University Physics I · Mathematical & Physical Foundations · 1.8
Estimation and Model Checks
Four cheap tests stand between a calculation and a result worth using: size, units, limiting cases, and physical plausibility. Each takes seconds, and each catches a mistake the others miss.
Build the model
Connect the measurement to the mechanism.
A physics result is a claim, and a claim needs testing before it is used. Four tests do most of that work, and none of them is a repeat of the calculation. Size: an order-of-magnitude estimate, built from factors you can each bound, says which power of ten the answer should live in.
Units: both sides must match, and so must every term added to another. Limits: push a parameter to zero, to infinity, or into a symmetry, and the formula must collapse onto a case you already trust. Plausibility: compare the number with something measured or memorised — g = 9.81 m s⁻², air at 1.2 kg m⁻³, a sustained human output near 10² W.
A result that fails any one of these is wrong. A result that passes all four is not proved, but it has earned enough confidence to build on, and you knew that in under a minute.
- Simple definition
- An estimate is a deliberately rough calculation aimed at the right power of ten. A model check is a fast test — units, a limiting case, or a known benchmark — that a candidate result must pass before it is believed.
- Example
- Drag on a cyclist at 8 m s⁻¹ gives P = ½ρCdAv³ ≈ ½(1.2)(0.3)(8³) ≈ 92 W, so the answer is 10² W — the order of a person's sustained mechanical output.
Report the exponent, not the digits: the claim is about scale, not precision.
decades meet at √10 ≈ 3.2, so a factor of 3 still counts as right
Split the unknowable into factors you can each bound; independent errors partly cancel.
logarithms add · bracket a stubborn factor and take √(qlow qhigh)
Check the symbols, not the number: a lost root or a stray mass shows up as a mismatch.
every added term must match · arguments of sin, ln and exp are dimensionless
One formula, two decades apart: the cube of the speed does almost all of the work.
cyclist 8 m s⁻¹ → 92 W · car, CdA 0.8 m², 30 m s⁻¹ → 1.3 × 10⁴ W
Push one parameter to an extreme; the formula must collapse onto a case you trust.
small-parameter checks: sin θ ≈ θ (radians) · (1 + x)ⁿ ≈ 1 + nx for |x| ≪ 1
Compare the answer with a number you already trust before you write it down.
walking 1.4 m s⁻¹ · atmosphere 1.0 × 10⁵ Pa · sustained human output ~10² W
What an order-of-magnitude claim actually says
An order-of-magnitude estimate names the power of ten an answer belongs to, and nothing finer. Writing Q ≈ 10ⁿ claims the true value lies within a factor of √10 ≈ 3.2 of 10ⁿ — between 10(n − ½) and 10(n + ½), because that is where consecutive decades meet. That looseness is the point. An estimate is cheap enough to make before the full calculation, so it works as an independent prediction rather than a summary of one. When the finished result lands a decade or more from the estimate, one of the two is wrong, and you now know to look.
Break the unknown into factors you can bound
Nobody can quote the power a cyclist spends pushing air aside at 8 m s⁻¹ (28.8 km h⁻¹), but every factor in P = ½ρCdAv³ can be bounded. Air density ρ is 1.2 kg m⁻³ near sea level. The drag area CdA — drag coefficient times frontal area — is near 0.3 m² for a rider on the hoods. Then P ≈ ½(1.2)(0.3)(512) ≈ 92 W, so the answer is 10² W, and the tens of watts of rolling resistance do not move it. No single factor had to be known well: multiply them and their logarithms add, so an over-estimate in one is partly cancelled by an under-estimate in another. When a factor resists bounding, bracket it and take the geometric mean √(qlow qhigh) — the multiplicative midpoint, not the arithmetic one.
Units are the cheapest lie detector
Both sides of an equation must carry the same dimensions, and so must every term added to another: [v] = L T⁻¹, [a] = L T⁻², [E] = M L² T⁻². Arguments of sin, cos, ln and exp are pure numbers, so a decay factor e(−bt/m) forces [b] = M T⁻¹, that is kg s⁻¹. Run the check on the symbols rather than the final number, because that is where a dropped square root, a mass written where a mass squared belongs, or an integration that lost its dt announces itself. The verdict is one-sided, though: a unit check can convict, never acquit.
Run the four together on one result
The tests are worth more together, because each is blind where another sees. Take the drag formula to motorway speed: a car with CdA ≈ 0.8 m² at 30 m s⁻¹ needs P ≈ ½(1.2)(0.8)(2.7 × 10⁴) ≈ 1.3 × 10⁴ W. Units: (M L⁻³)(L²)(L³ T⁻³) = M L² T⁻³, the dimension of power. Size: 10⁴ W. Limits: v → 0 kills the power, and doubling v multiplies it by eight — the cube law you can feel on a bicycle. Plausibility: a car rated near 10² kW is spending about a tenth of itself to cruise, which is why lifting off the throttle slows you so gently. Four tests, one result, under a minute.
