Layout Rules and Standards That Prevent Rework
A rule that nobody understands is followed until it becomes inconvenient and then abandoned. That is why a set of layout rules is worth more when it is short, specific and attached to a reason. The rules themselves are not complicated; what takes time is learning which one is protecting against what.
This article collects the layout rules that apply to almost every board, grouped by what they protect: the netlist, the ground system, the power distribution, the thermal path and the assembly process.
Rules Are Not Ceremony
Every rule in a layout standard exists because a class of defect was observed repeatedly. The three-times-width spacing rule exists because coupling between parallel traces falls predictably with distance. The twenty-H rule exists because a power plane that extends to the edge of the ground plane radiates from that edge. Understanding the mechanism is what allows the rule to be applied intelligently when the geometry does not match the textbook case.
It also prevents the opposite failure, which is following every rule rigidly on a board where most of them are irrelevant. A two-layer control board does not need a review of its plane-edge radiation; a dense high-speed board does. Applying the right subset is a judgement, and the judgement comes from knowing why each rule is there.
Schematic and Footprint Preparation
Two rules belong at the front of any standard. The first is that the schematic should use the real pin arrangement of the part rather than a convenient symbol, because a symbol that hides the physical pin order makes the layout harder and errors more likely. The second is that any footprint not already proven should be made by hand and checked against the package drawing before it is used, with particular care over pin numbering.
The reason for both is the same: a mistake in a footprint or a symbol cannot be corrected by good routing. It reaches the board and appears at first assembly, which is the most expensive place for it to appear.
Ground, Copper and the 20H and 3W Rules
The ground system deserves its own section in any standard. Signal and power returns should not share a conductor that carries significant current, the ground under a mixed-signal device should be partitioned deliberately, and the copper should be as wide as the layout allows. The rule of thumb for an ordinary signal trace is not less than about 0.3 millimetres, with power and ground considerably wider.
Two named rules belong here. The twenty-H rule keeps the power plane inset from the ground plane by twenty times the dielectric thickness between them, which suppresses fringing radiation from the plane edge. The three-W rule keeps parallel traces at least three times their width apart, which reduces crosstalk to a level that can be ignored on most digital nets. Copper on multilayer boards should also be sufficient in area to spread heat, since local overheating of an under-filled area is a fire risk as well as a reliability one.
Decoupling and Power Distribution
Every supply pin of an integrated circuit needs a local decoupling capacitor, placed as close to the pin as the layout permits, with the shortest possible connection to the ground reference. A bulk capacitor at the input of the board supply covers the slower transients, and the two work together rather than being alternatives.
The distribution path itself is the other half of the rule. Power should be routed as a network rather than as a single long trace to the far corner of the board, because a single path accumulates a voltage drop and couples noise along its length. Placing the regulators so that each serves a nearby group of loads is a placement decision that makes the distribution rule easy to satisfy.
Routing Rules That Apply to Every Board
Signals should run in the same general direction as the power and ground returns they use, so that the current paths remain predictable. Input and output lines should be separated by a ground conductor where they run together, and the same applies between functional blocks. Where a jumper is used to avoid a long route, it should not be placed under a large package, because it will be inaccessible for rework.
Where signals change direction, a curve or a chamfer is preferable to a sharp corner, and long parallel runs between two signal traces should be broken up by a change of layer or by a deliberate jog rather than left to couple along their whole length. The effect of right-angle routing on differential traces is a specific case of the same principle.
Thermal, Test and Assembly Provisions
Heat-producing components belong near a board edge or in a position with airflow, and they should be spread out rather than clustered, both to help each device and to keep the local temperature of the laminate within its rating. Test points should be defined at the layout stage, with enough space for a probe and a reference nearby.
Assembly provisions are a small set of rules that are easy to forget. Every board should carry several alignment holes, in the corners and at the centre, so that it can be located for drilling and assembly. Solder mask and copper should not form an undercut around a pad, and pads that land on a plane should use thermal relief where hand soldering is expected, so the plane does not draw heat out of the joint before it forms.
High-Speed and Mixed-Signal Rules
High-speed nets need a controlled impedance, a continuous reference plane and, where the edge rate is fast, series termination at the driver. The impedance should be maintained along the whole route, which means keeping the trace geometry and the layer stack constant from driver to receiver, without a change of reference in the middle.
Mixed-signal boards add the rule that analogue and digital routing must not run parallel and must not cross. Where the two must coexist in the same area, a ground conductor between them and a deliberate separation of the reference planes is the usual answer. The guidelines for mixed-signal layout expand this into a full set of placement and routing constraints for boards that contain both domains.
Reviewing Against the Rules
The rules are most useful as a review checklist rather than as a constraint set. A short review that walks the layout block by block, asking which rules apply and whether they are met, catches more than a design rule check alone, because the check verifies dimensions while the review verifies intent.
Where a rule has to be broken because the mechanical design leaves no alternative, the exception should be recorded with its reason and its expected effect. That record is what allows the next revision to revisit the decision rather than re-discover it, and it is what turns a set of rules into a design standard that improves from project to project.
FAQ
Are this many rules really necessary on a simple board? No. On a two-layer control board the relevant subset is the footprint check, the decoupling placement, the ground width and the thermal provisions. The high-speed rules only matter when there is a high-speed net.
What is the twenty-H rule for? It keeps the power plane inset from the ground plane edge by twenty times the dielectric thickness, which reduces the fringing field that radiates from the plane boundary. It matters most at high frequencies.
How do I decide whether to use a ground plane or a grid fill? Use a solid plane where a continuous reference is needed, and a grid where the copper is primarily mechanical or thermal. The trade is covered in the comparison of mesh and solid copper flooding.
