SMT Assembly, Step by Step: How Your Board Gets Built

SMT Assembly, Step by Step: How Your Board Gets Built Surface-mount technology (SMT) is how virtually every modern PCB is assembled. Understanding the five steps lets you spot a good EMS line — and the defects a sloppy one will ship. Nex-G SMT capability at a glance Before the detail, the numbers. Nex-G's line places down to 0402 passives and fine-pitch QFN and BGA on a 4-head machine holding 0.025 mm repeatability and placing up to 6,000 components per hour, fed from Yamaha 8, 16, 24 and 32 mm reels. A 5-zone reflow oven runs closed-loop PID control. Prototypes go up to 280×280 mm; mass-production boards up to 350×450 mm. Every board passes SPI and AOI, and X-ray where the package demands it. The SMT line, step by step 1. Stencil printing — a laser-cut stencil deposits solder paste onto the pads. The single biggest source of defects; paste volume and alignment decide everything downstream. 2. SPI (solder paste inspection) — a 3D camera checks paste volume and position before a single component lands. Shops that skip SPI inspect defects into the board. 3. Pick-and-place — placement heads mount components at thousands per hour. Placement precision (0.025 mm on a serious line) determines what package sizes the line can handle. 4. Reflow — the board passes through a temperature profile that melts the paste and forms the solder joint, then cools to solidify it. 5. AOI (automated optical inspection) — a camera verifies every joint after reflow, catching bridges, tombstones and missing parts. Why inspection is not optional SPI and AOI are the two gates that separate a repeatable line from a random one. SPI checks the paste deposit before a component lands, when a bad board is still worth its value in paste — clean it, reprint, and run again in seconds. After reflow that same board is rework or scrap. AOI checks every joint against a programmed image after reflow, catching bridges, tombstones, missing parts and polarity errors before the board reaches you. Neither gate sees inside a package, which is why BGA and QFN add X-ray. But the discipline of running both gates on every board — and being able to show the images and the programmed limits — is how you tell a shop that inspects defects out from one that ships them. Stencil printing: the highest-leverage step The stencil controls how much solder reaches each pad, and a pad that gets too little or too much becomes the field failure you chase months later. Three variables decide the outcome: the aperture design, the paste, and whether the line verifies the deposit before a component lands. A serious line matches stencil thickness to the finest feature and paste to the application, then lets SPI measure height, area and volume on every deposit so a worn stencil is caught in the first boards of a run, not in rework. Solder paste: the material under every joint Solder paste is a suspension of metal-alloy spheres in flux. Its job is to hold components in place through placement, then melt into the joint at reflow. For lead-free assembly the alloy is almost always SAC305 — 96.5% tin, 3% silver, 0.5% copper — which melts around 217–220°C. Metal load runs 88–90% by weight; the remaining flux cleans oxides off the pads and lets the solder wet them. Particle size decides what the paste can print. Type 3 (25–45 µm) covers 0402 and 0.5 mm pitch. Type 4 (20–38 µm) covers 0201 and 0.4 mm pitch QFN. Type 5 (10–25 µm) is reserved for 01005 and ultra-fine pitch. A line that prints 0402 with Type 3, then tries a 0.4 mm pitch BGA from the same jar, will fight shorts and skips all shift. Paste is perishable and temperature-sensitive. It ships and stores refrigerated, needs time to reach room temperature before printing so moisture does not condense into it, and has a finite stencil life once opened — typically 8–12 hours before the flux dries and release degrades. Viscosity is tuned to print speed and aperture: too thin and it slumps into bridges, too thick and it will not release from the stencil cleanly. Name the smallest pitch on your board when you brief the line — it dictates the paste type and whether fine-pitch parts force a step-down stencil. Stencil design: aperture, thickness, and the area-ratio rule The stencil is a stainless or nickel foil, laser-cut or electroformed, with an aperture over every pad. Two numbers define it: thickness and aperture size. Thickness sets paste volume — 100–150 µm for standard boards, dropping to 80 µm or less when fine-pitch parts share the board with larger pads. The number that predicts print quality is the area ratio: aperture area divided by aperture wall area. For consistent release the ratio should stay