Sunking PCB

Back Drill Stub Control: Depth, Reference Planes and Cross-Section Acceptance

Design Guide2026-09-11Sunking Circuits Technical Team
Back Drill Stub Control: Depth, Reference Planes and Cross-Section Acceptance

Back drill requirements usually reach the factory as a single line in the fabrication notes: "back drill signal vias, stub under 10 mil." That line is complete from the design side and incomplete from the manufacturing side. It does not say which surface to drill from, what plane the depth is referenced to, or who measures the stub and where. Without those three items the factory sets a depth using its own convention, and the number it measures is often not the same quantity the designer had in mind.

Back drilling is not a single "machine to depth" operation. It is a chain that runs from stack-up design through measured lamination thickness, drill depth compensation and finally cross-section verification. Wherever the chain has no agreement, stub length becomes a dispute after the fact. This guide follows that chain.

1. What back drilling does and does not solve

Back drilling removes the length of plated barrel in a through via that carries no signal, known as the stub. In a high-speed link the stub behaves as an open branch that resonates at a specific frequency and produces an insertion loss dip. Shortening the stub pushes that resonance above the working bandwidth.

That also defines what back drilling does not do. It does not fix impedance discontinuity itself, it does not replace anti-pad design, and it does not correct impedance deviation caused by dielectric thickness variation. If signal integrity is expected to come from back drilling alone, the channel often still misses its target after the stub has been driven to the process limit. Impedance control belongs to a separate discussion covered in impedance control PCB design; back drilling addresses only the stub.

The second common misconception is that back drilling can be added to any multilayer board. It requires an unambiguous target layer, enough interlayer thickness margin inside the board, and clearance between the enlarged hole and adjacent structures. Applicable structures and the engineering review entry point are on the multilayer PCB product page; feasibility for a specific stack-up is confirmed by engineering.

2. What the drill data and fabrication notes must state

A back drill requirement the factory can execute directly needs at least the following. Missing items rarely cause the data package to be rejected; they cause the factory to apply its own default.

  • Hole grouping: which holes are back drilled and how they are identified (separate drill layer, net names, diameter groups). "All signal vias" must clarify whether test points, power and ground vias, and tooling holes are included.
  • Drill surface: top, bottom, or both. Boards drilled from both sides need a target layer stated for each side.
  • Target layer and stub definition: the stub is measured from a specific surface of the last functional connection layer to the end of the plated barrel. "To layer 6" is not enough; state whether it is the top or bottom surface of the layer 6 copper.
  • Stub target and tolerance: give a target with upper and lower limits, not only a maximum. With a maximum alone, the factory drills conservatively deep and may cut into the functional connection layer.
  • Back drill diameter: the increase relative to the original hole, or the finished back drill diameter directly.
  • Acceptance method and sampling: cross-section locations, measurement definition, sample count per lot, and which document owns the pass criteria.

3. Depth control methods and the reference plane

How accurately a factory can hold back drill depth depends on what the machine treats as zero. Two approaches are common and their error sources are entirely different.

The first references the board surface. The spindle touches down on the panel and then machines a set distance. It is simple to implement, but every interlayer thickness deviation accumulates in the stub: if the actual dielectric build above the target layer is thicker than nominal, the stub comes out long.

The second uses feedback from the target layer itself. During drilling the machine detects contact with the target layer electrically or physically and stops after a set retract. This method is insensitive to lamination thickness variation because it references the real layer instead of a nominal thickness.

The design side does not need to specify which method the factory uses, but should understand the consequence: the tighter the stub tolerance, the more it depends on depth control method and lamination thickness consistency. A tolerance set beyond process capability produces systematic lot failures. SKPCB publishes back drill depth control of ±0.05 mm, which can be checked on the capabilities overview; the achievable tolerance for a given project still needs engineering review against the stack-up.

The reference plane also has to be stated on the drawing. Zeroing on outer layer copper, on the solder mask surface, or on the finished board surface introduces a systematic offset on the order of mask and finish thickness. That offset is not negligible when the stub budget is only a few tenths of a millimetre.

4. Back drill diameter versus original hole, annular ring and adjacent structures

The back drill diameter must exceed the plated hole diameter so the barrel copper is fully removed. Insufficient oversize leaves a ring of copper on the wall, which appears in cross-section as an uncut copper ring and means the stub was never truly removed.

The oversize cannot grow without limit either. As the back drill diameter increases, clearance to adjacent copper, to inner layer anti-pad edges and to neighbouring holes shrinks at the same time. The design side needs a dedicated clearance rule for back drilled holes during layout rather than leaving the judgement to the factory. SKPCB's standard back drill diameter is 0.5 mm, a usable starting point for reserving layout space; smaller back drill structures fall under engineering review.

One item is easy to miss: the section of barrel that gets removed also takes with it any inner layer pad it connected to. If a functional inner layer pad sits inside the back drill depth range, that connection disappears. This conflict is invisible on the netlist and only surfaces when hole groups and stack-up are reviewed together, which is an extra check beyond the multilayer stack-up drawing release checklist.

