PCB Prototype: Plan What the First Build Must Prove

A PCB Prototype is an actual PCB or PCBA built to solve specific engineering issues before the design goes into wider production. The value of a prototype is derived from the fact that it represents a known revision of that design, with representative features, controlled test conditions, and an acceptance rule stated prior to testing.

A PCB prototype can be used to confirm the circuit’s operation, confirm that parts and connectors will fit together, verify that the assembly is feasible, evaluate firmware interactions with installed components, assess the thermal characteristics of the design, and illustrate how the design will fit within the enclosure. For any tests performed, the PCB must contain the features relevant to the question being investigated; while a hand-modified PCB may yield useful information, the modifications and the test conditions must be clearly identified.

What Is a PCB Prototype Used For?

By providing a means to reduce uncertainty before making larger commitments to tooling, purchasing components, performing compliance tests, and managing production inventories, PCB Prototype development can bring to light schematic errors, incorrect footprints, unreachable test points, connector interference, power sequencing errors, signal integrity problems, thermal hot spots, or assembly limitations that can be changed while the design is still being developed.

A PCB Prototype does not automatically serve as a sample for production. Verification verifies that the product meets defined specifications; validation determines whether the product will meet the intended use and conditions of use. NASA uses this distinction as part of its systems engineering methodology, and it requires that test data collected on a test article be applied only within the defined pedigree of the test article. While not a requirement of PCB development in general, the same rationale for the classification of test articles can be applied to PCB development.

Which Prototype Matches the Engineering Question?

When engineers use any of the previous definitions for PCB prototypes, they often confuse the prototype classification with the level of maturity or decision point of the prototype. A bare PCB, assembled engineering PCBA, and production-representative unit are terms used to refer to the physical test articles, while terms such as visual model, POC, EVT, and DVT represent the various levels of maturity or decision points of the prototype. There is no universally accepted definition of what each of these terms means; the project must clearly define what materials, processes, components, firmware, and testing correspond to each of these classifications.

IPC-6012F specifies qualification and performance requirements for rigid printed boards. IPC-A-600M specifies acceptability for printed boards; IPC-A-610J specifies assembly acceptability of electronics. While all three of these standards provide specifications for different objects, they do not turn a visual inspection into a functional system test. IPC has identified the current revisions of these three standards as IPC-6012F; IPC-A-600M; IPC-A-610J.

FR-4 is a common glass-fiber-reinforced epoxy used to build PCBs, but FR-4 alone is not sufficient to completely describe how to build a PCB. Additional characteristics of the PCB (for example, thickness, copper weight, layer stack-up, dielectric thickness, surface finish, frequency, thermal load, and mechanical stress) will also determine if any given PCB prototype will answer the question it was designed to answer.

How Does PCB Prototyping Work?

The same key steps are outlined in more detail in Sierra Circuits’ description of their process for building PCB prototypes.

1. Define decision and acceptance criteria. Define any uncertainty related to electrical, mechanical, thermal, assembly, firmware, or interface that can impact your decision on the next step. Measurable output should have a clearly defined passing condition prior to placing your order.
2. Review the schematic and BOM. Perform electrical rule checks, validate power trees and protection devices, confirm pin mapping, validate that all the footprints agree with the selected MPN, document approved substitutions and DNP positions.
3. Create the layout and stack-up. First place connectors and mechanically constrained parts on the board; control return paths and power distribution, then thermal paths, sensitive analog regions, and finally any high-speed routing. The final impedance calculations can be confirmed with the fabricator before going to manufacture.
4. Perform DRC, DFM, and DFA checks. Check the design using the fabrication and assembly route you chose and not a generic minimum. Check for drill-to-copper clearance, annular rings, copper clearance to the edge, solder-mask features, footprints, polarity of parts and spacing between parts, fiducials, and access for testing.
5. Create your release package and inspect it for errors. Export all of the information necessary for fabrication and assembly: fabrication data, NC drill files, drawings, stack information, BOM, CPL, and documentation for the assembly of the selected article. Read the manufacturing data files in a third-party application and verify that the files are all associated with the same revision level.
6. Manufacture and assemble the controlled revision. Document all engineering questions, changes made to CAM files, substitutions for materials, changes of components, and any deviation that is allowed. If any correction was made to the generated manufacturing data, it must be returned to the CAD source to prevent the old error from generating again in the next release.
7. Bring up and test in an incremental manner, from the least risk to the greatest risk. Inspect the boards for defects prior to powering them up. If necessary, use current limiting when applying power. Check the voltage rails and reset behavior followed by testing of the modules before integrating them into a complete system. Document the equipment, firmware, loads, cables, environment, and results.
8. Only revise based on documented evidence. Categorize each issue as a manufacturing defect, an assembly defect, a setup issue, a component issue, a firmware issue, or a design-related change. Preserve evidence of valid results and only create new evidence for claims related to the change.