above 0.66 (IPC-7525); a fine-pitch aperture below it releases paste unevenly, the classic cause of skips and low-volume joints. That rule is why a 0.4 mm pitch QFN forces a thinner stencil or a modified aperture on its exposed pad. Aperture shape matters as well. Home-plate and rounded-corner apertures cut solder balls and bridging on tight pitches, and step stencils add paste only where a connector or QFN needs it. Nano or polymer coatings reduce paste sticking and extend wipe intervals. Whatever the design, SPI is the check — it measures height, area and volume on every deposit, so a worn or clogged aperture is caught in the first boards, not after reflow. Component packages your board will use PackageTypical sizeNotesWhat to watch Passives (resistors, caps)0402, 0603, 0805, 12060402 is the common small size; 01005 is high-densityBelow 0402, tombstoning and solder-ball risk rise as pad and paste volume shrink 0201 / 01005ultra-small passivesHigh-density boards onlyNeeds Type 4/5 paste, tight placement and a clean stencil; easy to tombstone SOT / SOPsmall-outline ICsSurface-mount transistors and small ICsKeep polarity oriented consistently; fine-pitch SOP wants local fiducials QFNquad flat no-leadExposed pad underneath — thermal vias matterVoids under the exposed pad; split the aperture and verify with X-ray BGAball grid arrayFine pitch; needs X-ray, not just AOI, to verifyHead-in-pillow, voids and warpage; control the profile and measure with X-ray QFP / TQFPquad flat packageGull-wing leads on four sidesLead coplanarity and bridging on fine pitch; AOI checks every lead Pick-and-place: speed is quoted, accuracy is what matters Everyone quotes placement speed, but placement accuracy is the number that decides whether your board builds. A head holding 0.025 mm repeatability handles tight pitches without drama, while a sloppier machine drops parts at fine pitch. Feeders matter as much as the head for the odd parts in a BOM, and so does verifying polarity on the fly. Name the smallest package on your board before you commit: 0201 passives, 0.4 mm pitch QFNs and fine-pitch BGAs each demand placement and inspection equipment specced for them. Nozzle selection is the last variable — a worn or wrong-sized nozzle drops parts or crushes small ones, and it surfaces as intermittent placement faults. Feeders: the quiet half of placement A placement head is only as productive as the feeders behind it. Nex-G runs Yamaha feeders in 8, 16, 24 and 32 mm widths, covering everything from 0402 passives on 8 mm tape to wide-body ICs on 24 and 32 mm reels. Feeder pitch and tape indexing must match the part exactly — a mis-indexed feeder feeds the wrong pitch and the machine places air or jams the lane. For prototype and short-run work the same feeders handle cut tape and partial reels, so a one-piece build does not need a full reel per part. Reflow soldering, explained in one paragraph Reflow heats the board through a controlled profile — preheat to drive off solvents, a soak to equalize temperature, a peak above the solder's melting point (typically 240–250°C for lead-free), then a controlled cool-down. Too fast or too slow and you get voids, cold joints or tombstoning. A line with PID closed-loop temperature control and multiple heating zones holds the profile that keeps joints reliable. The reflow profile, zone by zone A multi-zone oven with independent PID loops holds each zone to its setpoint and repeats the profile board after board; a cheap oven lets the profile drift with load. For lead-free paste the peak sits around 240–250°C, held above liquidus just long enough to wet the pads fully without growing brittle intermetallics. Nitrogen is the other lever, cutting oxidation and improving wetting on fine-pitch and hard-to-wet finishes. Preheat The board ramps from room temperature to roughly 150°C at 1–3°C per second. This drives off solvents and activates the flux without thermal-shocking the parts — ramp too fast and moisture in the laminate or a part can delaminate or pop. Soak The board holds near 150–180°C so large and small parts equalize in temperature and the flux finishes cleaning the pads. A soak that runs long dries the flux out before reflow; one that runs short sends the board into reflow with a temperature spread that leaves some joints under-heated. Reflow The profile peaks above liquidus — around 240–250°C for SAC305 — and stays above 217°C long enough, typically 45–90 seconds, for every joint to wet and form a sound intermetallic. Time above liquidus is the figure to watch: too little and joints are weak, too much and the intermetallic grows brittle. Cooling The board cools at a controlled 2–4°C per second to solidify joints with