CT Saturation and Relay Testing: What Steady-State Injection Hides

Most relay test records share a blind spot. The tests were run by injecting clean sinusoidal current from a test set straight into the relay’s current inputs. That proves the relay. It says nothing about the current transformer, because the current transformer was not in the loop.

That matters, because protection misoperations often trace back to the current measurement chain rather than to the relay’s logic — a CT that stopped reproducing its primary current faithfully at the moment it mattered most. This article covers what saturation is, what decides it, what it does to each protection function, why standard testing hides it, and what actually exposes it.

Key takeaways

• Saturation is a flux limit. Past the knee of the excitation curve the secondary current becomes a distorted, reduced version of the primary current, typically spike-and-collapse in shape.
• AC saturation comes from high symmetrical current against the CT’s accuracy limit; transient saturation comes from the DC offset of an asymmetrical fault and can bite well below the symmetrical rating.
• Remanent flux from a previous fault or a DC resistance test moves saturation onset much earlier. Demagnetise after any DC test.
• The failure mode that hurts most is a false differential current when only one side’s CT saturates during a through-fault.
• Standard secondary injection cannot show any of this. COMTRADE playback, EMT-derived records and system-based testing can, within the limits of the record and the amplifier.

What saturation is, in flux terms

A current transformer induces a voltage in its secondary winding through a changing core flux. The secondary current is driven around the loop formed by the secondary winding resistance, the leads and the relay burden; the voltage needed to drive it comes from the rate of change of flux. Higher secondary current, or a larger burden, means more voltage, which means more flux.

The core can only carry so much. The excitation characteristic — secondary voltage against exciting current, conventionally on log axes — is nearly vertical at low levels and bends over at the knee. Beyond the knee, a large increase in exciting current produces only a small increase in flux. Exciting current is current drawn away from the secondary circuit, so as the core saturates the secondary current falls short of the ratio-scaled primary current and the waveform deforms, typically into a narrow spike near each zero crossing of the primary that collapses towards zero for the rest of the half cycle.

The relevant distinction for testing is what pushes the core there.

AC saturation. A high symmetrical fault current, with no significant DC component, demands a secondary voltage above what the CT can support into its actual burden. This is the case the accuracy limit covers: a steady-state condition, reasonably predictable from the CT’s rating and the calculated fault current.

DC or transient saturation. A fault initiated away from the voltage peak produces an asymmetrical current with a decaying DC offset whose time constant is set by the system X/R ratio. The DC component drives flux in one direction and does not reverse it every half cycle, so flux accumulates until the core saturates — which can happen at a symmetrical current well inside the CT’s nominal capability. On a high-X/R circuit the DC time constant can be many cycles, so the CT may spend the whole fault duration in and out of saturation. This is the case that surprises people, because the nameplate looks adequate.

Remanence

When a fault is interrupted, the flux in the core does not necessarily return to zero. Depending on where in the cycle the current was interrupted and on the core material, substantial residual flux can remain, biasing the core towards one polarity. If the next fault drives flux the same way, the available headroom is much smaller and saturation onset is much earlier.

The same mechanism is created deliberately, and accidentally, in the test bay. A DC winding-resistance measurement, or any test passing direct current through a CT winding, leaves the core strongly magnetised. So does an interrupted excitation test. Return that CT to service and the next fault may be measured by a core with a fraction of its designed headroom. Demagnetising — typically by applying an AC excitation slowly reduced to zero, following the applicable standard and the test-set manufacturer’s guidance — is the remedy, and it belongs in the procedure rather than in someone’s memory. Note that some CT classes bound remanence by design, through gapped cores, and some do not.

What decides whether a CT saturates

Parameter	Effect
Turns ratio	Sets secondary current for a given primary current; higher ratio means less required voltage
Knee-point voltage	The practical measure of flux headroom, defined by a specified exciting-current rise for a specified voltage rise
Secondary winding resistance	Part of the voltage the core must supply; larger on higher-ratio CTs
Burden, including leads	Often dominated by the loop resistance of long cable runs, not by the relay
Accuracy class and limit factor	Bounds error at a stated multiple of rated current into the rated burden
Transient dimensioning factor	Margin so the core can also absorb DC flux for the specified fault and duty cycle

Burden is the term most often wrong in practice. A numerical relay presents a very small burden compared with an electromechanical one, which sounds like good news, but the secondary lead resistance for a long run does not shrink. When a CT circuit is extended, or a relay is replaced and the wiring rerouted, redo the burden calculation.

