Industrial Heavy Copper PCB Second Source: What Transfers, What Gets Re-decided

The drawing is not the board

You have one supplier for an industrial heavy copper PCB. It works. Then

procurement asks for a second source, or the first fab raises price, misses a

window, or stops quoting the part. You send the same data package to a new fab

and expect the same board back.

You will not get it, and the reason is not quality. It is that your data package

never fully described the board. It described your *requirements*. Everything

between your requirements and a manufacturable board was decided by the first

fab, once, probably years ago, by an engineer you never spoke to. Those

decisions were never written down anywhere you hold.

That gap is what this article calls portability. A board with high

portability can move because the decisions that make it work are declared in

documents you own. A board with low portability cannot move without someone

making those decisions again, from scratch, without knowing what the first

answer was.

Heavy copper is where this bites hardest. Thin-copper boards have narrow

decision space: most fabs will land in roughly the same place. Heavy copper

widens it. Etch behaviour, lamination fill, plating distribution, thermal

relief geometry, and surface finish adhesion all shift with copper weight, and

each fab has its own compensation habits tuned to its own equipment. Two

competent fabs reading the same drawing will build two different boards, both

of which pass your drawing.

> Framing note. "Portability" and "undeclared default" are terms this

> article uses to organise the decision. They are not industry-standard

> designations and they do not replace any specification clause. Nothing here

> is legal advice.

Why the second fab asks questions the first one never did

The first fab asked those questions too. It asked them internally, answered

them itself, and never sent them to you. By the time you were a repeat

customer, the answers were frozen in its own routing and tooling records.

The second fab has no such records. It has your drawing, your Gerbers, your

stackup sheet if you have one, and nothing else. So it either asks, or it

assumes. Asking produces an engineering query and a delay. Assuming produces a

board that meets your drawing and behaves differently.

The expensive version of this is the third possibility: it asks, you forward

the question to the first fab, and the first fab declines to answer, because

its compensation practice is the thing that makes it competitive. That is a

legitimate position and you should expect it.

What was never yours to begin with

Before listing what to declare, separate the board into three kinds of content.

Content you own and already hold. Netlist, outline, hole pattern, copper

geometry, layer count and order, material class, finish type, marking,

acceptance class, packaging. This is in your data package. It transfers

cleanly.

Content you own but never recorded. Decisions you made verbally or by

accepting the first fab's proposal in an email, which were never folded back

into a controlled document. Which deviations you approved. Which cosmetic

conditions you agreed to accept. Which test coverage you waived. This is yours.

It transfers only if you go find it.

Content the first fab owns. Its etch compensation values, its lamination

cycle, its plating current distribution, its drill feeds, its tooling and

fixtures, its process records and travellers. This does not transfer. Do not

plan around getting it. Plan around the second fab producing its own version.

Most failed transfers are failures to notice the second category. People assume

anything they agreed to is written down somewhere, and it usually is not.

The undeclared defaults that get re-decided

These are the items that most often have no written owner. Each one was settled

by the first fab in a way that worked. The second fab will settle it again,

differently, unless you declare it.

Copper weight placement per layer. Your drawing may state a copper weight

for the board, or for inner and outer layers as a pair. Heavy copper designs

frequently use different weights on different layers. If the drawing does not

state the weight *per layer*, the second fab assigns them, usually in the way

that is easiest to laminate.

Which features the stated line width refers to after etching. A conductor

width on a drawing can be read as the finished width or the artwork width. In

heavy copper the difference is not a rounding error. The first fab resolved

this by choosing a compensation and never telling you which reading it used.

Dielectric thickness between specific layer pairs. A total board thickness

plus a layer count does not determine where the dielectric goes. Heavy copper

boards often need thicker dielectric adjacent to heavy layers for resin fill.

The first fab chose a distribution. If you did not fix it, the second one

chooses its own, and your interlayer spacing changes.

Glass style and resin content used to achieve the stated thickness. Two

constructions can reach the same thickness with different glass and resin, and

they behave differently under thermal cycling and drilling.

