Aluminum PCB Manufacturer Quotes: Where Does the Scope Actually Stop? | PCB168

Three quotes, nearly 2x apart, and the board is not the reason

You have three quotes for an aluminum LED lighting board. The highest is close to double the lowest. So you start questioning the dielectric, the aluminum grade, the LED source, and whether the cheapest supplier is cutting something you cannot see.

The far more boring truth is usually this: the three suppliers did not quote the same thing.

The first quoted a bare board — fabricated, masked, legend printed, electrically tested, shipped. The second quoted a populated board with the LEDs on it. The third included the thermal interface material and, working from your fixture drawing, mounted the board onto the heat sink before shipping. All three quotes are honest. All three are accurate. All three are answering a question you never finished asking.

This bites harder on aluminum lighting boards than on general PCBA work, because this product has an extra layer in the middle of the chain that belongs neither to the board nor to the fixture: the thermal interface. It is not on the fabricator's bill of materials. It is not on the luminaire assembler's incoming inspection sheet. And the real thermal resistance of the whole path is decided by it. If your RFQ does not say who owns it, it reliably falls into the gap — and the first time anyone asks about it is usually after a sample has been mounted, powered for a couple of hours, and measured hotter than expected.

This article is not about which aluminum PCB parameters to choose. That is a separate problem, covered in declaring the thermal path in an aluminum PCB RFQ. Here we do one thing: turn the delivery boundary into a sentence that can be priced, inspected, and attributed.

1. "Aluminum LED board" is at least four different deliverables

Order what you might be buying by how far the boundary extends. There are four stops. They are not service tiers — they are four different objects of purchase.

Stop 1: bare board. Aluminum-clad laminate through imaging, etch, solder mask, legend and routing. No components. What you get is electrical connectivity and mechanical outline.

Stop 2: populated board (PCBA). Stop 1 plus LEDs, driver-side components and connectors, through reflow and electrical test. What you get is a board that lights up.

Stop 3: board with thermal interface applied. Stop 2 plus the thermal interface material applied to the back in a specified manner — grease, pad, phase-change or thermally conductive adhesive, which behave very differently and need separate ownership language. What you get is a board ready to be pressed onto a heat sink.

Stop 4: assembly mounted into the fixture. Stop 3 plus the board fastened to the fixture's heat sink with screws, clips or retainers, torque and compression per drawing. What you get is a sub-assembly with the thermal path already closed.

The difference between stops is not simply a few extra operations. Every stop forward swaps out the entire set: pricing object, inspection method, burden of proof on failure, and the cost of rework.

2. What each stop actually prices, and what it lets you inspect

Quotes refuse to line up because "acceptable" is not the same event at each stop.

The right way to read that table is not "stop 4 is safest." It is: whichever stop you intend to judge acceptance at, every acceptance input for that stop has to be handed over before it.

Buy a bare board and you need the ability to prove the board is fine and the problem is downstream. Buy stop 4 and you skip a lot of argument, but you must hand over the fixture drawing, heat sink flatness requirements, fastening method and torque, and the test method for the finished item — otherwise the supplier builds to its own interpretation, and "its own interpretation" is, for liability purposes, the same as no standard at all.

The mismatch that keeps recurring

The most common failure is this: negotiate at stop 1 pricing, inspect at stop 4 standards.

Buying led by bare-board comparison picks the cheapest, then acceptance introduces "temperature must be within spec once mounted." That requirement is entirely reasonable. It simply does not belong to the stop you bought. The fabricator does not have your fixture, has never seen your heat sink, does not know your interface material or your torque. The board it shipped is fully conforming inside its own inspection scope.

No amount of additional data resolves this, because what is missing is not measurement. What is missing is ownership of the acceptance basis.

3. The thermal interface: the layer that falls into the gap

If only one section of this article survives, make it this one.

Thermal interface material is usually a cheap line on the BOM and the least controlled segment of the thermal path. The reason is simple: its performance is not set by the material alone, but jointly by the material, the application process, the surface condition on both sides, and the assembly pressure. Those four belong to four different parties.

Which produces a very stable misalignment:

• the material part number may have been chosen by your hardware engineer at design review;
• the application method — brush, screen, place, dispense path — is decided by whoever physically performs that step;
• the heat sink side surface condition is decided by your fixture supplier;
• the assembly pressure follows from fastening method and torque, which usually live in an assembly work instruction rather than in the RFQ.

Leave any one of the four unowned, and the whole thermal path is unowned.

Four sentences the RFQ has to contain

No numbers required. Only ownership. Each sentence needs an explicit subject:

1. Who specifies the interface material, and who buys it. If you specify a part number, say in the same breath whether equivalents are allowed and whose approval an equivalent needs. For the written substitution workflow, see BOM risk and alternate part approval on turnkey PCBA.
2. Whose document defines the application process. Do you issue the work instruction, or does the supplier issue it for your approval? "Per standard practice" is not an answer.
3. Who supplies the interface acceptance method, and at which stop it is judged. Appearance? Coverage? Mass? Thickness? Leave the method unspecified and the supplier will use whichever one it can pass most consistently.
4. Which document carries the assembly pressure assumption. Interface performance depends on compression, and compression follows from your fixture design. That assumption has to be written down with a revision number, or every later temperature anomaly collapses into an unresolvable "bad material" versus "bad assembly" standoff.