The classes engineers actually meet

Under IEC 61869-2, protection current transformers are specified in families rather than by a single number. What each family is for:

• Class P guarantees a composite error not exceeding a stated percentage at a stated accuracy limit factor, into the rated burden, under steady-state symmetrical conditions. It says nothing about remanence and nothing about the DC transient. A “5P20” style designation belongs here.
• Class PR is a protection class with a limit on the remanence factor, achieved in practice with a small core gap. Use it where remanence-advanced saturation is unacceptable.
• Class PX (and PXR with a remanence limit) is the low-leakage-reactance class specified by its rated knee-point e.m.f. and excitation characteristic rather than by a composite error. It is used where the relay’s own dimensioning calculation needs the excitation curve, high-impedance busbar protection being the classic case.
• Transient classes TPX, TPY and TPZ are specified for duty cycles including the DC component, using a transient dimensioning factor and an error limit on the instantaneous rather than the RMS composite error. Broadly, TPX has no specified remanence limit, TPY limits remanence through a gapped core, and TPZ has a specified secondary time constant that strongly attenuates the DC component at the cost of accuracy for it. Where an application turns on the details of these definitions, read the standard rather than any summary, including this one.

Under IEEE C57.13, relaying accuracy is expressed as a C (or T) class with a secondary terminal voltage. A C-class rating states that ratio error will not exceed 10 percent for any symmetrical secondary current from one to twenty times rated, with the burden that produces the stated secondary voltage at twenty times rated. For a 5 A CT, twenty times rated is 100 A, so C100 corresponds to 100 V across a 1.0 ohm standard burden. The “C” indicates the ratio error can be calculated because leakage flux is negligible; “T” indicates it must be determined by test. IEEE C37.110 develops how these ratings apply to relaying, and IEC TR 61869-100 gives application guidance on the IEC side.

What saturation does to each protection function

Overcurrent (50/51). A saturated CT delivers a reduced RMS and fundamental current to the relay. On an inverse-time curve, a lower measured current means a longer operating time — sometimes much longer, because the curves are steep. In the worst case an instantaneous element set close to the available fault current does not pick up at all, and a delayed backup clears the fault. If an event produced a clearing time far longer than the curve predicts and the relay tested correctly afterwards, saturation is a prime suspect. The mathematics of the curves is a separate topic — see IEEE C37.112 curves explained.

Distance (21). The impedance the relay computes is a ratio of voltage to current. Distort the current and the computed impedance moves. Saturation is worst in the first cycles of a high-current close-in fault, which is precisely where Zone 1 has to make a fast decision. Depending on the filtering and the phase error introduced, the apparent impedance can move outward, delaying or preventing a Zone 1 trip, or inward. Steady-state boundary sweeps cannot show this; see distance relay reach testing.

Differential (87). This is the severe case. A differential element compares currents from both zone boundaries. During a heavy through-fault both CTs see the same primary current, but they may not be in the same condition: different ratios, different burdens (one panel a long cable run away), different remanent flux, different classes. If one saturates and the other does not, or one saturates earlier, the two secondary currents no longer match and the relay computes a large differential current with no fault behind it. The bias characteristic and the CT-saturation detection logic exist to hold the element stable through exactly this, and testing them requires distorted waveforms. See how to test a transformer differential relay.

Harmonic content. The spike-and-collapse waveshape is rich in harmonics, including even harmonics during the DC-driven phase. A transformer differential relay blocks on second-harmonic content to ride through inrush, so saturation-generated harmonics can interact with that logic — holding the element blocked when it should operate, or reaching the threshold when the current is not inrush at all. The interaction depends on the relay’s harmonic-measurement design, so verify it per relay rather than assuming a general rule.

Why standard testing hides all of it

This is the central point, and it deserves stating without qualification: a conventional secondary injection test does not include the current transformer. The test set’s amplifier connects directly to the relay’s current terminals and produces a clean sinusoid at the commanded magnitude and phase. No core is magnetised, no flux accumulates, no remanence exists, no distortion appears.

