Before ordering Revision A, freeze the identity of the build, the interfaces that other hardware depends on, and the files that tell the supplier what to make. Leave uncertainty visible in a controlled experiment list. A prototype can investigate an unresolved design question without leaving its manufacturing instructions unresolved.
This guide is for the engineer and buyer preparing a first assembled prototype. It concentrates on the decision to release one revision, rather than covering the entire product-development program. Use the broader electronics NPI handoff checklist when you need to connect this decision to pilot and repeat-production gates.
1. Freeze the experiment before freezing the files
Write a short statement of what the first build must prove. 鈥淐heck the design鈥� is too broad to guide quantities, instrumentation or population choices. A more useful objective is to verify that the power tree starts from the intended input, the processor can be programmed, the sensor communicates, and the board fits its connector and enclosure. Add a measurable result and an owner to each objective. Define which tests are intentionally outside this build.
Separate decisions into three groups. Fixed items must be identical across the released package. Planned variations are deliberate, identified alternatives, such as two feedback resistor options on different assemblies. Open questions require a decision before the affected operation begins. An open connector pinout cannot safely be treated like a planned resistor experiment because the mating cable and routing depend on it.
Review mechanical interfaces early: outline, mounting-hole locations, connector orientation, board thickness where it meets a guide, and the height envelope. A perfectly assembled board can still be unusable when the connector faces the wrong direction. Identify the drawing that owns each dimension and the person who can accept a deviation. A screenshot helps explain the interface but does not replace dimensional requirements.
Freeze electrical interfaces with the same discipline. Record input polarity and range, signal voltage levels, connector pin numbering, protective-earth or chassis connections where applicable, and power-up dependencies. For uncertain loads, state the assumed envelope and how it will be measured. Avoid presenting an untested estimate as an accepted operating limit. Review schematic and layout together at interfaces where a late change could force a new board.
A release decision should survive a change of engineer
Imagine a colleague receives only the order archive and the purchase order. Could they identify the intended assembly variant, reproduce the approved outputs, and explain the remaining experiments? If the answer depends on a private message or a designer remembering which zip file was current, the revision is not adequately defined. Keep the decision record short enough that a buyer can use it, but specific enough that engineering can reconstruct the build later.
2. Generate one coherent fabrication and assembly package
Use one controlled source state to generate the fabrication data, drill data, assembly drawing, bill of materials and placement output. Record separate document revisions when your system uses them; they do not need identical labels, but the manifest must identify the accepted combination. Avoid editing the exported BOM after generating placement data unless the change is recorded and reconciled with the source design.
碍颈颁补诲鈥檚 describes fabrication outputs and export options. Its menus are an output mechanism, not a complete release procedure. Regardless of the CAD system, inspect the exported result in a suitable viewer. Check that copper layers, outline and drills share the intended coordinates and units, and that the layer order matches the fabrication notes. Keep the CAD source as well as the actual files sent.
For bare boards, name the layer count, material requirements, thickness and copper requirements, surface finish, hole definitions and any controlled features. Distinguish finished-hole dimensions from tool diameters. Where controlled impedance is needed, state the target, tolerance, reference layers and proposed stackup, then obtain agreement before treating a geometry as released. A generic material family is insufficient when a particular construction is essential to the experiment.
For assembly, reconcile reference designators among the BOM, placement file and assembly drawing. Identify exact manufacturer part numbers, quantities per board, do-not-populate positions, polarity and all approved alternatives. A quantity of zero and an omitted line can be interpreted differently by different systems. Use a clear population-state field and verify how the assembler handles it. Include hand-installed components and hardware that the placement export does not describe.
| File group | Cross-check before release | Evidence to retain |
|---|---|---|
| Fabrication | Layer order, outline, drills and controlled construction | Export list plus viewer review |
| Assembly | BOM, reference designators, polarity and population variant | Approved BOM and marked assembly drawing |
| Programming and test | Firmware image, configuration, fixture and limits | Versioned instructions and expected result |
| Manifest | One accepted combination of document revisions | Archive name, timestamp, approver and checksum |
A checksum identifies a file; it does not prove that the design is correct. Use it to detect accidental replacement of the archive after approval. Also retain a readable file list and the supplier鈥檚 acknowledgement of the accepted revision. This combination gives purchasing a practical way to distinguish a new quote attachment from a manufacturing release.