Thermal relief and connection geometry at heavy pads. Spoke count, spoke

width, and whether a pad is flooded or relieved. These get adjusted for

manufacturability. The adjustment changes thermal and current behaviour.

Plated hole copper distribution in thick copper stacks. Where the plating

is thinnest, and whether that location is controlled or incidental.

Surface finish process variant and its interaction with thick copper. Not

just the finish name, but the specific process route and any pre-treatment.

Finish names cover several processes with different adhesion behaviour on heavy

copper.

Edge and cutout formation method. Routed, scored, punched, or a

combination, and how heavy copper at the board edge is handled.

Accepted deviations and cosmetic conditions. What the first fab was allowed

to ship: repairs, touch-up, discolouration, resin recession, minor voiding.

This was agreed, often informally, and it is almost never in the drawing.

Test and inspection coverage actually performed. Electrical test method and

coverage, cross-section frequency, any thermal or current verification. If the

drawing does not require it, the second fab will not do it, even if the first

one did it as standard practice.

What to declare before you send the package out

Turn the list above into document changes. Three rules make this work.

**Put it in the controlled drawing or a controlled stackup sheet, not in the

RFQ email.** An email specification binds nothing at incoming inspection. If it

matters enough to transfer, it matters enough to be in a revision-controlled

document.

Declare intent alongside the number, not instead of it. State the per-layer

copper weight and state that the value is the finished weight. State the

conductor width and state whether it is finished or artwork. The second fab can

work with either, but it cannot guess which one you meant.

Declare the construction, not only the outcome. For heavy copper, fix the

layer-by-layer dielectric distribution and the material class, not only the

total thickness. If you genuinely do not care, say so explicitly, in writing,

so the fab's choice is on the record as permitted rather than assumed.

Where you do not know the first fab's value, you have two honest options:

measure it from boards you already hold, or declare your own requirement and

accept that the second fab's board will differ from the first fab's board in

that respect. Both are better than silence. Silence produces a difference you

discover in the field.

Tooling, fixtures and records: who holds what

Transfers stall on this more than on technical content, because the commercial

answer and the physical answer differ.

You may own a tooling charge without owning tooling. Paying an NRE or

tooling line on an order does not by itself transfer ownership of the physical

tools, the digital tooling files, or the right to have them released. What the

charge bought depends entirely on the words in your purchase order and terms.

If those words are silent, the fab's default reading is that it bought

*capability to build your part at that fab*.

Fab-specific tooling has little value elsewhere. Drill programs, routing

programs, test fixtures, and panel layouts are built around a specific

machine set, panel size, and process route. Even if released, the second fab

will mostly rebuild them. Expect to pay tooling again at the second source and

budget for it rather than negotiating for a release that will not help.

Test fixtures are the exception worth asking about. A bed-of-nails or

custom electrical test fixture is more portable than a drill program, and it is

tied to your netlist rather than the fab's panel. Ask specifically about this

one.

Process records are the first fab's. Travellers, compensation values,

lamination cycles, and plating recipes are its process knowledge. You can ask;

expect refusal; do not design the transfer plan around receiving them.

What you should hold regardless. Your own approved data package at the

revision actually shipped, the approved first-article or qualification

documentation, your approved deviation list, and inspection reports for

delivered lots. These are deliverables you can require in terms, and they are

the raw material for the comparison in the next section. If you do not have

them for the incumbent, start requiring them now, before you need the

transfer.

Address all of this in the purchase order terms at the *first* source, not at

transfer time. By the time you want to move, you have no leverage left.

The comparison evidence that proves it is the same board

The second fab's board will pass your drawing. That is not the question. The

question is whether it matches the board your product was validated with.

Build the comparison on three layers of evidence.

Documentary comparison. Place the two fabs' stackup proposals and

engineering query responses side by side against the same declared

requirements. Differences in interpretation show up here, before any board is

built, and this is the cheapest place to find them.

Physical comparison on first articles. Cross-section both fabs' boards at

the same locations and compare construction: per-layer copper, dielectric

distribution, plated hole copper, and the geometry at heavy pads. Compare

surface finish condition. Where the incumbent board is no longer available,

use retained samples or delivered stock, and say in the report which you used.