5. How stack-up thickness tolerance consumes the stub budget

This is the most underestimated link. Stub length is not an independent variable. It is the actual position of the target layer minus the actual depth the drill reached. The first is set by post-lamination interlayer thickness, the second by machine accuracy and the reference plane.

Board thickness tolerance is normally expressed as a percentage. SKPCB publishes ±10% or ±0.005 inch, whichever is greater, on the manufacturing tolerance page. Put that magnitude next to a stub budget: on a thick multilayer board, the accumulated thickness deviation above the target layer can be the same order as the stub target itself. Setting drill depth from nominal stack-up thickness therefore lets the stub distribution drift with each lamination lot.

Two practical rules follow. First, stub tolerance has to be discussed together with stack-up thickness control, not only with the machine's depth repeatability. Second, when a tight tolerance is genuinely needed, the data package should authorise the factory to compensate depth per lot from measured thickness, while acceptance still rests on measured stub values. Forbidding depth adjustment and demanding a tight stub tolerance are mutually contradictory requirements.

6. Acceptance: cross-section locations, measurement definition and sampling

Stub length can only be confirmed on a destructive cross-section. Surface inspection, electrical test and X-ray cannot substitute for it, because each answers a different question: electrical test confirms the target layer connection survived, X-ray confirms concentricity between the back drill and the original hole, and only cross-section yields the distance from barrel end to target layer.

The cross-section plan needs three things in writing. First, sampling locations: include holes from different board regions such as near the edge and near the centre, because lamination thickness is not uniform across the panel. Second, measurement definition: from which surface of the target layer copper to which point on the barrel end, and whether the longest or average value is taken when the hole is skewed or off-centre. Third, sampling rules: how many holes per lot, per lamination batch or per working panel, plus the escalation rule when a sample fails.

Acceptance class comes from an industry standard while the numbers come from project documents. The referenceable acceptance class is IPC-6012 Class II and Class III; SKPCB's inspection stages and evidence formats can be checked on the quality assurance page. Specific stub limits, allowable copper residue criteria and measurement conventions are project-specific and belong in the customer approved documents or a bilateral technical agreement, not in a generic number of unknown origin.

Lot evidence is organised the same way as for other process items: cross-section images, measured values, hole locations and working panel identifiers must trace back to a specific lot. SKPCB maintains full-chain QR code traceability from raw material to finished board, which provides the basis for linking cross-section evidence to lots. The broader inspection framework is described in PCB quality inspection methods; back drill specific cross-section clauses have to be declared on top of it.

7. Back drill release checklist

Confirm each item before production, and make each one point to a controlled document rather than an email conclusion:

  1. Back drilled hole groups are defined by a separate layer or an explicit rule, including which holes are excluded.
  2. Drill surface (top / bottom / both) and the target layer for each surface are stated.
  3. Stub definition names a specific surface of the target layer copper, with a target value and upper and lower limits.
  4. The depth reference plane is declared (outer copper / solder mask / finished board surface).
  5. Back drill diameter or oversize is given, and layout clearance was reserved for that diameter.
  6. No functional inner layer pad sits within the back drill depth range.
  7. Stack-up thickness tolerance is accounted for in the stub budget, and per-lot depth compensation is authorised in writing.
  8. Cross-section locations, measurement definition and sample counts are agreed, with the source of pass criteria identified.
  9. Acceptance class and revision are stated, with document precedence agreed for conflicts.
  10. First article cross-section results are reviewed and signed off as the production baseline.

To confirm back drill feasibility and achievable tolerance for a specific stack-up, submit the stack-up table and drill files for engineering review through online quote.

FAQ

Q1: Is "keep the stub as short as possible" acceptable in fabrication notes? No. It is not executable. State the drill surface, the target layer with the specific reference surface, and a target value with limits, otherwise the factory applies its own default depth convention.

Q2: What is the risk of giving only a maximum stub value? To stay under the maximum, the factory drills conservatively deep, which can cut the target layer connection or make copper residue judgement ambiguous. A two-sided tolerance constrains that bias.

Q3: Can stub tolerance simply be tightened? No. Actual stub length depends on post-lamination interlayer thickness deviation and the depth control method together. A tolerance beyond process capability produces lot-level failures and should be fixed during engineering review alongside stack-up control.

Q4: Can back drilling replace via optimisation or impedance control? No. It removes only the non-functional stub. Anti-pad design, dielectric thickness deviation and impedance discontinuity have to be handled at the stack-up and trace width level.

Q5: Can stub length be verified non-destructively? Numerical acceptance requires cross-section. Electrical test confirms the target layer connection remains, X-ray confirms concentricity between back drill and original hole, but neither yields a stub length measurement.

Want this stackup and HDI process checked?

Layer count, blind/buried vias and impedance control drive both yield and cost. Send us the stackup and we will return DFM feedback with optimisation options.