Errors in Measurement Errors in Measurement Physics starts with measuring the world: a length with a scale, a time with a stopwatch, a mass with a balance. No instrument is perfect and no eye is perfect, so every reading you take includes some uncertainty. That uncertainty is not a mistake; it is a natural part of measurement. The goal is not to hide it but to quantify it. When you report a length as 12.3 cm, you are claiming something about how finely you could read the instrument (least count), how consistently you could repeat the reading (precision), and how close that value might be to the actual value (accuracy). This chapter gives you a clear toolkit: how to estimate a best value from repeated readings using the arithmetic mean, how to classify error types (gross, systematic, random), how to express uncertainty as absolute, relative, and percentage errors, and how to carry that uncertainty through calculations using simple propagation rules. You will also learn how to handle common lab tools like the vernier caliper and screw gauge, including least count and zero error corrections. Finally, you will practise the art of reporting: the correct use of significant figures and rounding so that your numerical answers honestly reflect the quality of your data. The mindset is scientific humility: we do not pretend to know more than our measurements can justify, and we do not throw away the information contained in the spread of repeated readings. remember Think of measuring your height against a wall with a pencil mark. If three friends label slightly different points, the marks cluster near a common level. The average of those marks is your best estimate, and the spread between marks is a measure of error. Better tools shrink the spread; they do not magically make it zero. The process of comparing a physical quantity with a chosen standard to assign it a numerical value and unit. Measurement The uncertainty in a measured value. It is the difference between the measured value and the (unknown) true value, or an estimate of that difference. Error How close your measurement is to the true value. High accuracy means small systematic offset. Accuracy Precision How closely repeated measurements agree with each other. High precision means small random spread. The smallest change in the measured quantity that an instrument can resolve. It limits the reading precision. Least Count A nonzero reading when the true value should be zero. It leads to a constant offset that must be corrected. Zero Error Not all errors are alike. Gross errors are blunders such as reading the wrong scale or writing 3.2 as 2.3. These are avoided by care and repetition. Systematic errors are consistent shifts that push all readings in one direction. They arise from calibration faults (a scale that starts at 0.2 cm), environmental conditions (thermal expansion, wind), or personal biases (parallax from eye position). Systematic errors reduce accuracy, and repeating the measurement does not remove them unless you correct the cause. Random errors are small unpredictable fluctuations from reading to reading, caused by human reaction time, tiny vibrations, or electronic noise. They produce a spread around the best value and can be tamed by averaging many readings. Your analysis must identify which type dominates, so you know whether to average more, correct a bias, or remeasure with a better instrument. Gross error: Stopwatch started late by 2 s due to distraction. Systematic error (instrumental): Screw gauge zero is 3 divisions ahead when fully closed; all diameters are overestimated unless corrected. Systematic error (environmental): Steel rod expands by heat; length readings are all larger. Systematic error (personal): Eye not at right angle to scale (parallax) causing a consistent offset. Random error: Fluctuation of the last digit when reading a thermometer or digital meter. Error types with examples We communicate uncertainty with three linked ideas. Absolute error a is the uncertainty written in the same unit as the measurement, e.g., 12.3 0.1 , cm . Relative error is the fraction of uncertainty compared to the magnitude, a/a (dimensionless). Percentage error is that fraction expressed in percent: % , error = ( a/a) 100 % . If the true value is unknown, we estimate it by the arithmetic mean of n repeated readings and then measure how far each reading deviates from the mean. In lab practice for NEET level, we often use the mean absolute error a mean = (1/n) |a i - a | as the uncertainty of the mean, and quote the final answer as a a mean . This honest reporting tells the reader both where the value lies and how trustworthy the digits are. When the true value is unknown, replace it by the mean of repeated measurements. Absolute error of a reading tip Propagation rules assume small errors: x x so that higher-order terms like ( x/x) 2 are negligible. If relative errors are large (say more than 10%), treat the result with caution. Least Count and Zero Error: Reading Instruments Right Least count sets the smallest step you can trust in a reading. On a metre scale with millimetre marks, least count is 1 mm = 0.1 cm. For a vernier caliper, least count equals the value of one main scale division (MSD) minus the value of one vernier scale division (VSD). For most common calipers, LC is 0.01 cm or 0.02 mm. For a screw gauge (micrometer), LC equals pitch divided by the number of circular (head) scale divisions. If the pitch is 0.5 mm and there are 100 divisions, LC = 0.5/100 = 0.005 mm. Zero error affects every reading by a constant correction. If the zero of the instrument is ahead of the reference (positive zero error), subtract the zero correction from observed readings. If it is behind (negative zero error), add the correction. Always write the corrected formula explicitly: true reading = main scale reading + (head/vernier reading) × LC ± zero correction. A screw gauge has pitch 0.5 mm and 100 circular divisions. With an