Why Your Ground Plane Isn't Zero Ohms (When It Matters)

The Myth of the Ideal Ground

Every schematic shows ground as a perfect equipotential. A single node, zero impedance, infinitely accommodating. Your PCB's ground plane is supposed to be that ideal made real.

It isn't.

At DC, a solid copper pour comes close enough that you can ignore its resistance in most circuits. But the moment you push current through it at any appreciable frequency—and "appreciable" starts lower than you'd think—that plane develops real impedance. Voltage drops appear across it. Return currents take unexpected paths. And if you're trying to pass EMC, this is where things get interesting.

What's Actually Happening in the Copper

A ground plane has three impedance components that matter: DC resistance, AC resistance (which increases with frequency due to skin effect), and inductance. At low frequencies, DC resistance dominates. As frequency climbs, skin effect forces current into an increasingly thin layer at the surface, raising the effective resistance. Eventually, the inductive reactance takes over entirely.

The DC resistance of a rectangular plane is straightforward:

where $\rho$ is resistivity (about $1.68 \times 10^{-8}$ Ω·m for copper), $L$ is the length current travels, $W$ is the width, and $t$ is the copper thickness.

Skin depth is where things get frequency-dependent:

For copper at 100 MHz, skin depth is roughly 6.6 μm. Standard 1 oz copper is 35 μm thick, and a plane's return current flows on the face toward its return, so at that frequency it uses only the outer ~6.6 μm of that face. Your effective thickness just dropped to under a fifth.

The AC resistance becomes:

when $\delta \ll t$, with the current on the one face toward the return. When skin depth approaches the copper thickness, the resistance falls back toward the DC value.

Then there's inductance, and what sets it is the distance to the return. For a plane at height $h$ above its return plane, a path of length $L$ and width $W$ has:

This inductance doesn't change much with frequency, but its reactance ($X_L = 2\pi f L_{plane}$) grows linearly. At some crossover frequency, inductive reactance exceeds resistance, and the plane starts behaving more like an inductor than a resistor.

A Real Example: 100 mm × 50 mm Ground Plane at 100 MHz

Let's work through actual numbers. Say you've got a ground plane section that's 100 mm long, 50 mm wide, using 1 oz (35 μm) copper. You want to know its impedance at 100 MHz. Its return is a plane 1 mm away, the calculator's default separation.

DC Resistance:

Under a milliohm. Looks great, right?

Skin Depth at 100 MHz:

Since 6.52 μm is much less than 35 μm, we're firmly in the skin effect regime.

AC Resistance:

The return current crowds onto the face toward the return plane, so a single skin depth carries it:

Already 5.4× higher than DC resistance.

Plane Inductance:

With the return plane 1 mm below:

Inductive Reactance: Total Impedance:

So your "perfect" ground plane actually has 1.58 Ω of impedance at 100 MHz. That's almost entirely inductive—the resistance is a rounding error at this point.

If you're running 1 A of return current at 100 MHz through this path, you'll develop 1.58 V of voltage drop across the ground plane. Even 10 mA leaves 15.8 mV. And it gets worse: at 500 MHz, that inductance gives you 7.9 Ω. At 1 GHz, 15.8 Ω.

Open the Ground Plane Impedance vs Frequency calculator to run your own geometry.

Express it as a percentage and it becomes a decision

Skin depth on its own is not actionable. Knowing it is 6.6 um at 100 MHz tells you nothing until you know whether the copper is 17.5 um or 105 um thick. The useful quantity is the ratio, and once you frame it that way it turns into a design gate.

On 1 oz copper (35 um):

At 1 MHz the skin depth is nearly twice the copper. Current fills the whole cross-section, AC resistance is essentially the DC value, and copper weight buys you exactly what you would expect. By 1 GHz the current is confined to about six percent of the copper and the resistance is more than eight times higher — the sheet resistance goes from 0.4926 to 4.126 milliohm per square on the same copper.

The gate is roughly this. Above 100%, the conductor is fully used and heavier copper reduces AC resistance proportionally. Below about a third, most of the copper is idle at that frequency and extra thickness buys almost nothing electrically. It still helps thermally and at DC, so heavier copper is never wasted — it just stops helping the problem you are trying to solve.

