1.1 Introduction
Measuring a physical quantity always means comparing it against some fixed, agreed-upon reference amount, a unit. Say a rod is “2 metres” long, and you’re really saying its length is twice the length everyone has agreed to call one metre. Every measurement is this pairing: a number, and the unit that number is counted in. Change the unit and the number changes with it, but the actual length of the rod obviously doesn’t.
Physics needs surprisingly few independent units. Quantities like length, mass, and time can’t be built from anything simpler, so they get their own base units. Everything else, speed, force, energy, pressure, is really just a combination of these base quantities, and its unit is a derived unit built the same way. A complete collection of base and derived units is called a system of units.
1.2 The International System of Units (SI)
Before everyone agreed on one system, different regions used different base units for the same quantities, centimetre-gram-second (CGS), foot-pound-second (FPS), and metre-kilogram-second (MKS) all coexisted. The system in use worldwide today is the SI (Système International d’Unités), built on seven base units. As of a 2019 revision, every one of them is defined by fixing the exact numerical value of a fundamental constant of nature, rather than by a physical object that could wear down or change over time.
| Base quantity | SI unit | Fixed by |
|---|---|---|
| Length | metre (m) | Fixed value of the speed of light, c |
| Mass | kilogram (kg) | Fixed value of the Planck constant, h |
| Time | second (s) | Fixed value of the caesium-133 hyperfine transition frequency |
| Electric current | ampere (A) | Fixed value of the elementary charge, e |
| Thermodynamic temperature | kelvin (K) | Fixed value of the Boltzmann constant, k |
| Amount of substance | mole (mol) | Fixed value of the Avogadro constant, Nᴀ |
| Luminous intensity | candela (cd) | Fixed value of luminous efficacy, K꜀d |
Two extra units are defined for angles, and both happen to be dimensionless: the radian (plane angle, arc length divided by radius) and the steradian (solid angle, area divided by radius squared). Being ratios of two lengths (or two areas), the units cancel out entirely, which is exactly why angles slot so cleanly into equations without dragging along a unit of their own.
1.3 Measuring Small Lengths: the Vernier Caliper and Screw Gauge
A metre scale is marked in millimetres, and that is the finest thing it can tell you: a length somewhere between 23 mm and 24 mm can only be guessed at by eye. The smallest length an instrument can actually resolve is called its least count (LC), and for a metre scale it’s 1 mm. Measuring the diameter of a marble or the thickness of a wire to a fraction of a millimetre needs a cleverer instrument, and the two found in every school lab are the vernier caliper (least count typically 0.1 mm or 0.02 mm) and the screw gauge (typically 0.01 mm). Neither is magic. Each uses one simple geometric trick to read a fraction of a millimetre that your eye could never judge directly.
1.3.1 The vernier scale and its least count
A vernier caliper has two scales side by side: the fixed main scale, marked in ordinary millimetres, and a short sliding vernier scale that moves along it with the jaw. The whole trick is in how the vernier is divided. Its divisions are made to span exactly main-scale divisions, so each vernier division (VSD) is very slightly shorter than a main-scale division (MSD):
Every vernier division falls short of a main-scale division by the same small amount, , and that shortfall is the least count:
With 10 vernier divisions spanning 9 mm, LC = 0.1 mm; with 50 vernier divisions spanning 49 mm, LC = 0.02 mm. A longer, more finely divided vernier gives a smaller least count, at the price of finer lines that are harder to line up by eye.
