Unstable Micro-Ohm Readings: How to Find the Measurement Error

A bolted joint measures 46 micro-ohms, then 63, then 49. The connection has not been altered, yet each retest tells a different story. Before condemning the joint, establish whether the changing result follows the test setup or remains with the asset under controlled conditions.

Low-resistance testing is sensitive to where voltage is sensed, how test current flows and what happens while the reading settles. A four-wire instrument removes an important source of error, but it does not make every connection geometry or test condition equivalent.

This troubleshooting sequence applies to isolated, de-energized equipment under an approved test procedure. Verify absence of voltage and stored energy before connecting. Inductive equipment also needs a suitable test instrument and its specified discharge process.

First define the piece of resistance being measured

Draw two boundaries around the joint, contact or conductor section of interest. The potential leads determine the voltage-measurement boundary. If their contact points move between tests, the included conductor length and connection resistance can change.

For an isolated joint, the usual arrangement places the potential contacts on the test object inside the current contacts. Use the terminal arrangement in the instrument manual. Do not clip a sense lead onto a current clip and assume it senses the same metal surface as a separate contact.

Inspect for oxide, paint, grease, loose probe pressure and unstable access. Prepare contact points only in a way permitted for the asset; removing protective plating or altering the joint is not a neutral measurement step. Mark the positions so another technician can repeat them.

The four-terminal principle separates current injection from voltage sensing. Megger's low-resistance Q&A explains how this avoids including current-lead voltage drops in the measured resistance.

Use repeatability as a diagnostic tool

Take several readings with the connections undisturbed, using the same current, duration and settling criterion. Then, after safely ending the test, remove and reconnect the leads at the marked positions and repeat.

If undisturbed readings are stable but reconnection changes them, contact placement or pressure deserves attention. If the value drifts even without touching the leads, look at heating, settling, interference and the instrument's current regulation. If a known reference also behaves poorly, examine the measurement chain before drawing conclusions about the asset.

Do not average incompatible conditions into a reassuring result. Preserve the sequence. A simple time-ordered record often reveals more than an average and a single final number.

The voltage signal can be very small

Ohm's law makes the sensitivity clear. In an illustrative test, a 50 micro-ohm resistance carrying 1 A produces only 50 microvolts. An unwanted 5-microvolt offset is one tenth of that signal. At 10 A, the same ideal resistance produces 500 microvolts, so the same offset represents a smaller fraction.

That calculation does not mean the highest available current is always the best choice. Increasing current also increases heating: power is I²R. Contact behavior, asset restrictions and the required procedure can limit the permissible current and duration.

Use the current specified for the task and record it. Results taken with different currents are not automatically interchangeable, particularly for contacts whose surface behavior changes with the applied test conditions. Small-signal contacts and high-current power contacts can require quite different methods.

Control temperature and thermoelectric offsets

A component brought from a hot plant room into a cooler test area may continue changing temperature during measurement. Leads warmed by a hand, dissimilar-metal junctions and uneven heating can also introduce small thermoelectric voltages. Allow suitable stabilization and avoid unnecessary handling of measurement junctions during a sensitive reading.

Where the instrument and approved method support it, current reversal or offset compensation can reduce a stable unwanted voltage. The underlying relation is V+ = IR + E and V- = -IR + E, so R = (V+ - V-)/(2I) when the offset E remains stable between readings. Tektronix discusses the limitations of this method in its low-resistance measurement note.

Use the instrument's documented process. Do not improvise live lead reversal, especially on inductive equipment. Record whether compensation was enabled. A method used for one set of measurements should not silently change for the next.

Temperature correction needs the material, reference temperature and a valid method. A generic correction should not be applied indiscriminately to a mixed-metal joint or nonlinear contact. Preserve the uncorrected reading and measured temperature alongside any calculated value.

Parallel paths can hide the resistance you want

Four-wire measurement does not force all the source current through a particular joint. In an assembled system, current can divide through connected conductors, bonds or other branches. The instrument's displayed V/I can therefore describe an equivalent network rather than the isolated component.

Megger's parallel-resistance application note discusses this limitation. The practical implication is to understand the circuit before deciding that a low value proves a sound individual connection.

For a controlled teaching example, two equal 100-micro-ohm paths in parallel have an equivalent resistance of 50 micro-ohms. A 50-micro-ohm reading does not show that either individual path is 50 micro-ohms. The same reasoning explains why a bypass can conceal a deteriorating joint.

Where separation is required, use an approved isolation and restoration procedure. Do not remove protective bonds or circuit connections during a test simply to force agreement with a previous result.

Treat windings as a separate measurement problem

A winding can take time to reach a stable DC measurement condition. Its inductance, applied current, instrument output capability and internal connections influence the settling behavior. A handheld contact-resistance tester is not automatically suitable for every large transformer winding.

Follow the asset and instrument procedures for connection, stabilization, discharge and, where required, demagnetization. Do not disconnect leads while test current is flowing or while the instrument indicates stored energy. A changing winding result should be assessed against those requirements before it is labeled a contact problem.

For test-equipment selection, distinguish the required current from the current the instrument can sustain into the actual circuit. Ask about output compliance, inductive-load suitability, discharge provisions and the applicable resistance range.

Match the EK model to the test, not to the lowest displayed digit

The EK-C366A DC Resistance Tester lists automatic and manual current selection, with manual settings from 1 mA to 10 A, along with storage and data communication. Its selectable current is useful when the approved task calls for different measurement conditions. Confirm the range-specific specifications and lead arrangement against the current manual.

The EK-C369 Loop Resistance Tester lists a four-wire method and 50 A or 100 A settings for low-resistance contact applications. The C369 and C369C have different listed current selections; specify the exact model. A 100 A capability does not by itself establish that every breaker or connection must be tested at 100 A.

Ask for a test demonstration using a suitable reference and the intended lead set. Confirm how current, duration and results are saved. Resolution, accuracy, repeatability and suitability for an inductive load are separate specifications.

A practical repeatability record

Use one row for each reading, with columns for asset, marked sense points, test current, test duration, settling criterion, temperature, compensation mode, result and whether leads were reconnected. Add a sketch showing potential and current contacts.

A useful comparison includes an undisturbed sequence, a reconnection sequence and, when the procedure allows it, a reference check. Keep different currents in separate groups. Report unexplained variation instead of choosing the result nearest the expected value.

For example, “The range narrowed after the potential contacts were fixed at marked locations” supports a measurement-setup explanation. “The joint is good because one reading was low” does not. Final acceptance still depends on the asset requirement and the responsible reviewer.

When to stop repeating and investigate the asset

If the setup and reference checks are stable but the component remains inconsistent, preserve that evidence. Further work may involve an isolated inspection, comparison with equivalent phases or joints, review of mechanical condition, or an asset-specific test approved by the responsible engineer.

Do not retighten a connection between undocumented readings and present the result as if nothing changed. Record maintenance separately, then repeat the accepted method. This makes the before-and-after comparison useful and avoids losing the original fault evidence.

The aim is a result that another technician can reproduce under the same conditions. When requesting support, send the circuit boundary, asset type, instrument model, lead sketch, current, duration and the full reading sequence. That is enough to begin distinguishing an unstable measurement from an unstable connection.