Different forms, different ownership language

The four common forms behave so differently in procurement that they must not be bundled into one sentence:

• Grease-like. Applied volume and uniformity depend heavily on process, and cleaning is a separate operation during rework. Who cleans, and whether cleaning counts as billable rework, has to be written.
• Pad-like. Size and position come from the drawing, misplacement is a visible defect, so it is comparatively easy to inspect — but cutting and placement tolerance ownership still needs stating.
• Phase-change. Performance establishes only after the first thermal cycle, which means the final state is not visible at outgoing inspection. The only honest acceptance point is after mounting, and that directly decides your boundary cannot stop at stop 3.
• Thermally conductive adhesive. Once cured, rework is essentially destructive. The RFQ must state who carries the consequence of a non-removable joint, and what the first-article confirmation flow is.

The last two are where a mis-drawn boundary costs the most. If you are using phase-change material or thermal adhesive, stopping delivery at "board with interface applied" is a hidden trap — you cannot inspect the final state, and the supplier's shipment genuinely does conform.

4. Assembly scope and the fixture interface: whose drawing speaks

Stop 4 looks like the least work for you. It introduces a new problem: the fixture was not designed by the board shop.

Asking a supplier to mount the board into the fixture is asking it to execute your or a third party's design and be answerable for the execution. Three things must be in place; if any is missing, this stop should not exist in your RFQ.

One: a revision-controlled assembly drawing. Not a sketch. A drawing with fastener locations, fastener specifications, tightening sequence and torque requirement, carrying a drawing number and a revision. The revision matters more than the content — without it, nobody can establish three months later which version was actually built.

Two: an acceptance basis for incoming fixture parts. The mounting face of the heat sink directly determines how the interface performs. Who states the flatness, roughness, burr and cleanliness requirements for that face; who has authority to reject non-conforming incoming parts; and who carries the cost and schedule when a rejection happens. All three need answers. If you buy the fixture separately and consign it to the assembler, you are the interface owner for that segment — the same class of problem, seen from a different side, as who owns the defect when fabrication and assembly are split.

Three: ownership of the finished-item test method. "Must light up normally and run at a normal temperature" contains two instances of "normal," and each has to point at a document. Test conditions — ambient, duration, mounting orientation, with or without the diffuser — change the result substantially. Without a stated method the data is not comparable, and a dispute has no way to converge.

One test for whether you should buy stop 4 at all

If you cannot supply a revision-controlled assembly drawing and a finished-item test method, you are not ready to buy stop 4. Stopping at stop 2 or stop 3 and doing the assembly yourself leaves a cleaner liability map.

5. How failure liability moves with the boundary

Take one symptom — the luminaire runs hotter than designed after a period of operation — and watch where it lands at each stop.

• Boundary at bare board. Supplier liability ends at the board's material and construction conforming to the order. The burden of proof on temperature is yours: you have to rule out placement, interface, assembly and fixture design first.
• Boundary at populated board. Joint and placement quality enter supplier liability. Voiding or offset under an LED that degrades conduction is theirs. Interface and assembly remain yours.
• Boundary at interface applied. Application quality enters supplier liability. The line here is fine: application quality is the supplier's; whether the chosen material suits the design belongs to whoever specified it. If you named the part number, the selection risk does not transfer.
• Boundary at mounted assembly. Assembly execution enters supplier liability. Fixture design, heat sink selection and overall thermal design are still yours, and outsourcing the assembly does not move them.

One principle runs through all four stops and deserves to stand alone:

> Moving the boundary moves execution liability. It does not move design liability.

You can hand more and more operations to a supplier, but as long as the design inputs come from you, the consequences of a poor design stay with you. A large share of procurement disputes starts from assuming that buying further down the chain buys peace of mind. Buying further down only sharpens execution liability; design liability has not shifted an inch.

The converse also holds: if you want the supplier to carry some design responsibility — asking them to recommend the interface material, for instance — that has to be a written handover of design input with an approval step, not a line in an email saying "use your judgement." For how that reads in contract language, see material ownership in outsourced assembly.

