. Then show how the same mechanism can honestly produce different large numbers under different definitions.
Investigation 2: Convert frequency and period
Make a table for 80, 100, 120 and 150 Hz. Calculate periods in milliseconds and cycles per minute.
Check reciprocity by multiplying fT. Explain why rounding period too early creates a noticeable error when converted back.
Investigation 3: Map angular path
For amplitudes 10°, 15° and 20°, calculate peak-to-peak angle and full-cycle angular path. Convert degrees to radians.
At two radii, calculate arc distance. State that bristle compliance and non-rigid motion are excluded.
Investigation 4: Differentiate a sinusoid
Use θ=A sin(2πft) to calculate maximum angular speed. Plot position and speed on aligned axes.
Mark that speed is zero at angular extremes and largest near centre. This corrects the intuition that the head moves fastest where displacement is greatest.
Investigation 5: Explore sampling
Generate a 12 Hz sine wave in a spreadsheet and sample it at 100, 30 and 20 samples/s. Plot only the sampled points.
Identify which trace preserves shape and which aliases. Transfer the explanation to video of rotating objects.
Investigation 6: Count noisy peaks
Add small random noise to a clean motion trace. Compare naive local-maximum counting with a rule requiring minimum prominence and separation.
Calculate false and missed detections against the known simulated cycles. Explain how a threshold changes both error types.
Investigation 7: Draw the timer state machine
Create OFF, RUNNING, PAUSED and COMPLETE states. Add start, pause, resume, reset and timeout transitions.
Simulate a session with three pauses. Calculate active and wall-clock time and list cue timestamps under each timer definition.
Investigation 8: Quantify clock drift
For errors of −0.2%, +0.1% and +0.5%, calculate displayed completion time over nominal 120 s.
Add a ±0.15 s observation error and decide which drift values can be distinguished in one trial. Do not confuse resolution with accuracy.
Investigation 9: Test a pressure threshold on paper
Use a supplied force trace and compare a 2.0 N single threshold with 2.0/1.7 N hysteresis. Count warning transitions.
Graph both outputs below the signal. Explain which changes come from data and which come from algorithm choice.
Investigation 10: Add validation delay
At 100 samples/s, require 10 consecutive samples above threshold. Test spikes lasting 5, 9, 10 and 20 samples.
Introduce one below-threshold sample within a long event and define whether the counter resets. Small logic choices can change results.
Investigation 11: Fit a sensor calibration
Fit voltage against known force using supplied points. Plot residuals and predict force for an intermediate voltage.
Calculate a prediction interval from repeated calibration data. Keep “estimated force” distinct from contact pressure.
Investigation 12: Calculate session energy
Integrate motor, indicator and standby power across a day. Convert joules to watt-hours and estimate sessions from usable energy.
Round completed sessions downward and include a sensitivity range for battery ageing. Do not treat nominal capacity as fully usable.
Investigation 13: Audit a specification table
For each row, classify the quantity as count, rate, angle, time, force proxy, energy or marketing term. Add missing units and event definitions.
Reject comparisons where two products count different events. A large number is not useful until the denominator and event are aligned.
Investigation 14: Read a patent mechanism
Use one cited patent to draw a neutral block diagram of drive, timing or sensing. Paraphrase and identify the publication date.
State what the disclosure proves and what it does not: it documents an invention, not adoption, clinical effectiveness or current product compliance.
Investigation 15: Write a bounded conclusion
Combine one motion calculation, one timer simulation and one threshold graph. Mark supplied, measured, assumed and calculated values.
End with a statement about the model and a separate statement about evidence needed for health or product-performance claims. Keep those domains distinct.
Turning the Laboratory into a Strong Report
Choose a precise question such as, “How does the counting definition change the advertised rate?” Place definitions before results. Use one diagram, one timeline and one graph so geometry, logic and data are not mixed.
Ask a partner to reproduce period, bristle arc and active time. If answers differ, inspect degrees versus radians, whether a cycle includes the return path, and how pauses were treated.
Run a sensitivity check on amplitude, radius, threshold and sampling rate. A robust conclusion survives reasonable changes; a fragile one should be reported conditionally.
Frequently Asked Questions
What is one oscillation?
Usually it is a complete repeating cycle, but a specification may count strokes or movements differently. Check the stated definition.
How are frequency and period related?
They are reciprocals: T=1/f.
Does a timer always measure wall-clock time?
Not necessarily. It may accumulate active motor time and pause under defined conditions.
Is force the same as pressure?
No. Pressure equals force divided by contact area. Many device sensors measure a proxy rather than distributed bristle pressure.
Can motion numbers prove cleaning effectiveness?
No. They describe mechanics. Health and effectiveness claims require appropriate clinical evidence and professional context.
