1. Why a bare outline tolerance does not hold up in production
Writing a tolerance band after an outline dimension carries an assumption: that the outline is a quantity you can measure directly. It is actually a relative quantity.
To measure board length you must first decide which of two opposing edges is the reference. To measure a mounting hole to the board edge you must first decide whether the hole references the edge or the edge references the hole. In manufacturing terms those are two entirely different jobs:
- Hole as datum. The router locates the hole, then cuts the outline relative to it. Hole-to-edge distance is controlled, but overall length and width float with laminate movement.
- Outline as datum. The router follows a fixed path from panel tooling features. Overall length and width are controlled, but hole-to-edge distance absorbs all accumulated error from drilling and layer registration.
Same machine, same cutter, two different numbers for what both parties call "outline tolerance." When the drawing is silent, production uses its own default, normally the second approach, because it is fastest and requires no per-board alignment. If the designer actually cared about the first, the deviation was determined at order placement, not at the machine.
So the first requirement for a workable outline tolerance is not that it be tight. It is that it be complete: which dimension carries the band, and where that dimension starts.
2. Three sources of outline deviation
When an outline disagrees with the drawing, the deviation comes from one of three places. They behave differently, and discussing them together makes diagnosis impossible.
Source one: datum establishment. Before cutting, the panel is clamped and the datum located. Error in locating the datum transfers to the outline one-for-one, as a systematic offset: every board in the batch shifts the same way. Signature: all boards consistent, all dimensions shifted in the same direction, the whole tolerance band translated.
Source two: the cutter or the die. Routing carries cutter diameter tolerance, wear-driven change in effective cut width, and deflection under load. Punching carries die clearance and the die's own manufacturing error. Signature: small in magnitude but distributed around each local contour feature, most visible at internal corners, narrow ribs, and small notches.
Source three: laminate movement and dimensional change. By the time a panel reaches profiling, lamination and baking have already changed its dimensions from the cut-to-size state. If clamping is inadequate, side loads from the cutter push the board as well. Signature: deviation differs by position on the panel, with edge locations worse than center.
To isolate an outline problem, first classify the deviation as batch-consistent, feature-local, or position-dependent. The fixes do not interchange: datum problems are solved on the drawing and in the datum agreement, cutter and die problems in process parameters and tool management, movement problems in clamping strategy and panel design.
3. Routing: where the accuracy budget goes
Routing handles the large majority of boards. It is flexible, needs no tooling investment, and can produce any contour. Its accuracy cost concentrates in two places.
First, cutter diameter governs internal corners, not outline accuracy. This is the most commonly misread point. A router bit is round, so any internal square corner comes out as an arc with a radius equal to the cutter radius. Draw an internal right angle and the part will have a radius. That is not poor accuracy; it is geometrically unavoidable. If that corner has to clear a mating part, the drawing must state the maximum acceptable internal corner radius. Otherwise the shop picks a common cutter size, and that radius may not be one assembly can live with.
Second, narrow ribs and small features are limited by cutter diameter, not by tolerance. Where the material between two contours is narrower than the cutter, it cannot survive. Nor is "use a smaller cutter and go slowly" a free fix: a smaller cutter sharply reduces the board thickness it can handle and raises breakage risk. SKPCB's registered thickness range is 0.2–6.0 mm (PCB-THICK), and the thick end demands markedly more cutter rigidity, so the smallest usable cutter diameter grows with thickness. In other words: the thicker the board, the larger the minimum internal radius and the minimum rib width become. That is a direct coupling between outline capability and thickness, and the design side needs it.
4. Punching: the cost of single-stroke forming
Punching forms the contour in one stroke with a die. Per-piece cycle time is far below routing, which suits high volume, fixed outlines, and simple shapes.
Its constraints are concrete:
- It requires a die, so any outline change means a die change. Not suitable for development or low volume.
- It is sensitive to material and thickness. Thinner boards and tougher laminates punch well; thick boards and hard, brittle materials show edge delamination, whitening, or chipping.
- Contour complexity is limited. Fine protrusions and very small internal notches are difficult to implement or give very short die life.
- Edge condition differs from routing. A punched edge is formed by tearing and compression, not cutting. Where edge appearance or insulation spacing matters, that has to be agreed in advance.
