Question 1 of 7: Instrumentation Fundamentals — Calibration, Resolution, and Dynamic Response

Question 1: Instrumentation Fundamentals — Calibration, Resolution, and Dynamic Response (20 marks)

Question text not reproduced: the examination questions are © Engineers and Geoscientists BC. Open the official past paper (linked at the top of this page) to read the question, then follow the worked solution below.

a) $R^2$ is a dimensionless goodness-of-fit statistic — it says how much of the output's VARIANCE the fitted line explains, and it says nothing about the fit's actual size of error in the instrument's own engineering units. A calibration can carry a high $R^2$ while still having a large absolute error (e.g. over a wide input range) or a systematic offset that barely dents $R^2$. RMS error is computed in the measured variable's own units and directly states the typical SIZE of the residual between the fitted curve and the data — it is what an engineer actually needs to know to state the instrument's uncertainty.

b) A one- or two-point calibration only fixes an offset (and, with two points, a slope) through those specific points, which silently assumes the response is perfectly linear everywhere in between. A third point, taken somewhere between (or beyond) the first two, is the minimum needed to actually TEST that linearity assumption — if the third point falls off the line drawn through the first two, the instrument has measurable non-linearity that a two-point calibration could never reveal.

c) With hysteresis the sensor's output for a GIVEN input value depends on whether that input was approached from above or below (its recent history), so the output is no longer a single-valued function of the input — two different true values can produce the identical reading, and the same true value can produce two different readings depending on direction of approach. Since the whole point of a measuring instrument is to invert output back to a unique input value, an output that cannot be uniquely inverted without also knowing the input's unknown recent history is useless as presented, unless the direction-dependent loop is itself characterized and corrected for.

d) The smallest reliably measurable value is set by the instrument's own NOISE FLOOR, not by its nominal resolution or display digits — a genuine signal must rise above the RMS noise/baseline fluctuation by some statistical margin (commonly taken as $3\sigma$ of the noise) before it can be distinguished from random background variation with confidence. Below that level the reading is indistinguishable from noise regardless of how many digits the display shows.

e) The wait time is governed by the sensor's own DYNAMIC (transient) response — its time constant $\tau$ (first order) or natural frequency/damping ratio (second order) — not by a fixed universal number. A first-order sensor is conventionally considered settled after about $4\tau$–$5\tau$ (98–99% of the final value); a second-order (underdamped) sensor needs the settling time $t_s\approx4/(\zeta\omega_n)$ for a $\pm2\%$ band, and any residual ringing must have decayed within that band before the reading is trusted.

f) Accuracy is generally the more important of the two, because a PRECISE but INACCURATE instrument gives a tight, confident-looking cluster of readings that is nonetheless offset from the true value — the operator has no way to know, from the readings alone, that they are all wrong in the same direction. A precise instrument that is also biased is more dangerous than an imprecise one, because its apparent repeatability creates false confidence in a wrong number; averaging repeated readings improves precision-limited scatter but does nothing for a systematic accuracy error.

g) The ZERO is the output reading corresponding to the lower range value (LRV) of the input — the instrument's baseline/offset point. The SPAN is the difference between the upper and lower range values of the input the instrument is calibrated to cover (equivalently, the output range corresponding to that full input range) — it sets the instrument's scaling slope (output change per unit input change) between the zero point and full scale.

h) The Gibbs phase rule, $F=C-P+2$, shows that for a single-component system ($C=1$) with two phases coexisting in equilibrium ($P=2$), the number of degrees of freedom is $F=1$: once one intensive variable (commonly pressure) is fixed, the temperature of that two-phase equilibrium is completely DETERMINED and cannot drift while both phases remain present (e.g. an ice–water bath sits at a fixed temperature at a given pressure as long as both ice and water remain). At the TRIPLE POINT ($P=3$ for $C=1$), $F=0$ — an absolutely invariant temperature and pressure with no free variable at all. This is exactly why phase-change points (ice point, steam point, and especially the triple point of water) make reproducible, self-buffering fixed points for calibration standards: the phase rule guarantees the temperature cannot move away from the defined value as long as the coexisting phases are present.

i) A mass (load) sensor is normally modelled as a second-order spring–mass–damper system, so its dynamic response is fully described by three parameters: the static SENSITIVITY (output per unit mass at steady state), the NATURAL FREQUENCY $\omega_n$ (how fast it would oscillate undamped), and the DAMPING RATIO $\zeta$ (how quickly oscillation decays — underdamped, critically damped, or overdamped). Together these three set the derived response characteristics — rise time, percent overshoot, settling time — that determine how fast and how cleanly the sensor reaches a new steady reading after a step change in load.

j) Any single calibration reading at a given point is itself corrupted by that instant's random noise, so a one-pass calibration curve can be biased by whatever noise happened to occur at each point. Repeating each point several times and averaging cancels the random component and reveals the true underlying input–output relationship, while the SPREAD between repeats at a given point is itself the only direct measure of the instrument's own repeatability/precision — information a single pass can never provide — and repeated passes also give an early warning of drift between calibration runs.