Measurement Errors Notes FE Sem 2 Measurements and Sensors | MU
On this page
Every engineering decision rests on a number that somebody measured, and no measured number is perfect. This first module of Physics of Measurements and Sensors (BSC2023, FE Semester 2, NEP 2020) gives you the vocabulary to talk about measurements honestly: what a measurement is, how far it can be from the true value, how instruments are verified and calibrated, which terms describe an instrument (range, sensitivity, resolution, response time, repeatability, accuracy and precision), where errors come from, and how a set of repeated readings is treated statistically with the mean, the standard deviation and the method of least squares. It maps to CO1 and carries 6 lecture hours, the largest share of the course, so expect at least one full question on it.
Summary
Principle of measurement: To measure a quantity is to compare it with a standard of the same kind and express the result as a number times a unit, for example 25.13 mm. The quantity being measured is the measurand. A measurement system usually has three stages: a sensing element (or primary transducer) that responds to the measurand, a signal conditioning stage that amplifies, filters or converts the signal, and a display or recording stage. The physical principle used for sensing may be mechanical (a spring stretches), electrical (resistance changes), optical (fringes move), thermal (a junction gives an emf) and so on. A good measurement disturbs the measurand as little as possible; a thermometer that takes heat from a small sample, or a voltmeter that draws current, changes the very thing it measures (loading effect).
Error of measurement and correction: The true value is the value that a perfect instrument would give; in practice it is replaced by a conventional true value obtained with a much better standard. The error is defined as \(E = V_m - V_t\), the measured value minus the true value. The correction is the quantity that must be added to the measured value to get the true value, \(C = V_t - V_m = -E\). So error and correction are equal in size and opposite in sign. Relative error \(E/V_t\) and percentage error \(100E/V_t\) let us compare errors on different scales. Correctness of measurement means that, after known corrections have been applied, the result agrees with the true value within the stated uncertainty.
Reliability, verification and calibration: A measurement is reliable when the instrument gives the same, correct reading every time under the stated conditions, over its working life. Verification is a check, against a standard, that an instrument still meets its specified limits of error; the result is a simple pass or fail. Calibration goes further: the instrument is compared with a standard of higher accuracy (traceable to national standards, in India maintained by CSIR-NPL, New Delhi) over its whole range, and the deviations are recorded as a calibration chart or curve, or the instrument is adjusted so that its readings match the standard. Calibration must be repeated at regular intervals because instruments drift with age, wear and environment.
Characteristics of measuring instruments: Measuring range is the interval between the lowest and highest values the instrument can measure within its specified accuracy; span is the difference between these limits. Sensitivity is the change in output per unit change in input, \(S = \Delta\theta_o/\Delta\theta_i\); a 0.01 mm dial gauge is more sensitive than a steel rule. Scale interval is the difference between the values of two adjacent scale marks; the resolution (or discrimination) is the smallest change in input that produces a detectable change in output. Response time is the time an instrument takes after a sudden change in input to show (a specified fraction, often 90 or 95 percent of) the final reading. Repeatability is the closeness of agreement between successive readings of the same quantity by the same observer, method and instrument under the same conditions; reproducibility is the same idea when conditions (observer, place, time) change. Accuracy is the closeness of the measured value to the true value; inaccuracy (or uncertainty) is the total error band, often given as a percentage of full scale. Precision is the closeness of repeated readings to each other, irrespective of the true value.
Accuracy versus precision: A dartboard picture makes the difference clear. Tight grouping far from the centre is precise but not accurate (a systematic error is present). Scattered darts centred on the bull's eye are accurate on average but not precise (large random error). The ideal instrument is both. Precision can be judged from the spread of repeated readings (the standard deviation); accuracy can be judged only by comparison with a standard.
