Measurement and Instrumentation
Probe loading, bandwidth, aliasing, uncertainty, and evidence practices for measurements that survive review.
A measurement is an inference from a physical system through a probe, fixture, instrument, processing chain, and calibration model. The setup changes the system. A credible result defines the measurand, loading, bandwidth, reference plane, uncertainty, and decision rule.
Define the measurement before choosing a tool
Record:
- the physical quantity and exact probe point,
- signal reference and expected common-mode voltage,
- amplitude, frequency, rise time, and duration range,
- acceptable loading and disturbance,
- required uncertainty and pass/fail rule,
- environmental and operating state,
- instrument, probe, fixture, and calibration identifiers,
- raw-data format and processing steps.
“Measure the 3.3 V rail” is incomplete. “Measure minimum voltage at the MCU supply pins during radio startup, with 20 MHz measurement bandwidth, and verify the lower uncertainty bound remains above the device limit” can drive a setup.
Quantify probe loading
At high frequency, probe capacitance can dominate its nominal megohm input resistance. A first-order pole from source resistance $R_s$ and total shunt capacitance $C_T$ is:
\[f_p = \frac{1}{2\pi R_s C_T}\]A $10\ \text{k}\Omega$ source observed with 10 pF total capacitance has $f_p \approx 1.59$ MHz. The probe can attenuate or destabilize behavior far below its advertised oscilloscope bandwidth.
The capacitive reactance is:
\[|X_C| = \frac{1}{2\pi fC}\]At 50 MHz, 10 pF is only about 318 $\Omega$. Evaluate loading at the signal’s harmonic content, not only its repetition frequency.
Keep probe ground loops short. A long ground lead adds inductance that can create ringing absent from the circuit. For power-rail measurements, use a ground spring, coaxial connection, or dedicated probe footprint appropriate to the bandwidth.
Bandwidth and rise time
For a Gaussian-response measurement system:
\[t_{r,\text{scope}} \approx \frac{0.35}{\mathrm{BW}}\]The observed rise time can be estimated by:
\[t_{r,\text{meas}} \approx \sqrt{t_{r,\text{signal}}^2 + t_{r,\text{scope}}^2}\]A 100 MHz system has an approximate 3.5 ns rise time. Applied to a true 2 ns edge, it reports about 4.0 ns. That setup can show the transition, but it cannot measure the original rise time accurately. Probe and scope must be treated as one measurement system.
Bandwidth limiting can improve noise repeatability, but the selected limit becomes part of the measurand. Record it.
Make differential and floating measurements safely
A floating node requires a differential measurement. Do not defeat protective earth on an earth-referenced oscilloscope to make its ground clip float. That can place the instrument chassis, connectors, and connected equipment at the circuit potential.
Before attaching a differential or isolated probe, check:
- maximum differential voltage,
- maximum common-mode voltage and its frequency derating,
- transient and measurement-category rating,
- input capacitance and resistance on each side,
- common-mode rejection ratio at the frequency of interest,
- isolation rating between the probe head, instrument, and operator-accessible connections.
Common-mode rejection is frequency dependent. If a probe has 40 dB CMRR at the switching edge frequency, its linear rejection ratio is only 100:1. A 400 V common-mode transition can therefore create a 4 V input-referred residual in the worst-case screening calculation, overwhelming a 100 mV differential signal.
Use a probe or isolated acquisition system rated for the working voltage and expected transients. Keep both input paths geometrically similar, minimize loop area, and verify the residual with both inputs connected to the same switching node. A differential probe is not automatically galvanically isolated, and isolation does not remove voltage, energy, or clearance limits.
Detect aliasing
For a tone sampled at $f_s$, an alias can appear at:
\[f_{\text{alias}} = |f_{\text{in}} - kf_s|\]where integer $k$ places the result in the observed Nyquist zone. A 105 kHz tone sampled at 100 ksample/s appears at 5 kHz. A stable trace is not proof that the displayed frequency exists at the input.
Use an analog anti-alias filter, sufficient sample rate, and a known acquisition bandwidth. Change sample rate as a diagnostic. A real input component remains fixed while an alias generally moves.
Digital multimeter and current-measurement traps
- Input resistance loads high-impedance nodes.
- Current ranges insert a shunt and fuse resistance that create burden voltage.
- Auto-ranging can disturb sensitive circuits or miss transients.
- Four-wire resistance removes test-lead resistance, not fixture contact variation or thermal EMF.
- Integrating apertures can reject line-frequency noise while hiding shorter events.
- Safety category, transient rating, and lead rating matter more than display digits.
For a shunt measurement:
\[I = \frac{V_{\text{shunt}}}{R_{\text{shunt}}}\]Check shunt tolerance, temperature coefficient, self-heating, amplifier offset, common-mode range, and bandwidth. A $10\ \text{m}\Omega$ shunt at 5 A drops 50 mV and dissipates 250 mW before tolerance and connection losses.
Carry uncertainty through the result
Report a result as:
\[Y = y \pm U\]with units, coverage factor, and method. Contributors can include calibration, resolution, noise, repeatability, loading, drift, temperature, fixture geometry, and model correction. Instrument accuracy specifications are inputs to an uncertainty budget, not the complete budget.
Use repeated setups to measure repeatability when cable, probe, connector, or fixture placement matters. Correct recognized systematic effects before treating the remaining terms statistically.
Measurement plan
| Question | Probe point | Tool | Expected result | Failure interpretation |
|---|---|---|---|---|
| DC rail value | load pins | calibrated DMM | inside static limit with uncertainty | regulator setpoint, path drop, load |
| Rail transient | load pins | scope and low-inductance connection | inside dynamic limit at defined bandwidth | control loop, bypass, layout, load step |
| Clock at receiver | receiver pin | low-capacitance active probe | valid levels, edge, and timing | source, loading, termination, return path |
| Current step | supply path | shunt plus differential probe | peak and duration inside budget | firmware state, load, current limit |
| Sampled sensor tone | ADC input and digital output | scope plus captured samples | source frequency maps as predicted | aliasing, clock error, filtering |
| RF transfer response | calibrated planes | VNA | trace and uncertainty meet limits | DUT, fixture, calibration, repeatability |
Preserve the evidence
Save:
- raw samples before filtering or unit conversion,
- instrument state and scripted configuration,
- probe and fixture photographs,
- calibration and check-standard results,
- hardware, firmware, and configuration identifiers,
- environmental conditions,
- analysis code and version,
- reconnect and rerun results when repeatability matters.
A screenshot without raw data, scaling, probe identity, bandwidth, and configuration is context, not a complete measurement record.
Related notes
PCB Design and Bring-Up, RF Measurement and VNA Workflow, and Design Verification and Test.
Sources
- NIST, Technical Note 1297.
- Tektronix, ABCs of Probes.
- Tektronix, How Oscilloscope Probes Affect Your Measurement.
- Tektronix, Isolation Addresses Common Sources of Differential Measurement Error.
- Analog Devices, Fundamentals of Sampled Data Systems.
- Analog Devices, Filter Basics: Anti-Aliasing.
- Keysight, Signal Analysis Measurement Fundamentals.