1.3.2 Taking a reading, and correcting for zero error
Close the jaws on the object and the zero mark of the vernier lands somewhere along the main scale, usually between two millimetre marks. The last main-scale mark it has passed gives the main scale reading (MSR). What remains is some fraction of a division beyond that mark, and this is the part the vernier is built to measure. Look along the vernier for the one line, say line , that lines up exactly with a main-scale line. Why does that single coincidence pin down ? Vernier line sits vernier divisions beyond the vernier zero, so its position is:
Compare it with the main-scale line divisions past the MSR mark, which sits at . The vernier line is ahead of it by , so the two coincide exactly when . Each successive vernier line “catches up” one least count on the main scale, and the line where the catching-up is complete tells you how many least counts the vernier zero was ahead of the MSR mark. So:
One check has to come before any measurement: close the jaws on nothing at all and see whether the vernier zero sits exactly on the main-scale zero. On a worn or knocked caliper it often doesn’t, and the reading with the jaws shut, called the zero error , is then built into every measurement the instrument makes. If the vernier zero sits slightly to the right of the main-scale zero, the caliper already shows a small positive reading with nothing between its jaws, and the ordinary reading rule gives a positive zero error, , where is the coinciding vernier line with the jaws closed. If it sits slightly to the left, the vernier zero lies somewhere between one main-scale division behind zero and zero itself, so the same reading rule, applied with as the main scale reading, gives a negative zero error:
using from §1.3.1. Either way, the instrument adds (with its sign) to every length it reports, so the correction is always a subtraction:
For a negative zero error, subtracting a negative number puts the missing length back. Zero error is the textbook example of a systematic error (§1.5.1): it is the same, in size and sign, on every reading, so no amount of repeating the measurement will average it away. It has to be found and subtracted.
Worked example
Reading a 0.02 mm vernier caliper with a zero error
A caliper’s main scale is marked in millimetres, and its vernier has 50 divisions spanning 49 mm. With the jaws closed, the vernier zero sits just right of the main-scale zero and the 7th vernier line coincides with a main-scale line. Measuring a glass marble, the vernier zero lies just past the 23 mm mark and the 31st vernier line coincides. Find the least count and the marble’s true diameter.
Zero error (positive, since the vernier zero is to the right):
Observed reading, then the correction:
Skipping the zero-error check would have overstated the diameter by 0.14 mm, seven whole least counts: an instrument’s fine resolution is wasted if its zero is wrong.
1.3.3 The screw gauge
For still smaller lengths, the thickness of a wire or a thin metal sheet, a screw gauge uses a different trick: a screw. Its movable face, the spindle, is the end of a finely cut screw. Every full turn advances the spindle by a fixed distance, the pitch , which is the spacing between neighbouring threads and is marked off as the divisions of a linear scale along the sleeve (commonly 0.5 mm or 1 mm). The rotating thimble carries a circular scale with equal divisions around its rim. Turning the thimble through one circular division is of a full turn, so it moves the spindle through of a pitch:
The screw is acting as a magnifier: a movement of the spindle far too small to see is turned into a clearly visible rotation of the thimble rim. A reading is taken just like a vernier reading. The last linear-scale mark left uncovered by the thimble gives the linear scale reading (LSR), and the circular-scale division sitting on the sleeve’s reference line supplies the fraction of a pitch beyond it, since the thimble has turned divisions past that mark and each division is one least count of advance:
Zero error works the same way too. Bring the spindle gently up against the anvil with nothing between them, and the circular-scale zero should sit exactly on the reference line. If the faces touch while the reference line still points at some division just past zero, the gauge reads for zero thickness, a positive zero error. If instead the thimble edge has crept slightly past the linear-scale zero mark by the time the faces touch, the gauge is sitting below zero. Counting back one full pitch from the zero mark, with the reference line at division , its reading is:
using , a negative zero error, and the correction is the same signed subtraction as for the vernier, .
Worked example
Measuring a wire with a screw gauge that has a negative zero error
A screw gauge has a pitch of 0.5 mm and 50 divisions on its circular scale. With the faces in contact, the thimble edge has just crept past the linear-scale zero and the 46th circular division sits on the reference line. Measuring a wire, the linear scale reads 1.5 mm and the 27th circular division is on the reference line. Find the least count and the wire’s true diameter.
The negative zero error means the gauge under-reads, so the correction adds 0.04 mm back. The single rule handled the sign automatically.
1.4 Measuring Large Distances: the Parallax Method
At the other end of the scale, no tape is long enough to reach a hilltop across a valley, let alone the Moon or a star. Large distances are measured indirectly, using an effect you can see for yourself. Hold a finger up at arm’s length and look at it with one eye closed, then the other: the finger seems to jump sideways against the distant background. That apparent shift, caused purely by changing the point you look from, is called parallax. Your two eyes are the two observation points, and the distance between them is the baseline. Bring the finger closer and the jump gets bigger; move it farther away and the jump shrinks. That link between distance and shift is exactly what the parallax method turns into a measurement.
