HDI PCB Supplier Qualification: Capability Evidence to Request Without a Site Visit

Send the same HDI package to ten suppliers and you will usually get ten replies that look alike: all claim HDI capability, all attach a parameter table, all describe a mature quality system. The real difference is not in those statements. It is in whether they are willing, and able, to produce the records behind them.

An on-site audit is more direct, but many projects have no travel window: the sample schedule is tight, the supplier is in another country, or the evaluation stage does not yet justify trip cost. The only workable alternative is to convert "what I would look at in the plant" into "documents I can request by email and cross-check." That conversion is what this article provides: what to ask for, how to read it, and which answers should immediately downgrade a candidate. No PCB168 layer counts, trace widths, aspect ratios, yields, lead times, or certifications are stated here, and no specific supplier is evaluated. Public standards are cited only as external methodology references.

1. Separate capability claims from capability evidence

"We can build 10-layer HDI with single-level blind vias" is a claim. Claims cost almost nothing to make, which is why they carry almost no information during screening. Capability evidence is what makes that sentence falsifiable: the structures actually built, the process windows used at the time, the cross-sections or test records that correspond to them, and who signed those records and when.

Three questions will tell you which one you received:

• Is it traceable? Does the data belong to a specific lot on a specific date, or is it a summary table with no origin?
• Is it re-checkable? Does it name the acceptance basis it used (a class in IPC-6012, a visual criterion in IPC-A-600), or does it just say "pass"?
• Is it repeatable? Is it the best result from one sample build, or does it show how that structure behaved across several lots?

A reply that fails all three does not deserve deep investigation. Put it on a watch list and move on. The goal of screening is not to fully evaluate everyone; it is to isolate the few candidates worth a second round as quickly as possible.

2. Ask the capability matrix at the right granularity

Most off-the-shelf capability sheets are maximum-value tables: every parameter is set to its historical limit and stacked on one page. The problem is that these limits usually cannot hold simultaneously. The finest trace width rarely occurs on the heaviest copper; the highest aspect ratio rarely occurs at the smallest drill diameter. So the question has to shift from "what is your maximum?" to "what is achievable on my specific combination?"

In practice, break your board into its constrained dimensions and ask about their combination in one pass:

• Layer count and lamination cycles: how many press cycles are routine at this layer count, and does my microvia layer arrangement add one?
• Microvia structure: what is routine for stacked versus staggered, and what maximum level count is accepted?
• Trace width and spacing: the routine floor at my finished copper weight, and what extra process or acceptance step applies beyond it.
• Hole size and aspect ratio: routine ranges for mechanical and laser-drilled holes separately.
• Material: is my specified laminate a stocked grade, or does it need additional procurement lead time?
• Do surface finish and impedance requirements affect any of the above?

A good answer comes with conditions, such as "this combination works but the trace width floor moves up one step" or "possible, but it needs an extra lamination cycle." Conditional answers deserve more trust than a flat "no problem," because they show the supplier actually walked through your structure. For how these structural dimensions are described on our side, see our manufacturing capabilities page and the HDI PCB product page.

3. Structural evidence: coupons, cross-sections and plating records

HDI risk concentrates at the interconnect: whether the microvia bottom connects reliably, whether blind and buried via registration stays stable, and whether dielectric thickness remains within expectation after multiple lamination cycles. None of that is covered by visual inspection. Only destructive verification records can speak to it.

Structural evidence worth requesting:

• Cross-section reports with images, where the image makes the judgement location visible: the microvia-to-target-pad interface, the barrel plating, the dielectric layer thickness. A sentence saying "cross-section passed" is worth nothing.
• Test coupon design and disposition: by what rule are border coupons placed, do they cover the hardest structure on my board, and do coupon results ship with the lot?
• Plating thickness records: how barrel copper is measured, at which positions, and against which class.
• Interconnect reliability practice: whether thermal stress or comparable interconnect verification has been performed, and against what work instruction.

When reading these reports, the point is not that the numbers look good. The point is whether the acceptance basis is named. IPC-6012 (rigid board performance), IPC-6016 (HDI layer structure performance) and IPC-A-600 (visual acceptance) are publicly available sources of criteria. A report that states which document and which class it judged against is far more verifiable than one that only says "OK." These documents are named here purely as public methodology references and imply no certification claim by PCB168.

If a supplier refuses all cross-section imagery on confidentiality grounds, offer a fallback: a historical example with customer identification removed, or performing destructive verification on your own first article and delivering the report. A supplier who accepts no form of structural verification is not a fit for HDI work.

4. Process control and change notification: better predictors than a single report

One good sample build demonstrates engineering capability. Consistency across lots demonstrates process control. Remotely, you can confirm the latter from three directions:

• How critical steps are controlled: for laser drilling, desmear, electroless copper and plating — the steps that most affect microvia reliability — are there parameter windows and first-article confirmation, and who authorises release?
• Change notification: if the laminate grade, an outsourced process step, or a key piece of equipment changes, will the buyer be notified proactively? Are the trigger conditions and notice period written into the contract or quality agreement? This item is frequently skipped, and it is where production-phase surprises originate. The part number has not changed; the inputs have.
• Nonconforming material handling: what are the segregation, disposition and rework rules when a lot deviates, and does rework require buyer consent?