Sources of error: Static errors arise when the measurand is not changing and include reading errors (parallax, interpolation), characteristic errors and environmental errors. Characteristic errors are the deviations of the instrument's actual characteristic from its ideal one: zero error, non-linearity, hysteresis, backlash and drift in calibration. Environmental errors come from external conditions such as temperature, humidity, pressure, vibration, magnetic and electric fields; for example a steel gauge block expands about 11.5 μm per metre per degree, so length standards are referred to 20 °C. Dynamic errors occur when the measurand changes with time and the instrument cannot follow it because of inertia, damping or thermal lag; a thermometer plunged into hot water lags behind the true temperature during its response time. Errors are also grouped as gross errors (human mistakes), systematic errors (constant or predictable, can be corrected) and random errors (unpredictable, reduced by averaging many readings).
Statistical treatment of errors: When a quantity is measured many times, the readings scatter because of random errors. The arithmetic mean of n readings, \(\bar{x} = \frac{1}{n}\sum x_i\), is the best estimate of the true value. The deviation of each reading is \(d_i = x_i - \bar{x}\), and the sum of deviations is always zero. The spread is described by the standard deviation. If the n readings are the entire population, the population mean is \(\mu\) and the population standard deviation is \(\sigma = \sqrt{\sum (x_i-\mu)^2/n}\). In practice we have only a finite sample, so we use the sample mean \(\bar{x}\) and the sample standard deviation \(s = \sqrt{\sum (x_i-\bar{x})^2/(n-1)}\); dividing by \(n-1\) corrects for the fact that \(\bar{x}\) itself was calculated from the same data. For large n the two become practically equal. For normally distributed random errors about 68 percent of readings lie within \(\pm\sigma\) of the mean, 95 percent within \(\pm 2\sigma\) and 99.7 percent within \(\pm 3\sigma\). The standard deviation of the mean (standard error) is \(s/\sqrt{n}\), which is why averaging more readings improves precision.
Principle of least squares: The principle states that the most probable value of a measured quantity is the one for which the sum of the squares of the residuals (deviations of the observations from that value) is a minimum. For a single quantity measured repeatedly this most probable value turns out to be the arithmetic mean. For two related quantities, such as instrument output y against input x, the best straight line \(y = mx + c\) is the one that minimises \(\sum (y_i - mx_i - c)^2\). Setting the partial derivatives with respect to m and c to zero gives the normal equations and hence the slope and intercept. This is exactly how a calibration curve is fitted to calibration data.
Mind map
Introduction to Measurements
- Principle
- Compare with a standard
- Measurand, sensor, conditioning, display
- Loading effect
- Error terms
- Error E = Vm − Vt
- Correction C = −E
- Correctness, reliability
- Checking instruments
- Verification: pass or fail
- Calibration: compare over full range
- Traceability to national standards
- Instrument terms
- Range, span, sensitivity
- Scale interval, resolution
- Response time, repeatability
- Accuracy, precision, inaccuracy
- Sources of error
- Static, characteristic
- Environmental
- Dynamic
- Gross, systematic, random
- Statistics
- Sample and population mean
- Standard deviation s and σ
- Least squares, best fit line
Deriving the least squares straight line
- Take n pairs of readings \((x_i, y_i)\), for example calibration input and output.
- Assume the relation is a straight line \(y = mx + c\). The residual of the i-th point is \(r_i = y_i - (mx_i + c)\).
- By the principle of least squares, choose m and c so that \(S = \sum r_i^2 = \sum (y_i - mx_i - c)^2\) is a minimum.
- Set \(\partial S/\partial c = 0\): \(-2\sum (y_i - mx_i - c) = 0\), giving \(\sum y_i = m\sum x_i + nc\).
- Set \(\partial S/\partial m = 0\): \(-2\sum x_i(y_i - mx_i - c) = 0\), giving \(\sum x_iy_i = m\sum x_i^2 + c\sum x_i\).
- These two are the normal equations. Solving them: \(m = \dfrac{n\sum x_iy_i - \sum x_i\sum y_i}{n\sum x_i^2 - (\sum x_i)^2}\) and \(c = \dfrac{\sum y_i - m\sum x_i}{n}\).
- The fitted line passes through the point \((\bar{x}, \bar{y})\).
