Electronics Manufacturing Services & PCB Assembly

Match the door to the scope. Bare-board fabricators build the unpopulated board from your fabrication data; a photochemical and drilling business with its own quick-turn and volume tiers. Assembly houses and EMS providers populate boards, and at scale they procure the components, build the box, load the software, test the product, and ship it under your label. The practical sort: board files alone go to a fabricator; boards plus a BOM go to an assembler, most of whom will arrange the bare boards; a product goes to an EMS partner scoped for box build and test. Two more filters decide who can actually bid: the technology floor, the finest pitch, smallest passives, and any HDI construction on your design, and the volume-and-mix band, since prototype shops, high-mix mid-volume houses, and high-volume lines are different businesses wearing one industry's name.

They are the industry's three-step answer to how good the work must be. Class 1 is general electronics, where cost rules and a cosmetic defect is tolerable. Class 2 is dedicated-service equipment, products that must work reliably for years, and it is the sensible default for most industrial hardware, which is also what you get when nobody asks. Class 3 is high-reliability work, medical, aerospace, defense, anything where failure injures someone or ends a mission, and it prices accordingly through tighter criteria, slower lines, and more inspection. Specify by writing it down: the standard, the revision, and the class on the drawing or order, remembering that the full Class 3 string is written out in full, IPC-6012 Class 3, J-STD-001 Class 3, IPC-A-610 Class 3, because the classes do not flow automatically from one document to the others, and that exceptions can run per reference designator when one circuit is critical and the rest of the board is not.

A complete package, because the quote is priced from files, not descriptions. The BOM with manufacturer part numbers, approved manufacturers and alternates, reference designators, and do-not-place notes. The design data: fabrication files in Gerber or ODB++, drill data, assembly drawings, and pick-and-place centroids. The acceptance class in writing, with any per-designator exceptions. Quantities and the release pattern: prototypes, pilot, production bands, since price breaks and process choices hang on them. The test expectation, even roughly stated. And the commercial stance: turnkey or consigned, target markets for regulatory compliance, and whether bare boards are in the assembler's scope. Incomplete packages do get quoted, slowly, with assumptions attached, and every assumption is a change order in the making.

Turnkey, where the assembler buys everything against your approved list, buys you their purchasing leverage, their incoming inspection, and one throat to choke, at a material markup that is usually cheaper than running board-level procurement yourself. Consignment keeps component sourcing, cost, and risk in your hands, and makes sense when you already buy the parts at scale, hold allocation relationships, or face controlled-sourcing rules. The working answer for most programs is the hybrid: consign the critical, allocated, or customer-directed parts, and let the assembler buy the commodity tail. Whatever the split, write down the risk terms that travel with it: who owns excess and end-of-life buys, what attrition allowance applies, and whose problem a lot failure is, because those questions only feel theoretical until the first constrained part.

Because assembly is fast and parts are not: the line turns boards in days, while a single constrained or dying component sets the calendar for everything. The protections are unglamorous and early. Screen the BOM for lifecycle before quoting; end-of-life and not-recommended parts are redesigns waiting politely. Approve alternates line by line in advance, so a substitution is a lookup instead of an engineering hold. State the channel policy: franchised distribution first, independent sources only under a written inspection and traceability practice, because the gray market is where counterfeits live, and allocation is when buyers visit it. And decide the risk-buy questions: who authorizes lifetime buys, who owns the excess, before scarcity decides them for you. Allocation cycles recur in this industry; programs that survive them are planned for them at RFQ time.

Buy it as a strategy, not a checkbox. In-process verification comes with a serious line: automated optical inspection against the design data, and X-ray wherever hidden-joint packages put solder where eyes cannot go. Electrical test is an economic choice you should see the arithmetic for: flying probe needs no fixture and owns prototypes and low volumes, while in-circuit test through a bed-of-nails fixture is fast per board and pays for its fixture at production quantities, provided test points were designed in. Functional test, the product exercised as a product against limits you define, catches what structure cannot. State the expectation in the RFQ, ask what coverage the proposed strategy actually reaches, and treat fixtures and test programs as owned, documented tooling, because that is what they are.

New product introduction is the engineered path from files to repeatable manufacturing: the DFM review that reconciles design with process, prototype builds, test development, first articles against the drawing and class, and the documentation that makes build ten identical to build two. What changes at production is the center of gravity: from engineering attention and fast turns to process stability, yield, and cost, sometimes at the same supplier, sometimes not, since prototype heroics and production discipline are different talents. The buyer's discipline is the gate: first-article approval, tied to acceptance and payment, marks the handoff, and the honest conversation about whether your NPI partner is your production partner happens before you commit volumes, not after the first late delivery.

Whoever the paper says, which is why the paper should say. Stencils, in-circuit fixtures, functional test rigs, and test programs are per-design engineering, usually billed as one-time charges. Under this platform's standing rule, the buyer who pays for tooling owns them, with their location and data recorded. Ownership without the data is decorative: the design package, test programs, and process documentation are what make a program portable to a second source or a successor supplier, so data rights travel in the same clause. None of this signals distrust; it is the same discipline every tooled sector on this platform runs, and suppliers who work at any scale expect the conversation and respect the buyer who has it.

Split the question the way the industry does. Assembly itself is quick: quick-turn prototypes run in days, production assembly in days to a few weeks once parts are staged. Bare boards run from days on domestic quick-turn to several weeks for complex or offshore fabrication. Components are the honest answer's center: stocked parts arrive this week, and constrained parts are quoted in months, which is why the parts strategy in the questions above is really the schedule strategy. Fixtures and NPI add their own first-build weeks. A supplier's quote should show these clocks separately: staged material, board fabrication, assembly, test, and a quote that shows one blended number has answered a simpler question than the one you asked.