Characteristic Impedance Control in PCB Design: Width, Height, Dk

Impedance is the property that separates a board that merely conducts from a board that transmits. A design that passes continuity and isolation testing can still fail to carry a fast signal, because what is being conveyed along the trace is not a current but a wave, and a wave is affected by the geometry of the conductor and its surroundings. Characteristic impedance control is the practice of holding that geometry within limits so the signal arrives intact, and it is the reason a finished board can be rejected for a manufacturing deviation that leaves the copper electrically continuous.

Understanding the terms in order avoids confusion. Resistance, impedance, and characteristic impedance are related but distinct, and each one describes a different situation.

Resistance and Impedance

Resistance is the opposition a conductor presents to direct current. It is denoted R, measured in ohms, and relates to voltage and current through R equals V divided by I. It also depends on the material and the geometry: resistivity multiplied by length, divided by cross-sectional area. A narrow or long trace has more resistance, and the resulting voltage drop is a DC concern that matters mainly for power and return paths.

Impedance is the opposition presented to alternating current, denoted Z and also measured in ohms. It differs from resistance because it includes the effects of inductance and capacitance, so the value depends on frequency. The magnitude is the square root of the sum of the resistance squared and the square of the difference between inductive and capacitive reactance.

At high frequency the reactive terms dominate, which is why a trace that is effectively a short circuit to direct current becomes a significant element in the signal path. The DC resistance of a trace and its impedance at the operating frequency are not interchangeable figures, and designing for one while ignoring the other is a common source of unexpected behaviour.

A related consequence is that the copper cross-section sets the resistance and the current capacity, while the dielectric structure sets the impedance. Both are properties of the same trace, so changing the width to correct one alters the other.

Why Signal Transmission Needs Its Own Measure

As signal rates increased, it became clear that the transmission of a fast edge along a trace is not a flow of charge in the ordinary sense. What propagates is a wave travelling in the dielectric and along the conductor, guided by the surrounding structure. The opposition that wave encounters is a different quantity from the resistance or the impedance of the conductor alone, and it is called the characteristic impedance, written Z0.

Because the wave is guided by the trace geometry and the dielectric, Z0 is not measured with a multimeter. It is a property of the transmission line, and it is defined by the conductor width, the conductor thickness, the dielectric thickness between the conductor and its reference plane, and the dielectric constant of that material.

This is why testing a high-speed board for opens, shorts, and feature defects is not sufficient. A controlled-impedance board must have its characteristic impedance measured, and the measured value must fall within the specified tolerance. Where it does not, the board is normally rejected rather than reworked, because the geometry that caused the deviation cannot be corrected after fabrication.

Typical Targets and Tolerances

The targets in common use are 50 ohms, 75 ohms, and 28 ohms, each with a tolerance typically stated as plus or minus ten percent. A 50 ohm environment dominates digital and RF work because it is a reasonable compromise between loss and power handling and because the test equipment ecosystem is built around it. The 75 ohm target is associated with video and broadcast interfaces, and lower impedances appear where a specific interface or a narrow trace requirement dictates them.

The tolerance is as important as the nominal figure. Without a stated tolerance there is no way to decide whether a measured board is acceptable, and the fabricator has no target to design the artwork compensation against. The tolerance should be quoted in the fabrication drawing along with the layers to which it applies.

The Four Variables

Four parameters determine the characteristic impedance of a trace: the conductor width, the conductor thickness, the dielectric thickness between the trace and its reference plane, and the dielectric constant of the material. Each one is controlled by a different party and each one varies during production.

Width and thickness are properties of the copper. Width is set by the artwork and modified by the etch process; thickness is set by the starting foil and the plating applied. Dielectric thickness is set by the stackup and modified by the resin flow during lamination. The dielectric constant is a material property that varies slightly between lots and with frequency.

<img src="https://www.gopcba.com/wp-content/uploads/2024/09/Layer-up-Lamination5-1.png" alt="Impedance test coupon measured by time domain reflectometry” />

For a conventional surface microstrip, the calculation uses the trace width, the trace thickness, the height from the trace to the reference plane, and the dielectric constant. For a stripline the same variables apply with the trace buried between two planes, and for a differential pair the spacing between the two conductors becomes an additional variable.

Which Variable Matters Most

Sensitivity is not evenly distributed. The dielectric thickness has the largest effect, because impedance is approximately proportional to the logarithm of the height-to-width ratio, so a small change in height produces a proportionally larger change than an equivalent change in any other variable. The dielectric constant follows, then the trace width, and finally the trace thickness, which has the smallest influence of the four.

The practical difficulty is that the two variables with the largest influence are the two that are hardest to hold tightly. After the material is chosen, the dielectric constant varies little, and the dielectric thickness varies only as much as the lamination process allows. Thickness of copper is comparatively easy to control.

Width is where the problem concentrates. Holding the finished width to within ten percent of nominal is difficult with a subtractive etching process, because the etchant removes copper laterally as well as vertically, and the amount removed depends on the copper thickness, the etchant condition, and the local pattern density. Defects such as pinholes, nicks, and depressions further reduce the effective cross-section.

Controlling Width in Production

Because the finished impedance depends so heavily on width, and because width is the variable the fabricator can influence most directly, impedance control in practice becomes a width control problem. The artwork is compensated for the expected etch loss, and the compensation is verified by measuring test coupons from the production panel rather than by assuming the process behaves as modelled.

Coupon design matters here. A coupon that reproduces the same trace width, the same dielectric thickness, and the same layer arrangement as the product will predict the product’s impedance. A coupon with a different geometry measures the process, not the design, and it can be acceptable while the product is out of tolerance.

The measured value is normally obtained by time domain reflectometry, which reports the impedance along the length of the line and shows where any discontinuity occurs. This makes the technique useful for diagnosing a problem as well as for accepting a lot, because a local deviation points to a specific feature such as a connector footprint or a layer transition. The relationship between impedance and the structures used to carry high-speed signals is described in PCB routing with microstrip and stripline, and the stackup that defines the dielectric thicknesses is covered in layer stackup from one to eight layers.

Impedance on Simple Constructions

Single-sided boards are not normally impedance controlled, because there is no reference plane on the opposite face to define the return path. A microstrip requires a reference plane beneath the trace, which a single-sided construction does not provide, so a trace on such a board has an undefined impedance determined largely by its surroundings and by whatever conductor happens to be nearby.

Where a single-sided design needs a defined impedance, the practical solutions are to move to a double-sided construction with a ground plane on the opposite face, or to keep the connections short enough that the transmission line behaviour does not dominate. For power and low-frequency signals, this is adequate. For fast digital or RF signals, a two-layer construction with a ground plane is the minimum, and the current-carrying requirements of the copper are calculated separately, as described in trace width and current calculation.

FAQ

Can characteristic impedance be measured with an ordinary meter? No. It is a property of the transmission line rather than of the conductor, and it is measured with time domain reflectometry or a vector network analyser on a test coupon or on the board itself. A resistance measurement says nothing about it.

Why is the tolerance usually ten percent? Because that is the range the fabrication process can hold reliably across a panel when width, dielectric thickness, and material variation are combined. A tighter tolerance requires tighter process control and a correspondingly higher cost, and it should be specified only where the interface genuinely requires it.

Which change affects impedance most? A change in the dielectric thickness between the trace and its reference plane. The relationship is logarithmic in the height-to-width ratio, so a given percentage change in height moves the impedance more than the same percentage change in width, thickness, or dielectric constant.