- Special case: for repeated readings of one quantity, minimising \(\sum (x_i - X)^2\) gives \(X = \bar{x}\), so the arithmetic mean is the most probable value.
Accuracy, precision and the error picture
Think of the total error of a single reading as two parts: a systematic part that is the same for every reading (it shifts the mean away from the true value) and a random part that changes from reading to reading (it spreads readings around the mean). Calibration finds and removes the systematic part; repetition and averaging reduce the random part. An instrument with a zero error of +0.02 mm gives readings that are all 0.02 mm too large: precise but not accurate until the correction of −0.02 mm is applied. A cheap balance in a draughty room gives readings that scatter widely: even if their average is right, a single reading is not trustworthy.
| Point | Accuracy | Precision |
|---|---|---|
| Meaning | Closeness to the true value | Closeness of repeated readings to each other |
| Spoilt by | Systematic errors (zero error, wrong calibration) | Random errors (noise, vibration, reading scatter) |
| Judged by | Comparison with a standard | Standard deviation of repeated readings |
| Improved by | Calibration and correction | Better instrument, averaging, steady conditions |
| Example | Mean of readings equals the true value | All readings agree to 0.001 mm |
Important definitions
| Term | Meaning |
|---|---|
| Measurand | The physical quantity that is being measured. |
| True value | The value a perfect measurement would give; in practice the value given by a much more accurate standard. |
| Error | Measured value minus true value, \(E = V_m - V_t\). |
| Correction | The amount added to the measured value to get the true value, \(C = V_t - V_m = -E\). |
| Calibration | Comparing an instrument with a more accurate, traceable standard over its range and recording or adjusting its deviations. |
| Verification | Checking that an instrument still works within its specified error limits. |
| Measuring range | The interval between the minimum and maximum values that can be measured within the stated accuracy. |
| Sensitivity | Change in output per unit change in input. |
| Scale interval | Difference between the values represented by two successive scale marks. |
| Resolution | Smallest change in the input that produces a detectable change in the output. |
| Response time | Time taken after a step change in input for the output to reach a stated percentage of its final value. |
| Repeatability | Agreement between successive readings of the same quantity under identical conditions. |
| Accuracy | Closeness of a measured value to the true value. |
| Precision | Closeness of repeated measurements to one another. |
| Hysteresis | Difference in output for the same input depending on whether the input is increasing or decreasing. |
| Drift | Slow change of an instrument's output over time with no change in input. |
| Standard deviation | Root mean square of deviations from the mean; a measure of scatter. |
Formulas
| Formula | Meaning |
|---|---|
| \( E = V_m - V_t,\quad C = V_t - V_m \) | Error and correction. |
| \( \%E = \dfrac{V_m - V_t}{V_t}\times 100 \) | Percentage (relative) error. |
| \( S = \dfrac{\Delta \text{output}}{\Delta \text{input}} \) | Static sensitivity. |
| \( \bar{x} = \dfrac{1}{n}\sum_{i=1}^{n} x_i \) | Sample mean (best estimate of the true value). |
| \( d_i = x_i - \bar{x},\quad \sum d_i = 0 \) | Deviation from the mean. |
| \( \sigma = \sqrt{\dfrac{\sum (x_i-\mu)^2}{n}} \) | Population standard deviation (μ = population mean). |
| \( s = \sqrt{\dfrac{\sum (x_i-\bar{x})^2}{n-1}} \) | Sample standard deviation. |
| \( s_{\bar{x}} = \dfrac{s}{\sqrt{n}} \) | Standard deviation of the mean. |
| \( m = \dfrac{n\sum xy - \sum x\sum y}{n\sum x^2 - (\sum x)^2},\ c = \dfrac{\sum y - m\sum x}{n} \) | Least squares slope and intercept of \(y = mx + c\). |
Key points to remember
- Error = measured − true; correction = true − measured; they are equal and opposite.
- Verification gives pass or fail; calibration gives a correction chart over the whole range.
- Calibration standards should be several times (typically 4 to 10 times) more accurate than the instrument and traceable to national standards (NPL in India).