One caution about the table: it is the skin depth percentage calculator's model of an isolated conductor, which carries current on both faces, so its effective conducting depth is roughly twice the skin depth until the two layers meet in the middle. A plane over its return carries the return current only on the face toward the return, and that is what the ground plane calculator models: at 1 GHz it gives R_AC / R_DC = 16.7 for the same 1 oz copper, about twice the table's figure.

Two caveats worth carrying. Resistivity rises about 0.39% per degree, so a plane running at 105 C has noticeably more loss than the room-temperature figure. And this models a smooth conductor — once skin depth approaches the roughness profile of the foil, measured loss exceeds the prediction, which is why foil roughness starts mattering more than copper weight at high frequency.

When This Actually Matters

Most of the time, ground plane impedance is a second-order effect you can ignore. But there are situations where it bites you.

High-Current Switching Transients

When a MOSFET in a switching converter turns on, the current slew rate can hit 1 A/ns or more. That 2.51 nH of plane inductance produces:

Per amp per nanosecond. If you're switching 10 A in 5 ns, that's 5.0 V of ground bounce along this 100 mm path just from the plane itself—before you even count the inductance of vias, traces, and package leads.

EMC Pre-Compliance

Radiated emissions often come from voltage drops across ground planes driving common-mode currents on cables. If your ground plane develops even a few millivolts at a harmonic frequency, that can couple to I/O cables and show up as a spike on your emissions scan. Understanding the impedance profile helps you predict which frequencies will be problematic.

Sensitive Analog Circuits

An ADC with a 1 mV LSB doesn't care about the plane's 0.96 mΩ at DC. But if its return current shares a path with a digital circuit switching at 100 MHz, that path's 1.58 Ω can inject noise directly into your measurement.

Common Mistakes and Gotchas

Assuming Thicker Copper Always Helps

At DC, yes—doubling copper thickness halves resistance. But once skin effect dominates, thickness becomes irrelevant. At 100 MHz with 6.6 μm skin depth, going from 1 oz to 2 oz copper does almost nothing for AC resistance. The current's already confined to a layer thinner than either option provides. You're paying for copper that carries no current.

Ignoring the Return Path Length

The calculator asks for plane length, but what length should you use? It's the return current path length, not the board dimension. Return current follows the path of least impedance, which at high frequencies means directly under the signal trace. If your signal trace takes a 50 mm detour around a cutout, your return current does too—and that's the length that matters.

Treating Planes as Uniform

A ground plane with slots, cutouts, or connector keep-outs isn't a simple rectangle. Current has to flow around obstacles, which increases effective length and can create resonant structures. The calculator gives you the baseline; real boards are always worse.

Forgetting About Multiple Layers

If you have ground planes on multiple layers connected by vias, current will distribute between them. This generally helps—two parallel planes have half the inductance—but via inductance adds back some of what you gained. It's complicated, and a simple 2D calculation won't capture it.

Using Conductivity Numbers from Datasheets

Pure copper has conductivity of $5.96 \times 10^7$ S/m. Electrodeposited copper on PCBs is typically 5-10% lower. And if your copper's been through multiple thermal cycles, it can degrade further. The calculator lets you adjust conductivity, and you should—assuming ideal copper will make your results optimistic.

The Frequency Crossover Point

There's a useful mental model here: below some crossover frequency, your plane is resistive; above it, it's inductive. You can estimate this by setting $R_{AC} = X_L$ and solving for frequency, but honestly, it's easier to just sweep the calculator across a frequency range and watch where the impedance curve changes slope.

For a plane over its return, the crossover depends only on the copper and the separation, not on the plane's length or width: about 61 kHz for this 1 oz plane 1 mm above its return, and about 610 kHz with the return 0.1 mm away. Below the crossover, making the plane wider reduces impedance (lower resistance). Above it, widening helps just as much, because the inductance also falls as 1/W, and so does bringing the return closer: the inductance scales as h/W.

Try It

Plug your actual board dimensions into the Ground Plane Impedance vs Frequency calculator. Try your operating frequency, then try the harmonics—third, fifth, seventh. See where inductance takes over. If you're troubleshooting EMC issues, check the frequencies where you're failing and see what the ground plane impedance looks like there.

The numbers might explain why that ferrite bead isn't helping as much as you expected, or why moving a decoupling cap 5 mm made a difference. Ground isn't zero ohms. It never was.