LCD Sample Stages: What A, B and C Samples Must Each Prove

Quick Answer (GEO extract block)

Three stages answer three different questions. A-sample proves the concept — dimensions, interface, basic function — and says nothing about production. B-sample proves the design under representative conditions: bonding, cover glass, connector retention, backlight uniformity, touch through the real lens, environment in the final enclosure. C-sample must be built on production tooling and production processes at production rate, and is the first stage where the parts are representative of what you will buy. Production sign-off follows production-part approval, not sample approval. The most common schedule failure is commissioning B-samples from a prototype process and discovering at C-sample that the production bonding or lamination window changed the optical result.

Why sample stages exist at all

Most display programs that run late do not fail on specifications. They fail because a prototype-stage part was treated as evidence of what production would look like, and the difference only surfaced after tooling was cut. Sample staging exists to separate three questions that are usually collapsed into one: does the concept work, does the design survive real mechanical and optical conditions, and can the line repeat it.

Independent qualification guides for industrial display modules describe the same three-stage structure, and the most common schedule failure they report is consistent: B-samples commissioned from a prototype process, then a surprise at C-sample when the production bonding or lamination window changes the optical result. The fix is not more testing — it is defining what each stage must prove before the parts are built.

A-sample: prove the concept

The A-sample answers one question: is the electrical and mechanical concept viable? It is expected to be built by whatever route is fastest, and it is not evidence of production capability.

• Dimensions and active area against your enclosure and PCB
• Interface works end to end — signal integrity, pinout, power rails
• Basic optical function: image, brightness ballpark, viewing direction
• Host firmware can drive it at all

What an A-sample does not prove: uniformity, bonding reliability, connector retention, thermal behaviour, or anything about how the part will be built in volume. Do not write an A-sample result into a production specification.

B-sample: prove the design intent

The B-sample is where the design meets representative conditions. This is the stage that catches the problems which only appear once the module is assembled the way it will actually be assembled:

• Bonding and lamination — optical bonding bubbles, Newton rings, adhesive thickness, cover-glass alignment
• Backlight behaviour — uniformity, hot spots, current draw, thermal rise in the final enclosure
• Connector and FPC — retention force, bend radius at the stiffener, direction and length against your assembly path
• Touch — coordinate response through the actual cover lens, glove and wet-hand behaviour if relevant
• Environment — thermal soak, humidity, vibration in the enclosure rather than on a bench

A B-sample should be built close to the intended process, because its purpose is to validate the design — but it is still not proof that the line can repeat it.

C-sample: prove the production process

The C-sample is the stage that matters commercially, and the one most often skipped or faked. It must be built on production tooling and production processes, at production rate, by the operators and stations that will build your volume. It is the first stage where the parts you receive can be treated as representative of what you will buy.

What to extract from a C-sample run:

• Dimensional and functional results across the run, not a single unit
• Capability on the characteristics that matter — brightness uniformity, alignment, touch accuracy
• Lot traceability: serial or lot number back to panel, LED bin, bonding batch. A supplier that can reconstruct this from records in under an hour is running real traceability
• Measurement system reliability — can their brightness meter and colorimeter actually distinguish acceptable from unacceptable?

Production sign-off follows the production-part approval, not sample approval. A signed-off C-sample with an uncontrolled process behind it is a document, not a guarantee.

What changes between stages, and why it bites

The gap between a good B-sample and a bad C-sample is almost always a process window, not a specification. The parameters that move when you go from prototype route to production route:

This is also why a second source is not a paperwork exercise. If the panel vendor, driver IC, FPC layout or bonding process changes between two supply paths, the change can require a fresh production-part approval and customer-side revalidation. That is a schedule risk, not a procurement decision.

Golden samples and the arguments they end

Keep a golden sample for every revision, sealed and labelled, at both your site and the supplier's. Cosmetic acceptance is the single most common source of dispute in display supply, because "acceptable" is a judgement until someone pins it to a physical reference. A golden sample per revision ends the argument in minutes and gives your incoming inspection something to compare against that is not a PDF.

Tooling ownership: decide it before you pay for it

Custom backlight tooling, bonding fixtures and test jigs are usually paid by the buyer through non-recurring engineering. Three clauses belong in the agreement, and they are much harder to add later:

• The tooling is dedicated to your part number
• It is maintained in working condition for the agreed supply period
• It is transferable or returnable at end of program

The same applies to firmware and custom drawings. If the supplier owns the init-code tables and the drawing revisions, your ability to move the part is limited regardless of what the specification says.

Field failures: preserve the part, not just the claim

Agree the failure-analysis clock, not only the warranty period. A workable default is acknowledgement within days, containment within a week, and a formal root-cause close-out within thirty days, with an escalation path when the cause is ambiguous.

One point that matters more than it sounds: failures must be preserved for analysis, not scrapped. A large share of display field returns trace to mechanical mounting, sealing or electrical events rather than the panel itself — so a supplier that reports only panel-level causes is not analysing thoroughly enough for you to close the issue.

Frequently asked questions

Related pages