object placed, the main scale reading (MSR) is 5.00 mm and the circular reading (CR) is 64 divisions. When fully closed (no object), the zero of the head is 4 divisions ahead of the reference (positive zero error). Find the corrected diameter. easy Corrected diameter d Pitch p = 0.5 mm CSD = 100 LC = p/CSD = 0.005 mm MSR = 5.00 mm CR = 64 div Zero error = +4 div ⇒ zero correction = −4 × LC = −0.020 mm Observed reading R obs = MSR + CR × LC. Corrected reading d = R obs + zero correction. mm Sign of zero correction matters. A positive zero error means the zero mark on the circular/vernier scale is ahead when it should coincide, so the observed reading is too large; apply a negative correction. A negative zero error means the zero is behind; apply a positive correction. Always compute least count first, then compute observed reading, then add the zero correction with its sign. For repeating measurements, keep the correction consistent across all readings before averaging. Arithmetic mean (best estimate) Estimate the true value by averaging n repeated measurements of the same quantity. Random errors are symmetric about the true value. Systematic errors have been corrected (or are negligible). All readings measure the same quantity under identical conditions. Arithmetic Mean as Best Estimate Define the mean as the candidate estimate. We seek the "best " estimator by minimizing the spread around it. Stationary condition for minimum. Thus the mean minimizes S( ) and is the best estimate under symmetric random errors. a = a i n The mean pulls information from all readings, reducing random error. If random errors are independent with similar spread, the uncertainty in the mean shrinks roughly like 1/ n . But averaging cannot fix systematic bias; you must calibrate or correct for that. In lab reporting, pair the mean with an uncertainty estimate such as the mean absolute error or a standard-deviation-based measure. At NEET level, using the mean absolute error is sufficient, provided you show how you computed deviations from the mean and round the final result to match the uncertainty. Non-zero digits Always significant 235 Zeros between non-zeros Significant 2003 Leading zeros Not significant (placeholders) 0.0052 2 (5 and 2) Trailing zeros (decimal present) Significant 12.300 Trailing zeros (no decimal shown) Usually not significant by default 1200 2 (unless specified by bar/decimal) Exact counts and definitions Infinite significant figures 10 students; 1 m = 100 cm Significant figures rules for decimal numerals Case Rule Example Sig figs Rounding keeps digits consistent with uncertainty. When adding or subtracting, match decimal places to the least precise term. When multiplying or dividing, match significant figures to the factor with the fewest significant figures. Round 5 up only if it is strictly 5 with no following nonzero digit; in practice, use the conventional rule: 4 or less, round down; 5 or more, round up. Carry extra digits through intermediate steps and round only at the end, so that rounding does not compound your computational error. Always ensure that the reported uncertainty has one, at most two, significant figures, and align the measured value’s last digit with the uncertainty’s last digit. neet-alert Do not round intermediate results to the final number of significant figures. Keep guard digits during calculation, then round the final reported value and its uncertainty together. For Z = A B , absolute errors add for worst-case estimate. Propagation: Sum or difference Propagation: Product or quotient For Z = AB or Z = A/B , relative errors add for worst-case estimate. Propagation: Power rule For Z = A n , multiply the relative error of A by |n| . Best estimate of a quantity from repeated measurements is the arithmetic mean. For addition/subtraction, absolute errors add for the maximum possible uncertainty. For multiplication/division, add relative (fractional) errors to estimate the worst-case uncertainty. For a power, multiply the relative error by the absolute value of the exponent. Convert differentials to finite small errors and sum magnitudes for worst case. Z Z = A A + B B Errors are small: A A , B B . Worst-case estimate: errors add in magnitude. Variables are uncorrelated. Propagation for Product/Quotient Choose the propagation rule to match the calculation. For sums and differences, keep absolute errors in the same unit and add them. For products and quotients, add fractional (or percentage) errors. For powers, multiply the fractional error by the absolute value of the exponent. These rules estimate the maximum possible error if all individual errors conspire in the same direction. The approximation relies on small fractional errors so that cross terms like ( A/A)( B/B) are negligible. If relative errors are large, remeasure with a better instrument or present a conservative bound instead of a precise figure. tip Boundary checks: If A 0 or B 0 , the relative error A/A or B/B can blow up. Avoid dividing by near-zero quantities or treat the result as highly uncertain. Four measurements of a wire length (cm) are: 12.4, 12.3, 12.5, 12.2. Find the mean, mean absolute error, relative error, and percentage error. Report the result with proper significant figures. medium Final reported length with uncertainty Readings (cm): 12.4, 12.3, 12.5, 12.2 n = 4 Compute a , the deviations |a i - a | , mean absolute error a mean , then relative and percentage errors. cm cm 3 hard r = 2.50 ± 0.01 cm h = 10.0 ± 0.1 cm V = r 2 h Use propagation: for V = r 2 h , V V = 2 r r + h h . V and ΔV A solid cylinder’s radius r and height h are measured as r = (2.50 ± 0.01) cm and h = (10.0 ± 0.1) cm. Compute the volume V = π r 2 h with its absolute error and report with proper significant figures. Relative random error (qualitative) Number of readings n control dependent Relative error ~ 1/sqrt(n) Curve decreasing like 1/sqrt(n), showing diminishing returns: each 4× increase in n halves the