Measurement Instruments and Tools

The board resets intermittently, yet the multimeter consistently reads 3.3 V: the power supply sag may have been too short for the meter's update rate to capture, or connecting the probe may have altered the node, temporarily masking the issue. Measurement is not just about reading numbers; it requires explaining the reference point, loading effects, bandwidth, time window, and measurement uncertainty.

Multimeter: Determine Connection Method and Measurement Load First

Voltage mode is connected in parallel; current mode is inserted in series into the branch under test. The current input is typically a low-resistance path and must not be connected across a power source as if it were a voltage measurement. After measuring, return the test leads to the correct jacks and use the meter/leads according to their rated specifications and application environment. Resistance and continuity measurements should be performed with power off and stored energy properly dissipated; measurements on-board may still be influenced by parallel paths, but this does not mean components must always be desoldered.

The beep threshold, response time, and test current for the diode mode vary by model. A beep does not automatically equate to a direct copper connection, and capacitor charging may cause a momentary beep. High input impedance reduces loading but can also read potentials formed by capacitive coupling. Do not interpret "phantom voltage" as a zero-risk state that requires no verification.

If the node is equivalent to 3.3 V in series with 100 kΩ, connecting a 1 MΩ meter yields a reading of 3.3×1M/(100k+1M)=3.000 V; connecting 10 MΩ yields approx. 3.267 V. Actual probes also include capacitance, and high-frequency loading may be more significant than DC input resistance.

Current mode measures via a shunt resistor, introducing a burden voltage into the branch. Assuming a total equivalent series resistance of 2 Ω and measuring 100 mA, an additional 0.2 V is lost, which may cause low-voltage equipment to change its operating state. When measuring sleep microamps and wireless pulses, consider range switching and peak capability; do not pursue the lowest range resolution exclusively.

Whether the AC mode measures AC components, AC+DC true RMS, and the effective frequency band and crest factor limits should be read from the instrument manual; not all AC modes provide true RMS for arbitrary waveforms. If the specification is ±(0.5% reading + 2 counts), for a 3.000 V reading with a least significant digit of 1 mV, the uncertainty under these conditions is ±17 mV, not ±2 mV.

Oscilloscope Bandwidth: Understand via Edge and Amplitude Error

Bandwidth is typically defined at a specific attenuation point, not as a hard cutoff where signals disappear completely. For a single-pole approximation, the sine wave amplitude at the −3 dB bandwidth is only about 70.7%; the fundamental frequency of a square wave does not represent the bandwidth required for its edges. "Bandwidth greater than five times the clock frequency" serves as a rough heuristic but does not guarantee accurate measurement of any rise time.

For applicable response models, the instrument's rise time can be approximated as tr≈0.35/BW; a common approximation for combined systems is tr_measured² ≈ tr_signal² + tr_instrument². A real 2 ns edge paired with a 100 MHz oscilloscope (approx. 3.5 ns rise time) system might measure approx. 4.03 ns. The coefficients and square-sum relationships depend on the response model; ultimately, combine probe and instrument specifications. Keysight: The Impact of Probing on Measurement Accuracy

Sample rate determines discrete time resolution capability; memory depth roughly satisfies record length = samples / sample rate; 1 Mpoint at 1 Gsample/s yields only 1 ms. During long-duration observations, the instrument may reduce the sample rate; multi-channel modes may also share resources. The Nyquist condition also relies on the signal being band-limited; do not treat double the sample rate as a guarantee to observe all narrow glitches.

Probes: Confirm Reference Relationship, Then Compensate and Connect

The BNC shell, probe reference lead, and references for multiple channels of a standard benchtop oscilloscope are typically grounded together and connected to protective earth. Clipping the ground lead to a switching node alters the circuit and may cause a short circuit; subtracting two channels mathematically does not make them isolated inputs. Floating measurements should use appropriately rated differential or isolated measurement solutions; verify differential mode, common mode, frequency derating, and measurement environment. Do not attempt floating measurements by disconnecting the oscilloscope's protective earth. Tektronix: Probes and Floating Measurements

10× passive probes typically offer lower input capacitance and higher bandwidth than their corresponding 1× modes, but they attenuate small signals; low-amplitude ripple or high-impedance nodes may require other probing solutions. Compensation adjustment matches the input time constant; it is not a complete instrument calibration, and probe and channel attenuation settings must also be consistent.

10 pF has a capacitive reactance of approx. 1.59 kΩ at 10 MHz, which is sufficient to disturb some high-impedance circuits. Crystal oscillators are particularly sensitive; prioritize measuring available buffered clock outputs. Probing crystal nodes directly requires appropriately low-loading probes; there is no essential difference between probing a crystal pin versus a pad with the same net load capacitor. Short reference connections reduce noise pickup and parasitic inductance, but they must still be connected to the correct reference point.

Does attaching a probe change the signal?

