Electric Field Stress and Estimation

Picture two flat metal plates facing each other across a thin air gap, like the inside of a parallel-plate capacitor. Apply a voltage and the field between them is beautifully even — every field line runs straight from one plate to the other, and the stress at any point is just the voltage divided by the spacing. This is the world of Chapter 1's \(E = V/d\), and it is genuinely useful for a first estimate. But it describes an idealisation that real high-voltage equipment almost never matches.

The moment an electrode curves — and a conductor, a cable core, a bushing, a sphere or a busbar is nothing but curved surfaces — the field stops being uniform. Field lines crowd together where the surface is most sharply rounded and spread apart where it is flat. Because the local stress is set by how densely those lines pack, the field is no longer \(V/d\) but something larger near tight curvature and smaller in the open spaces. The whole game of high-voltage design is to find the single highest stress anywhere in the geometry — because that is the point that breaks down first — and \(V/d\) simply cannot see it.

So we need estimation methods that respect geometry. For a handful of clean, symmetric shapes we can solve the field exactly with pen and paper; those solutions are the backbone of this chapter, because they build the intuition every numerical tool later confirms. For the messy real shapes that resist analysis, we turn to the computer — the subject of Chapter 3. The two approaches are partners: the exact solutions tell us what to expect, and the numerical ones tell us exactly how much.

Estimation rests on the two relationships we already met. The field is the spatial gradient of potential, and the potential is what the boundaries (the electrodes) fix. Recover one from the other and the stress falls out:

In a charge-free insulating region the potential obeys Laplace's equation, \(\nabla^2 V = 0\); solving it subject to the electrode voltages gives \(V\) everywhere, and the gradient gives the stress. For arbitrary shapes that is hard — Chapter 3's territory. But when the geometry has a clean symmetry, a far shorter path opens up through Gauss's law: enclose the charged electrode in an imaginary surface that mirrors its symmetry, and the field comes out almost for free.

The recipe is always the same: choose a Gaussian surface on which \(E\) is constant and points straight through, read off \(E\) in terms of the enclosed charge \(Q\), then tie \(Q\) to the applied voltage \(V\) by integrating the field across the gap. Two geometries reward this trick completely — the cylinder and the sphere — and between them they cover an astonishing fraction of real high-voltage hardware: every cable, bushing, capacitor and sphere gap. We take them in turn.

Strip a single-core power cable and you find the most important geometry in the subject: a round central conductor of radius \(a\), a concentric metallic sheath of inner radius \(b\), and the insulation filling the space between. The same shape is a capacitor bushing, a coaxial test cell, a gas-insulated busbar. Solve it once and you have solved a thousand devices.

Wrap a cylindrical Gaussian surface of radius \(x\) and length \(\ell\) around the conductor (with \(a \le x \le b\)). By symmetry the field is purely radial and constant over that surface, so Gauss's law gives \(\varepsilon\,E(x)\,(2\pi x \ell) = q\ell\), where \(q\) is the charge per unit length. Hence the field falls off as \(1/x\):

To replace the awkward charge \(q\) with the measurable voltage \(V\), integrate the field from conductor to sheath:

Substitute back and the charge vanishes, leaving the field written entirely in terms of the applied voltage and the two radii. This single expression is the one to memorise:

Because the field varies as \(1/x\), it is largest where \(x\) is smallest — at the conductor surface, \(x = a\) — and smallest at the sheath, \(x = b\). These two bounds are worth carrying around:

The sphere is the cylinder's three-dimensional cousin, and it appears wherever an electrode is rounded into a ball — high-voltage terminals, sphere gaps used for measurement (Chapter 16), and the stress-relief electrodes on test equipment. Take an inner sphere of radius \(a\) inside a concentric outer sphere of radius \(b\). A spherical Gaussian surface of radius \(x\) now has area \(4\pi x^2\), so the field falls off faster — as \(1/x^2\):

Eliminating \(Q\) exactly as before gives the field and its maximum at the inner sphere:

The pattern repeats: the curved inner electrode again carries the highest stress, and the faster \(1/x^2\) decay makes the sphere's field even more lopsided than the cylinder's. The same Gauss-law machinery, the same lesson — only the geometry's exponent changes.

