Earthing, insulation and generator testing
An installation can be well built and still be dangerous: a poor earth connection or insulation that has slowly degraded can be neither seen nor smelled, it has to be measured. This topic brings together the three measurements that close a commissioning and underpin maintenance: earth resistance, insulation resistance and the testing of a generator.
01What is measured in an earthing system
An earthing system (grounding system) connects the metal frames of the installation to the ground through a protective conductor and an electrode: a driven rod, a mesh, a plate or a buried cable. What is measured is not the resistance of the metal, which is negligible, but that of the soil around the electrode: from there the fault current spreads out in all directions, and since the cross-section it flows through grows quickly with distance, almost all of the resistance lies in the first few metres around the electrode.
For a vertical rod of length L and radius a driven into homogeneous soil of resistivity ρ, the resistance is estimated with Dwight’s formula:
A 2 m rod, 16 mm in diameter (a = 0.008 m), in 100 Ω·m soil. The factor ρ / (2πL) = 100 / 12.57 = 7.958 Ω. The logarithm is ln(4 · 2 / 0.008) = ln 1000 = 6.908, and minus 1 gives 5.908. So R ≈ 7.958 · 5.908 = 47.0 Ω.
If the length is doubled (4 m), R drops to 26.3 Ω. If instead the diameter is doubled (32 mm), it drops only to 41.5 Ω. To improve an earth connection it pays to lengthen or multiply the electrodes, not to thicken them. Several rods in parallel help only if they are spaced apart by at least the length of one rod: closer than that, their dispersion zones overlap and the benefit is lost.
Soil resistivity
The resistivity ρ (in Ω·m) changes with soil type, moisture, temperature and dissolved salts: the same electrode can read twice as much in a drought as after rain. That is why the value is recorded together with the conditions of the day. It is measured separately, before designing the earthing, with the Wenner method: four spikes in a line and equally spaced by a distance a; the instrument injects current through the two outer ones and measures the voltage between the two inner ones.
| Type of ground | Resistivity (order of magnitude) |
|---|---|
| Wet clay, marshy ground | tens of Ω·m |
| Farmland soil, dry clay | tens to a few hundred Ω·m |
| Dry sand and gravel | hundreds to thousands of Ω·m |
| Rock, limestone ground | thousands of Ω·m |
In a TT system, when a fault to a frame occurs the current returns through the ground. The frame is left at a voltage RA · Ifault with respect to earth, and that voltage must not exceed the limit touch voltage. Since the protection is the RCD (residual current device), the condition is:
With UL = 24 V (the value that, as explained in Electrical protection, the AEA 90364 adopts) and a 30 mA RCD, the maximum resistance is 24 / 0.03 = 800 Ω; with a 300 mA one, 80 Ω. That is the mathematical floor, not the goal: the value required in each case is set by the regulations in force, and an earth of a few ohms remains the best guarantee that the system works if the RCD fails.
02The earth tester and the fall-of-potential method
The resistance of an earthing system cannot be measured with an ordinary ohmmeter: there is no “other end” to connect the second probe to. The earth tester solves this with two auxiliary spikes driven into the ground. Between the electrode under test E and the current spike C it injects a current I; with the potential spike P, placed between them, it measures the voltage V that this current produces between E and P. The resistance is calculated as R = V / I.
Two details of the instrument matter in practice. The current is alternating, not direct, because with direct current the electrolyte in the soil polarizes and the reading drifts; and its frequency is chosen to differ from the 50 Hz of the mains and its harmonics, so that the instrument’s filter rejects the stray currents that always flow through the ground. Four-terminal earth testers separate the current terminal from the potential terminal at E, which cancels the resistance of the test leads, important when the earth measures only a few ohms.
The 62 % rule
Around E and C there is a “zone of influence” where the ground potential changes quickly. If P falls inside one of them, the reading is shifted. In homogeneous soil, the effects of E and C cancel exactly when P is at (√5 − 1) / 2 ≈ 0.618 of the distance between E and C, measured from E: at that point the curve has its flat stretch and the measured R = RE. Hence the practical rule:
Real ground is not homogeneous. That is why three positions of P are measured: at 62 % and 10 percentage points on either side (52 % and 72 %). If the three readings agree within a few percent, the measurement is in the flat zone and is valid; the average is reported. If they differ widely, the current spike was too close: move it farther away and repeat. As a guide, C goes no closer than 5 to 10 times the largest dimension of the electrode: for a 2 m rod, 20 m to 30 m away is usual. P, C and E must be aligned and kept away from other buried metal (pipes, other electrodes), which bend the current.
