Dielectric Loss and Tan-δ Measurement

A perfect capacitor is a perfect energy store: it draws a current that leads the voltage by exactly \(90^\circ\), takes energy in during one quarter-cycle and returns all of it in the next, and consumes no power. Real insulation falls short of this ideal. Inside a real dielectric, charges do not respond instantly and without friction — molecular dipoles rub as they reorient, a little leakage current trickles through, and any moisture, contamination or void adds its own dissipation. The result is that the current leads the voltage by slightly less than \(90^\circ\), and the small shortfall corresponds to real power turned into heat.

That heat is the dielectric loss, and the angle of the shortfall is the key to measuring it. The beauty of the idea is that the loss is a property of the whole insulation system, integrated over its entire volume — so a single measurement of that angle reports on the average condition of all the insulation at once. Where the partial-discharge test of Chapter 19 hunts for a localised defect, the loss measurement takes the insulation's overall temperature, as it were, in one number.

Resolve the current a real dielectric draws into two parts. The large quadrature (capacitive) component \(I_C\) leads the voltage by \(90^\circ\) and carries the stored, recoverable energy. The small in-phase (resistive) component \(I_R\) is aligned with the voltage and carries the lost energy. The total current leads \(V\) by the phase angle \(\varphi\), which is just short of \(90^\circ\); the small remainder, \(\delta = 90^\circ - \varphi\), is the loss angle. The dissipation factor is its tangent — the ratio of the loss current to the capacitive current:

A closely related figure is the power factor, \(\cos\varphi = \sin\delta\). For the very small loss angles of good insulation — where \(\delta\) is a fraction of a degree — \(\sin\delta\), \(\tan\delta\) and \(\delta\) (in radians) are all nearly equal, so the dissipation factor and the power factor are practically the same number. Engineers use both names; "tan δ" and "power factor" and "dissipation factor" all point at the same small quantity, the signature of a lossy dielectric.

To compute and measure \(\tan\delta\), the lossy dielectric is represented by an ideal capacitor with a resistor attached. Two equivalent forms exist. In the parallel model, a leakage resistance \(R_p\) sits across the capacitance \(C\); the loss current is \(V/R_p\) and the capacitive current is \(V\omega C\), so

In the series model, a resistance \(R_s\) sits in series with \(C\); the loss voltage is \(IR_s\) and the capacitive voltage is \(I/\omega C\), giving \(\tan\delta = \omega C R_s\). Both descriptions are equally valid — they yield the same \(\tan\delta\) and the same loss at a chosen frequency, differing only in how the resistance is connected. The series form is natural when the loss behaves like a resistance in the current path; the parallel form suits a leaky dielectric. Notice the opposite frequency dependence: parallel \(\tan\delta\) falls with frequency, series \(\tan\delta\) rises — a reminder that a single \(\tan\delta\) figure always belongs to a stated frequency.

The power dissipated follows directly. The real power is \(P = V I\cos\varphi = V I_R\), and since \(I_R = I_C\tan\delta = V\omega C\tan\delta\), the dielectric loss is

This little equation carries two warnings. First, the loss rises with the square of the voltage, so high-voltage insulation dissipates far more than its low-voltage behaviour would suggest, and that heat must escape or the insulation warms. Second — and more dangerous — a warmer dielectric is often a lossier one: tan δ tends to rise with temperature, which raises the loss, which raises the temperature further. If this feedback runs away, the result is thermal breakdown, met in the solid-dielectric chapter, where the insulation cooks itself to destruction. A low tan δ is therefore not just a sign of quality but a guard against thermal instability, which is the deeper reason it is watched so closely.

What makes \(\tan\delta\) so useful is that it is small and stable for healthy insulation and grows whenever something goes wrong. Clean, dry, well-made insulation has a very low loss tangent — a few thousandths or less. Let it absorb moisture, gather contamination, or age and oxidise, and the leakage and friction rise, lifting \(\tan\delta\). A single measurement compared against the standard's limit, or against the object's own history, is therefore a sensitive verdict on overall condition — and because it integrates over the whole volume, it catches diffuse, distributed deterioration that a localised test might miss.

An especially powerful refinement is to measure \(\tan\delta\) as a function of voltage. For sound insulation the loss tangent is essentially flat — the same at low and high voltage. But if voids begin to ionise (the partial discharge of Chapter 19), each discharge dissipates energy, adding loss that switches on only above the inception voltage. The loss tangent then tips up with rising voltage, and the increase \(\Delta\tan\delta\) between two voltages — the tip-up — is a direct measure of internal ionisation. This ties the two great diagnostic tests together: PD and tan δ are two windows onto the same deterioration, one counting the discharges, the other weighing the heat they shed.

Measuring a loss angle of a fraction of a degree demands a precise instrument, and the classic one is the Schering bridge — a four-arm AC bridge that balances the unknown lossy capacitance against a near-perfect standard. The four arms are: the test object (its capacitance \(C_x\) with its loss); a loss-free standard capacitor \(C_s\); a fixed resistor \(R_3\); and a resistor \(R_4\) in parallel with a variable capacitor \(C_4\). A null detector bridges the two midpoints, and the operator adjusts \(R_3\) and \(C_4\) until it reads zero.

At balance the products of opposite arms are equal, and separating the real and imaginary parts gives a wonderfully clean pair of results. The capacitance comes from the resistive ratio, and — the prize — the dissipation factor depends only on the single \(R_4 C_4\) arm:

Putting the bridge to work at high voltage takes care, because two of its arms — the test object and the standard capacitor — sit at full test potential while the operator must adjust \(R_3\) and \(C_4\) by hand. The standard arrangement places the two capacitive arms at the top, at high voltage, and the two adjustable resistive/capacitive arms at the bottom, near earth, where the detector and the controls are safe to touch; the supply is applied across the high-voltage diagonal. The standard capacitor itself is a compressed-gas (often \(\mathrm{SF_6}\) or nitrogen) unit whose loss is utterly negligible, so it is the trustworthy zero-loss reference the whole measurement leans on.

One subtle enemy remains: stray capacitance from the bridge arms and leads to nearby earthed objects injects unwanted currents that corrupt the tiny loss measurement. The classic cure is the Wagner earth — an auxiliary arm that holds the detector's potential at earth so that the strays no longer flow through it — together with careful guarding and screening of the components. With these in place, the high-voltage Schering bridge resolves loss angles to a remarkable precision, which is why it remained the reference dielectric-loss instrument for the better part of a century.

The manual Schering bridge taught the principle, but in service it has largely been replaced by automatic capacitance-and-tan-δ test sets that balance electronically, read out both quantities on a display, and need no hand-adjustment. Many use a transformer-ratio-arm design for stability, and modern instruments measure \(\tan\delta\) at several voltages automatically to produce the tip-up curve in one sweep. Some now sweep frequency as well, since dielectric frequency response (also called frequency-domain spectroscopy) spreads the single \(\tan\delta\) number into a curve across frequency that separates moisture from ageing far more sharply than a single point can.

In the field, the same measurement is made on installed transformers, bushings, cables and rotating machines, often using the apparatus's own bushing tap as the connection point. Because \(\tan\delta\) integrates the whole insulation's health into one trendable number, repeated measurements over a machine's life — watching for the slow upward creep that signals moisture ingress or ageing — make it a cornerstone of the condition-monitoring strategy of Chapter 21. Together with partial discharge, it forms the two-handed grip by which the health of high-voltage insulation is judged: PD for the local fault, tan δ for the global condition.

For each item, first identify what it tests — the loss-angle idea, a model formula, the loss power, the diagnostic interpretation, or the Schering bridge — then apply it. Difficulty rises down the list.