To find the distance to a far object S, observe it from two points A and B, a known baseline apart. The lines of sight AS and BS are not quite parallel, and the small angle between them at S, , is the parallax angle. (In practice it is measured as the change in the object’s direction between A and B, judged against a much more distant background whose own direction barely changes at all.) Because S is so far away compared with the baseline, A and B lie almost exactly on a circle of radius centred on S, and the straight baseline is almost exactly an arc of that circle. By the very definition of the radian in §1.2, arc length = radius × angle:
How good is the “baseline ≈ arc” step? Take the case where S lies on the perpendicular bisector of AB, so triangle ASB is isosceles. Splitting it down the middle gives two right triangles, each with opposite side , adjacent side , and angle at S, so exactly:
For a small angle measured in radians, , so and the exact formula collapses back to . The worked example below puts a number on just how small the difference is for the tiny angles parallax involves. Astronomers use exactly this method with far longer baselines: two observatories on opposite sides of the Earth for the Moon and the planets, and opposite sides of the Earth’s orbit (observations six months apart) for the nearest stars.
Worked example
How far away is the tower on the far hill?
A surveyor wants the distance to a transmission tower on a hill across a valley. She marks two points 120 m apart, roughly perpendicular to her line of sight, and from each one measures the tower’s direction against a far-off mountain peak. The tower’s direction shifts by 1.50° between the two points. How far away is the tower?
First convert the parallax angle to radians, since came from arc = radius × angle:
Checking against the exact isosceles-triangle formula:
The two answers differ by about a quarter of a metre in four and a half kilometres (under 0.01%), far smaller than any realistic error in measuring the angle itself.
1.5 Errors in Measurement
Every measuring instrument, in anyone’s hands, gives an answer that is slightly off. The difference between the measured value and the true value of a quantity is the error in the measurement, and it isn’t a mistake to be embarrassed about: it is an unavoidable part of measuring anything. What physics demands is that you know roughly how big it is, and report it honestly alongside the answer. Two words get used constantly here and are worth keeping apart. Accuracy is how close a measurement is to the true value; precision is how finely it resolves the quantity, and how closely repeated measurements agree with one another. The caliper in §1.3.2, used without its zero correction, would give readings that agree beautifully to 0.02 mm (precise) and are all wrong by the same 0.14 mm (inaccurate).
1.5.1 Systematic errors and random errors
Errors come in two kinds, and the difference between them decides how you fight them. Systematic errors push every reading the same way, always too high or always too low, by an amount that doesn’t change from one reading to the next. They come from three sources.
Instrumental errors come from the instrument itself: imperfect design, manufacture, or calibration. The zero error of §1.3.2 is one. A steel tape that has stretched over years of use is another, since its markings are now spaced too far apart and every length it reports comes out too short.
Imperfection in experimental technique or procedure means the method itself is biased, even with a perfect instrument. Dip a thermometer into a cup of hot tea and pull it out before it has had time to warm all the way up, and it reads low, every time, however good the thermometer is.
Personal errors come from the observer: a habit, a bias, or plain carelessness. Someone who always reads a scale with their eye slightly to one side of the pointer picks up the parallax shift of §1.4, this time working against them, on every single reading. Someone who always stops a stopwatch a fraction of a second late makes every time too long.
Because a systematic error repeats itself identically, taking more readings does nothing to it. It has to be hunted down and removed: better instruments, improved technique, careful habits, and, where the error is known (like a zero error), a direct correction.
Random errors are the opposite: irregular, unpredictable, and just as likely to make a reading too high as too low. They come from small fluctuations nobody controls, a draught across a balance pan, a flicker in the supply voltage, a vibration of the bench, the observer’s judgement of a reading wobbling a little each time. You can’t predict the next one, but you can beat them in bulk. Since they scatter on both sides of the true value, they tend to cancel when many readings are averaged, and the more readings you take, the more complete the cancelling. That is the whole reason experiments are repeated.
