The conclusion first: the hard part of depth-controlled features is not "how deep" but "measured from where." Board thickness carries tolerance and the laminated surface is not an ideal plane, while the machine controls depth from the actual top surface of the actual board. So the number you put on the drawing and the dimension you actually care about — remaining thickness, sink depth, distance to a layer — are usually not the same dimension. The drawing has to say which one it is.
1. "Add a countersink" is half a requirement
Conventional PCB drilling is binary: a hole either goes through or does not exist. The drill enters one face and exits the other, and depth needs no control because board thickness defines it. This is one of the few PCB operations with no Z-axis precision requirement at all.
Depth-controlled features remove that convenience. A countersink has to let a tapered screw head sit flush. A counterbore has to let a pan-head screw or bushing drop in. A milled cavity has to open a space inside the board without breaking through. All three require the cutter to stop somewhere inside the thickness.
The difficulty is that "somewhere inside" has no natural datum. Finished board thickness is itself a toleranced dimension, and after lamination the surface is not perfectly flat. The machine can position the tool relative to a reference surface, but that surface is the actual board in front of it, not the nominal board on the drawing.
This produces a mismatch that shows up in drawing reviews over and over. What the designer wants to control is remaining thickness (how much material is left under the hole) or sink depth (how far the head drops below the surface). What the machine directly controls is depth from the entry face. Those two differ by one board-thickness tolerance. When thickness runs high, remaining thickness runs high. When thickness runs low, remaining thickness runs low. Same programmed depth, different result per board — that is not machining error, it is the direct consequence of datum choice.
This article does not answer "can this countersink be made." Usually it can. The questions that matter are: which face the depth is referenced to, which dimension owns the tolerance, how the layers under the feature are handled, whether depaneling will damage it, and when the feature should not be on the PCB at all.
2. Three feature classes, three different sources of precision
| Feature | Typical purpose | How it is machined | What governs the precision | Most common failure |
|---|---|---|---|---|
| Countersink (conical) | Tapered screw head flush with surface | Form cutter, or drill then conical mill | Concentricity of cone to hole; feed depth | Cone not concentric with hole; sink depth drifts with board thickness |
| Counterbore (cylindrical) | Pan-head screw or bushing recessed | Two stages: small hole, then larger depth-controlled bore | Alignment of the two stages; Z stop of the larger bore | Step face not flat; the two stages run eccentric |
| Depth-controlled cavity | Internal space, local thinning | Milling cutter in passes, Z stops mid-thickness | Surface flatness; remaining-thickness datum | Uneven floor thickness; local overcut |
The column worth reading twice is "what governs the precision." It says that one depth requirement is not one difficulty across the three, and is not owned by one process step. Countersink risk concentrates in concentricity. Counterbore risk concentrates in two-stage alignment. Cavity risk concentrates in floor-thickness consistency. Calling all three "depth-controlled machining" and governing them with a single tolerance is the most common simplification error on these drawings.
One further distinction that gets blurred: back drilling is also depth-controlled, but it is not the same problem. Back drilling exists to remove excess barrel copper for signal reasons and its reference object is a copper layer, for which the process has usable stop criteria. Countersinks and cavities reference mechanical dimensions, and there is no layer inside the board to stop against. Different purpose, different datum, no transfer of logic between them.
3. Depth datum: which face is that number measured from
This is the single item most worth the drawing's ink. There are only a few datum options, and the consequences differ sharply.
- Depth from the entry face. This is what the machine executes most directly. Highest machining certainty; the cost is that remaining thickness floats with board thickness tolerance.
- Remaining thickness (floor to opposite face). This is what many structural designs actually care about, for example guaranteeing load-bearing material behind the feature. It requires the shop to back-calculate programmed depth from each board's measured thickness — an added requirement, not default behaviour.
- Sink depth (part head position relative to the surface). This is the assembly view. It involves more than the hole: surface finish thickness and surface flatness enter the chain too.
These three cannot all be tightened at once. For a given board thickness tolerance, tightening one loosens the others. That is arithmetic in the dimension chain, not a shop declining to cooperate. So the correct drawing practice is not to tolerance all three, but to declare which one is the critical dimension and mark the others as reference.
The same logic appears in the general practice described on the tolerance page: put tolerance on the dimension you actually need to control, rather than spreading it evenly across every dimension on the sheet.
4. The region under the feature: keep-out, clearance and inner-layer integrity
A depth-controlled feature consumes volume inside the board, and that volume may contain copper.
The cone or step face of a countersink or counterbore cuts through several layers below the outer surface. Where those layers carry copper in that region — traces, planes, inner pads — machining will cut it or expose it. An exposed copper edge is a reliability concern in subsequent processing: it has no solder mask coverage and none of the plating protection a normal barrel wall gets.
A cavity does the same thing over a larger footprint. Every layer above the floor is removed across the entire cavity outline.
So the design input must include one explicit item: which layers are kept clear within the projected footprint of the depth-controlled feature. If the designer does not supply it, the shop can only infer it from the Gerbers, and a wrong inference is a batch-level problem. The keep-out should be larger than the machined outline, because the machining itself has positional tolerance and the cutter has a radius.
