Via Fence Shielding Between RF Blocks

A via fence is a row of grounded vias placed along a line between two circuits, or around the edge of a radio frequency block, to reduce the coupling between them. It works like a wall: it does not block the field completely, but it interrupts the surface currents that would otherwise carry energy from one area to the other.

This article covers how a fence reduces coupling, how the via spacing is chosen, and where the space it costs is justified.

How A Fence Works

The energy that couples from one part of a board to another travels partly through the air and partly along the conductors and the reference planes. The plane component is often the larger of the two, and it is the one a fence addresses. A row of vias that ties the surface ground to the internal ground plane creates a low impedance path across the plane boundary, so the return current that would have spread sideways is confined to the region it belongs to.

The fence is effective where the ground plane would otherwise be continuous across the boundary. If the plane is split, and the split is arranged so that the two circuits reference different areas, the fence is redundant. The problem arises where a single plane serves both circuits, because the plane is then a shared conductor. The plane arrangements that avoid this are described under ground routing and power trace planning.

Spacing And Its Effect

The spacing between the vias is the parameter that sets the isolation. The fence behaves as a wall only for wavelengths longer than about twice the spacing; above that frequency, energy passes between the vias as though the wall had holes. A useful starting rule is to keep the spacing below a tenth of the wavelength at the highest frequency of interest, and to reduce it further where the isolation requirement is demanding.

For ground stitching of this kind, the spacing is measured centre to centre, and the via diameter is largely irrelevant to the isolation as long as the vias are plated and connected. Where the space is limited, the choice is between a wider spacing with a shorter fence and a closer spacing over a shorter distance: isolation depends on the continuity of the wall along the boundary, so a fence with a gap in it is much less effective than a shorter fence with no gap. The general rules for controlling coupling are set out under <a href="https://www.gopcba.com/crosstalk-3w-rule/” title=”crosstalk and the 3W rule”>crosstalk and the 3W rule.

Where A Fence Is Used

The classic application is between a radio frequency transmitter and a sensitive receiver on the same board, where the isolation requirement is tens of decibels and the two circuits sit a few millimetres apart. A second is around a crystal or a clock circuit, to keep its harmonics out of the surrounding analogue circuitry. A third is around a connector or a cable entry, where the fence, together with a shield, forms a barrier against energy entering or leaving the board.

A fence is also used at the boundary between a digital and an analogue area, where the return currents of the two are different in character. In that case the fence is usually combined with a plane split, and the vias are placed along the split so that the two grounds are tied together at one point rather than over a whole area. Deciding which of those arrangements is correct is a system question, and the stackup that supports it is described under the advantages of multilayer construction at high speed.

Cost And Layout Considerations

A fence consumes routing space. Each via needs an antipad on every layer it passes through, and the antipads remove copper from the planes and the routing layers. A dense fence through several layers can remove a significant fraction of the plane area along the boundary, which itself changes the behaviour of the plane. The pattern also has to avoid traces, so a fence often forces the routing to be arranged around it.

The cost is normally paid where the isolation requirement is real. Adding a fence everywhere to a design with no isolation problem adds vias, reduces the plane, and consumes routing space for no benefit. Where the requirement is marginal, the alternative measures, such as increasing the physical separation or reducing the bandwidth of the aggressor, are often cheaper and more effective than a fence.

Designing The Fence

The fence should follow the boundary as closely as possible, because the isolation comes from the continuity of the wall rather than from its position. It should be connected to the ground plane on every layer it crosses, and it should have no gaps where a trace passes through, which means the trace either goes around the fence or the fence is routed around the trace with a maintained spacing. A trace that crosses a fence through a gap creates the very coupling the fence was meant to prevent.

The fence should also be tied to the ground of the area it protects, and the connection should be a low impedance one. Where the fence is tied to a plane that itself carries noise, the fence spreads the noise rather than confining it. The check is a continuity check of the ground connection along the fence, performed on the finished layout rather than assumed from the schematic.

Verification

The verification of a fence is a measurement, either of the coupling between the two areas on a test board or of the isolation of the finished product in an anechoic chamber. A useful intermediate check is a board that duplicates the geometry with and without the fence, measured at the frequencies of interest, which gives a direct figure for the improvement. Where the fence is fitted to meet a compliance requirement, that measurement is part of the evidence for the design.

The fence also has to be checked for its effect on the circuits it separates. Vias that pass through a plane remove copper, which changes the impedance of any trace above them, and a fence placed under a controlled impedance line changes the impedance locally. Where the fence runs beside a high speed line, the two requirements should be checked together rather than one after the other.

The fence is also worth extending around the corners of the protected area rather than stopping at the edges of the block. Energy that leaves the block at a corner finds the shortest path around the end of the wall, and a fence that ends abruptly leaves that route open. Continuing the row for a short distance past the corner, or bending it to follow the outline of the block, closes the path at a modest cost in vias and space.

FAQ

How close should the vias be spaced? A common starting rule is a spacing below a tenth of the wavelength at the highest frequency of interest. The requirement becomes tighter as the isolation figure rises.

Does a fence replace a shield? No. A fence reduces coupling through the plane and along the surface, while a shield blocks radiated energy. Where the requirement is a radiated one, the shield does the work and the fence supports it.

Can a fence be added after the layout is complete? It can usually be inserted, but it consumes routing space and it removes plane copper, so the effect on the routing and on the impedance has to be rechecked. Adding it early is easier.