Finally, every instrument has a least count error, the uncertainty that comes from its resolution, since nothing finer than one least count can ever be read off it. It can be reduced by using an instrument with a smaller least count, which is precisely what the vernier caliper and screw gauge of §1.3 are for, and by repeating the measurement and averaging.
1.5.2 Absolute, relative, and percentage error
Suppose a quantity is measured times, giving the readings . Since random errors are as likely to push a reading up as down, they tend to cancel in a sum, and the best available estimate of the true value is the arithmetic mean:
How far each individual reading strays from this best estimate is its absolute error:
Some of these are positive and some negative. To summarise them in a single number, the natural instinct is to average them, but the plain average turns out to be useless, and it’s worth seeing exactly why:
The signed errors always cancel to exactly zero, for any set of readings whatsoever, because that is precisely what defines the mean. So the signs are thrown away and the magnitudes are averaged instead, giving the mean absolute error:
The result is reported as , a statement that the true value very likely lies somewhere between and .
An absolute error on its own doesn’t say whether a measurement is any good. An error of 1 mm is disastrous for the diameter of a 5 mm bead and utterly negligible for the length of a 100 m running track. What matters is the error compared with the size of the quantity, the relative error (or fractional error), and the same thing expressed per hundred, the percentage error:
For the bead, 1 mm is a 20% error; for the track, 0.001%. The same absolute error, and a wildly different quality of measurement.
Worked example
Five timings of the same run
A toy car is released from the top of a ramp and its run is timed five times with a stopwatch: 1.85 s, 1.80 s, 1.84 s, 1.81 s, and 1.85 s. Find the mean time, the mean absolute error, and the relative and percentage errors.
The absolute errors, , are:
which add to zero, as they must. Averaging their magnitudes:
So the run time is , known to about .
1.5.3 Combining errors: sums and differences
Most results in physics aren’t measured directly. They are calculated from two or more measured quantities, and each of those brings its own error into the calculation, so the question is how the errors carry through. The approach throughout is to take the worst case: we don’t know which way each individual error actually went, so we ask how far off the result could possibly be, which guarantees the quoted uncertainty is never an understatement.
Start with a sum, , where the measured values are and . The calculated result then carries some error :
The two signs are independent, since each error could have gone either way. In the lucky case they point in opposite directions and partly cancel; in the worst case both point the same way and pile up, leaving the result off by the full . Now try a difference, :
The subtraction flips the sign in front of , but since that error could have gone either way anyway, this changes nothing. The worst case is now measured too high while is measured too low (or the other way round), and the result is once again off by . Subtracting the values does not subtract their errors. Both cases give the same rule:
When quantities are added or subtracted, their absolute errors add. This has a sharp consequence for the difference of two nearly equal quantities: the difference itself is small, but its absolute error is the full sum of both, so its relative error can be enormous.
Worked example
The extension of a spring
A spring hangs with length cm. With a mass attached, it stretches to cm. Find the extension and its uncertainty.
Each length on its own was known to about 0.4%, but the extension, a small difference between two nearly equal lengths, is uncertain by about 11%. The cure is to make the quantity you actually care about larger: a heavier load gives a bigger extension, and the same 0.2 cm uncertainty becomes a much smaller fraction of it.
1.5.4 Combining errors: products and quotients
For a product, , multiply out the measured values:
Divide the left side by and the right side by the same thing written as :
The last term is a product of two small fractions. If each measurement is good to 1%, it comes to , a hundredth of a percent, negligible next to the other two terms, so it is dropped. What remains says the relative error of is made of the two individual relative errors, each of which could have either sign.
For a quotient, , one extra step is needed, because the error now sits in a denominator:
For any small number , , since is negligible; in other words, . Dividing by is therefore the same, to this accuracy, as multiplying by :
(again dropping the product of two small fractions), and dividing through by :
Product or quotient, the relative error of is the sum of two relative errors whose signs are unknown. Whatever those signs actually were, its size can never exceed the case where both push the same way:
and that worst case is the one quoted, as the maximum relative error:
When quantities are multiplied or divided, their relative (or percentage) errors add, and once again, dividing does not mean the errors subtract. Notice the contrast with §1.5.3: for sums and differences it is absolute errors that add, for products and quotients it is relative errors. This is the rule behind the rectangle example in §1.6.3.
