EMI Shielding Change Cost Ownership: Who Pays After an EMC Failure | PCB168

1. Why compliance-driven changes always land at the latest possible point

EMC is not an attribute you can fully close out at design time. Layout, stackup, enclosure, cabling,

load behaviour and firmware all shape the final emissions and immunity result, and roughly half of

those cannot be observed until the board sits inside the actual product. So the real verdict on EMC

necessarily arrives after the system-level build exists — and by the time the system-level build

exists, bare boards are in fabrication, components are purchased, and assembly is scheduled.

That lag creates the category of spend buyers least want to face: **the change itself may be small,

but the point at which it lands forces every investment already made to be re-settled.** Adding a ring

of ground vias is nearly free at design time. After release and fabrication it means a revision, a

re-order of laminate, a disposition decision on work in progress, and a fresh application for a

certification slot.

Worse, the party triggering the change is usually neither the buyer nor the manufacturer, but a

third-party test result. The contract says "build to print". The print is not wrong. The boards are

not wrong. The product still fails. All three parties stand correctly inside their own obligations,

and the invoice hangs in the space between them.

So this article makes a single argument: **for compliance-driven changes, the size of the cost is

determined by the trigger point, and the ownership of the cost is determined by the ownership of the

decision right.** If decision rights were never assigned at order time, ownership has been left to

post-hoc negotiation.

2. Four remediation routes, and the change level each triggers

There are only four classes of action available after an EMC failure. They are often discussed as

peers in technical reviews, but in procurement terms they are not remotely equivalent, because the

change level each triggers decides which already-incurred investments get written off.

Route one: add a board-level shielding can. No electrical nets change, but the board surface does

— you need the can's pad ring, ground via pattern and keep-out, plus clearance checks against

neighbouring component heights. This is a combination of **board-surface revision + new material +

an added assembly operation**. It is chosen most often because it appears "not to touch the design".

It is under-costed most often for exactly that illusion.

Route two: change the stackup. Adjust reference planes, layer order or dielectric thickness to

improve return paths and interlayer coupling. Not a single net may move, yet this is a **fundamental

change to the manufacturing input**: material combination, impedance results, overall thickness and

even mechanical fit must all be re-confirmed. Laminate already cut is generally not reusable.

Route three: reroute. Touching routing touches net topology, which means **drawing revision +

possible component position changes + total write-off of bare boards already fabricated**. It is

frequently the cleanest answer technically and almost always the most expensive one commercially.

Route four: add filtering or suppression components. Ferrites, capacitors, common-mode chokes or

ESD parts at the power entry, interfaces or clock paths. If the original design reserved the

footprints, this is only a BOM change + assembly program update. If it did not, route four

immediately degrades into route three. That is why "was a footprint reserved" is a procurement

question before the order, not merely a design question.

Ranking the four routes by what they write off — rather than by technical elegance — is the only

workable starting point for an ownership discussion:

3. Shielding cans: the most chosen route, and the most under-costed

The shielding can gets first consideration after an EMC failure for a practical reason: it does not

touch electrical nets, so in a negotiation it sounds like "adding a part". From a procurement view it

is a change that pulls three parties in at once, and the quote that follows usually contains only one

of the resulting cost lines.

The lines that get missed, roughly in the order they surface. Supply ownership of the can itself:

is it buyer-supplied, procured by the assembly house on your behalf, or supplied by the board shop as

part of the package? Incoming inspection responsibility, shortage liability and the price basis are

all different across those three models, and this question is routinely skipped in the urgent

conversation that follows a failure. Authorship of the board-side features: who draws the pad ring

and the ground via array? A buyer revising the file, an assembly house suggesting it, and a board shop

executing it produce different results — and different liability when something is wrong.

The knock-on effect on surface finish: how the can is attached imposes requirements back onto the

soldering surface, and once surface finish is changed, the effect is not confined to that one ring

— it reaches whole-board solderability and the inspection baseline. This is the connected expense

buyers least anticipate. The change in inspection method: once the can is fitted, the area beneath

it moves from "visually inspectable" to "evidenced only by pre-assembly records". In other words,

adding a can changes the acceptance method, not just the part count. Rework authorisation: who is

permitted to lift the can for repair? Once lifted, does that board count as reworked, and must that be

disclosed? Leave this unwritten and every subsequent repair reopens the same argument.

