Heavy Copper PCB for Power Supply: The Voltage Inputs Your RFQ Is Missing

Open a typical heavy copper RFQ for a power supply board and almost every line is about current: finished copper weight in oz, peak and continuous current, acceptable temperature rise, trace width, whether vias need filling. Those inputs are correct and necessary — we cover how to turn them into quotable terms in current and thermal budget versus copper weight and trace width and spacing limits at heavy copper weights.

But a power supply board differs structurally from a general heavy copper board: high and low voltage, input and output, live and operator-accessible sections all coexist on the same panel. Whether the board is usable depends not only on whether the copper is thick enough to carry the current, but on whether the gaps are far enough apart and insulating enough. In most RFQs, that second half is simply absent.

The consequence is concrete. A board passes visual inspection, passes dimensional inspection, measures correct on copper weight, clears flying-probe open/short testing — and then breaks down during safety certification hipot testing or in a humid field environment. Reviewing the documents afterward: the drawing never marked which gap was safety-critical, the RFQ never stated working voltage or pollution degree, the purchase order never listed a hipot test. On every attribute that was actually agreed, the board conformed.

That is not a gap in the supplier's inspection. It is a gap in the buyer's declaration. Attributes nobody declared do not get inspected, and they do not get attributed when they fail.

Why Thick Copper Breaks the Assumption That Drawn Spacing Equals Built Spacing

On thin copper, treating the gap between two conductors as a flat two-dimensional distance works well enough. The copper is thin, sidewall height is negligible, and the difference between drawn spacing and built spacing is too small to discuss separately.

Heavy copper breaks that approximation, for three reasons that act at once.

One: sidewalls are not vertical. Etching proceeds from the top down, so copper near the surface is exposed to etchant longer than copper near the base. The resulting conductor cross-section is trapezoidal rather than rectangular — narrower at the top, wider at the base. Which means the gap between two adjacent conductors is at its widest near the board surface and at its narrowest down at the foil base. The spacing you measure under a microscope from above is the widest part of that gap, not the tightest. Fabricators apply etch compensation to offset the loss, but compensation is applied from process experience, and the actual resulting profile still has to be measured to be known. The thicker the copper, the taller that profile and the more pronounced the top-to-bottom difference.

Two: the minimum spacing occurs at a height you never specified. If the gap is a profile rather than a line, then "minimum spacing" is a quantity that requires a stated measurement height. What does the number on your drawing refer to — the distance between conductor top surfaces, between base surfaces, or an average? Default interpretations differ between fabricators, and on heavy copper those three numbers already differ enough to change a pass/fail result.

Three: solder mask coverage on a thick copper sidewall is a process outcome, not an automatic fact. Getting mask ink to climb a tall step and hold adequate thickness at the corner is a known heavy copper difficulty — we treat it on its own in heavy copper solder mask coverage declarations. Why does it matter for voltage? Because insulation standards apply different spacing requirements to coated surfaces than to bare ones. If the sidewall of a safety-critical gap is not genuinely covered, that gap is not, electrically, the category your drawing assumed it was.

Stack those three together and the conclusion is that insulation distance on a heavy copper power board is a combined profile-geometry plus surface-condition problem. It is not settled by a single dimension on a drawing. It has to be declared, assigned a measurement method, and accepted.

Creepage and Clearance Are Two Different Things, and Buyers Must Declare Them Separately

These two terms get used interchangeably in RFQs. They are different physical paths, driven by different factors, with different sensitivity to copper weight.

• Clearance: the shortest distance between two conductors through air. It concerns air breakdown, so it is affected by altitude (air density) and essentially unaffected by surface contamination.
• Creepage: the shortest path between two conductors along an insulating surface. It concerns whether contaminants, moisture, or carbonized tracking will progressively form a conductive path, so it is affected by pollution degree and the material's tracking resistance, and essentially unaffected by altitude.

Why must a buyer separate them? Because for the same conductor pair on the same board, the two distances can be set by entirely different geometry. If a slot runs between two conductors, the air path may cut straight across the opening while the surface path must travel down and up the slot walls — far longer. On a flat featureless surface, the two may be nearly equal. If you put one number on the drawing, the supplier cannot tell which path you intended to constrain, and cannot tell which features (slots, routed board edges, mask openings) you deliberately placed to extend creepage versus which you drew incidentally.