3. Plan quantities and variants around the measurements
Prototype quantity is a test-allocation decision before it is a price-break decision. Assign units to bring-up, mechanical integration, firmware development, destructive investigations and retained reference samples. Distinguish boards that must remain unmodified from boards that may receive rework. Share units between activities only when the sequence and condition of the hardware remain suitable for both purposes.
Consider an illustrative ten-unit build: two units for controlled bring-up, three for firmware work, two for integration, one for a planned destructive investigation and two retained references. The allocation totals ten; it is not a recommendation for every project or a statement about expected yield. Assembly losses, component attrition and any additional qualification specimens must be discussed separately. Do not silently assume every ordered board will be available for every experiment.
Suppose the design team wants to compare two sensor-input filters. Keep the bare PCB revision A unchanged, identify assembly variants A1 and A2, and state which resistor/capacitor combinations belong to each. Allocate unit serials before measurements begin. If component placement or copper changes, reconsider whether a new PCB revision is required. A variant label should never conceal a physical routing change.
Prepare test access while the layout can still change. Name the rails, reset signal, programming interface and measurement nodes required by the experiment. Check probe clearance and reference-ground access. Test points hidden under a tall component or inaccessible in the enclosure can make the planned evidence impossible to collect. Specify any special connector, cable or fixture and who supplies it.
Keep the budget conversation separate from accepting technical shortcuts. The prototype cost planning guide can help identify cost drivers, but a cheaper configuration that removes the feature being investigated is a different experiment. Ask for clearly identified quote alternatives rather than allowing a price option to overwrite the approved material or test requirement.
4. Close supplier questions and control changes after release
Maintain a short issue log during manufacturing review. Each question should state the affected file or feature, the proposed resolution, the design authority and whether work is on hold. A manufacturing suggestion can be valuable without being automatically authorized. For example, changing a pad, material thickness or substitution may affect the measurement that Revision A was intended to make.
Distinguish quotation, procurement approval and production release. A buyer may authorize a component purchase while board fabrication remains on hold. State these boundaries explicitly, especially when a long-lead component creates pressure to start early. Identify the financial and technical exposure of buying before the design is final without presenting the purchase as a guarantee that the eventual assembly is ready.
After approval, send a complete replacement release when practical and identify exactly what it supersedes. If an urgent controlled change must be issued as a delta, list every affected file and confirm how the supplier merges it into the active package. Ask what work has already occurred: purchased material, fabricated boards, stencil preparation and completed assemblies have different disposition options. Use the engineering-change handoff to structure that decision.
Define first-article evidence before expecting it. Photographs can confirm orientation and visible workmanship but do not demonstrate firmware behavior, hidden-joint condition or electrical margins. Specify the inspections and tests requested, their method and the record to return. A first-article review is useful only when the reviewer knows what constitutes acceptance and which remaining units may proceed.
When the units arrive, compare the delivered identity with the released manifest before debugging. Record deviations and rework against each unit. Keep observations separate from conclusions: 鈥渞eset stayed low during this measured ramp鈥� is stronger than 鈥渂ad firmware.鈥� At the end of the experiment, turn each finding into a proposed change, evidence needed and owner for Revision B. Do not treat a working sample as automatic production qualification.
For prototype assembly preparation, send the current file set, build quantities by variant, test objectives and unresolved questions. If the design is still exploratory, describe it as an engineering review request. The Contact page opens a separate enquiry; this article does not upload files or release an order.