The practical split is clear: development, low volume, complex contours, and thick boards go to routing; high volume with simple contours on thin material can be evaluated for punching. Note that punching does not automatically mean lower accuracy. A stable high-volume punching process can repeat very well; its weaknesses are flexibility and edge condition, not repeatability.
The design side should not specify a forming process as the only option on the drawing. State the constraints instead (whether the outline may change, volume band, edge quality requirement, whether internal slots are needed) and let the process side propose what is feasible.
5. Three features that fail before the outer edge does
Outline disputes focus on the outer contour, but in production the higher risk sits in three other places.
Internal slots and shaped openings. Every slot, rectangular opening, and irregular notch inside the board is routed, so both rules from section 3 apply: cutter diameter sets the internal corners, and material narrower than the cutter cannot survive. Worse, material around an internal slot has no surrounding support, so local rigidity is low and deflection and chatter are more pronounced than on the outer contour. The closer slot width is to the cutter diameter limit and the longer the slot, the larger the deviation. Slot width tolerance and end radii must be called out separately; they must not inherit the general outline tolerance.
One distinction matters here: a slot is not a hole. Drilled round holes follow the hole specification. SKPCB's registered mechanical minimum hole is 0.2 mm standard and 0.15 mm advanced (PCB-HOLE-MECH); hole diameter tolerance is ±0.075 mm for PTH, ±0.05 mm for NPTH, and ±0.05 mm for press-fit (PCB-HOLE-TOL); hole position accuracy is ±0.05 mm on the normal basis (PCB-HOLE-POS). None of those apply to a routed slot. Calling out a long slot as a "hole" and applying hole tolerance to it is a frequent drawing error.
Panel break-off features (mouse bites and V-cut). Units are held in the panel by break-off features, so after depaneling the residual burrs and break ridges sit on the board edge. That deviation belongs to the post-depanel outline, while most drawings only define the as-routed outline. The two differ:
- Mouse bites leave a visible perforation remnant with an uneven edge.
- V-cut leaves a beveled edge and introduces stress near the edge during separation.
If the board edge slides into a slot, mates to a housing, or acts as an assembly datum, the drawing must state that the edge is a break-off edge and what is required of it. Otherwise the shop applies normal practice, under which break-off residue is an accepted condition.
Thin boards and heavy copper boards. Thin boards are pushed more easily by side loads during forming. Heavy copper boards carry large copper areas at the edge whose cutting resistance differs substantially from the dielectric, so edge condition is not uniform. Neither type should inherit the default treatment used for conventional boards; both need to be flagged at order placement. Other copper-thickness trade-offs are covered in heavy copper manufacturing limits and not repeated here.
6. How thickness, copper weight, and material change achievable outline accuracy
Outline capability is not a standalone shop parameter. It moves with the board itself, in one direction and predictably:
| Input | Direction of change | Effect on the outline |
|---|---|---|
| Thickness increases | upward within 0.2–6.0 mm | smallest usable cutter grows → minimum internal radius and rib width grow; side cutting load rises |
| Layer count increases | standard 40 → maximum 64 | thickness accumulates, same as the row above; accumulated registration error also grows, which shows up when the hole is the datum |
| Copper weight increases | toward >12 oz | uneven cutting resistance at the edge; edge condition differs from dielectric areas |
| Material becomes harder or more brittle | ceramic types, some RF laminates | punching not applicable; routing requires relaxed edge-quality expectations |
| Unit size approaches the minimum | toward 5×5 mm | insufficient clamping area; panel and break-off design become the dominant factor |
| Board size approaches the maximum | above standard 600×570 mm | absolute dimensional movement is larger on big panels; above standard size requires engineering review |
Numeric values in the table come from SKPCB's registered capability set: layer count 40 in standard production and 64 maximum (PCB-LAYERS, with the maximum tier being an advanced capability subject to engineering review); thickness 0.2–6.0 mm (PCB-THICK); maximum copper weight >12 oz (PCB-COPPER, conditional); board size 600×570 mm standard and 610×1100 mm advanced (PCB-SIZE, advanced tier requires engineering review); minimum board size 5×5 mm (PCB-SIZE-MIN).
