Identify thermoelectric voltage separately from resistive voltage
A low-resistance observation can contain a voltage contribution that is not produced by resistance under the applied test current. Keithley describes thermoelectric EMFs associated with dissimilar conductor junctions and temperature differences within the circuit. The physical cause is the junction and thermal condition. Treating every observed voltage as the resistive response would combine that separate source with the quantity the measurement intends to characterize.
A stable lead-resistance contribution and a thermoelectric voltage consequently have different meanings. The former belongs to a current-carrying path; the latter is an additional voltage associated with the thermal arrangement. Separating current and sense connections can address the contribution of connection resistance without making the thermal source vanish. A four-terminal label therefore cannot serve as a complete account of the voltage present in the measurement circuit.
The handbook also relates these thermal voltages to changes around sensitive circuitry. That relationship explains why a record’s thermal context matters even when its connection topology is known. It does not identify the size of an offset in an existing setup or establish which junction dominates. The useful technical distinction is between the desired current-related signal and a thermally generated voltage, without a coefficient table, assembly prescription or numerical error calculation.
Retain offset drift relative to the measurement cycle
The behavior of an offset over time is another independent part of its meaning. A thermal voltage that remains effectively constant during a measurement cycle presents a different condition from one changing while the observations are taken. Keithley retains that distinction when discussing current-reversal compensation. The method name does not remove the need to consider how the offset behaves over the relevant measurement interval.
The handbook connects the effectiveness of reversal with the instrument response being fast relative to the circuit’s thermal change. If the thermal condition changes appreciably during the cycle, the observations no longer contain the same offset contribution. The compensation premise has changed. This is a temporal relationship between the measurement and the unwanted source, rather than a universal statement that changing polarity cancels every thermal error.
Keithley discusses delta compensation separately for changing thermoelectric voltages. The important distinction here is the retained temporal premise, not a sequence to reproduce. A record identifying a compensation method should not be interpreted as unconditional proof of removed drift. Constant and changing offsets remain different cases even when they originate from the same thermal mechanism. No instrument setting, algorithm, measurement order or achieved cancellation result is inferred from this explanation.
Separate a non-ohmic contact from a fixed series resistance
A connection can also behave differently from a fixed series resistance. Keithley describes a non-ohmic contact through a potential difference that is not linearly proportional to the current passing through it. Oxide films or other nonlinear connections can introduce that behavior. The issue is the shape of the electrical relationship, rather than simply an additional constant resistance in an otherwise unchanged measurement path.
The handbook further explains that such a nonlinear connection can rectify radio-frequency energy present in the circuit, creating a DC offset. An observation interpreted as a resistance-related voltage can therefore include a contribution produced by unwanted high-frequency energy acting through a nonlinear contact. This is different from the thermoelectric account, whose source is the material-junction and temperature condition rather than rectification of picked-up RF.
Keeping these paths separate makes a resistance record more precise without turning it into a diagnosis. An offset does not uniquely identify contact nonlinearity, and current dependence does not automatically establish a failed calibration. The technical point is why a fixed lead-resistance model may be incomplete when the contact relationship itself is nonlinear. No cleaning procedure, RF suppression circuit, contact test or finding about the condition of particular hardware is supplied here.
Recognize test-current heating as a change of DUT state
Test current can change the subject being measured rather than merely add an external voltage contribution. Keithley’s device-heating discussion identifies dissipation in a temperature-sensitive device as a reason its resistance can change during measurement. In that case the thermal condition of the device under test is part of the resistance observation. The source of heat belongs to the measurement interaction itself, not only to the surrounding environment.
This differs from thermoelectric EMF and from nonlinear-contact rectification. Those mechanisms add or alter voltage contributions associated with junctions or interference paths. Self-heating changes the device’s own resistance through its temperature sensitivity. A consistent connection arrangement cannot make these physical responsibilities identical, and a resistance value recorded under one measurement condition need not represent a thermally unchanged state under another condition.
The distinction explains why test-current context belongs with an existing record without selecting a current or declaring the record wrong. A changed observation could involve the device state, an offset or another mechanism, and the qualitative account does not choose among them for actual hardware. It supplies no power formula, pulse strategy, calibration result or temperature correction. The contribution is recognizing measurement-induced state change as separate from connection topology and external voltage errors.
Sources and further reading
Keithley Low Level Measurements Handbook, seventh edition: bounded physical116,135,138–139 prose. Equations, coefficients, compensation sequences, current selection, cleaning or wiring instructions, actual readings, calibration and diagnoses are excluded. Kelvin connection topology remains a separate related topic.