C is driven 30 m from the rod. P goes at 0.62 · 30 = 18.6 m; the check positions at 0.52 · 30 = 15.6 m and 0.72 · 30 = 21.6 m. The earth tester reads 41.2 Ω – 42.0 Ω – 43.1 Ω.
The largest difference is 43.1 − 41.2 = 1.9 Ω, 4.5 % of the average. The average is (41.2 + 42.0 + 43.1) / 3 = 42.1 Ω: a valid measurement. Against the 800 Ω floor of the previous example with a 30 mA RCD, it complies by a wide margin; against 80 Ω with a 300 mA one, it also does. The 42.1 Ω is recorded together with the date, the ground condition (dry, wet) and the distances used.
Other ways to check an earth
When spikes cannot be driven —in a city, on concrete— there are the clamp-on earth tester (it measures the loop formed by the electrode and the rest of the earths in parallel, and is valid only if other earths close the circuit) and the fault-loop impedance measurement with an installation tester. Both complement the earth tester but do not replace it on a new earth. In the Argentine workplace, the result is entered in the protocol established by Res. SRT 900/2015 (measurement of earthing and continuity of frames).
If a fault occurs while you are working, the electrode and its conductor can become live, and the test leads carry that voltage to the auxiliary spikes. Do not measure during a thunderstorm or with a known fault in the installation, do not touch the spikes during the reading, and wear insulating footwear and gloves when the procedure requires it. To measure the electrode on its own, the test link (disconnectable terminal) must be opened: while it is open, the frames of the installation have no protection, so this is done with the installation de-energized, or with authorization and notice to its users.
03Insulation resistance and the megohmmeter
No insulator is perfect: when a voltage is applied, a small current always passes through. The insulation resistance is the ratio between that voltage and the leakage current, and is expressed in megohms (MΩ) or gigohms (GΩ). Over time, heat, moisture, dirt and vibration degrade it, and poor insulation ends in a fault to frame, a short circuit or an electric shock. Measuring it before that happens is predictive maintenance in its purest form: see Maintenance of electrical components.
The megohmmeter (insulation tester) generates a high direct voltage (250, 500, 1000, 2500 or 5000 V depending on the model), applies it between the conductor and earth, and measures the current on a scale reaching gigohms. It normally has three terminals: L (line), which goes to the conductor or winding; E (earth), which goes to the frame or casing; and G (guard), which diverts the surface leakage current —through dirt on an insulator— so that it does not add to the bulk measurement. Direct current is used because with alternating current the capacitive current of the cable or winding would also flow, which says nothing about the condition of the insulation.
Which test voltage to use
The test voltage is chosen according to the rated voltage of the circuit: it must be high enough to stress the insulation without damaging it. The reference values of IEC 60364-6, on which part 6 of the AEA 90364 (verification of installations) is based, are:
| Circuit | Test voltage (DC) | Reference minimum resistance |
|---|---|---|
| Safety extra-low voltage (SELV, PELV) | 250 V | 0.5 MΩ |
| Up to 500 V | 500 V | 1 MΩ |
| Above 500 V | 1000 V | 1 MΩ |
These are reference values, not a recipe: the minimum required in each installation is set by the regulations in force, and new insulation measures far more. For rotating machines, the criterion of the IEEE 43 standard is also used, which sets a minimum as a function of the rated voltage Un of the winding:
The polarization index
The current delivered by the megohmmeter is not constant: when the voltage is applied, first the capacitive charging current flows (dying out in seconds), then the absorption current (the insulation polarizes, decaying over several minutes) and, always, the conduction leakage current, which does not change. In good insulation the total current falls with time and the resistance read rises; in damp or dirty insulation leakage dominates and the reading hardly changes. The polarization index (PI) measures that difference, and it does not depend on temperature or on the size of the machine:
| PI | Insulation condition (indicative) |
|---|---|
| below 1 | Dangerous: do not energize. |
| 1 to 2 | Doubtful: dry, clean, repeat and follow the trend. |
| 2 to 4 | Good. |
| above 4 | Excellent. |
The IEEE 43 standard asks for a PI of at least 2.0 for thermal classes B, F and H. If the 1-minute resistance already exceeds about 5000 MΩ, the PI is no longer meaningful and is not required: the insulation is so good that there is hardly any absorption to measure.
