Backdrilling and Stub Removal
The Stub Problem
A plated through hole connects every layer it passes through, so a via that is used to connect an inner layer still has a length of barrel below the connection. That unused length is a stub, and at high frequency it behaves as an open circuited transmission line branch. It resonates at a frequency related to its length and the material’s dielectric constant, and near that resonance it distorts the signal, adds loss and creates a notch in the channel’s response. The effect is negligible at low frequency and significant at multi gigabit rates, which is why stub removal appears in high speed designs and nowhere else.
How Backdrilling Works
Backdrilling re-enters the hole from the side opposite to the connection and drills away the stub with a bit that is slightly larger than the original, so that the plating is removed without touching the layers that the via is meant to connect. The depth is controlled so that a small amount of stub remains, since removing the plating exactly at the target layer risks damaging the connection. The process is performed after the plating and before, or sometimes after, the surface finish, and its registration and depth control are the two parameters that decide the result. The hole’s entry is enlarged, which affects the pad’s annular ring and the routing clearance around it.
Designing for Backdrilling
The design has to allow for the larger entry hole. The pad on the side that is backdrilled has to provide an annular ring for the larger diameter, and the routing around the via has to clear it. The via’s position relative to the connector and the layer it connects to determines the stub’s length, so the stackup and the layer assignment decide how much stub there is to remove. Where the design has several high speed vias, the depth control and the repeatability matter more, and the fabricator’s capability should be confirmed before the stackup is fixed rather than after the design is complete.
The Remaining Stub and Its Resonances
The residual stub is a design parameter. A shorter residual pushes the resonance higher, and the target is usually to move it above the frequency band of interest or to reduce its magnitude sufficiently. The achievable residual depends on the process, and asking for an unrealistically short one increases the risk of damaging the connection. The design should therefore calculate the stub’s length from the stackup, estimate the resonance and decide whether backdrilling is needed at all, because a via whose stub is short enough may not require the extra process.
Process Effects and Damage
The backdrilling process removes material and generates heat, and it can leave a rough surface or a burr at the enlarged entry. The plating’s adhesion at the transition from the drilled region to the remaining barrel is a location worth examining, and the section should be checked for a crack or a separation there. The enlarged entry also collects the finish and the solder mask differently, which can affect the assembly. The process control includes the depth, the bit’s condition and the number of hits, and the verification includes a section and, where possible, a measurement of the stub.
Alternatives to Backdrilling
The alternatives are a blind or a buried via, which connects only the layers needed and has no stub by construction, and a design that positions the connection so that the stub is short. A blind via requires a sequential lamination, which adds cost and lead time in a different way. A stackup with the high speed layers near the surface reduces the stub’s length without an extra process. Choosing between them is a comparison of the cost, the lead time and the electrical benefit, and the answer depends on how much the stub actually affects the channel, which should be calculated before the more expensive option is selected.
Verification
The verification is by a microsection, which shows the remaining stub, the enlarged entry and the condition of the plating, and by a measurement of the via’s electrical behaviour, which can be done with a test coupon that includes a comparable via. The coupon is the practical way to confirm the process’s capability, since the section is destructive and a product board cannot be sacrificed. Where the design’s margin is small, the coupon’s measurement should be compared with the simulation’s prediction, and the difference is the evidence that the design’s assumption about the stub is valid.
Cost, Yield and When It Is Justified
Backdrilling adds a process step, an extra drilling operation and a yield risk, so it carries a cost that a design without it does not. The justification comes from the electrical requirement: where the simulation shows that the channel’s eye is closed by the stub, the cost is part of meeting the specification; where the margin is comfortable, the step buys nothing. The comparison should be made between the backdrilled design and the alternatives, which are a different stackup, a blind via or a lower data rate, and it should use the measured behaviour of the process rather than a nominal figure. The decision is an engineering one and it belongs in the design review rather than in the fabrication quotation.
FAQ
Why is a via stub a problem? It behaves as an open circuited branch, resonating at a frequency and distorting the signal at high data rates.
Why not remove the stub completely? Removing the plating at the target layer risks damaging the connection, so a small residual is left.
What does the design have to allow for? A larger entry hole, so the annular ring and the routing clearance must accommodate it.
What is the alternative to backdrilling? A blind or buried via, which has no stub but requires a sequential lamination.
How is the process verified? By a microsection and by a coupon that measures the via’s electrical behaviour.
Conclusion
Backdrilling removes the barrel that is not needed, and it is worth doing only where the stub affects the channel, so calculate before you specify. Check the coupon. High speed via design belongs to PCB design and layout, the process is performed in PCB manufacturing, and the acceptance is part of quality management. High speed stackups are first built during prototype PCB assembly in 2026.