Heavy copper also affects the two differently. Sidewall taper mainly shifts where the narrowest point of the air path sits; mask coverage quality mainly determines whether the surface counts as coated. Collapsing both into one declared number compresses two independent risks into one unverifiable quantity.

The governing basis follows the end product's standard family: industrial control and measurement equipment typically falls under IEC 61010, information and audio/video equipment under IEC 62368, with insulation coordination principles in both deriving from the IEC 60664 series. On the board side, conductor spacing design and fabrication judgment are framed by IPC-2221 and IPC-6012. Standard names only here, with no values reproduced — because the actual required distances depend on working voltage, pollution degree, insulation type, and material group, and those inputs are known only to the buyer. Which is the entire content of the next section.

The Voltage Inputs the Buyer Must Supply

A supplier cannot decide your insulation requirements, because every input to those requirements comes from the end application, and only you know the end application. The following must be declared by the buyer and cannot be inferred.

1. Working voltage, and its nature. Not just a number. DC or AC, RMS or peak, whether transient overvoltage is expected, and between which two nets the voltage appears. A bare "400V" does not locate the requirement: 400V DC bus to chassis and 400V AC line to line are different problems.

2. The end product standard family this board belongs to. As noted, insulation coordination rules differ between IEC 61010 and IEC 62368. Your supplier cannot see your finished equipment and does not know which certification lab you are headed to. Omit this and the supplier can only infer from general industrial practice — and an inference carries no accountability.

3. Pollution degree assumption. Is the board inside a sealed enclosure, or in an open environment with dust and condensation risk? This directly changes creepage requirements, and it is the same body of information as the temperature, humidity, and contamination declarations in operating environment inputs for industrial heavy copper boards. Do not issue contradictory versions across two documents.

4. Altitude assumption. Affects clearance, not creepage. Leave it out for a high-altitude application and you have defaulted to sea level.

5. Whether conformal coating, potting, or secondary insulation exists. If the finished equipment is coated or potted, the real insulation boundary extends beyond the bare board. But coating is your assembly step, not the fabricator's deliverable — so the purpose of this declaration is to tell the supplier that the gap will be protected downstream, not to make the supplier warrant that protection. If the board shop is to perform the coating, that is an independent scope item and must be written into scope explicitly; see how to bound assembly supplier scope.

6. Insulation type. Functional, basic, or reinforced — the same gap carries different requirements under each. This is a property only your equipment-level safety design can determine.

Not one of those six can be filled in by a board shop. With all six missing, the only thing a supplier can do is treat the board as a conventional industrial board — and conventional industrial default spacing treatment is not the same thing as power supply insulation requirements.

Safety-Critical Gaps Must Be Marked on the Drawing, Not Inferred From Net Names

This is the most actionable point in the article.

Suppose you did declare all six inputs above. The supplier now knows working voltage, standard family, and pollution degree. What it is looking at is a Gerber set with hundreds of nets and thousands of gaps. How is it supposed to know which gap between which two nets is the safety-critical one?

Common wrong assumptions:

• "The net names make it obvious." Net names are chosen by the designer. Conventions like HV_IN, PRI_L, or AC_N vary completely between companies, and Gerber data typically carries no net information at all.
• "The spacing size makes it obvious." Plenty of places on a board have generous spacing; some are insulation requirements and some are just relaxed routing. A supplier cannot distinguish "3mm here because of safety" from "3mm here because nothing needed to go there."
• "They should just build to the tightest requirement." Applying the strictest treatment board-wide degrades manufacturability and cost sharply, and a supplier will not unilaterally raise your cost — it is more likely to apply conventional treatment.

The correct approach is to mark insulation-critical regions explicitly in the fabrication data. The form can vary: list the constrained net pairs in the assembly drawing or a specification document, add a dedicated Gerber layer marking the critical gaps, or write per-item fabrication notes stating "spacing between net A and net B is a safety-critical dimension, governed by standard XX, measured by method YY." The form does not matter. Explicitness does.