Read the table backwards. Do not fix an outline tolerance and then ask whether it is achievable. Identify which rows the board sits in first, then judge whether the outline requirement is realistic. A 6 mm thick board with 16 oz copper and a 2 mm slot cannot share an outline specification with a 1 mm board in conventional copper.
7. The five things a drawing must state explicitly
The previous six sections compressed into an actionable list. Omit any item and the shop applies a default, which may not be the one intended.
- The profiling datum. State explicitly what the outline references: panel tooling holes, a specified mounting hole, or a specified board edge. Without this item, the other four lose their meaning.
- Tolerance assignment. Which dimension carries the band, and whether it is bilateral or unilateral. Overall length and width, hole-to-edge, slot width, and notch depth each need their own; a single "unspecified tolerance" note is not enough.
- Internal corner requirements. Call out the maximum permitted radius on every internal corner, or state that the radius is unrestricted. This is a hard geometric constraint of routing; leaving it blank hands the decision to whatever cutter is common.
- Panel and depanel requirements. Whether mouse bites are allowed, whether V-cut is allowed, which edges are break-off edges, and what residue is acceptable on them. An edge used as an assembly datum must be declared separately.
- Critical dimension identification. Mark the two or three dimensions that actually drive assembly. Tightening everything is the same as prioritizing nothing — a shop cannot optimize all dimensions at once and will end up distributing margin evenly, leaving the critical ones no better off.
Two further items do not belong on the drawing but must accompany the order: volume band (which determines whether punching is worth evaluating) and edge quality requirement (which drives process selection and any extra treatment).
State the acceptance basis as well. SKPCB works to IPC-6012 Class II and Class III (QC-IPC), and the judgment criteria for appearance and structure differ between classes. All orders receive 100% electrical test (Q-ETEST, flying probe or fixture). Electrical test covers continuity only, not geometry — the outline requires dimensional inspection, and the two cannot substitute for each other. The defect classification framework is covered in defect classification and IPC acceptance class.
8. FAQ
How tight can profiling tolerance be?
There is no single answer, and SKPCB's published capability register contains no standalone "outline tolerance" value. Achievable accuracy depends on how the datum is defined, plus thickness, material, unit size, and panel design — the table in section 6 lists every contributing factor. The productive way to ask is to supply the board parameters and the two or three dimensions that genuinely matter, and let the process side answer for that specific combination. Asking for a number first and fitting the board to it later is the wrong order.
What happens if the drawing omits the outline datum?
The shop applies its default, normally the panel tooling holes, because that requires no per-board alignment and is fastest. If the design actually cares about a mounting-hole-to-edge distance, that distance absorbs all accumulated upstream error, and the drawing provides no basis for judging pass or fail. This is the single most common cause of outline disputes.
Why did my internal square corner come out rounded?
A router bit is round, so the minimum internal corner radius equals the cutter radius. That is geometry, not an accuracy shortfall. Thicker boards need larger cutters, so the radius grows with thickness. Where the corner matters, state the maximum permitted radius on the drawing.
Can I call out a long slot using hole tolerance?
No. Round holes are drilled and follow hole diameter tolerance (PTH ±0.075 mm, NPTH ±0.05 mm) and hole position accuracy (±0.05 mm, normal basis). A long slot is routed and is affected by cutter diameter, deflection, and low local rigidity — a different process path. Slot width, length, and end radii need separate callouts.
Does the post-depanel edge from mouse bites or V-cut count against outline tolerance?
It depends on how the drawing defines it. Most drawings only define the as-routed outline, so break-off residue (perforation remnants, V-cut bevels) is treated as a normal condition. If the edge must mate to a housing or serve as an assembly datum, state that it is a break-off edge and specify the requirement.
Is punching less accurate than routing?
Not as a general rule. A stable high-volume punching process can repeat very well. Its real weaknesses are flexibility (an outline change means a die change) and edge condition (a torn and compressed face rather than a cut one), plus unsuitability for thick boards and hard, brittle materials. Development, low volume, complex contours, and thick boards belong on a router.
Will 100% electrical test catch outline problems?
No. Electrical test verifies continuity; it does not measure geometry. The outline must be confirmed by dimensional inspection. The two cover different failure modes and cannot replace each other.