Measured at 500 V between a winding and the casing: 85 MΩ at 1 minute and 190 MΩ at 10 minutes, with the motor at 28 °C. The PI is 190 / 85 = 2.24: it meets the 2.0 minimum.
Corrected to 40 °C: a 12 °C difference is 1.2 times the 10 °C step, so R40 = 85 · 0.51.2 = 85 · 0.435 = 37.0 MΩ. The minimum for a 0.38 kV motor is 0.38 + 1 = 1.38 MΩ. It is 27 times above it: approved, and the value is recorded to compare with the next test.
Before connecting it: installation de-energized, locked out and absence of voltage verified. Electronic equipment, drives, surge protectors and capacitors are disconnected or bridged, because they would be damaged by the test voltage or would falsify the measurement. During the reading do not touch the conductor or the terminals. When finished, the cable or winding remains charged like a capacitor: it is discharged to earth, for at least as long as the test lasted, before touching it or disconnecting the terminals. On a long cable the charge can be dangerous.
04Generating sets and generator testing
A generating set is an assembly of an internal-combustion engine (almost always diesel), a synchronous alternator, a speed governor, a voltage regulator (AVR), a control panel and, in installations with backup supply, a transfer board that connects it to the load when the mains fails. The engine sets the frequency, and the AVR controls the alternator’s field current to hold the voltage. Frequency depends only on speed and on the number of pole pairs p:
| What is checked | With what | What is verified |
|---|---|---|
| Voltage and frequency | True-RMS voltmeter and frequency meter | At no load and at each load step: stability, and how much the frequency drops when taking load. |
| Current and power | Clamp ammeter or power analyzer | Current per phase, kW, kVA and cos φ; balance between phases. |
| Phase sequence | Phase sequence indicator | That it matches the mains before closing the transfer. |
| Insulation and windings | Megohmmeter and low-resistance ohmmeter | Insulation and PI of stator and rotor; resistance of each winding in direct current. |
| Engine parameters | Panel instruments | Oil pressure, coolant temperature, running hours, fuel level. |
| Transfer times | Stopwatch or recorder | Time to start, to stabilize and to take the load. |
The alternator tests
The test method for alternators is standardized (IEC 60034-4 describes how to obtain their parameters from measurements). The basic tests are four:
- Winding resistance in direct current, with the alternator cold. The temperature is noted, because the resistance of copper grows with it. This gives the armature resistance Ra.
- Insulation with a megohmmeter, between each phase and frame and between phases, and the PI (section 03).
- Open-circuit test: the alternator is turned at rated speed with the terminals open and the field current Iexc is raised in steps, noting the voltage E0 at each one. The result is the open-circuit characteristic: it first rises in a straight line (the air-gap line) and then saturates because the iron cannot carry more flux.
- Short-circuit test: at the same speed, with the three terminals shorted through clamp ammeters, the excitation is raised from zero until the short-circuit current Isc reaches the rated value, at most slightly more. The characteristic is a straight line, because with the armature shorted the iron works unsaturated.
The two tests give the synchronous impedance per phase, which summarizes armature reaction and leakage reactance. At the same excitation, take the per-phase voltage from the air-gap line and the short-circuit current. Since Ra is much smaller, the reactance is almost the same:
The load test
To see how it really behaves, the set is loaded in steps (for example 25, 50, 75 and 100 % of its power) with a load bank: resistors, and also inductors if the 0.8 power factor on the nameplate is wanted. At each step, once it has stabilized, voltage, frequency, current, power, temperatures and fuel consumption are recorded. The voltage regulation compares the no-load voltage with the full-load voltage:
Three-phase alternator, 400 V, 80 kVA, 50 Hz and cos φ = 0.8. The phase voltage is 400 / √3 = 230.9 V and the rated current is 80 000 / (√3 · 400) = 115 A. The resistance of one winding, measured in direct current, is Ra = 0.05 Ω.
In short circuit, 1.5 A of excitation is needed to circulate 115 A; at that excitation, the air-gap line gives E0 = 110 V per phase. Then Zs = 110 / 115 = 0.957 Ω and Xs = √(0.957² − 0.05²) = 0.955 Ω.
At full load, with cos φ = 0.8 lagging, the current lags the voltage by 36.9°: I = 115 · (0.8 − j 0.6) = 92 − j 69 A. The internal voltage is E0 = U + (Ra + j Xs) · I = 230.9 + 4.6 + 65.9 + j (−3.5 + 87.9) = 301.5 + j 84.4 V, with magnitude 313 V. The regulation is (313 − 230.9) / 230.9 ≈ 36 %. It is a pessimistic value (because the unsaturated impedance is used) and in a real set the AVR corrects it by raising the excitation, but it shows why, without a regulator, the voltage falls so much on loading.