What Files Should Be Checked Before Ordering?

The contents of the release package should indicate a single configuration. KiCad documents the procedures for checking the board and generating Gerber, IPC-2581, drill, and placement outputs, and Altium separates fabrication, assembly, and component output documentation. While both software programs are optional, the revision relationships among the various outputs are not.

Current examples of quick-turn services include service A, which has a minimum drill size of 0.20 mm and offers 4/4 mil (0.102/0.102 mm) trace and space, and service B, which has a 0.203 mm finished drill and offers 5/5 mil (0.127/0.127 mm) trace and space. They do not represent a universally accepted “safe” default for all materials, copper weights, layer counts, or yield targets.

Which DFM and Test-Access Checks Matter Most?

Bare-board data

Verify that the PCB outline is made up of one continuous path. Verify that all plated holes, non-plated holes, routed slots, and internal cutouts are clearly defined. Verify that the selected process supports the required minimum trace width, minimum trace spacing, drill size, annular ring, copper-to-edge distance, solder-mask web width, width of the legend, and registration allowance. For controlled-impedance PCBs, you must mutually agree upon the stack-up, material family, finished copper, target impedance, and tolerance before freezing the routing.

Assembly data

Compare each line of the BOM with its respective land pattern and placement record. Verify that pin 1, LED and diode polarity, the orientation of the electrolytic capacitor, keying information for connectors, rotation information for bottom-side components, all DNP references, exposed pads, thermal pads, and package variants are all correct. Verify that, for dense components, sufficient clearance exists for placement, inspection, and rework, and remember that a valid Gerber set cannot compensate for a footprint or placement rotation that is incorrect or unclear.

Observation and isolation

Identify all required measurements for the test plan as outlined above. Ground references and power-rail test points should be readily available, along with programming and debugging access. Observation points for clocks, reset, buses, and analog nodes should also be included when those signals affect acceptance. Device-specific guidance may also provide 0-ohm links or shunts for isolation or current measurement, as stated by Texas Instruments in its AM64x and AM243x Hardware Design Guidelines documents; however, these examples of device-specific guidance are not universal requirements.

When Is In-House Prototyping Appropriate?

The ability to perform in-house milling can help shorten the design, test, and adjust cycle; allow you to keep your detailed layouts confidential; and allow you to quickly tune various RF structures, connectors, fixtures, and simple circuit modifications. One of the reasons LPKF cites for using an internal laboratory to create PCBs is that LPKF can provide same-day milling, drilling, routing, optional through-hole plating, multilayer lamination, and adjacent SMT equipment as part of its internal laboratory process.

Whether to use the option of in-house milling would depend on the circumstances of that specific project. There may be times when an internally milled board would be great for topology, placement, antenna tuning, or rapid debugging, but it would not be an exact copy of the intended laminate, plated-via construction, solder mask, surface finish, controlled impedance, or production assembly method. As such, when those attributes will have an impact on the result or require formal fabrication evidence, you may want to use an external contracted facility for fabrication. Lead-time comparisons should take place after defining all aspects of your product, including construction, assembly, testing, and delivered records.

