Q1Power System Instrumentation

(a) Statistical Analysis of Voltmeter Readings

Given readings: x1=99.7V, x2=99.8V, x3=100.0V, x4=100.2V, x5=100.3V (n=5 observations).

(i) Arithmetic Mean

(ii) Average (Mean) Deviation

The deviation of each reading from the mean is: d1=-0.3, d2=-0.2, d3=0.0, d4=0.2, d5=0.3 (all in volts). The average deviation is the arithmetic mean of the absolute values of these deviations:

(iii) Standard Deviation

For a finite sample of observations, the standard deviation is calculated using (n-1) in the denominator (Bessel's correction, providing an unbiased estimate of the true population standard deviation from a limited sample):

(iv) Variance

Variance is simply the square of the standard deviation:

Summary of results: mean=100.0V, average deviation=0.2V, standard deviation approximately 0.255V, variance approximately 0.065 V-squared. These four statistical measures together characterize both the central tendency (mean) and the dispersion/spread (average deviation, standard deviation, variance) of the five independent observer readings, with the standard deviation being the most widely used dispersion measure in precision instrumentation analysis since it is directly related to the normal (Gaussian) error distribution model commonly assumed for random measurement errors.

(v) Why (n-1) is Used Instead of n

When the standard deviation is computed from a limited sample of readings rather than from the entire (infinite, in principle) population of all possible readings, using the sample mean itself in place of the true (unknown) population mean introduces a small, systematic downward bias if the divisor n is used, because the sample mean is, by construction, the value that minimizes the sum of squared deviations for that particular sample, so the computed sum of squared deviations from the sample mean is always slightly smaller than the sum of squared deviations that would be obtained from the true, unknown population mean. Dividing by (n-1) instead of n (Bessel's correction) exactly compensates for this loss of one independent degree of freedom (since one degree of freedom is effectively 'used up' in estimating the mean from the same data before the deviations are computed), yielding an unbiased estimator of the true population variance. For a small sample such as the five readings in this problem, this correction is numerically significant: dividing by (n-1)=4 instead of n=5 raises the computed standard deviation, appropriately reflecting the greater uncertainty inherent in estimating dispersion from only a handful of observations rather than from a very large data set, where the distinction between n and (n-1) becomes negligible.

Practical Significance of Variance in Calibration and Quality Control

Beyond its role as a purely mathematical dispersion measure, the variance (and its square root, the standard deviation) has direct practical importance in instrument calibration and quality-control practice. When a calibration laboratory repeatedly measures a known reference standard using a test instrument, the resulting variance quantifies the instrument's repeatability, which is reported on the calibration certificate as part of the stated measurement uncertainty budget, alongside systematic corrections for any known bias. A calibration process with a lower variance across repeated trials indicates a more stable, more repeatable instrument, allowing a tighter uncertainty figure to be stated for any single measurement made with that instrument in the field, whereas a larger variance signals that the instrument (or the measurement procedure/environment) requires improvement before it can be relied upon for precision work. In a production/quality-control setting, tracking the variance of a measured parameter across a batch of manufactured items (such as the calibration error of a batch of newly manufactured voltmeters before dispatch) allows statistical process control charts to be constructed, with control limits typically set at some multiple of the standard deviation around the process mean, enabling early detection of a drift in the manufacturing or calibration process before individual instruments fall outside their specified accuracy class.

(b) Types of Errors in Measurement

Systematic error: an error that is consistent, repeatable, and predictable in nature, arising from a known or identifiable cause, and which therefore affects all readings taken under similar conditions in essentially the same way (same magnitude and direction) rather than varying randomly from reading to reading. Systematic errors are further sub-classified into instrumental errors (arising from the measuring instrument's own imperfections, such as friction, calibration drift, or component ageing), environmental errors (arising from external influencing factors such as temperature, humidity, or stray magnetic/electric fields affecting the measurement), and observational errors (arising from the observer's own reading technique, such as parallax error in reading an analog needle-and-scale instrument). Since systematic errors are consistent and identifiable, they can, in principle, be corrected for through careful calibration, environmental compensation, or improved measurement technique.

Random error: an error that varies unpredictably in both magnitude and sign from one measurement to the next, even when the measurement is repeated under seemingly identical conditions, arising from numerous small, uncontrollable disturbing factors (minute fluctuations in friction, small electrical noise, minor variations in observer reaction time, and similar factors) that cannot be individually identified or eliminated. Because random errors are unpredictable in direction, they cannot be corrected for on any single reading, but their statistical effect can be characterized and reduced (in terms of the resulting uncertainty in the estimated true value) by taking multiple repeated measurements and applying statistical treatment (calculating the mean, standard deviation, and probable error, as illustrated in part (a) of this question).

Normal (Gaussian) error: a specific statistical model describing the distribution pattern typically followed by the combined effect of many small, independent random error sources acting together, in accordance with the central limit theorem — normal error follows the familiar symmetric, bell-shaped Gaussian probability distribution curve, characterized entirely by its mean (representing the true value, in the absence of systematic error) and its standard deviation (representing the spread/dispersion of individual readings around this mean), and forms the fundamental theoretical basis for the standard statistical treatment of random measurement errors (including the concepts of probable error and confidence intervals) used throughout precision measurement and instrumentation practice.

Gross error: a large, generally isolated error arising from human mistake or procedural oversight, such as misreading an instrument's scale, incorrectly recording an observed value, using an instrument with an incorrect range/multiplier setting, or a computational mistake in subsequently processing the recorded data — unlike systematic and random errors, which are inherent to the measurement process itself, gross errors are avoidable mistakes attributable to carelessness or human oversight, and are typically identified and eliminated through careful, repeated observation, cross-checking of results, and rejection of outlier readings that deviate implausibly from the rest of a data set (using statistical outlier-rejection criteria such as Chauvenet's criterion in more rigorous measurement analysis).