- Sensitivity is output per unit input; resolution is the smallest detectable input change.
- Accuracy needs a standard to judge; precision is judged from repeated readings alone.
- Static error: steady measurand. Dynamic error: changing measurand that the instrument cannot follow.
- Systematic errors are removed by correction; random errors are reduced by averaging.
- Use n − 1 in the sample standard deviation and n in the population standard deviation.
- About 68, 95 and 99.7 percent of normal readings lie within 1σ, 2σ and 3σ of the mean.
- Least squares: minimise the sum of squared residuals; the best fit line passes through the mean point.
Solved examples
Example 1. A standard 50.20 mm slip gauge is measured with a micrometer which reads 50.35 mm. Find the error, the correction and the percentage error.
Error \(E = V_m - V_t = 50.35 - 50.20 = +0.15\) mm (the instrument reads high). Correction \(C = -E = -0.15\) mm. Percentage error \(= 0.15/50.20\times100 = 0.299\%\), about 0.3 percent.
Example 2. Eight readings of the diameter of a shaft (mm) are 25.12, 25.15, 25.10, 25.14, 25.13, 25.16, 25.11, 25.13. Find the mean, the sample standard deviation and the population standard deviation.
Sum \(= 201.04\), so \(\bar{x} = 201.04/8 = 25.13\) mm. Deviations: −0.01, +0.02, −0.03, +0.01, 0, +0.03, −0.02, 0 (sum zero, as a check). Squares: 0.0001, 0.0004, 0.0009, 0.0001, 0, 0.0009, 0.0004, 0, total \(\sum d^2 = 0.0028\) mm². Sample standard deviation \(s = \sqrt{0.0028/7} = \sqrt{0.0004} = 0.020\) mm. Treating the eight readings as the whole population, \(\sigma = \sqrt{0.0028/8} = 0.0187\) mm. Result: \(25.13 \pm 0.02\) mm; standard deviation of the mean \(= 0.020/\sqrt{8} = 0.0071\) mm.
Example 3. A temperature sensor is calibrated and gives these outputs: x (°C) = 0, 20, 40, 60, 80, 100 and y (Ω) = 100.0, 107.8, 115.5, 123.4, 131.0, 138.6. Fit the best straight line by least squares.
n = 6, \(\sum x = 300\), \(\sum y = 716.3\), \(\sum xy = 38520\), \(\sum x^2 = 22000\). Slope \(m = \dfrac{6(38520) - 300(716.3)}{6(22000) - 300^2} = \dfrac{231120 - 214890}{132000 - 90000} = \dfrac{16230}{42000} = 0.3864\) Ω/°C. Intercept \(c = (716.3 - 0.3864\times300)/6 = (716.3 - 115.93)/6 = 100.06\) Ω. Best fit: \(R = 100.06 + 0.3864\,T\). Sensitivity of the sensor is 0.386 Ω/°C, and \(m/c = 0.00386\) per °C is its temperature coefficient.
Example 4. A pressure gauge of range 0 to 10 bar has an inaccuracy of ±1 percent of full scale. What is the possible percentage error when it reads 2 bar?
Maximum error \(= 0.01\times10 = \pm0.1\) bar at any point of the scale. At 2 bar the relative error can be \(0.1/2\times100 = \pm5\%\). Lesson: an instrument specified as a percentage of full scale should be used in the upper part of its range.
Example 5. The pointer of a meter moves 20 mm when the input changes by 4 mV. Find its sensitivity. If the smallest pointer movement that can be read is 0.5 mm, what is the resolution?
Sensitivity \(S = 20/4 = 5\) mm/mV. Resolution \(= 0.5/5 = 0.1\) mV.
Important questions and answers
Q1. Define error and correction. How are they related? (3 marks)
Error is the difference between the measured value and the true value, \(E = V_m - V_t\). Correction is the value that must be added algebraically to the measured value to obtain the true value, \(C = V_t - V_m\). Hence \(C = -E\): they have the same magnitude and opposite sign. A reading of 10.05 mm on a 10.00 mm standard has error +0.05 mm and correction −0.05 mm.