Declare the measurement convention at the same time: does the marked spacing refer to conductor top surfaces, base surfaces, or the narrowest point? Those three answers differ on heavy copper, and leaving it unstated hands the judgment to the other party's default. The nature of this problem is identical to aligning finished versus base copper thickness on the PO: same number, different datum, different verdict.

Also make sure the fabrication notes and the drawing itself do not contradict each other. If a safety-critical dimension appears in both a drawing callout and a text note with different conventions, the board's basis for judgment is indeterminate. For how to hunt down that class of conflict, see the fabrication notes conflict checklist.

Does Solder Mask Count as Insulation? That One Sentence Changes Which Spacing Class Applies

As noted, insulation standards apply different spacing requirements to coated versus bare surfaces. That makes the role of solder mask something you must settle explicitly.

Buyers need to separate three distinct claims:

1. "The board has solder mask." A delivery fact, true of essentially any conventional board.
2. "The sidewalls and base of the critical gap are continuously covered by mask." A process outcome, and not automatic on heavy copper — coverage thickness over tall steps and at corners is a known difficulty.
3. "Solder mask is recognized as an insulating layer, so the coated spacing requirement applies to this gap." A safety design determination, which only the equipment-level safety designer can make, never the board shop.

Collapsing those three into "there's mask, so it counts as coated" is the most common error. The correct handling:

• If your safety design does not rely on mask as insulation, design the spacing to bare-surface requirements. Mask coverage quality is then an ordinary cosmetic and manufacturability matter.
• If your safety design does rely on mask as insulation, then this must be escalated on the purchase order into a requirement with an acceptance method: declare which gaps have critical mask coverage, state how it will be confirmed (visual judgment? cross-section?), and define the disposition when it fails.

In the second case, the basis follows IPC-A-600 for visual acceptability criteria and IPC-6012 for finished board requirements. If cross-sectioning is required, that is a separate per-lot evidence item and the order must state who produces it, when, and who retains it. Do not assume a board shop will treat mask coverage as a safety item on its own initiative — absent a declaration, it is a cosmetic item.

Four Elements That Put Hipot Testing on the Purchase Order

Hipot (dielectric withstand) testing is the direct way to verify insulation. Its actual status in procurement is often: the buyer assumes the board shop does it, the board shop assumes the customer did not ask for it, and neither is wrong, because the order never listed it.

To make it a real deliverable, the order needs four things.

1. Test conditions. Applied voltage, voltage type (DC/AC), duration, and between which nets or net groups it is applied. That last element is the one most often omitted — "perform a hipot test" without specified test points specifies no test content at all. The actual values come from your end product standard and safety design, not from the board shop.

2. Test coverage and sampling. One hundred percent or sampled? If sampled, sampled per what unit — per lot, per panel, per shipping unit? The cost difference is large, and this is the same class of decision as agreeing electrical test and coupon terms: inspection intensity is something you can buy or decline to buy, but you have to buy it at order time, not claim it retroactively after a problem.

3. Acceptance criteria and failure disposition. What constitutes a pass (no breakdown? leakage below a limit?), whether a single failure rejects the lot or just removes that board, whether removed boards are replaced, and whether replacement affects lead time. Worth noting: hipot is a stress test with destructive potential — even a board that passes has been through one electrical stress event. So "one hundred percent testing" as a requirement is something you should confirm is acceptable to you.

4. Records and retention. Report format, delivered per lot or held on file, retained how long, whether anomalous data is included. The weight of this item only becomes visible during a dispute: without original records, "these boards passed hipot at shipment" is an unsupportable statement. We cover the decisive role of evidence retention in failure attribution and evidence retention terms.

With all four written, hipot becomes an order line item with a price, a deliverable, and an acceptance basis. With only "hipot test required," it may or may not have been priced, and you cannot tell which at receiving.

Dimensionally Conforming but Hipot Failing: Where Responsibility Was Actually Decided

Back to the opening scenario. The board passed every routine inspection and failed hipot after installation. How does responsibility divide?

The answer lies before the order, not after the failure. Work down this order of precedence:

First: was the safety-critical gap marked on the drawing, with a measurement convention? Not marked — the board was never processed as a safety item on the manufacturing side; the supplier judged it by conventional spacing and delivered, and it conforms on every agreed attribute. Marked with a stated method — proceed.