a fine grain structure. Cooling too slowly coarsens the joint; too fast stresses parts and laminate. A line that holds all four zones to spec, board after board, is what separates a repeatable process from one that drifts with load. The defects a sloppy line ships Tombstoning — a chip resistor stands on end. Usually a paste or placement imbalance. Solder bridging — a short between adjacent pads. Caught by AOI if the line runs it. Cold joints — dull, weak joints from insufficient heat. They pass visual inspection and fail in the field. Voids — gas pockets inside BGA joints. Only X-ray sees them. This is why you ask an EMS what it inspects, not just what it claims. Here is the full vetting checklist. Defect causes: tombstoning, bridging, and voiding Nearly every SMT defect traces to one of three root causes: paste, placement, or profile. Knowing which is which tells you whether the fix belongs in the stencil, the machine, or the oven. Tombstoning A chip stands on one end when its two pads wet at different moments — the wetting force on one side lifts the part before the other side reflows. Uneven paste deposits, a pad tied to a heavy copper pour that heats late, or a part placed slightly off-center all trigger it. Fixes: balanced apertures, thermal relief on ground pads, and a soak that equalizes temperature. Bridging Too much paste, paste that slumps, or a part pressed into a paste pool shorts adjacent pads. Fine pitch and the wrong stencil thickness are the usual culprits. Fixes: a thinner stencil, smaller or rounded apertures, and paste with the right viscosity. AOI catches bridges when the line runs it. Voiding Gas trapped in a joint — from flux, paste solvents, or the exposed pad under a QFN — leaves voids that weaken thermal and electrical paths. Fixes: a split or windowed aperture on large pads, a profile that lets gas escape before the joint solidifies, and X-ray to measure what remains on BGA. Testing beyond AOI: X-ray, ICT, and flying probe For a quality lead. AOI sees the surface, not inside a BGA or under a QFN, and not the electrical behavior of the finished board. X-ray looks through the package at voids, shorts and open balls — the only non-destructive way to verify those joints. ICT checks each net against its expected value through a bed-of-nails fixture; flying probe does the same job with moving probes and no fixture, which suits prototypes and low-volume runs where a dedicated fixture does not yet pay for itself. Match the test to the stage: prototypes lean on flying probe and X-ray, production adds ICT and functional test. A partner that raises the fixture question early — and says plainly what each test can and cannot catch — saves you from finding the coverage gap late. DFM for SMT: design choices that decide yield For a manufacturing engineer. Most SMT defects are designed in before a stencil is ever cut. Five choices do the heavy lifting. Footprints and land patterns Use the manufacturer's recommended land pattern or an IPC-compliant one (IPC-7351). Oversized pads float parts and pull joints off-center; undersized pads starve the joint of solder. For QFNs, match the exposed-pad size and split it when it is large so paste does not trap gas under the part. Fiducials These are the reference marks the machine uses to locate the board and compensate for stretch and shrinkage. A 1 mm round pad on bare copper with soldermask clearance gives the vision camera a crisp, high-contrast target. Place three global fiducials — one near a corner, two on the diagonal — plus local fiducials beside fine-pitch and BGA parts. A missing fiducial forces the line to place a 0.5 mm pitch part off a board-edge reference, and placement error follows. Panelization Panelize with routed slots or V-scores and keep a 3–5 mm handling rail clear of components. Breakaway tabs belong away from fine-pitch parts so depanelization stress does not crack joints. Hand depanelization favors V-score; machine depanelization favors routed tabs with mouse bites. Thermal relief and copper balance Pads tied directly to large copper pours sink heat during reflow and produce cold joints. Thermal-relief spokes on ground pads let the joint reach temperature evenly, and balanced copper across the panel keeps the profile uniform from edge to center. Component spacing and orientation Keep small parts a safe distance from larger ones so the placement head and rework iron can reach them, and orient polarized parts — diodes, tantalum caps, ICs — the same way across the board. It cuts placement errors and speeds inspection. These rules are cheap in layout and expensive on the line. The line is only as good as the BOM it is fed For a sourcing engineer. An SMT line