Useful Next Reading
The automatic soap dispenser mathematics article develops sensing thresholds and state logic in another everyday device. Continue through the eduKate Mathematics Learning Hub for more connections among mathematics, technology and daily life.
The Bigger Answer to “Why Mathematics?”
An electric toothbrush turns repeated motion into a language of cycles, angles, paths and samples. Its timer turns elapsed seconds into states and transitions. Its warning system turns a changing sensor signal into a threshold decision.
Mathematics makes those mechanisms auditable while setting a boundary around what they can prove. A motion rate is not a health outcome, a warning is not a calibrated medical measurement, and a timer is only meaningful when its clock is defined. That disciplined clarity is one of the most useful benefits of mathematics.
Extended Case Studies for Deeper Transfer
Case 1: Reconcile three movement claims
A mechanism completes 110 full cycles each second. It can be described as 6,600 cycles/min, 13,200 extreme-to-extreme strokes/min or 13,200 direction changes/min if each cycle contains two of each. The equal last two counts arise from the chosen definitions, not a law that always makes strokes and reversals identical.
Write all three labels beside one waveform. A comparison table must align the counted event before ranking products. Converting a larger number to the same denominator can remove an apparent advantage entirely.
Case 2: Compare points across one brush head
A head oscillates with 12° amplitude. Points 3, 6 and 9 mm from the axis have extreme-to-extreme arc lengths 1.26, 2.51 and 3.77 mm. Their path rates at the same frequency differ in direct proportion to radius.
Therefore, no single “bristle speed” describes the whole head. A published value must identify the reference point and whether it is mean, maximum or accumulated path rate. Geometry creates variation even before bristle flex is considered.
Case 3: Bound frequency from a short video
A high-speed recording shows 59 to 61 cycles in a 0.500±0.002 s clip, depending on whether partial cycles at the ends are counted. Lower frequency is 59/0.502≈117.5 Hz; upper is 61/0.498≈122.5 Hz.
Longer clips reduce relative endpoint-count uncertainty, provided the frame rate and dropped frames are known. A precise timestamp display cannot fix ambiguous cycle boundaries. Report the interval rather than selecting the most attractive central value.
Case 4: Detect aliasing with two cameras
Camera A records at 240 frames/s and Camera B at 300 frames/s. A head near 120 Hz may look almost stationary in A because there are roughly two frames per cycle, while B samples at 2.5 frames per cycle and shows a different pattern.
If estimated motion frequency changes dramatically with frame rate, suspect aliasing. Use a known timing reference or sufficiently high-rate sensor. Visual slow motion is an observation tool, not automatic proof of true speed.
Case 5: Audit cue-time accuracy
Ten automated cue detections for a nominal 30 s interval have mean 30.06 s and standard deviation 0.08 s. The microphone and algorithm together add an estimated systematic delay of 0.04 s.
Subtracting that delay gives 30.02 s, but its own uncertainty should be included. A one-sample t interval can describe repeatability around the mean; it does not establish long-term clock stability across battery levels and temperature.
Case 6: Compare pause policies
Policy A pauses the timer whenever motor current drops below a threshold. Policy B continues unless the user presses pause. In a session with four five-second low-load intervals, Policy A completes after 140 s wall time while Policy B completes after 120 s.
Neither timer is arithmetically wrong. They implement different state definitions. User experience, instructions and validation goals determine which policy is appropriate; mathematics makes the distinction visible.
Case 7: Tune hysteresis and delay together
Three simulated signals contain baseline noise ±0.15 N, short 2.2 N spikes and one sustained 2.3 N event. Compare a 2.0 N threshold alone, 2.0/1.7 N hysteresis, and hysteresis plus 100 ms validation.
Count false warning starts, missed sustained events and warning latency. Stronger filtering can reduce false positives while increasing delay. Optimisation needs a cost for each error type, not a blanket preference for “more filtering.”
Case 8: Convert force to a pressure range
If estimated force is 1.8 N and active bristle contact area plausibly ranges from 40 to 90 mm², average pressure range is 1.8/90 to 1.8/40 N/mm², or 20 to 45 kPa.
This wide range shows why force warning and local pressure are not interchangeable. Contact is distributed unevenly and bristles flex. The calculation is a sensitivity illustration, not a safe-use recommendation.
Case 9: Account for daily battery energy
Two 2-minute sessions at 1.1 W use 0.0733 Wh per day. Standby at 0.004 W for the remaining 23.933 h uses about 0.0957 Wh, more than the brushing energy in this hypothetical device.
If charging conversion is 75% efficient, wall energy exceeds stored energy. The example demonstrates why tiny standby power matters over long durations and why a battery-runtime model should include every state.
Case 10: Create a traceability table
List each conclusion beside its evidence: frequency from cycle count and duration; amplitude from calibrated angle; cue time from audio timestamps; warning state from the published threshold algorithm; energy from integrated power.