Second: does measured spacing match the callout, using the agreed method? It does not — that is a manufacturing deviation and sits with the supplier. It does — proceed.

Third: does the callout value itself satisfy the end standard requirement? It does not — that is a design issue sitting with the buyer's safety design; the supplier built to drawing without fault. It does — proceed.

Fourth: what was the actual failure path? Surface tracking where the design relied on mask coverage returns you to whether mask was declared as a critical item with acceptance. Reliance on equipment-level coating or potting with an assembly deviation puts it on the assembly side. Insufficient material tracking resistance returns you to the material declaration; see heavy copper laminate and Tg declarations.

The use of this table is not litigation. It is the reverse map of what to write at order time: every "not stated" at any level becomes "you absorb it" at failure time.

A Usable RFQ Supplement Block

Collapsing the article into a declaration skeleton you can paste into an RFQ (fill the brackets; an unfilled item means you waived that constraint):

• End product standard family: [ ]; insulation type: [functional / basic / reinforced]
• Critical net pairs and working voltage: [net A]–[net B], [voltage and nature]
• Pollution degree assumption: [ ]; altitude assumption: [ ]
• Coating/potting at equipment level: [yes/no]; if yes, performed by [buyer/supplier]
• Location of safety-critical gap markings: [layer name / assembly drawing number / note item number]
• Spacing measurement convention: [top surface / base surface / narrowest point], [method]
• Solder mask as insulating layer: [yes/no]; if yes, coverage confirmation method: [ ]
• Hipot test: [conditions] / [sampling] / [criteria and failure disposition] / [records and retention]
• Document precedence for adjudication: [drawing > specification > quote confirmation], or your own order

That last line matters more than it looks. When the drawing, the specification, and the quote confirmation disagree, which one governs? The general handling is in how a fabrication specification sheet binds the quote. On an attribute like insulation, where a failure scraps the finished equipment, leaving document precedence unstated hands adjudication to the dispute.

With those inputs prepared, request a quote or review the heavy copper board capability and specification page and the resulting quotes will actually be comparable.

FAQ

Q: Once copper gets thicker, is it enough to scale up the drawn spacing proportionally?

A: The relationship is not proportional. Enlarging spacing consumes routing space and may force added layers, while the three things that actually need resolving are where the narrowest point of the profile sits, what the measurement datum is, and whether the surface is effectively covered. Enlarging the callout without agreeing a measurement method still leaves the basis for judgment indeterminate.

Q: Will a board shop apply safety requirements to a power board on its own?

A: Absent a declaration, the supplier can only apply conventional industrial treatment. It cannot see your finished equipment, does not know working voltage or pollution degree, and cannot infer from Gerber data which gap is safety-critical. Undeclared attributes are not inspected.

Q: Can the RFQ state creepage and clearance as a single number?

A: Not advisable. For the same conductor pair those two distances may be set by entirely different geometry — a slot can greatly extend the surface path while leaving the air path unchanged. With one number, the supplier cannot tell which path you meant to constrain, nor recognize which features you placed deliberately to extend creepage.

Q: Can solder mask count as insulation?

A: That is a determination by the equipment-level safety designer, not a board shop delivery claim. If your design relies on it, escalate "mask coverage on critical gaps" into an order requirement with a stated confirmation method. If it does not, design spacing to bare-surface requirements.

Q: Should hipot be one hundred percent or sampled?

A: It depends on your risk tolerance and cost, but confirm one thing first: hipot applies electrical stress, and boards that pass have still been stressed once. Settle what full testing means between the parties before negotiating sampling.

Q: I already wrote "shall comply with applicable IEC safety requirements." Is that enough?

A: No. That sentence supplies none of the working voltage, pollution degree, insulation type, or critical net pairs — and those four are the entire input to the requirement. A supplier cannot reverse-engineer specific spacing requirements from a compliance statement, nor determine which two points to test.

Q: What if we are already in production before discovering the critical gaps were never marked?

A: First establish the basis of judgment for the current lot using the precedence order above, then decide whether to disposition the current lot or change documents for subsequent lots. Any document change must run through change control, or the next lot reverts to the old files; see change control terms for repeat orders.