Both exact solutions whisper the same warning, and it is time to say it plainly: the tighter the curvature, the higher the stress. Isolate a single sphere of radius \(r\) far from anything and raise it to potential \(V\); its surface field is simply

Halve the radius and you double the surface stress for the same voltage. Shrink \(r\) toward zero — a sharp edge, a burr, a stray wire whisker, a contamination particle — and the local field rockets upward without the voltage rising at all. This is field enhancement, and it is the villain of practical high-voltage design. It explains why corona ignites first at thin conductors and pointed hardware, why a microscopic protrusion inside an insulator can seed a discharge, and why a scratch on a polished electrode can decide where a gap breaks down.

The cure follows straight from the cause: where you cannot avoid a high-voltage terminal, round it generously. Corona rings on insulator strings, toroidal stress shields on impulse generators, and the smooth radii machined onto every test electrode all exist to enlarge \(r\) and so pull \(E\) back down. Field estimation and field control are two sides of one coin: once you can predict where the stress concentrates, you know exactly where to add metal to relieve it.

We now have a way to find \(E_{\max}\) in several geometries, but a designer also wants a single, dimensionless verdict: how good is this electrode arrangement at using its insulation? A perfectly uniform gap stresses every cubic millimetre of dielectric equally and so wrings the most withstand out of the least material. A peaky, non-uniform gap wastes insulation — most of it idles while one small region does all the suffering. The number that captures this is the field-utilization factor, also called the Schwaiger factor or degree of uniformity:

Here \(d\) is the gap length, so \(E_{\text{mean}} = V/d\) is the stress the gap would have if it were uniform. A value \(\eta = 1\) means a perfectly even field; the smaller \(\eta\) becomes, the more sharply the stress peaks above the average and the more inefficient — and breakdown-prone — the geometry. For the coaxial cylinder, with \(d = b - a\), the factor follows directly from Section 2-3:

The "best-case" column gives the most uniform each geometry can be made; sphere gaps used for voltage measurement become steadily less uniform as the gap opens relative to the sphere diameter, which is exactly why their calibration tables (Chapter 16) are restricted to modest spacings. The utilization factor turns the vague wish for "a uniform field" into a number you can design toward.

Here is one of the most satisfying results in the whole subject. Suppose the sheath radius \(b\) and the voltage \(V\) are fixed — set, say, by the duct the cable must fit into — and you are free to choose the conductor radius \(a\). What \(a\) gives the smallest peak stress \(E_{\max}\), and so lets the insulation hold the most voltage?

It is tempting to assume a fat conductor is always gentler, but the algebra disagrees. Write \(E_{\max} = V/[a\ln(b/a)]\) and minimise it by maximising the denominator \(f(a) = a\ln(b/a)\). Differentiating,

The optimum is reached when the sheath-to-conductor ratio equals Euler's number, \(b/a = e \approx 2.718\). Substituting back, the lowest achievable peak stress is

The two failure modes either side of the minimum are equally instructive. Make the conductor too thin and the \(1/a\) crowding at its surface dominates; make it too thick and you squeeze the insulation into so narrow a gap that \(\ln(b/a)\) collapses. The sweet spot at \(b/a = e\) balances the two. Real cables cannot always sit exactly there — conductor size is also set by current rating and cost — but the result tells the designer which direction helps and roughly how much room there is to move.

Optimising \(a\) lowers the peak, but it does nothing about the underlying unfairness of the coaxial field: the inner insulation still works far harder than the outer, in the ratio \(b/a\). For high-voltage cables that imbalance becomes the limiting factor — the conductor-adjacent layer reaches its strength while the outer layers are barely stressed. Grading is the art of redistributing that stress so every layer pulls closer to its share.

Two classical schemes do this. Capacitance grading wraps the conductor in concentric layers of progressively lower permittivity moving outward. Since the radial flux \(D = \varepsilon E\) is continuous across the layers, a higher \(\varepsilon\) near the conductor pulls the local \(E\) down where it was worst; choosing the permittivities so that \(\varepsilon\,x\) stays constant would, in the ideal limit, flatten the stress to a single value across the whole insulation. Intersheath grading reaches the same goal differently: thin metallic foils are buried in the insulation and held at carefully chosen intermediate potentials, effectively splitting one steep coaxial field into several gentler ones stacked in series.

Both methods buy a higher voltage rating from the same overall dimensions, at the price of manufacturing complexity. They are the practical pay-off of everything in this chapter: once you can estimate where stress concentrates, you can deliberately re-engineer the geometry and the materials to even it out. Field estimation is never an end in itself — it is the map that tells the insulation designer where to act.

For each item, first decide which tool it calls for — the coaxial formula, the spherical formula, the \(V/r\) enhancement rule, the utilization factor, or the \(b/a=e\) optimum — then apply it. Difficulty rises down the list.