Where the can was in the plan at order time, these are ordinary RFQ inputs and are handled separately

(see shielding can RFQ inputs). This article

deals with the harder case: the can was not in scope at all, and appears only after the test fails.

At that point no existing document assigns responsibility for it, so all six lines hang unowned.

4. Disposition ownership for purchased material and WIP

This is the largest real number and the least contracted. Once the change is decided, three asset

classes need a disposition decision, and the ownership logic differs for each.

Bare boards fabricated or complete. The question is not "can they still be used" but **whose

approved document were they built to**. If the board shop built exactly to a buyer-released file with

no manufacturing deviation, those boards are delivered or deliverable goods, and the write-off sits

with the ordering party by default. There is one exception: where the supplier completed the file on

your behalf and warranted the compliance outcome — that is a different transaction, and it has to be

stated at order time.

Purchased components. Split these in two. Commodity parts can usually move to a later build or

return through the channel, and the loss is restocking plus a possible price delta. Long-lead,

custom, or minimum-order-constrained parts bought specifically for this design are close to fully

sunk once a reroute invalidates their position. **This is precisely why "was a filter footprint

reserved" matters more commercially than it does technically** — a reserved footprint locks route four

into the BOM-change tier and stops it sliding into the drawing-revision tier. Alternate-part approval

paths are handled separately (see

alternate part approval).

Work in progress and consumed process investment. Material already cut, laminated or drilled

cannot step back to the prior operation; its value is counted by how far it got. Stackup changes have

the least transparent cost structure of the four routes precisely because what they write off is

mostly this invisible intermediate investment — and intermediate investment was amortised into unit

price in the original quote and never itemised.

The disposition decision itself also needs an owner. Who is authorised to decide whether this WIP is

completed, held, or scrapped? Every day the decision is delayed, storage and scheduling costs keep

accruing. **Disposition authority and a decision deadline belong in the same sentence as cost

ownership**, otherwise you get the familiar situation where both parties wait for the other to speak

while the invoice grows on its own.

5. Retest cost and the certification slot

After the change you must test again, and retesting carries two costs: the test itself, and **the

knock-on loss from the certification window slipping**. The first is a clear number and easy to

discuss. The second is usually an order of magnitude larger and appears in almost no contract.

The logic for test cost is relatively clean: it follows the ownership of the failure cause. If

non-compliance traces to a manufacturing deviation — actual layer order or finish not matching the

released file — retest cost sits with the manufacturer. If it traces to a design choice, it sits with

the designer. If it traces to no deviation by either party, which is the most common case because EMC

is a system property, it is project risk and the sharing method has to be agreed at order time.

Otherwise it gets settled afterwards by relative bargaining power.

The certification slot loss is harder, because no party "caused" it — the calendar did. Lab capacity

is finite, requeuing means waiting, and during the wait the product cannot ship. Contracts handle this

one of three ways: booked as project risk to the ordering party, following technical responsibility

for the failure, or capped and shared. All three are acceptable. **What is not acceptable is leaving

it unwritten** — the result of silence is not a fair split, it is that whoever needs the goods more

absorbs it.

One connected item gets forgotten: after a change, **do the existing test and inspection requirements

still apply?** Add a can and the visual inspection scope changes. Change the stackup and the

electrical test baseline may need re-confirmation. Add filtering and functional test items need

supplementing. These requirements should have been defined at RFQ stage (handled separately in

defining PCBA test and inspection requirements),

and after a change they need defining once more. The effort of that redefinition, and who performs

it, is itself an ownership question.

6. Three sentences before you order

Every dispute above traces to the same gap: **decision rights were never explicitly assigned at order

time.** Closing that gap does not require a long annex. Three sentences in the RFQ and the PO are

enough. They change no technical solution; they change where the later negotiation starts.

Sentence one: assign the final decision right over the compliance outcome. State which party is

responsible for achieving EMC compliance on this project: the ordering party carries it (the supplier

builds to print only), or the supplier warrants the outcome (in which case the supplier holds veto

over design choices). These two models cannot be mixed — you cannot require a supplier to guarantee a

pass while denying it the right to change the design. Once the model is fixed, sentences two and three

follow from it.