The question here is not "do you have documented procedures." It is "who signs for each of these." A supplier who can name the role and the trigger condition immediately is usually running the process as described.

5. Inspection and test scope: translate "full inspection" into line items

"We inspect everything" is another statement with near-zero information content. Inspection scope has to be broken into selectable line items, with standard versus separately agreed clearly marked:

• Electrical test: flying probe or dedicated fixture, and whether net coverage is 100%.
• Which layers and which defect classes automated optical inspection covers.
• Impedance test: whether a report ships with the lot, and how coupons or test points are selected.
• Sampling rules for dimensional and hole position checks.
• Which records the shipping document set contains, retention period, and whether records can be retrieved by lot.

Sending that list and asking the supplier to mark each item "standard / by agreement / not offered" is far more effective than asking "how do you inspect?" Line up three suppliers' markings side by side and the differences become obvious. For background on the inspection methods themselves, see PCB quality inspection methods.

6. Prototype-to-production consistency evidence

Many problems appear not during sampling but at the transition into production. A sample may have run on a smaller panel, with more manual attention, watched by senior staff. Production runs under different conditions. To assess that gap, ask the supplier to state:

• Whether samples and production run on the same line under the same process documentation; if not, where the differences are.
• Whether production panelisation differs from the sample, and whether that changes any manufacturability judgement.
• Whether the first production lot has a dedicated release action, and which records it is released against.
• Whether they will freeze key process inputs (laminate grade, lamination structure, surface finish) in the contract or quality agreement, with changes requiring prior written confirmation.

That last item matters most. It converts all the earlier evidence from one-off evaluation material into an enforceable term. A supplier willing to freeze key inputs is effectively confirming that the evidence provided represents normal practice rather than a best-case demonstration.

7. Communication and engineering response interface

HDI projects almost always generate engineering questions before production: stackup adjustment, drill size consolidation, impedance scheme confirmation. How fast those get resolved often affects schedule more than unit price does. Remotely, you can observe this directly during the RFQ, because the RFQ is a sample of the future working relationship:

• Did the reply raise specific engineering questions, or only return a price?
• Is there a named engineering contact, or does everything pass through sales?
• Do the clarifications point at real risk areas (a particular microvia layer arrangement, a trace width and copper weight combination), which shows the files were actually reviewed?
• Are DFM comments graded as "must change" versus "recommended"?

A supplier who flags problems in your files during the RFQ usually costs less overall than one who returns a low price immediately, because the problems surface before the order rather than after delivery. To move straight to that step, submit files through the online quote form or agree an engineering contact through the contact page.

8. Shortlist scorecard and record keeping

Collapsing all of the above into one archivable table is what makes the method reusable. A workable column structure:

When scoring, distinguish "not provided" from "provided without a stated acceptance basis." The risks differ in kind: the first may simply be a process gap, while the second suggests the record itself may not be verifiable. Archive the completed table together with the original attachments. The next project can reuse it, and if a quality event occurs you can retrace why a supplier was shortlisted.

The table does not need to cover every supplier. In practice, filter quickly with the three questions from section 1, then run the full table on the remaining two or three. Evaluation resources are always limited; spend them on candidates with a real chance.

FAQ

Q1: Can you really shortlist an HDI supplier without a site visit?

You can build a defensible "worth a second round" list, but it does not replace final production qualification. Remote evidence verification narrows ten candidates to two or three and gives any later step — on-site audit, third-party audit, or first-lot release — specific targets to check. It reduces evaluation cost, not all risk.

Q2: A supplier refuses cross-section images for confidentiality reasons. Disqualify them?

Not automatically, but require a fallback: a historical example with customer identification removed, or destructive verification performed on your own first article with the report delivered. If both fallbacks are refused, the supplier accepts no form of structural verification, which is a material risk on an HDI build.

Q3: Can the minimum trace width and maximum aspect ratio on a capability sheet be taken at face value?

Not recommended. Those tables usually list historical limits per parameter with no guarantee they hold together. Send your actual combination — layer count, finished copper weight, hole size, microvia levels, material — and ask for the boundary on that combination, plus whether extra process steps apply.

Q4: Why single out change notification?

Because most production-phase surprises come from unannounced changes: a laminate grade substitution, an outsourced step moved, a key machine adjusted. The part number is unchanged; the inputs are not. Writing trigger conditions and notice periods into the quality agreement in advance costs far less than tracing the cause afterwards.

Q5: Does this evidence list cover automotive or other projects with additional system requirements?

It works as the technical capability layer, but industry system requirements need separate confirmation. This article covers only general evidence related to HDI structure and process capability, and makes no compliance determination for any industry scheme.

Q6: Should this evidence be requested before or after the RFQ?

In parallel. Requesting during the RFQ lets you observe engineering response quality, which is itself a piece of evidence, and it prevents unit price from anchoring the decision before capability has been checked.