Define the Acceptance Rule Before Testing

Break down any broad goal into one or more observable results to be achieved under a controlled set of stated conditions. NASA categorizes verification methods into Analysis, Inspection, Demonstration, and Test categories. A PCB plan can use the same categories without importing a large systems-engineering process.

1. Identify a single point of uncertainty. This will indicate the next design, sourcing, testing, or release decision that might be affected.
2. Choose an observable result. An observable result could include voltage, waveform, current, temperature, distance or measurable clearance, an enumerated event, an inspection report, or anything else that records an output.
3. Determine which method you will use for obtaining that result. You may use inspection methods for physical attributes, analysis for calculated results, demonstration for observable operation, or test for a measurable response.
4. Determine the conditions under which you will be performing your tests. When performing your tests, be sure to accurately record supply type or voltage, load type or resistor values, fixture used, cable used, firmware used, environmental factors (temperature, humidity, etc.), sequence of testing, configuration of samples, and the instrument settings used when collecting results.
5. Document the acceptance criteria prior to collecting and reviewing your results. Criteria to establish acceptance levels can originate from project requirements, manufacturer’s specification (Data Sheet), approved Interface Specifications (Drawing), or another approved process. Once you have established an acceptance level, you cannot change it after reviewing your results.

Example: “Check USB” is not an acceptance rule. You could create a specific acceptance level for the same example that states, “PCB Revision B2, BOM Variant A, Firmware Version 0.9, shall complete all Mandatory Test Points in TP-USB-04 using the released connector, specified cable, Host configuration, and Supply Conditions without any unexplained reset or Protocol Errors.” All identifiers used in this example are provided for illustration purposes only.

How Do You Create an Evidence Matrix Before Ordering?

Create one line in your evidence matrix for each question that could change your decision. The evidence matrix presented below is an adaptation for PCB prototype planning; it is not an official NASA or IPC evidence matrix.

Some requirements for configuration control include a) the documentation of every revision or version of all prototypes and inputs to production that have the potential to impact a result; b) any substitutions to prototypes (deliberate mismatches or temporary); c) testing with hardware assembled by hand (non-automated); and d) testing using firmware specifically for test purposes only. These differences should be documented so that the test results can be reproduced or correctly limited.

How Should a New PCBA Be Brought Up?

1. Inspect the product while not powered, taking note of any polarity issues; orientation problems; solder bridges; unpopulated required parts; connector alignment; damaged components; and measuring resistance between power rails and ground where that measurement is meaningful.
2. Protect the board at first energization. When doing this, you must use the appropriate input source and current limit (if applicable) and start without connecting any external loads (those not needed) for the first round of tests. If current, temperature, smell, or sound are out of the ordinary, unplug the board immediately.
3. Check the power tree, including input (source) current; voltage on each main power rail; enable, power-good, and sequencing signals. At this point, you should not be programming complex firmware or attaching expensive peripherals.
4. Check oscillator activity, reset release, boot straps, and programming access for the clock, reset, and boot states. The first firmware image should be small enough to separate hardware bring-up from application behavior.
5. Test modules one at a time. Test regulators, sensors, memory, communications, and outputs independently before full integration. Use known loads and fixtures, and make sure failures can be localized.
6. Test interface and system function. Once the prototype has passed the previous tests, you may connect released cables, enclosures, loads, and realistic operating conditions, and test the boundary conditions and failure cases required by the acceptance plan.
7. Document.

Ensure that the results of all testing are preserved, including: set-up pictures, captured data from instruments, logs of the results; firmware hashes or versions; operator; sample ID; acceptance pass/fail result; and any deviation or rework performed prior to retest.