Every result on such a record is therefore a statement about the relay’s measuring elements, filters, logic and outputs — and nothing else. The record can be perfect while the installed protection cannot clear a close-in fault in the expected time, because the failure lives upstream of where the test starts.

Primary injection does include the CT, and proves ratio, polarity and the wiring path. But a portable primary injection set delivers a small fraction of fault level, symmetrically, so it verifies the chain without exercising the saturation regime either.

What actually exposes it

COMTRADE playback. A transient record — from a relay’s own disturbance recorder during a real event, or from a simulation — is replayed through the test set’s amplifiers into the relay. The record is a sampled time series of instantaneous voltage and current values in the format defined by IEEE/IEC C37.111: a configuration file declaring channels, scaling, sampling rates and timing, plus a data file of samples. The test set reconstructs each channel and drives its amplifiers to follow it. If the record contains the DC offset and the saturated waveshape, the relay sees them and responds, and its response can be compared with what actually happened. Playing back a record from a real misoperation is the most direct diagnostic available: if the relay reproduces its original behaviour, the waveform explains it rather than an intermittent fault or a setting change.

Its limits are real and belong in the report:

• The record bounds the content. A record captured at a modest sampling rate, or with anti-alias filtering, contains no harmonic content above its own bandwidth. Detail that was never sampled cannot be replayed.
• The amplifier bounds the reproduction. Following a steep, distorted, DC-offset waveform into a real burden demands bandwidth, slew rate and compliance voltage. Where the amplifier cannot keep up, the relay sees a smoothed version — which may be enough, or may hide the effect being chased.
• A record is one event. It carries the fault incidence angle, source impedance and remanence state that happened to exist. That is evidence, not coverage.

EMT-simulated records. An electromagnetic transient model of the circuit, with a CT model including the excitation characteristic and an initial remanent flux, generates records for conditions you have not experienced: worst-case incidence angle, worst-case X/R, worst-case remanence, reclose onto a fault. The result is only as good as the CT and source model, so the model parameters belong with the results.

System-based testing. A network model drives the test iteratively, with the relay’s trip output fed back into the simulation so that breaker operation and post-fault conditions follow from the relay’s own decisions. This tests scheme rather than element behaviour; IEEE C37.233 is the guide for how such testing is structured.

Practical guidance

Review the dimensioning before blaming the relay. When an operation looks slow, or a differential element operated for what appears to have been a through-fault, the first step is arithmetic, not injection. Take the calculated fault current and X/R at that location, the CT ratio, class and secondary resistance, and the actual burden including leads, and check whether the CT was ever going to reproduce that current faithfully.

Check the burden as installed. Measure the loop resistance of the secondary circuit rather than taking it from the design drawing. Wiring gets rerouted, terminals get added, panels move.

Treat CT tests as CT tests. Ratio testing and excitation (knee-point) testing are separate measurements with their own records. An excitation curve taken at commissioning is the baseline that makes a later curve interpretable — and taking one is how you discover that the CT you inherited is not the class the drawing claims. Demagnetise afterwards.

Document the CT data with the relay record. A misoperation review needs the CT ratio, class, knee point or C rating, secondary resistance, measured burden and the date of the last excitation test, in the same place as the relay results. Assembling that after an event, from three filing systems, is where investigations lose their first week. The PRC-005 evidence checklist covers what a defensible maintenance record needs.

Where software helps

The practical work is unglamorous: read the record, plot the channels, look at the waveshape, measure the harmonic content, compare the relay’s response against the record. Tooling that makes that quick changes how often it gets done.

GridAPM ProtectionAI reads and plots COMTRADE records, compares records against each other, and performs harmonic analysis on selected windows. Replay runs through its driver layer against the built-in simulator today, which is useful for checking that a record and a test plan are coherent before field time. Playback into physical amplifiers depends on the vendor drivers on the roadmap — ProtectionAI does not drive physical test sets, and no COMTRADE playback into a real relay happens through it today. See ProtectionAI capabilities for the boundary between built and planned, and relay testing software for how this sits alongside the test-set manufacturers’ own playback tools.