A related point that gets missed: this region also must not host panel tie-bars or rail features. A depth-controlled feature thins the board locally, and a thinned region tolerates depaneling stress poorly.
5. Panelization and depaneling: the constraints these features add
There is a direct conflict between depth-controlled features and depaneling method, and it deserves its own section.
A locally thinned board cracks or deforms more readily during depaneling. If a cavity sits near the board edge and depaneling is by punch or V-cut, the stress path runs through that thin region. The usual constraints for outline routing and V-cut are not sufficient here — there is an extra question to ask: how far is the thinned region from the separation line, and does the depaneling method load it.
Practical directions: increase the distance from the feature to the board edge and separation line; switch to a depaneling method that is gentler on stress; or reorient the units in the array so the thin region sits away from the load path. Which of these is available depends on board shape and array layout, so it is an engineering review item, not something the design side settles alone.
6. What the drawing has to state
Collecting the preceding sections into a drawing checklist. Any missing item becomes either a guess or a clarification round:
- Feature class: countersink / counterbore / cavity, with its geometric definition (upper and lower diameters, cone angle or step position, cavity outline).
- Depth datum: which face the depth is measured from. State "from top surface" or "from bottom surface" explicitly; do not supply a bare depth number.
- Critical dimension declaration: among programmed depth, remaining thickness and sink depth, which one carries the tolerance and which two are reference.
- Keep-out areas: per-layer keep-out within the projected footprint, with margin.
- Mating part information: what is being recessed (screw head type, bushing, connector), because that defines what "acceptable" means.
- Acceptable edge condition: whether visible layer interfaces are acceptable, whether deburring is required — stated as general practice, not as numeric limits.
- Relationship to depaneling: if the feature sits near an edge, state the preference or limitation on depaneling method.
In the context of the manufacturing capabilities page, all of these fall into the "requires engineering confirmation" category of inputs. They are not information a quoting step can derive on its own.
7. When the feature should not be on the PCB
The most practically useful section. In several situations, implementing a depth-controlled feature in a PCB is using the wrong process for the problem:
- Depth tolerance far tighter than board thickness tolerance can support. Such a requirement means you actually need machined-part precision, not board precision.
- Geometry approaching a machined component (multi-level steps, angled faces, precision mating surfaces). The PCB process chain was not designed for that geometry.
- Electrical connection required inside the depth-controlled region. Depth control severs layers there; requiring conduction in the same place is two contradictory requirements.
- Thinned area is a large fraction of the board. That affects overall stiffness, post-lamination flatness and depaneling feasibility at the same time.
The usual alternatives are to move the mechanical function onto a separate bracket or structural part and leave the PCB with electrical function plus ordinary mounting holes, or to change board type altogether. Discovering this during drawing review is far cheaper than discovering it on the first production lot.
8. Putting it back in the process chain
Where the feature sits in the process chain determines how it gets inspected. It is produced at drilling or routing, while the usual hole quality criteria — wall roughness, barrel copper, nail-heading — are mostly defined for through-holes and plated holes and cannot be applied directly to a depth-controlled feature. What has to be checked here is depth, concentricity, step-face condition and floor-thickness consistency: a set of dimensions that must be declared on the drawing before they can be measured in the quality workflow.
Put plainly: what the drawing does not declare, the line does not measure. That is the usual loss-of-control path for these features — not insufficient capability, but a requirement that was never fully expressed.
FAQ
Q1: Are countersinks and counterbores the same thing?
No. A countersink is conical and seats a tapered screw head flush with the surface. A counterbore is a cylindrical recess for a pan-head screw or bushing. They are machined differently (form cutter versus two-stage) and carry different precision risks (concentricity versus two-stage alignment). The drawing must say which one, not just "recess."
Q2: I put a depth on the drawing. Why does the shop still ask?
Usually because the datum is missing. Whether depth is from the top or bottom face, and whether you actually need programmed depth or remaining thickness, is not contained in a single number. Board thickness carries tolerance, so different datums give different results.
Q3: How tight can the depth tolerance be?
It depends on board thickness tolerance, board flatness and feature class, so there is no answer independent of the specific board. The productive question is to state your assembly requirement and critical dimension and let engineering review decide which datum scheme satisfies it. If the requirement approaches machined-component magnitude, the answer may be to use a structural part instead.
Q4: Back drilling is depth-controlled, so why not handle countersinks the same way?
Because the reference object differs. Back drilling targets a stop relative to a copper layer, for which the process has usable criteria. Countersinks and cavities target mechanical dimensions, and no internal layer is available as a stop reference. Different datum, no transferable logic.
Q5: Can layers under a depth-controlled feature carry routing?
No. Layers cut by the feature must be kept clear across its projected footprint, with margin beyond the machined outline. Exposed copper edges lack solder mask and plating protection and are a reliability concern.
Q6: Will these requirements affect lead time?
Depth-controlled features are an engineering-confirmation input. When the drawing is complete, review is quick; the clarification rounds caused by missing information are the real time cost. Stating the items in section 6 on first submission is the fastest path.