Backfeed: a set connected to an installation that is still tied to the utility network can energize the street cables and kill a lineman who believes they are dead. The connection always goes through a transfer switch with mechanical interlock, never through a plug or a jumper. Carbon monoxide: the exhaust is not tested in closed rooms. Short-circuit test: it is done with the planned protections, for a few seconds, taking care not to exceed the current the winding can carry. Also: rotating parts, hot surfaces, fuel and noise.
05Logic controllers, sensors and actuators in measurements
On a test bench or in a generating set, much of what is measured is measured by a programmable logic controller (PLC). It does not replace the laboratory instrument, but it does what an instrument does not: it reads continuously, compares against limits, decides and acts. Its basic functions are four: reading inputs (digital and analog), running a cyclic program (read, solve, write, and start again), timing and counting (running hours, number of starts) and driving outputs. How the unit is structured is explained in Structure of a programmable controller and how it is programmed in PLC programming and PLC automation.
What feeds the PLC are sensors and what carries out its orders are actuators; Sensors, actuators and transducers covers them in detail. Applied to the measurements of this topic:
| Quantity | Sensor | Signal reaching the PLC |
|---|---|---|
| Current | Current transformer with transducer | 4–20 mA proportional to the current |
| Winding temperature | Pt100 (100 Ω at 0 °C; 138.5 Ω at 100 °C) | Resistance, through an RTD input module |
| Speed | Inductive sensor on a gear tooth, or encoder | Pulses: the PLC counts them and calculates rpm |
| Oil pressure | Pressure transmitter | 4–20 mA |
| Mains condition | Voltage and phase-sequence monitoring relay | Contact: yes / no |
Typical actuators are contactors that connect the load bank step by step, solenoid valves that cut the fuel, alarm and start relays, and drives or servomotors when something has to be moved accurately.
An oil-pressure transmitter measures 0 to 10 bar and delivers 4 mA at 0 bar and 20 mA at 10 bar: a 16 mA span. The PLC reads 12 mA. Pressure = (12 − 4) / 16 · 10 = 5 bar. If the reading falls below 4 mA (say, 1 mA), the most likely cause is not “negative pressure”: it is a broken wire, and the program must treat it as a sensor fault, not as zero pressure.
Introduction to servo systems
A servo system is a feedback system whose goal is for a quantity —the position or speed of a shaft— to follow a setpoint. A sensor (encoder, resolver or tachometer) measures the real value and subtracts it from the setpoint; that difference is the error, and the controller, generally a PID (see PID control), turns it into a command for the drive (a servo amplifier and a servomotor). In open loop the command is sent and assumed to be carried out; in closed loop it is checked and corrected, which is why the servo holds position even when the load changes. In measurements, a servo system positions a probe, a test rig or the brake that loads the motor under test.
06In the laboratory
Materials: earth tester, two auxiliary spikes, two cable reels of at least 30 m, tape measure, insulating gloves, data sheet. Steps: with the teacher, open the test link of the electrode; drive C at 30 m; drive P at 18.6 m, in line; measure. Repeat with P at 15.6 m and at 21.6 m. Record the date, the ground condition and the distances. What is measured: three readings. Acceptance criterion: the three agree within 10 % of each other (if not, move C farther away and repeat) and the average meets the condition RA · IΔn ≤ UL with the installation’s RCD, in addition to the regulations in force.
Materials: 500 V megohmmeter, 380 V three-phase motor with the terminal box open, thermometer, stopwatch, data sheet. Steps: de-energize and verify absence of voltage, open the links in the terminal block, connect L to the winding and E to the casing, apply the voltage and read at 15 s, 30 s, 1 min and 10 min; discharge. Repeat phase by phase. What is measured: R1 min, R10 min and the temperature. Acceptance criterion: R1 min corrected to 40 °C greater than 0.38 + 1 MΩ and PI ≥ 2.0 (or R1 min above 5000 MΩ). Plot R against time and compare with Figure 2.
