PCB Trace Corner Design And Acid Traps

Every change of direction in a trace is a small discontinuity. At low speed and at moderate current a right angle corner makes no measurable difference, and the practice of rounding corners is often justified by tradition rather than by analysis. At high frequency and at high current the corner does matter, and the reason is partly electrical and partly chemical.

This article covers what happens at a corner, how etching and current crowding behave, and how corners should be routed.

What Happens At A Trace Corner

The current in a trace follows the path of least impedance, which at high frequency means the path that encloses the least area. At a corner, the current does not turn through the outer edge evenly; it crowds towards the inside of the bend, and the effective width of the conductor is reduced there. The result is a local rise in resistance and inductance, and a small reflection of the signal.

The magnitude of the effect depends on the ratio of the trace width to the wavelength and on the sharpness of the corner. A trace corner that is small compared with the wavelength behaves as a lumped impedance discontinuity, and its effect is usually negligible. A corner that is a significant fraction of the wavelength produces a reflection that appears in the return loss and in the eye diagram, and it becomes visible when several such corners occur in one line. The transmission line behaviour behind this is described under PCB routing with microstrip and stripline.

Acid Traps And Etching

The chemical effect is more definite than the electrical one. During etching, the etchant has to reach the copper and the reaction products have to leave. In an acute angle between two conductors, the etchant circulates poorly and the reaction products accumulate, so the etch proceeds at a different rate from the open areas. The result is a corner that is either over-etched, leaving a thin neck, or under-etched, leaving a residual sliver of copper.

An acute angle is therefore avoided not for electrical reasons but because it produces a feature that varies from board to board. A ninety degree corner etched in a dense area behaves similarly, because the copper around it restricts the circulation. The measurement that catches it is the trace width next to the corner on a section or under a microscope, and the rule that prevents it is a minimum angle in the design rule check, usually ninety degrees or greater.

Electrical Effect At High Frequency

At high frequency the corner behaves as a small capacitance and a small inductance in series, and the reflection it produces rises with frequency. The classic mitigation is a chamfer or a curved corner, which spreads the change of direction over a distance. The improvement is real but modest: for a corner that is a small fraction of a wavelength, the difference between a right angle and a well chamfered one is measured in fractions of a decibel.

Where the corner matters more is where the trace also changes width or passes a reference plane edge, because the two discontinuities then add. A signal that changes layer, changes reference and turns a corner within a short distance sees the sum of all three, and the return loss at that point can be much worse than any one of them would produce. The routing practices that keep such transitions controlled are described under routing high frequency traces and data buses.

Current Crowding

At high current the corner reduces the effective cross section, and the local current density rises. For a trace operating near its thermal limit, the corner becomes the hottest point, and the failure that starts there is a fusing or an electromigration failure rather than a signal problem. The effect is small for a ninety degree corner and larger for a sharp acute angle, where the inner radius approaches zero.

The practical rule is to size the corner for the current rather than to round it for the signal. A wide power trace carrying many amperes is better with a chamfered or a curved corner, because the change of direction then happens over a distance and the current has room to redistribute. The calculation of the width for a given current is described under trace width and current calculation.

Routing Practice

In practice, three conventions cover most cases. Ordinary signal traces are routed with ninety degree corners, which is what the layout tool produces by default and what the fabrication process can hold. High frequency traces use chamfered or curved corners where the trace width is a significant fraction of the wavelength, and the chamfer is typically one trace width on each leg. Power traces use a curve or a chamfer where the current is high, and the corner is widened rather than merely rounded.

What every case has in common is that an acute angle is avoided. An acute angle concentrates the etch problem, concentrates the current and is difficult to inspect, and there is no application in which it is the right choice. Where a routing constraint seems to demand one, the answer is usually to move the via or to change the layer rather than to accept the angle.

Design Rule Checks

The checks are a minimum angle between any two segments of the same trace, a minimum width everywhere including at the corner, and a minimum clearance to any neighbouring conductor. A tool can flag an acute angle directly, and it can also flag a trace whose width after etch modelling falls below the minimum at a corner.

Where the board is built to a controlled impedance, the check should be applied to the same net that carries the impedance requirement, and the corner treatment should be consistent along the line. A line that is chamfered in one part and square in another has two different behaviours, and the inconsistency makes the measurement harder to interpret than a uniform line of either kind. The fabrication side of these rules is covered under PCB design and fabrication.

The checks are worth running even on a board with no impedance requirement, because a corner that is over-etched is a reliability problem as well as a performance one. A trace that is thinner at the corner than elsewhere is the point at which an overcurrent or a thermal cycle eventually opens the line, and the defect is invisible at normal magnification. Reporting the minimum width over the whole trace, rather than the design width, is the check that catches it.

FAQ

Do right angle corners cause EMI? They do not radiate on their own at the frequencies used on most boards. The effect is a small reflection and a small change in impedance, and it matters only where the corner is a significant fraction of a wavelength.

Should every corner be rounded? No. Rounding every corner costs routing space and adds nothing where the trace is short compared with the wavelength. The convention should be applied where the geometry or the current justifies it.

Why is an acute angle worse than a right angle? Because the etch chemistry circulates poorly in a narrow wedge, so the resulting feature is variable, and because the current crowds into a very small area at the tip. Both effects are more severe as the angle becomes smaller.