random error of the mean. custom Averaging reduces random error roughly as 1/√n. Single reading: full random spread 1/2 n = 4 halves error 16 n = 16 quarters error 1/4 2D PLOT Random error of the mean vs number of readings err = 1 / sqrt(n) err remember Averaging is powerful but with diminishing returns. To make random error 10 times smaller, you need about 100 times more readings. A tight cluster of readings can be far from the true value if there is a systematic error. Precision reduces random spread; calibration removes bias. High precision guarantees high accuracy. When subtracting quantities, subtract their errors too. For worst-case estimates, absolute errors always add for both addition and subtraction: Z = A + B . Note instrument’s least count and check for zero error; correct all readings first. Take multiple readings under similar conditions; reject obvious blunders. Compute the mean value and an uncertainty (e.g., mean absolute error). Propagate errors through any derived calculation using the correct rule. Round the final value and its uncertainty consistently; align last digits. State units and any corrections or assumptions made. Protocol for reporting a measurement Z = A + B or A − B Absolute error ΔZ = ΔA + ΔB Length difference Z = AB or A/B Relative error (ΔZ/Z) = (ΔA/A) + (ΔB/B) Area = l × b Z = A n Relative error (ΔZ/Z) = |n|(ΔA/A) Volume ∝ r 3 Percent form Percentage error Multiply relative error by 100% 4% + 2% = 6% Quick propagation cheat sheet (worst-case) Operation Quantity Rule Example tip Keep units consistent before propagation. Convert all lengths to the same unit first; relative errors are unit-free but absolute errors are not. Beware of small denominators. If you compute a quantity like resistance R = V/I and the current I is very small with an uncertainty comparable to its value, the relative error in R becomes very large. In such cases, improving the measurement of the small quantity (e.g., using a more sensitive ammeter, averaging longer to reduce noise, or changing the circuit to increase current) has the biggest impact on the final uncertainty. Percentage error vs percentage value: 2.0% is the uncertainty, not the same as saying the quantity is 0.02. Also, report uncertainty with 1–2 significant figures and match the measured value’s last digit with it. neet-alert For repeated measurements, a simple and accepted estimate of uncertainty at this level is the mean absolute error. It balances robustness and simplicity. Compute the arithmetic mean first, then find the absolute deviation of each reading from the mean, and average those deviations. Quote the final result as a a mean . This communicates a likely range for the true value without assuming any specific error distribution. If a single blundered reading stands out by a huge margin, justify discarding it and recompute. Mean absolute error (of readings) Use this with the mean to report a a mean . This method provides a robust estimate of the average uncertainty by averaging the absolute deviations from the calculated mean. Time period T (s) of a pendulum measured 5 times: 2.05, 2.00, 2.03, 2.02, 2.00. Find T ± ΔT using mean absolute error. Then compute g = 4π 2 L/ T 2 with L = (1.000 ± 0.001) m and report g with its uncertainty. medium T ± ΔT, then g ± Δg Compute T and T mean . Then propagate: ( g/g) = ( L/L) + 2( T/ T ) . T readings (s): 2.05, 2.00, 2.03, 2.02, 2.00 n = 5 L = 1.000 ± 0.001 m g = 4 2 L/ T 2 m/ s 2 Propagation rules memory aid ADP Power: Addition → Delta adds; Division/Product → Percent adds; Power → multiply percent by |n|. Systematic errors require strategy, not averaging. Calibrate instruments against standards, align your eye to avoid parallax, control temperature and drafts, and use proper technique (e.g., measure diameter at multiple orientations to reduce shape bias). Record instrument IDs and calibration dates. If you suspect a constant offset, measure a known standard to estimate and apply a correction to all readings. When reporting, state the source of major uncertainties and what you did to minimize them. Reducing errors: quick remedies Random errors: take many readings, average, stabilize the setup, increase signal (e.g., longer timing intervals). Instrumental systematic: recalibrate or apply zero correction; use higher resolution tools. Environmental systematic: control temperature, humidity, level surfaces, and shield from drafts. Personal systematic: avoid parallax by eye alignment; use steady timing methods (e.g., average multiple oscillations). tip Before recording, check zero and least count. For vernier and screw gauge, take readings at two or three orientations and average to mitigate asymmetry of the object. In summary, measurements come with uncertainty that can be stated clearly and handled consistently. Identify error types and correct what you can, estimate the best value with the arithmetic mean, quantify uncertainty as absolute and percentage errors, and propagate those errors through any calculation using the appropriate rule. Report your final answer with properly matched significant figures so that every digit you write has a reason to be trusted. Key Terms Recap Arithmetic mean Sum of readings divided by their count; best estimate of the true value for symmetric random errors. Uncertainty in the same unit as the measurement. Absolute error Relative error Fractional uncertainty: absolute error divided by the value. a/a Relative error expressed in percent. Percentage error ( a/a) 100 % Least count Smallest resolvable interval of the instrument. Nonzero reading when true value is zero; must be corrected. Zero error Precision Closeness of repeated readings with each other. Closeness of a measurement to the true value. Accuracy Error propagation Rules to estimate uncertainty in a derived quantity from uncertainties in measured inputs. Mean absolute error Average of absolute deviations from the mean; a practical uncertainty estimate for repeated readings.