3. Measurement system components

A measurement system in physical testing is not a single instrument but a chain of interdependent elements. Every CAE validation program ultimately rests on the credibility of that chain: if the sensor, the signal conditioning, the data acquisition (DAQ) hardware, or the calibration state is compromised, the resulting test data cannot serve as a trustworthy reference for simulation correlation. This section examines each link in the chain, how the links interact, and what engineers must verify to keep the chain intact.

The Measurement Chain as a Signal Path

The most useful mental model is a signal path that transforms a physical quantity into a digital number. A strain gauge converts mechanical strain into a small resistance change; a bridge circuit and amplifier convert that into a voltage; an anti-aliasing filter limits bandwidth; an analog-to-digital converter (ADC) samples and quantizes the voltage; and the acquisition software stores counts that are later scaled to engineering units. Each stage contributes gain, offset, noise, and bandwidth limitations. The overall measurement uncertainty is a combination of all stages, not just the sensor specification printed on the datasheet.

Consider a typical validation case: measuring dynamic strain on a suspension control arm during a road-load test. The expected strain amplitude is on the order of 200–800 µε with frequency content up to roughly 150 Hz. A foil strain gauge with a gauge factor of 2.1 in a quarter-bridge configuration produces a resistance change of about 0.04% at 200 µε — a few tenths of an ohm on a 350 Ω gauge. That tiny change must be amplified and digitized without being buried in noise, which is precisely why signal conditioning and DAQ selection matter as much as the sensor itself.

Sensors: Selection and Installation

Sensor selection begins with the measurand and its expected range, bandwidth, and environment. For the control-arm example, a 350 Ω constantan foil gauge bonded with cyanoacrylate is appropriate for moderate temperatures and good fatigue life. If the surface is aluminum and the test duration is long, a gauge with a matched thermal expansion coefficient reduces apparent strain. Installation quality dominates sensor performance: adhesive cure time, clamping pressure, lead-wire routing, and moisture protection all affect the result. A poorly bonded gauge can show hysteresis or drift that no amount of downstream correction will fix.

For accelerometers, the choice between piezoelectric (charge output) and IEPE (integrated electronics) types depends on cable length and environment. IEPE sensors are convenient for long cable runs because the charge is converted to a low-impedance voltage at the sensor, but they require constant-current excitation and are limited in maximum temperature. Mounting method matters: stud mounting gives the best high-frequency fidelity, while adhesive mounting is acceptable up to a few kilohertz, and magnetic mounting should be avoided for quantitative work above roughly 1 kHz.

Signal Conditioning: Making the Signal Usable

Signal conditioning performs the transformations needed before digitization. For strain gauges, a Wheatstone bridge converts resistance change to voltage, and bridge completion (quarter, half, or full) determines sensitivity and temperature compensation. Excitation voltage must be stable; a 1% drift in excitation produces a 1% gain error. Amplifiers must have sufficient common-mode rejection ratio (CMRR) — typically above 80 dB for bridge measurements in electrically noisy environments — and their gain should be set so the expected signal uses a large fraction of the ADC input range without clipping.

Filtering is equally important. An anti-aliasing low-pass filter set below half the sampling rate prevents high-frequency content from folding into the measurement band. For the suspension example with content up to 150 Hz, a sampling rate of at least 1 kS/s per channel and an anti-alias cutoff near 400 Hz provide margin. Engineers should verify the filter's actual response rather than trusting the nominal setting, since some DAQ modules apply fixed analog filters plus configurable digital filters with different roll-off characteristics.

Data Acquisition: Resolution, Rate, and Synchronization

ADC resolution determines the smallest resolvable increment. A 16-bit converter over a ±10 V range has a nominal least significant bit (LSB) of about 305 µV; a 24-bit converter reduces this to roughly 1.2 µV, but effective resolution is limited by noise. The effective number of bits (ENOB) derived from a noise test is a more honest indicator than the datasheet resolution. For the strain example, a 200 µε signal through a gain of 100 might produce about 4 mV at the amplifier output; with 16-bit resolution and ±10 V range, that is roughly 13 LSBs — workable but not generous, which argues for higher gain or a 24-bit module.

Channel-to-channel synchronization matters when correlating multiple signals. Simultaneous-sampling DAQ hardware avoids the skew introduced by multiplexed architectures, which can be significant when comparing, for example, wheel force and body acceleration at the same instant. Sample rate should be chosen from the highest frequency of interest, not from convenience, and the acquisition clock should be verified against a known reference when timing accuracy is part of the validation objective.

Calibration: Establishing Traceability

Calibration is the process of establishing the relationship between the system output and a known input, with documented traceability to a national standard. Two levels are common in physical testing. End-to-end calibration applies a known stimulus — a shunt resistor across a bridge arm, a dead weight on a load cell, or a calibrated signal source — and records the system response. Component-level calibration verifies individual instruments against a reference. For strain measurement, shunt calibration simulates a known strain by switching a precision resistor across one bridge arm; the resulting reading is compared with the theoretical value, and the gain correction is applied to the data.

A practical calibration record should capture the reference value, the indicated value, the deviation, the environmental conditions, and the date. As an example, a shunt resistor of 59.7 kΩ across a 350 Ω arm with 2 V excitation and gauge factor 2.1 simulates approximately 200 µε; if the system indicates 196 µε, the gain error is about −2%, and the correction factor 200/196 ≈ 1.020 should be applied. This is an illustrative calculation, not a measured result, and the actual values depend on the specific bridge and instrumentation.

Common Failure Symptoms and Their Sources

Bringing the Chain Together

The practical lesson is that measurement system quality is a property of the whole chain, not of any single component. A high-grade sensor with a poorly set gain, an inadequate anti-alias filter, or an uncalibrated amplifier will produce data that looks plausible but fails under scrutiny. Before any validation test, engineers should document the sensor type and mounting, the conditioning settings (excitation, gain, filter), the DAQ configuration (resolution, sample rate, synchronization), and the calibration evidence. That documentation is what allows a CAE analyst to defend the test data as a correlation reference — and what allows a reviewer to trace any discrepancy back to a specific link in the chain.