With Rs=10 kΩ, increase probe capacitance from 10 pF to 100 pF. DC barely changes while high-frequency amplitude and edges do. Increasing Rs makes the same probe more intrusive. Rise time runs from 10% to 90% of the final amplitude, not fixed 0.1 V and 0.9 V thresholds.

$$H(s)=\frac{R_p}{R_s+R_p}\frac{1}{1+s(R_s\parallel R_p)C_p},t_{10\to 90}=\tau \ln 9$$

Organizing a Capture for Intermittent Resets

You can use the reset edge as the trigger and retain a sufficient pre-trigger window to observe the power supply. If the power supply sags first, followed by the reset, continue to cross-reference with load current and undervoltage thresholds; if the power supply is stable, check the watchdog, software, and other reset sources. Triggering aligns the record to the event; it does not mean "the waveform only exists from this point," nor does it alone prove the cause.

DC coupling allows simultaneous observation of operating voltage and transients. AC coupling can conveniently observe ripple in certain ranges, but it is a high-pass filter that alters the shape of low-frequency sags and recovery; therefore, AC coupling is not mandatory for ripple measurement. Settings such as 20 MHz bandwidth limiting should be declared according to the measurement objective: it can reduce high-frequency noise but may also filter out spikes that need to be observed.

Measuring switching losses requires synchronous measurement and integration of VDS and drain current, p(t)=v(t)i(t); measuring VGS and VDS simultaneously is insufficient. Time delay mismatch between voltage and current channels amplifies power error on fast edges; probe delay, zeroing, and measurement boundaries must be calibrated.

Logic Analyzer: Successful Decoding Does Not Guarantee Sufficient Electrical Margin

The analyzer quantizes inputs into logic states based on thresholds, making it suitable for observing the sequence, response, and cross-bus correlations of UART/I²C/SPI. It does not see the complete analog waveform; slow rise times, overshoot, or critical levels may still exist even when the decoded result appears normal; some mixed-signal devices have additional analog sampling channels, so distinguish their capabilities.

The sampling frequency must be able to resolve the shortest high/low pulse widths, not just exceed the nominal bus bit rate; protocol mode, bit order, address representation, chip select polarity, and thresholds must also be correct. First, verify the decoding configuration against a short segment of raw edges, then use long records to track anomalies, avoiding mistaking decoder configuration errors for device errors.

DC Power Supplies and Signal Generators Are Not Ideal Sources

The power supply maintains the set voltage in the CV (Constant Voltage) region and enters CC (Constant Current) mode, lowering the output voltage when the current limit condition is reached. If the inrush current is insufficient, the board may repeatedly reset due to undervoltage; do not assume a short circuit on the board just because CC mode is engaged, nor increase the current limit without analysis. For the first low-voltage power-up on the board, configure limits based on normal and startup requirements of the design; confirm polarity and wiring before observing the status.

Current limiting has a response time, and output capacitors may discharge energy before the control loop reacts; reverse current absorption capability may not exist. When measuring load transients, also include cable voltage drop, remote sensing, and electronic load dynamics in the error budget; the voltage displayed on the power supply panel may not equal the voltage at the board.

Many signal generators have a 50 Ω output impedance. If the displayed amplitude is defined for a matched 50 Ω load, but the actual connection is a high-impedance input, the voltage will be approximately twice the displayed value; if the display load has already been set to High-Z, do not multiply by two again. The load setting usually changes the display conversion but does not necessarily change the actual source impedance; also verify Vpp/Vrms and DC bias. Keysight: Output Termination and Amplitude

Soldering and ESD Tools: Select Based on Package and Process

Soldering temperature and dwell time should be adjusted according to the solder, pad thermal capacity, iron tip, and component process limits; do not uniformly specify 1–3 s for all solder joints. Exposed pins can be handled with appropriate hand-soldering techniques; QFN bottom pads and BGA joints cannot be proven to be well-connected solely by drag soldering or appearance; hot air, preheating, or reflow solutions depend on the package and rework conditions; not all situations have only one tool available.

Post-soldering visual inspection, short-circuit checks, supply current, and functional verification provide different evidence; normal continuity does not prove the quality of hidden joints. ESD wrist straps, table mats, and storage packaging should form a verified electrostatic control path; temporarily touching a metal object cannot replace continuous protection. The electrostatic grounding solutions here cannot be used as operational protection methods for live high-energy circuits.

Recalculating Instrument Impact

from math import isclose, sqrt, pi
assert isclose(3.3*1e6/(1e5+1e6), 3.0)
assert isclose(3.3*10e6/(1e5+10e6), 3.267326732673267)
assert isclose(0.1*2, 0.2)
assert isclose(3*0.005+2*0.001, 0.017)
assert 4.03 < sqrt(2**2+3.5**2) < 4.04
assert isclose(1e6/1e9, 0.001)
assert 1591 < 1/(2*pi*10e6*10e-12) < 1592
assert isclose(2*50/(50+50), 1)
print('Input loading, burden voltage, reading limits, bandwidth, record length, and termination budget passed')

These calculations quantify instrument loading and readout limits. Record the probe, range, bandwidth, reference connection and calibration status in actual measurements.

Continue Reading

• PCB Design Fundamentals: Return paths, test points, and layout.
• Debug and Programming Interfaces: Firmware download, breakpoints, and recovery entry points.