6. A scope declaration you can copy

Put the block below on page one of the RFQ. It prevents more rework than the ten pages of specification behind it. Anything in brackets is yours to fill; leaving it blank is the same as not writing it.

```

[SCOPE OF SUPPLY]

1. Delivery boundary: the deliverable for this RFQ is Stop [1/2/3/4], i.e.

[bare board / populated board / board with thermal interface applied /

assembly mounted into fixture]. Operations beyond this boundary are

not included in this quotation.

1. Material authority:

- LEDs: [buyer-specified, no substitution / buyer-specified, equivalents

require buyer's written approval / supplier-proposed, buyer-approved].

- Thermal interface material: form is [grease / pad / phase-change /

adhesive]; specified and purchased by [buyer / supplier]; substitution

approver is [ ].

- All other parts: per BOM rev [ ]; substitution rules per that revision.

1. Process document ownership:

- Interface application process is defined by [buyer work instruction,

doc no. [ ] rev [ ] / supplier document approved by buyer].

- Fastening method and tightening requirements are defined by

[buyer drawing, no. [ ] rev [ ]].

1. Acceptance basis:

- The acceptance basis for this deliverable is [document / drawing no. /

revision]; the judging method is [ ].

- Joint acceptance per IPC-A-610 class [ ] (class stated by buyer;

supplier does not self-select).

- Interface layer is judged on [appearance / coverage / other];

method supplied by [ ].

1. Explicitly out of scope (list it; do not leave it to default):

[luminaire thermal performance / optical parameters / incoming fixture

inspection / finished-item test / packaging / freight / ...]

1. Scope change: any work beyond the above boundary requires a

[written change order] before execution, recording the change, the

price impact and the schedule impact.

```

Item 5 is the one most often skipped and the one most worth writing. If you do not state what is excluded, default ownership gets interpreted along each party's interest at the moment of dispute.

7. Three things to align before comparing prices

When the quotes arrive, do these three before comparing unit price. Do not reorder them.

First, normalise every quote to the same stop. Have each supplier state which stop it quoted and what it excluded. If one quoted stop 2 while the others quoted stop 3, do not estimate the difference — ask for a requote at the same stop. Estimated adjustments are the single largest source of distorted comparison.

Second, confirm whether the interface material line is in the price at all. It is frequently excluded silently as "customer supplied," and the quote sheet does not show it. Ask every supplier the same question: who buys the interface material, whose document governs the application, and is that step inside the quoted price. Only when all three have answers is that line aligned.

Third, verify that the acceptance revision numbers match. All three may write "per customer drawing," but if they hold different revisions, the prices are not comparable and neither will the acceptance be. Require each supplier to state the drawing revision its quote is based on. The cost of asking is near zero; the payoff is not.

With those three done, the remaining spread is real spread. From there you can run normal supplier evaluation, or keep refining with the normalisation table in aligning scope boundaries and hidden charges in PCBA quotes.

When aluminum is not the answer

One note, not expanded: once the boundary is clear, if you find the cost is concentrated in assembly and interface rather than in the board, the thing to revisit is the thermal design route, not the board price. That is a design-stage question and outside this article's scope; the aluminum PCB thermal path declaration and extra RFQ inputs for heavy copper on an aluminum base each handle one of those directions.

Next step

To apply the scope declaration above to a live enquiry, start from the product overview and pick the matching product page, or submit directly through the quote request. For anything involving placement and assembly scope, see PCB assembly (PCBA); for the aluminum board itself, see aluminum PCB.

---

FAQ

Q: The supplier never asked about the boundary and just sent a quote. Is that normal?

Common, but not a good sign. A quote issued without asking is a quote against the supplier's own default scope. The first move on receiving one is to ask back which stop it covers and what it excludes — not to start comparing.

Q: I am only buying bare boards and handling interface and assembly myself. Do I still need interface language in the RFQ?

You need one sentence saying it is excluded. Explicit exclusion and silent non-inclusion carry very different weight in a dispute. It is also worth stating who owns the required surface condition of the board's back face, since that face is the interface contact surface.

Q: Why can phase-change material not stop at "board with interface applied"?

Because its final state establishes only after the first thermal cycle and is not inspectable at shipment. Stopping there puts your acceptance point and its performance point on different sides of an event. Either push the boundary to stop 4 with an agreed post-mount test method, or switch to a form that is inspectable at outgoing.

Q: Biggest procurement difference between thermal adhesive and a pad?

Reversibility. A wrong pad gets replaced. Cured adhesive makes rework essentially destructive to the board or the heat sink. With adhesive, the first-article confirmation flow and the "who owns a non-removable joint" clause must be in the order up front, not negotiated after a failure.

Q: I buy the fixture separately and consign it for assembly. How does liability work?

You own that interface. Three things minimum: the acceptance basis for incoming fixture parts, the assembler's authority to reject, and who carries loss and schedule after a rejection. Missing any one turns every fixture-related problem into mutual blame.

Q: Does nailing the scope down cost me flexibility?

No. Flexibility comes from the written change mechanism in item 6, not from a vague boundary. A vague boundary does not produce flexibility; it produces two parties reading the same sentence along their own interests when something goes wrong.

Q: With all this written in, will the RFQ get so long that suppliers decline to quote?

Experience runs the other way. A clearly scoped RFQ gets quoted faster, because nobody has to guess and nobody has to price in an uncertainty buffer. What actually slows quoting is rounds of clarification, and those are a product of vague scope.