Sentence two: agree the cost tiers for compliance-driven changes. Three tiers by failure

ownership: traceable to manufacturing deviation goes to the supplier; traceable to a design choice

goes to the ordering party; systemic failures traceable to no deviation are shared by an agreed method

(booked as ordering-party risk, or capped). State which cost lines the tiers cover: bare board

write-off, sunk material, WIP disposition, retest cost, certification slot loss.

Sentence three: agree disposition authority and a deadline. Name who decides whether WIP is

completed, held or scrapped, and how long after notification that decision must arrive. Storage and

scheduling costs past the deadline sit with the party that missed it. The purpose of this sentence is

not to save money — it is to stop the invoice growing while both parties wait on each other.

One further item is not a clause but a pre-order action: **require the design side to reserve filter

footprints**. It is the only pre-emptive step that locks an entire remediation route into the low

change tier, and it costs essentially nothing. Change control for repeat orders is handled separately

(see repeat order change control).

7. FAQ

Is an EMC failure the board shop's problem?

By default, no. If the shop built to the released file with no traceable manufacturing deviation, the

boards conform and the product-level failure is a system outcome. Only when the measured layer order,

material or surface finish differs from the released file is there a manufacturing deviation to own.

So the first step is always comparing the physical article against the released file, not opening

the cost discussion.

Is adding a shielding can the cheapest remediation?

It is usually the lowest change level, which is not the same as the lowest total cost. It pulls in

five cost lines: can supply, board-side feature authorship, surface finish knock-on, inspection method

change, and rework authorisation. Comparing only the can's unit price understates the real spend, and

those five lines are the ones most often skipped in the urgency after a failure.

Can purchased components be returned?

Commodity parts can usually move to a later build or return through the channel, at the cost of

restocking and a possible price delta. Long-lead, custom and minimum-order-constrained parts bought

for this design specifically are largely sunk. That is why a reroute is nearly always the most

expensive answer commercially.

Who normally pays for retesting?

It follows the failure cause. Manufacturing deviation to the supplier, design choice to the ordering

party, and systemic failure with no traceable deviation must have its sharing method agreed at order

time. Absent an agreement, the real outcome is set by bargaining power rather than fairness.

Can the loss from a delayed certification slot be claimed?

Only if ownership for that loss was assigned at order time. No party's operational error caused it,

so without a clause it is very hard to attribute. The workable approach is choosing one treatment in

advance: ordering-party risk, following technical responsibility, or a shared cap.

EMC testing has not happened yet at order time — does writing these clauses look like distrust?

These three sentences assign decision rights, not blame. Clear decision rights help the supplier too:

they establish that the supplier does not pay for design choices it had no veto over. An ambiguous

responsibility boundary is a risk to both sides when a dispute arrives.

8. Sources and basis

• The cost-ownership logic here rests on a general contractual principle: **change cost follows the

decision right**, not the implementing party. No standard clause, limit or figure is quoted.

• EMC standards (CISPR, the EN 55032 family and similar) are referenced by name only, to establish

that compliance is judged at system level. They are not used as the basis of any technical claim.

• The change-level classification of the four remediation routes is based on the procurement view of

which class of incurred investment each route writes off, not on any technical comparison of

their suppression effectiveness.

• This article makes no claim about PCB168's EMC test capability, laboratory, accreditation,

equipment, capacity, lead time, yield or customer cases.

• No frequencies, limits, attenuation figures, cost amounts or lead-time days appear anywhere.

9. Putting the three sentences to work

If your project has not been ordered yet, now is the cheapest point at which these three sentences can

be written. Put the compliance decision right, the change cost tiers, and disposition authority with a

deadline into the RFQ, and ask the design side to reserve filter footprints while you are there.

If the order is placed and the test has already failed, the sequence is: compare the physical article

against the released file to establish whether a manufacturing deviation exists; assess what each of

the four routes writes off; then make a time-bound decision on the three asset classes — **the

decision itself is more urgent than its content**.

To settle these inputs during quoting, see the

PCBA product page for how assembly-stage responsibilities divide, or

submit your files through the quote request noting that you need cost tiers agreed for

compliance-driven changes. Our capabilities overview marks which stages we hold

decision rights over and which require your release.

Layout-side EMC techniques and the physics of shielding selection are outside this article's scope;

see PCB EMI shielding design and the

EMC shielding design guide.

---