Limiting cases audit the structure
Take a finished formula and push one parameter to a value where the answer is already known. For an ideal Atwood machine, a = (m₁ − m₂)g/(m₁ + m₂). Equal masses give a = 0, as a balanced system must. Letting m₂ → 0 gives a → g, which is free fall. Exchanging m₁ and m₂ reverses the sign, as swapping which side falls should. Series expansions do the same work from the other end: sin θ ≈ θ for θ in radians, and (1 + x)ⁿ ≈ 1 + nx for |x| ≪ 1, reduce an intimidating result to a simpler model you can already check. A formula that fails a limit is wrong, however cleanly its units match.
Plausibility is a physical claim, not a feeling
The last test asks whether the number could exist. Anchor it against values you trust: g = 9.81 m s⁻², walking is 1.4 m s⁻¹, atmospheric pressure is 1.0 × 10⁵ Pa, water is 1.00 × 10³ kg m⁻³ and air 1.2 kg m⁻³, and a person sustains roughly 10² W of mechanical output. A 1500 kg car at 30 m s⁻¹ carries ½(1500)(900) = 6.8 × 10⁵ J, as much as lifting it 46 m — surprising, but defensible. Numbers that fail here violate something specific: an efficiency above 1, a speed above 3.00 × 10⁸ m s⁻¹, a two-body Atwood beating free fall. Then report only the precision you earned. A Fermi estimate written as 92.16 W claims a measurement nobody made.
Change one variable at a time
Make the relationship visible.
Push the speed slider up and watch the shaded decade jump only once P has moved by a factor of 3.2 — doubling v multiplies P by eight, almost a whole band.
DRAG POWER P92 W
log₁₀ P1.96
ORDER n2
P ÷ 10ⁿ0.92
Live interpretationDRAG POWER P: 92 W. log₁₀ P: 1.96. ORDER n: 2. P ÷ 10ⁿ: 0.92
Catch the common trap
Explain before calculating.
A projectile is launched at speed v₀ and angle θ above level ground. Which range formula survives a unit check, the θ → 0 limit, and the θ → 90° limit?
Choose an answer to test the model.
Practice & worked examples
Reason from the model, then test the result.
EasyA finished calculation returns 4.7 × 10³ W. State its order of magnitude, and say what range that claim actually covers.
- Take the base-10 logarithm: log₁₀(4.7 × 10³) = 3 + log₁₀ 4.7 = 3 + 0.672 = 3.672.
- Round to the nearest integer: n = 4, so the claim is P ≈ 10⁴ W. The decade boundary sits at √10 × 10³ = 3.2 × 10³ W, and 4.7 × 10³ W is above it.
- The claim covers 103.5 to 104.5, that is 3.2 × 10³ W up to 3.2 × 10⁴ W — anything from 3.2 kW to 32 kW would have satisfied it.
AnswerOrder 10⁴ W, a claim covering roughly 3.2 × 10³ W to 3.2 × 10⁴ W.
MediumA delivery van has drag area CdA = 1.2 m² and cruises at 25 m s⁻¹ through air of density 1.2 kg m⁻³. Estimate the power it spends pushing air aside, then run the unit, size and plausibility checks on the answer.
- Fermi decomposition: P = ½ρCdAv³, and every factor is bounded — ρ = 1.2 kg m⁻³, CdA = 1.2 m², v = 25 m s⁻¹.
- Prefactor: ½ × 1.2 × 1.2 = 0.72 kg m⁻¹. Speed cubed: 25³ = 1.5625 × 10⁴ m³ s⁻³.
- P = 0.72 × 1.5625 × 10⁴ = 1.125 × 10⁴ W, so P ≈ 1.1 × 10⁴ W — two figures is all the inputs earn.
- Units: (M L⁻³)(L²)(L³ T⁻³) = M L² T⁻³, the dimensions of power. Size: log₁₀(1.125 × 10⁴) = 4.05, so the order is 10⁴ W. Plausibility: 11 kW is about a tenth of a van's rated output, which is what steady cruising should cost.
AnswerP ≈ 1.1 × 10⁴ W, order 10⁴ W; units, size and plausibility all pass.
HardAn object falling with linear drag obeys v(t) = (mg/b)(1 − e(−bt/m)). Audit the formula with a unit check and two limiting cases, then put numbers on it for m = 0.145 kg and b = 0.020 kg s⁻¹.
- Units: the exponent bt/m must be a pure number, so [b] = M T⁻¹, that is kg s⁻¹. Then [mg/b] = (M · L T⁻²)/(M T⁻¹) = L T⁻¹, a speed — the prefactor is dimensionally allowed to be v.
- Limit t → ∞: e(−bt/m) → 0, so v → mg/b, a constant. That is terminal velocity, where the drag bv has grown to balance the weight mg.
- Limit bt/m ≪ 1: expand e(−x) ≈ 1 − x, so v ≈ (mg/b)(bt/m) = gt. Early on, before drag matters, the formula collapses onto free fall.
- Numbers: mg = 0.145 × 9.81 = 1.4225 N, so mg/b = 1.4225/0.020 = 71 m s⁻¹, and the time constant is m/b = 0.145/0.020 = 7.25 s.
- Plausibility: 71 m s⁻¹ is tens of metres per second, the decade a small dense object falls in, and the 7.25 s time constant says it needs several seconds to get near it — both numbers a real drop could settle.
Answer[b] = kg s⁻¹; v → mg/b = 71 m s⁻¹ as t → ∞; v → gt as t → 0. Terminal speed 71 m s⁻¹, drag time constant 7.25 s.