Functional comparison in your assembly and your application. Run the new

boards through your actual assembly process and your actual thermal and

electrical conditions. Heavy copper differences frequently show up as assembly

behaviour rather than as board defects: solder thermal demand, warpage during

reflow, and hole wetting.

Two practical disciplines make this evidence usable. First, decide what counts

as equivalent *before* you see the results, otherwise you will rationalise

whatever comes back. Second, document the differences you decide to accept as

permitted differences in a controlled document, so a later lot is judged

against the new baseline rather than the old one.

Prevent the next transfer from being this expensive

Most of the work above is retroactive documentation of decisions you were not

present for. The way out is to not create that debt again.

At the first source, on the first order: declare the construction per layer,

state whether dimensions are finished or artwork, write the deviation and

cosmetic acceptance terms into a controlled document, require the inspection

and test records as deliverables, and settle in the purchase order terms what

the tooling charge buys and what happens to data and fixtures if you stop

ordering.

Then review your incumbent heavy copper parts and ask one question per part:

if this fab stopped quoting tomorrow, what would I have to re-decide? The parts

with long answers are your real supply risk, and they are usually not the parts

with the single-source flag in the ERP.

• Compare construction options on the heavy copper PCB page
• Review multilayer construction choices on the multi-layer PCB page
• Check process scope on the capabilities page
• Submit a declared package for review via quote
• Scope a prequalification before award: prequalification scope and expiry
• Request process evidence before award: heavy copper supplier process evidence
• Separate transfer from reorder tooling reuse: reorder tooling reuse decision
• Decide what counts as a change on repeat orders: repeat order change control
• Resolve conflicts between drawing and data: specification conflict precedence
• Plan the receiving check: incoming inspection acceptance plan

FAQ

Can I just send the second fab the same files the first one used?

You can, and it is the correct starting point, but it is not sufficient. The

files describe your requirements. They do not describe the manufacturing

decisions the first fab made to satisfy those requirements. Send the files, and

separately declare the per-layer construction, the finished-versus-artwork

reading of dimensions, your accepted deviations, and your required test

coverage.

Will the second fab's heavy copper board be worse?

Not necessarily, and that framing will mislead you. The risk is not lower

quality, it is *difference*. A competent second fab produces a board that meets

your drawing and differs from the incumbent board in ways your drawing never

constrained. Whether that difference matters depends on how much margin your

design has in those areas.

Do I own the tooling if I paid a tooling charge?

Depends on what your purchase order and terms say. A tooling line on an invoice

does not transfer physical tools, digital tooling files, or a release right

unless the agreement says so. The common default reading is that you paid for

capability to build the part at that fab. Settle this in writing at the first

source, before you need it.

Should I ask the incumbent for its etch compensation values?

You can ask, and you should expect refusal. Compensation practice is tuned to a

fab's own equipment and is often treated as competitive knowledge. More

usefully, the values would not transfer well anyway, since the second fab's

equipment differs. Declare your finished requirement instead and let each fab

reach it its own way.

Is a second source worth it if the board has low portability?

That is a cost comparison, not a yes-or-no. Low portability means the transfer

cost includes re-declaring the construction, re-paying tooling, and running a

full comparison and requalification. Weigh that against your exposure if the

single source stops. The parts worth transferring first are usually the ones

where the declaration debt is smallest, not the ones with the highest spend.

What if the incumbent board is no longer available for comparison?

Use retained samples, remaining delivered stock, or your own qualification

records from the original validation. Whatever you use, state it in the

comparison report, because it changes how much weight the comparison carries.

Comparing against a drawing alone tells you the new board is compliant; it does

not tell you it is the same.

Can I run both sources in parallel on the same part?

Yes, and it is the most durable outcome, but it requires the same declaration

work plus one more thing: a documented decision about which differences between

the two sources are permitted. Without that, every lot gets judged against

whichever source arrived first, and incoming inspection becomes inconsistent.