Worked example
Timing a cyclist
A cyclist covers a measured stretch of road, m long, in a hand-timed s. Find her average speed and its uncertainty.
Look at where the error comes from: the distance contributes 0.33%, the hand timing 1.6%, nearly five times as much. Measuring the road more carefully would barely help; a better timing method (light gates instead of a thumb on a stopwatch) is the improvement that actually counts.
1.5.5 Combining errors: powers
Many formulas raise measured quantities to powers: the volume of a sphere goes as the cube of its diameter, the period of a pendulum as the square root of its length. The product rule handles these directly. A square is just a quantity multiplied by itself, so setting in §1.5.4:
Multiplying in one more factor of adds one more , and so on, so for any whole number carries times the relative error of . Fractional powers follow by running the same rule backwards. If , then , so , which means : taking a square root halves the relative error. (In general, if then , so , and .) Either way:
Now a general formula, , is just a product and quotient of the three factors , , and . Applying the product-and-quotient rule twice (first to , then to that result divided by ), the relative errors of all three factors add, and each one is , , or times the relative error of the measured quantity inside it:
Two lessons hide in this formula. The term for is added, even though sits in the denominator, exactly as division added errors in §1.5.4. And the quantity raised to the highest power dominates the final error, so it is the one that deserves your most careful measurement. Pure numbers in a formula, like the and the 6 in the example below, are exact and contribute no error at all.
Worked example
The density of a steel ball bearing
A steel ball bearing has mass g, and its diameter, measured with a 0.02 mm vernier caliper like the one in §1.3.2, is mm. Find its density and the uncertainty in it.
The density of a sphere is its mass over its volume, , so with cm:
This is the general rule with to the power 1 and to the power 3 (in the denominator):
The diameter was measured to a tiny 0.16%, yet after cubing it contributes 0.47%, four times as much as the mass does. A given percentage improvement in measuring is worth three times the same improvement in .
1.6 Significant Figures
Every measurement carries some uncertainty (§1.5), and significant figures are simply the digits worth trusting: every digit you’re confident about, plus one final digit that’s a reasonable estimate. A stopwatch reading of 1.62 s claims three significant figures, the 1 and 6 are solid, the 2 is the instrument’s best guess at the next decimal.
The counting rules, stated plainly:
- Every non-zero digit counts.
- A zero sitting between two non-zero digits counts.
- Leading zeros (before the first non-zero digit) never count, they just mark the size of the number.
- Trailing zeros count only if there’s a decimal point in the number.
That last rule is exactly why scientific notation is the safest way to report a measurement: write 4700 m as , and there’s no ambiguity left about whether those zeros were measured or just placeholders.
1.6.1 Arithmetic with significant figures
The result of a calculation can never be more precise than the least precise measurement that went into it:
- Multiplying or dividing: keep as many significant figures as the least-precise input has.
- Adding or subtracting: keep as many decimal places as the least-precise input has.
Worked example
Reporting a density measurement correctly
A block has mass 12.6 g (3 sig figs) and volume 4.1 cm³ (2 sig figs). Raw division gives . Since the volume only justifies 2 significant figures, the answer must be rounded to 3.1 g/cm³, not reported with five decimal places that the measurement never earned.
1.6.2 Rounding off
When the digit being dropped is anything other than exactly 5, round the usual way: 5 or above rounds up, below 5 rounds down. The one special case is when the dropped digit is exactly 5 with nothing but zeros after it. Convention then rounds to whichever choice leaves the preceding digit even (this is called round-half-to-even, and it avoids a systematic upward bias that would build up over many roundings). So 2.735 rounds up to 2.74 (making the last digit even), while 2.745 rounds down to 2.74 as well (4 was already even, so it stays).
