Measurements, Instrumentation, Sensors, and Uncertainty
Key Takeaways
- Measurement questions test whether the device fits the variable, range, response time, environment, and required accuracy.
- Accuracy, precision, resolution, repeatability, sensitivity, bias, and calibration are distinct terms.
- Full-scale accuracy specifications can produce large relative error at low readings.
- Signal conditioning includes amplification, filtering, isolation, linearization, and analog-to-digital conversion.
- Uncertainty propagation should follow the operation being performed, not a generic percentage rule.
- Sensor selection on the FE is usually solved by matching the physical principle to the measured quantity.
Think in measurement chains
A sensor is only one part of a measurement system. A complete chain begins with the physical variable, converts it to a signal, conditions that signal, digitizes or displays it, and then uses the value for a decision. FE Mechanical questions may name a device, but the underlying test is whether the device fits the variable, range, environment, response speed, and uncertainty requirement.
| Variable | Common sensor or instrument | Watch for |
|---|---|---|
| Temperature | Thermocouple, resistance temperature detector, thermistor, infrared sensor | Range, accuracy, response, contact method |
| Pressure | Bourdon tube, diaphragm transducer, strain-gage bridge, manometer | Gauge vs absolute, dynamic response, calibration |
| Flow | Orifice plate, Venturi, rotameter, turbine, ultrasonic, Coriolis | Pressure loss, fluid type, mass vs volume flow |
| Force or weight | Load cell, strain gage | Bridge completion, temperature compensation |
| Position | Linear variable differential transformer, encoder, potentiometer | Resolution and linearity |
| Speed | Tachometer, encoder, strobe | Sampling and aliasing |
Vocabulary that decides answers
Accuracy is closeness to the true value. Precision is tight repeatability, even if the readings are biased. Resolution is the smallest displayed or encoded change. Sensitivity is output change per input change. Bias is a systematic offset. Calibration compares the instrument against a known standard and adjusts or documents the relationship.
Do not treat these as synonyms. A digital indicator can have fine resolution and poor accuracy. A pressure gage can be precise but biased by zero offset. A sensor can be accurate at midrange but unsuitable near its limit because noise or nonlinear behavior dominates.
Uncertainty and propagation
Uncertainty should be tied to the specification. If an instrument is rated at plus or minus 0.25 percent of full scale on a 0 to 200 lbf range, the uncertainty is plus or minus 0.5 lbf anywhere in that range. At a 10 lbf reading, that is a large relative uncertainty. If an instrument is rated as percent of reading, the absolute uncertainty changes with the reading.
For independent measurements, sums and differences combine absolute uncertainties. Products and quotients usually combine relative uncertainties. When the handbook or problem gives a root-sum-square method, use it instead of worst-case addition. Always keep units attached to the uncertainty value.
Signal conditioning and sampling
Signal conditioning makes the raw sensor output usable. Strain gages often need a Wheatstone bridge and amplification. Thermocouples require cold-junction compensation. Noisy signals may need filtering. A high-frequency vibration measurement needs adequate sampling to avoid aliasing. The Nyquist rule says the sample rate must be at least twice the highest frequency component, but engineering practice usually uses more margin.
| Problem cue | Likely concept |
|---|---|
| Output offset at zero load | Bias or zero calibration |
| Scatter around a mean | Random error or precision |
| Smallest display increment | Resolution |
| Output voltage per degree | Sensitivity |
| Signal appears at false low frequency | Aliasing |
| Noisy high-frequency component | Low-pass filtering may help |
Exam workflow
First identify the measured variable. Then choose the device family. Next check range, accuracy basis, response time, and whether the output needs conditioning. Finally, calculate uncertainty in the same units requested by the answer choices. Many FE misses in this domain come from choosing a familiar sensor without checking whether its physics matches the job.
A 0 to 500 N load cell is specified as plus or minus 0.2 percent of full scale. What is the absolute uncertainty from that specification?
A rotating shaft vibration signal contains meaningful content up to 600 Hz. What is the minimum sampling rate required by the Nyquist criterion?
A pressure instrument gives nearly identical readings on repeated tests, but every reading is about 4 kPa above a calibrated standard. Which term best describes the dominant error?