Add columns for unit, uncertainty, assumption and excluded claim. A statement such as “the warning triggered after 120 ms in this trace” can be well supported while “the brush protects every user” remains outside the evidence. Traceability prevents a correct calculation from carrying an unjustified conclusion.
Quantitative Design Challenge: Reconstruct One Two-Minute Session
A teacher supplies synchronised motor-angle, current, cue-audio and force-proxy traces sampled at 1,000 Hz. The first task is not to calculate a headline. It is to verify timestamps, identify dropped samples and confirm that every channel uses the same clock. A one-second offset could make a pressure event appear to cause an unrelated cue.
Estimate motion frequency in ten-second windows. Use both zero crossings and spectral peak, then compare. A disagreement may indicate waveform asymmetry, noise or changing frequency. Report window-to-window range and do not average away a genuine mode change.
For amplitude, convert sensor voltage through a supplied calibration, find positive and negative peaks after filtering, and preserve filter settings. A filter that is too aggressive can shrink peaks; no filtering can count noise. Validate on a simulated signal with known amplitude before applying it to the session.
Reconstruct timer state from motor current and cue events. Define when active time starts, how pauses are detected and when a pause becomes long enough to count. Sum time intervals rather than counting samples if gaps exist. Compare reconstructed active time with wall-clock duration.
Apply three warning algorithms to the force proxy: single threshold, hysteresis, and hysteresis with delay. For each, calculate warning starts, total warning duration and latency for a labelled sustained event. There is no universally superior algorithm without a cost for false alarms and missed events.
Integrate electrical power to estimate session energy. If only current and battery voltage are available, multiply them cautiously and note that voltage may vary. Separate motor, indicators and pauses if the channels allow. Extrapolating one session to a month assumes usage and battery condition remain stable.
Prepare four panels: angle waveform, frequency by window, timer state and warning state. Link each conclusion to one panel and a formula. A legitimate conclusion might state, “Under the stated counting rule, frequency remained 118–122 Hz across active windows, and the delayed warning rejected two brief threshold crossings.”
End with exclusions. The trace cannot establish plaque removal, gum safety, user technique, waterproofing or charger compliance. Those require different evidence. The project succeeds when it explains mechanics and control precisely without borrowing authority from unrelated health outcomes.
Final Reasonableness Checks
First confirm that rate conversions preserve the counted event. Multiplying cycles per second by 60 yields cycles per minute, not automatically strokes or directional movements. If a full cycle includes two extreme-to-extreme paths, write that factor explicitly. Hidden factors of two are common in oscillation problems.
Second, keep degrees and radians separate. Arc length and calculus formulas require radians, while diagrams may be labelled in degrees. A head amplitude of 15° is 0.262 rad, not 15 rad. A unit-aware spreadsheet can prevent an enormous but visually plausible error.
Third, compare sampling frequency with signal frequency. A sensor that records only two or three points per cycle may estimate count under ideal conditions but cannot describe waveform and peaks reliably. Check for aliasing by changing sampling rate or using a known simulated signal.
Fourth, audit timer definitions. Sum actual intervals, state whether pauses count and distinguish cue-detection delay from clock error. One human stopwatch trial mixes device timing with reaction time. Repeats improve random uncertainty but do not automatically remove a systematic audio delay.
Fifth, treat the warning algorithm as a classifier. Report threshold, hysteresis, validation delay and the labelled events used to judge it. Reducing false warnings can increase missed events or latency. The chosen balance belongs to design evidence, not intuition alone.
The final report should therefore use restrained language: “the supplied trace supports this mechanical and logical description.” It should not turn an oscillation count, timer result or force proxy into a clinical recommendation. Keeping those claims separate is mathematical maturity.
Add a synchronisation check before finalising. Choose one event visible in two channels, such as motor start in current and motion, and calculate their timestamp difference. Repeat at the end of the session. A changing offset suggests clock drift; a constant offset suggests a fixed alignment correction. Either must be documented before causal timing claims are made.
Also test reproducibility across sessions rather than only within one. Compare the distribution of frequency, amplitude, cue time and warning latency for at least several supplied traces. Between-session variation can exceed within-session noise. Report both levels so a stable-looking two-second window is not mistaken for universal behaviour.
Preserve raw signals before filtering and keep every processing parameter. A moving average changes peak height and timing; a high-pass filter can remove slow drift but distort transitions. Process a simulated signal with known events first, then show how results change under a reasonable alternative filter. If the headline depends on one undocumented setting, it is not reproducible evidence.
Finish with a unit audit. Every angular quantity should say degrees or radians, every rate should name its counted event, every time should identify active or wall-clock basis, and every sensor value should distinguish raw voltage from calibrated estimate. This final pass is quick, yet it prevents the most convincing-looking errors in motion and control reports.