Q2. Differentiate between accuracy and precision with an example. (5 marks)
Accuracy is the closeness of a measured value to the true value; precision is the closeness of repeated readings to each other. Accuracy is spoilt by systematic errors and needs a standard to check; precision is spoilt by random errors and is measured by the standard deviation. Example: a shaft of true diameter 20.00 mm read five times as 20.11, 20.10, 20.11, 20.12, 20.11 mm is precise (spread 0.01 mm) but not accurate (about 0.11 mm high, probably a zero error). Readings 19.95, 20.06, 19.98, 20.04, 19.97 average 20.00 mm (accurate on average) but are not precise. Draw the four dartboard (target) diagrams: neither, precise only, accurate only, both.
Q3. Explain calibration and verification. Why must instruments be calibrated periodically? (5 marks)
Verification checks, using a standard, whether an instrument still meets its specified accuracy limits; the outcome is accept or reject. Calibration compares the instrument with a standard of higher accuracy at several points over its range, records the deviations as a calibration chart or curve, and if possible adjusts the instrument. The standard must be traceable through a chain of comparisons to national standards. Periodic calibration is needed because wear, ageing of springs and electronic components, shocks, temperature cycling and contamination cause drift in zero and sensitivity. Calibration intervals depend on use and required accuracy, and a calibration certificate records date, standard used, results and uncertainty.
Q4. Explain the following characteristics of measuring instruments: measuring range, sensitivity, scale interval, response time, repeatability. (5 marks)
Measuring range: the interval from the minimum to the maximum value the instrument can measure within its stated accuracy, for example 0 to 25 mm for a micrometer. Sensitivity: ratio of the change of output to the change of input, for example 2 mm of pointer movement per micrometre in a comparator. Scale interval: the difference between values corresponding to two consecutive scale marks, for example 0.01 mm on a micrometer thimble; it limits how finely you can read. Response time: time taken by the instrument to show a stated fraction (90 to 95 percent) of the final reading after a sudden change of input; important for changing quantities. Repeatability: the ability to give the same reading for repeated measurements of the same quantity by the same person, method and instrument under the same conditions.
Q5. Explain the different sources of error in measurement. (6 marks)
Static errors occur when the measured quantity is steady; they include reading errors such as parallax and interpolation, and errors from friction or loading. Characteristic errors are deviations of the real instrument from its specified characteristic: zero offset, non-linearity, hysteresis, backlash in gears and screws, and drift of calibration. Environmental errors are due to conditions differing from those of calibration: temperature (thermal expansion of the part and the instrument; standard temperature 20 °C), humidity, pressure, dust, vibration and stray electric or magnetic fields. They are reduced by controlled rooms, compensation and shielding. Dynamic errors occur when the measurand varies with time; the instrument's inertia, damping or thermal capacity make it lag, so the indicated value differs from the instantaneous true value. They are reduced by choosing an instrument with a short response time. Besides these, gross errors are blunders by the observer, systematic errors are repeatable and correctable, and random errors are irregular and treated statistically.
Q6. Define sample mean, sample standard deviation, population mean and population standard deviation. Why is n − 1 used for the sample? (5 marks)
For a sample of n readings, the sample mean \(\bar{x} = \sum x_i/n\) and the sample standard deviation \(s = \sqrt{\sum (x_i-\bar{x})^2/(n-1)}\). For the whole population of N possible readings (in theory infinitely many), the population mean \(\mu = \sum x_i/N\) and the population standard deviation \(\sigma = \sqrt{\sum (x_i-\mu)^2/N}\). Since \(\bar{x}\) is calculated from the same sample, the deviations from \(\bar{x}\) are on average smaller than deviations from the true mean μ, and only n − 1 of them are independent (they must sum to zero). Dividing by n − 1 (degrees of freedom) removes this bias, so s is a fair estimate of σ. For large n the difference is negligible.