1.6.3 Uncertainty in combined measurements
The significant-figure rules are really a quick shorthand for the error-combination rules derived in §1.5.3 to §1.5.5, and the two should agree. If a rectangle’s sides are cm and cm, that’s roughly and uncertainty respectively. By the product rule of §1.5.4 the percentage errors add, so the area, cm², is uncertain by about , or 2.6 cm², and is honestly reported as cm²: three significant figures, exactly what the multiplication rule of §1.6.1 would have given.
1.7 Dimensions of Physical Quantities
A quantity’s dimension describes what kind of thing it fundamentally is, independent of which units you happen to measure it in. Mechanics only ever needs three: mass , length , and time . Volume, being length × length × length, has dimension , regardless of whether you measure it in cubic metres or cubic centimetres.
Force is mass times acceleration, and acceleration is length divided by time squared, so:
Only the kind of quantity matters here, not its size: velocity, average velocity, and speed are all dimensionally identical, , even though their values differ from situation to situation.
1.8 Dimensional Formulae and Dimensional Equations
Writing a quantity’s dimensions out explicitly, like for speed, is its dimensional formula. Setting the quantity equal to that formula, , is a dimensional equation. A handful worth having on hand:
1.9 Dimensional Analysis and Its Applications
Only quantities with matching dimensions can be added, subtracted, or set equal to each other, and that single fact turns out to be a genuinely useful check on physics equations.
1.9.1 Checking dimensional consistency
Every valid physical equation must have identical dimensions on every term, on both sides. This is called the principle of homogeneity, and it catches a huge fraction of algebra mistakes before you even reach for a calculator.
Worked example
Checking s = ut + at²
Is dimensionally sound? Check each term against :
Every term matches, so the equation passes. That doesn’t prove it’s correct, the real kinematic equation has a in front of the last term, and no dimensional check can ever catch a missing pure number. It only proves the equation isn’t obviously wrong.
1.9.2 Deducing relations among physical quantities
If you can guess which quantities something depends on, you can often recover the shape of the formula, though never the constant out front, purely from matching dimensions.
Worked example
Deriving centripetal acceleration by dimensions
Suppose centripetal acceleration for a particle moving in a circle depends only on its speed and the radius : . Matching dimensions:
Comparing powers of : . Comparing powers of : . So , and the full derivation from circular motion (not available from dimensions alone) shows .
The method has real limits: it can never recover a dimensionless constant, it can’t handle a sum of several same-dimensioned terms, and it says nothing about relations built from sines, logarithms, or exponentials. The concept page walks through exactly where this breaks down, with the pendulum formula as the running example.
— From the NCERT exercises
A few of the chapter’s own exercise questions, with original worked solutions.
NCERT Exercise 1.6
Which of the following is the most precise device for measuring length: (a) a vernier callipers with 20 divisions on the sliding scale, (b) a screw gauge of pitch 1 mm and 100 divisions on the circular scale, (c) an optical instrument that can measure length to within a wavelength of light?
Solution
Compare least counts directly, using for the vernier (§1.3.1) and for the screw gauge (§1.3.3):
The optical instrument has by far the smallest least count, so it’s the most precise of the three.
NCERT Exercise 1.10
State the number of significant figures in: (a) 0.007 m², (b) 2.64×10²⁴ kg, (c) 0.2370 g cm⁻³, (d) 6.320 J, (e) 6.032 N m⁻², (f) 0.0006032 m².
Solution
Applying the counting rules from §1.6: (a) 1 (only the 7 counts), (b) 3, (c) 4 (the trailing zero follows a decimal point), (d) 4, (e) 4 (the zero sits between non-zero digits), (f) 4 (leading zeros don’t count, but the zero between 6 and 3 does).
NCERT Exercise 1.17
The Sun's inner core exceeds 10⁷ K and its outer surface is about 6000 K, hot enough that nothing stays solid or liquid. Given the Sun's mass (2.0×10³⁰ kg) and radius (7.0×10⁸ m), is its density closer to a solid/liquid or a gas?
Solution
Treating the Sun as a sphere:
That’s comparable to water (10³ kg/m³) and ordinary rock, not to a gas (roughly 1 kg/m³) despite being a super-heated plasma. Under gravity that strong, matter gets crushed to liquid-like densities even without ever solidifying.