A 5 V controller was fabricated as a prototype. The controller was specified to draw less than 60 mA before application firmware was loaded. The controller’s supply was limited to 100 mA and the first unit immediately drew that limit, with the 3.3 V rail staying at 0 V. As designed, the unit had a 0-ohm isolation link, and opening the link eliminated the overcurrent on the input, thereby localizing the issue to the downstream 3.3 V rail. Once inspected, a solder bridge was found. After the bridge was removed, the input current was measured at 42 mA and the 3.3 V rail was measured at 3.29 V. Since the repair modified the main power path, the power-up and downstream interface tests were repeated. The enclosure-fitting evidence was still valid; the numbers are project-specific and do not represent universal limits.

What Does a Passing Prototype Not Prove?

Just because a prototype has passed tests does not mean that production capability has been proven or that there is sufficient yield; nor does it guarantee sourcing continuity, that long-term reliability has been established, that all applicable regulations have been followed, or behavior outside the tested conditions. However, if the configuration and test conditions of the prototype do not match the configuration of a later revision or production, the result should not be transferred to the later revision or production.

A POC built using a hand-assembled PCBA can verify that the corrected interface circuit has functioned properly in the lab under recorded bench conditions, but it cannot by itself prove that the assembly process that will manufacture additional units will yield the same results. If the variance you observed is not relevant to your claim, keep the evidence and explain your reasoning. If the variance had an effect on the test paths or conditions that were established for your claim, obtain new evidence to support this claim.

How Should Rework and Revision Changes Be Recorded?

Document all jumpers, cut traces, component replacements, substitutions, setup corrections, and firmware changes that are made prior to using the next round of results. The current IPC document for rework, modifications, and repairs to electronic assemblies is IPC-7711/21D; however, this document does not determine whether a project may transfer a test conclusion.

What Controls Prototype Cost, Lead Time, and Quantity?

The price and timeline for prototypes are based on more than just the quantity of boards. As requirements change, the same design can move between a shared-panel standard route and a dedicated engineering route.

• Board geometry and panel use. Irregular-shaped or larger boards will take more panel space. Standard build types may share panel space and lower setup costs, but these build types limit available options for stack-up, materials, finishes, and lead times.
• Layer count and stackup. Extra laminations, non-standard dielectric thicknesses, controlled impedance, and tight registration will add to engineering and production timelines.
• Material and copper. High-Tg, low-loss, heavy copper, rigid-flex, flex, and metal-core may all require different raw materials, tooling, and production routes.
• Via and feature technology. Blind or buried vias, laser microvias, via-in-pad, filled vias, narrow spacing, and tight tolerances may make standard quick-turn capability impossible.
• Assembly and sourcing. Fine-pitch packages, BGAs requiring inspection, double-sided placement, through-hole projects, programming, rare components, and approved alternates all add to the schedules and prices.
• Inspection and test. Also to be included when comparing quotes are electrical tests, impedance coupons, AOI, X-ray, functional testing, environmental tests, reports, and traceability.
• Review, logistics, and clarification. Fabrication time does not encompass the total lead time. The time required for the review of files, engineering questions, purchasing of components, assembly, dispatch, shipping, and customs, represents different timelines to be tracked separately.

There is no single first-order quantity. Assign each unit a role before ordering. One example of a six-unit engineering run could be: one for controlled bring-up, one for enclosure fit, two for parallel firmware work, one for destructive analysis or thermal investigation, and one unmodified control unit for confirmation. The risk profile of the hardware may justify a different distribution of units, meaning the quantity of units should follow the evidence plan rather than adhere to a fixed industry number.

Related engineering guides: SUGA-PCBA custom PCB design guide and SUGA-PCBA low-volume assembly guide.

Conclusion

Develop a plan for the PCB prototype from the decision backward. Determine the uncertainty and pass/fail conditions, choose an article that demonstrates all of the relevant features, produce and release one cohesive set of files, inspect the exported data, and bring up the assembly in a controlled manner. Prior to providing results, document the configuration of the assembly, measurements taken, deviations, and rework performed. A prototype does not require production fidelity on every characteristic; however, a unit must possess sufficient fidelity to answer the assigned question without implying evidence that was never collected.