Q7. State the principle of least squares and use it to show that the arithmetic mean is the most probable value. (5 marks)
Principle: the most probable value of a quantity is the one that makes the sum of the squares of the residuals a minimum. Let readings be \(x_1, \dots, x_n\) and let X be the most probable value. Residuals \(r_i = x_i - X\). \(S = \sum (x_i - X)^2\). For a minimum, \(dS/dX = -2\sum (x_i - X) = 0\), so \(\sum x_i = nX\) and \(X = \sum x_i/n = \bar{x}\). Since \(d^2S/dX^2 = 2n > 0\), this is a minimum. Hence the arithmetic mean is the most probable value. The same principle applied to \(y = mx + c\) gives the normal equations for the best fit line used in calibration.
Q8. Distinguish between static and dynamic errors. (3 marks)
Static error occurs when the measured quantity is constant; it is the difference between the steady reading and the true value, caused by instrument imperfections, environment or the observer. Dynamic error occurs when the quantity changes with time; it is the difference between the true instantaneous value and the indicated value due to the instrument's lag, inertia or damping. Static errors are handled by calibration; dynamic errors by using instruments with fast response and proper damping.
Exam tips
- Always write the sign of the error and correction; examiners check that \(C = -E\).
- For accuracy versus precision, a four-target diagram earns easy marks.
- In statistics numericals, tabulate x, d and d² neatly; state whether you used n or n − 1.
- Show the check \(\sum d_i = 0\) to catch arithmetic slips.
- In least squares problems, write the sums \(\sum x, \sum y, \sum xy, \sum x^2\) first, then substitute.
- Give one practical example for each term (micrometer, dial gauge, thermometer); it makes definitions convincing.
- Link environmental error to the standard reference temperature of 20 °C for length measurement.
FAQs
Is resolution the same as sensitivity?
No. Sensitivity is how much the output changes for a unit change of input. Resolution is the smallest input change you can actually detect. A high sensitivity usually helps resolution, but noise and scale marks also limit it.
Can an instrument be precise but not accurate?
Yes. If it has a systematic error such as a zero offset, all readings agree with each other but are shifted from the true value. Calibration and correction make it accurate.
Why do we use n − 1 in the standard deviation formula?
Because the mean was calculated from the same readings, only n − 1 deviations are independent. Dividing by n − 1 gives an unbiased estimate of the population scatter.
How many marks does this course carry in the end semester exam?
The theory end semester exam is 45 marks. The course table gives its duration as 2 hours, though the semester overview table lists 1.5 hours, so confirm with your college. The paper has five questions of 15 marks each, Q1 compulsory, any three to be answered in total. Internal assessment tests add 30 marks.
These notes follow the University of Mumbai first-year engineering syllabus (NEP 2020, Physics of Measurements and Sensors) and were prepared with AI help and checked.
More in Physics of Measurements and Sensors
See all →- Nanotechnology Notes FE Sem 2 Measurements and Sensors | MU MU FE Sem 2 Measurements and Sensors Module 6: nano properties, surface to volume ratio, van der Waals bonding, SEM, TEM, AFM, gas sensing,...
- Wave Interference Notes FE Sem 2 Measurements and Sensors | MU MU FE Sem 2 Measurements and Sensors Module 2: monochromatic light, optical flats, air wedge fringes, flatness testing and surface contour t...
- Transducers Notes FE Sem 2 Measurements and Sensors | MU MU FE Sem 2 Physics of Measurements and Sensors Module 3: transducer classification, potentiometer, strain gauge, RTD, LVDT and thermal phot...
- Solid State Sensors Notes FE Sem 2 Measurements and Sensors | MU MU FE Sem 2 Measurements and Sensors Module 4: Hall effect and Hall voltage, piezoelectric ultrasonics, ultrasonic flow and distance meters.
- Temperature Notes FE Sem 2 Measurements and Sensors | MU MU FE Sem 2 Measurements and Sensors Module 5: bimetallic, Pt-100, thermocouple and NTC thermistor thermometers, selection, Pt-100 calibrati...