Materials: laboratory motor–alternator set with adjustable excitation, tachometer, voltmeter, three clamp ammeters, load bank. Steps: measure Ra in direct current with the equipment cold. At no load and rated speed, raise the excitation in 6 steps and record E0. Bring it back to zero, short the terminals with the planned protections and raise the excitation up to rated current, in 3 points. What is measured: E0(Iexc), Isc(Iexc), Ra. Acceptance criterion: the short-circuit curve is a straight line (the scatter of the points does not exceed the instrument error) and the Zs calculated with two different excitations agrees within 5 %. If no alternator is available, the teacher provides a data table and the plotting and calculation are done with it.
Materials: PLC with a 4–20 mA analog input, loop calibrator or adjustable source, pilot lamp. Steps: program the conversion to engineering units of a 0 to 10 bar transmitter, a comparator with an alarm above 8 bar and another below 2 bar, and a “sensor fault” condition below 3.6 mA. Inject 4, 8, 12, 16 and 20 mA and 1 mA. What is measured: the calculated value at each point. Acceptance criterion: the five values match the formula with an error below 1 % of full scale, and at 1 mA the sensor-fault flag is activated and not the low-pressure alarm.
07Common mistakes
| Mistake | Consequence |
|---|---|
| Driving the potential spike by eye or over other buried metal | The reading falls outside the flat zone or is bent by the pipe. A bad earth is accepted, or a good one rejected. |
| Leaving the current spike too close | The zones of influence overlap, there is no flat stretch and the three readings differ. It has to be moved away and the test repeated. |
| Measuring the earth only once and at any time of year | The value changes with moisture. Without the ground condition recorded, it cannot be compared with the next measurement. |
| Measuring with the electrode connected to the installation and without opening the link | Parallel earths and the neutral change the result. The whole assembly is measured, not the electrode. |
| Using the megohmmeter without disconnecting electronics or capacitors | The equipment is damaged or the reading is false. |
| Reading the insulation as soon as the voltage is applied | The capacitive charging current is measured: it gives a low value that says nothing about condition. Wait at least 1 minute. |
| Comparing readings taken at different temperatures | Resistance halves every 10 °C: sound insulation looks bad when hot. |
| Forgetting to discharge the cable or winding after the test | It stays charged and discharges into whoever touches it. |
| Connecting a generating set to the installation without a transfer switch | Backfeed to the network: risk of death for the utility’s personnel. |
| Calculating regulation with Zs as if it were exact | It is an unsaturated value: it gives a higher regulation than the real one. It is checked in the load test. |
08Self-assessment
Why is the resistance of a rod concentrated near the electrode, and what should be done to lower it?
Because the current spreads in all directions and the cross-section it flows through grows with distance: almost all of the resistance is in the first few metres. Lengthen the rod or place several well spaced apart; thickening it barely helps (in the example, from 47 Ω to 41.5 Ω, against 26.3 Ω when the length is doubled).
Where is the potential spike driven and why there?
On the E–C line, at about 62 % of the E–C distance measured from E. There the effects of E and C cancel and the R–distance curve is flat, so the reading is the true value. It is checked by also measuring at 52 % and 72 %.
Three readings give 55 Ω, 56 Ω and 70 Ω. What is concluded?
The third departs widely from the other two: P was no longer in the flat stretch or C is too close. The measurement is not valid: move C farther away and repeat. The readings are not averaged.
Why does the megohmmeter use direct voltage, and why is it read at 1 and at 10 minutes?
With direct current, once the capacitance has charged only the leakage current remains, which is the one that tells about the condition of the insulation. The reading rises with time because the charging and absorption currents decay; the ratio between the 10-minute and the 1-minute reading is the polarization index, which does not depend on temperature.
A 0.38 kV motor measures 40 MΩ at 1 minute and 44 MΩ at 10 minutes. How is it rated?
PI = 44 / 40 = 1.1: doubtful. The absolute value exceeds the 1.38 MΩ minimum, but the low PI points to moisture or dirt. Clean, dry and repeat before putting it into service, and follow the trend.
What do the open-circuit and short-circuit tests give, and why is the calculated regulation pessimistic?
They give the synchronous impedance: Zs = E0 / Isc at the same excitation, and with it the internal voltage at full load. It is pessimistic because the unsaturated value from the air-gap line is used; under real load the iron saturates and the effective impedance is lower.
A PLC reads 16 mA from a 0–10 bar transmitter. How many bar is that, and what about 2 mA?
16 mA: (16 − 4) / 16 · 10 = 7.5 bar. At 2 mA, below 4 mA, the value has no physical meaning: it is a sensor or wiring fault (broken wire) and the program must flag it.