cannot assemble parts it does not have, and a counterfeit or single-source IC stops the line faster than any process defect. Source through authorized channels, verify incoming parts, and keep a documented list of drop-in alternates. On landed cost, finished assemblies and bare boards classify under different HTS codes, so confirm the declared code and any Section 301 exposure before you price the duty. The full sourcing playbook. From prototype to production: EVT, DVT, PVT For a program manager. The line serves a program moving through three gates: EVT proves the design works, DVT proves it survives real conditions, and PVT proves the line builds it repeatably. Prototype runs deserve the same discipline as volume — a no-MOQ line lets you validate a board at low volume and scale without re-qualifying a new supplier. How the three gates work. SMT plus CNC under one roof: the box build A bare PCBA is rarely the deliverable. Enclosures, brackets, heatsinks and standoffs are machined, and they have to fit the board the first time. When the EMS also runs CNC, the enclosure and board come from one process owner — a tolerance problem is fixed in one conversation, not a chain of emails between factories. Nex-G runs 80+ machining centers — 50+ 3-axis, 20+ 4-axis and 20+ 5-axis mills, 25+ turning and 7+ grinding — on brands including Mazak, Brother, TSUGAMI and Sodick, holding ±0.005 mm on milling and turning and ±0.002 mm on grinding and wire EDM, and the same shop handles box build. Behind it: a 6,800 m² facility in Hengli, Dongguan, running since 2006 (Zhuohang) with 100+ staff. MOQ is one piece, with lead times of roughly 3 days simple, 7 days standard, and up to 30 days for complex builds. The quality system behind the line Certificates are the floor, not the ceiling. Nex-G holds ISO 9001:2015, IATF 16949:2016 and ISO 14001:2015, issued by URS and valid to 2027. The IATF scope excludes product design under clause 8.3, so no design responsibility is claimed — an honest scope for a build-to-print partner selling disciplined manufacture of your design. The discipline shows in capability. On an automotive production program, SPC runs to the 1.67 threshold for key characteristics. The same control that holds a machined dimension holds a paste deposit or solder joint when the line runs to a control plan. On the floor the checks behind that system are concrete rather than aspirational. Every BOM line is sourced through AS6081-aware channels with at least one cross-checked second source; the panel-house rule profile is used for DRC before the first build; and every first article gets an FAI per AS9102 plus an IPC-A-610 acceptance verdict from a certified trainer before the run scales. Frequently asked questions What package sizes can you place?Down to 0402 passives and fine-pitch QFN and BGA, verified by AOI and X-ray rather than by eye. Do I have to supply the components?Either works — we source through authorized channels with lot traceability, or build from your kit. Is there a minimum order?No. MOQ is one piece, from prototype through mass production. What tests do you run?SPI, AOI, X-ray for BGA and QFN, plus in-circuit or flying-probe and functional test matched to your stage. What files do you need to start?Gerber files, a BOM, the centroid (pick-and-place) file, and any assembly drawing or notes. We run a DFM review on those and flag risks before the first prototype. Do you review my design for manufacturability?Yes — footprint, fiducial, panelization and thermal-relief checks come back with the quote, so layout issues are fixed before parts are ordered. What are your lead times?Roughly 3 days for simple builds, 7 days standard, and up to 30 days for complex builds. Can you also machine the enclosure?Yes — 80+ CNC machines handle enclosures, brackets and heatsinks under the same roof, so the board and the box come from one process owner. A well-run SMT line makes the process invisible — the paste deposits match the stencil, the placement is repeatable, and the profile is dialed in once, so the only thing that varies is the part number, not the quality. If a board fails in the field, the defect almost always traces to one of three places — a footprint that did not match the pad, a paste deposit that was too thin, or a profile that heated too fast. Getting those three right is the entire discipline of SMT, and it is why process control matters more than machine brand. Designing a board for SMT?Send the BOM and Gerber files — we will flag the SMT risks before you order the first prototype.Request a quote Related articlesPCB Assembly in China: How to Vet an EMS PartnerDFM for Electronics: Design Rules That Cut CostEMS vs OEM vs ODM vs JDM