Introduction: Post-wave soldering inspection decisions depend on defect profile, conveyor integration, and false-call control rather than a generic AOI specification.
Post-wave soldering produces larger solder joints and busier board surfaces than reflow. For automotive and medical electronics lines, post-wave inspection has to catch bridges, insufficient fill, and lifted leads before assembly while keeping the line moving and the operator trusting the flags. Color variation from flux residue, silkscreen near pins, and mixed component heights can create many flags with few worth a rework stop. The decision comes down to defect type, conveyor integration, and false-call control.
Wave soldering pushes molten solder through through-hole pins and forms joints on the underside of the board, so the resulting geometry looks nothing like a reflow joint on a fine-pitch component. Solder volume is larger, the fillet shape depends on pin length and board warp, and thermal gradients move from panel edge to panel middle. IEC 61187 provides useful background on attachment geometry and spacing, which explains why the joint is physically bigger and why the same lighting recipe that works on a fine-pitch QFN may not translate to a wave-soldered connector pin. IEC 61188-5-1 offers general soldering verification and defect identification background. Those physical differences drive downstream choices about camera resolution, lighting design, and algorithm behavior at the inspection station.
A wave-soldered joint covers more surface area and sits higher off the pad than a reflow joint, which helps visibility but complicates classification. Shadowing from tall pins, curved fillets that slope away from the camera, and solder that cools at different rates across a panel mean that a bridge on pin one may not look like a bridge on pin twelve. Heating variation adds a second layer. If the preheat zone runs unevenly, some joints come out smooth and shiny while others look grainy or dull. A traditional 2D rule-based system may flag the dull joints as insufficient solder even when the joint is electrically fine. That is why post-wave AOI has historically required rule tuning rather than stable detection.
Post-wave boards rarely arrive at the inspection station clean. Flux residue scatters light, solder mask color shifts between board suppliers, silkscreen sits right next to pin fields, and some boards already carry clipped leads or heat-sink hardware. The camera needs enough resolving power to see fine fillet edges; the T7's 5 μm pixel size covers standard 0201 parts, with 01005 supported as an option. The lighting must also separate silkscreen and solder mask from solder. That contrast helps an AI model distinguish a marginal joint from a cosmetic mark on the mask. Optional 3D detection becomes useful when coplanarity or joint height is on the defect list.
Traditional AOI relies on fixed thresholds: brightness, area, and position tolerance. Every new board color, flux batch, or supplier change forces a rule update, and each update can shift detection on unrelated features. An AI-driven system learns the appearance pattern of an acceptable joint, fillet, and lead protrusion, then blends that with feature-based measurements. Adaptive learning loops use operator feedback to refine the model. On the T7, the nominal false-call reduction of up to 80% after adaptive learning is built around that feedback cycle, and the nominal defect detection accuracy of ≥ 99.5% is measured against the same pattern-matching approach. Results depend on lighting setup and board condition, so complex or unusual panels may need extra training passes. Conveyor integration matters as much as algorithm quality. An inline AOI at the post-wave station has to synchronize with the wave machine's exit and the next downstream step, so entry and exit track height, board edge clearance, and transport speed all need to match the line. The T7 supports boards from 0.3 to 6 mm thick, runs inspection speed up to 100+ mm/s, and provides Ethernet, USB 3.0, and RS-232 for signal and data exchange. The Hermes Standard provides protocol background for machine-to-machine communication in inline SMT equipment, so reviewing it early helps frame integration questions; the actual handshake still depends on supplier configuration. Before install, confirm the handshake sequence, confirm track speed matches the wave output, and confirm that optional barcode scanning and MES connection are wired into the line controller. A nominal 5-minute CAD-free teach on a single good board keeps multiple changeovers per shift inexpensive.
Offline sampling has a legitimate place: first-article checks, low-volume runs, and audit pulls. It only sees the boards you carry to it. If your process can drift between sampling points—wave height changes, flux specific gravity shifts, preheat zones reacting to room temperature—then sampling catches problems after they have reached assembly. At the post-wave station, online AI AOI optical inspection equipment sees every board, which shifts the conversation from how often to sample to what the system can classify and how much rework traffic it will generate. Three signals usually point toward an online system. First, a defect mix dominated by bridges, insufficient fill, and lifted leads that already force rework after assembly; these are defects you pay for twice when they escape the wave station. Second, traceability. Automotive and medical customers increasingly ask for per-board records tied to a serial number, which is where optional barcode scanning and MES connection have real value on an inline platform. Third, changeover frequency. If your line runs multiple designs per shift, CAD-free programming keeps the inspection step from becoming a bottleneck. When all three apply, offline sampling is the wrong tool for process control; it sees only sampled boards and leaves gaps between checks.
Post-wave soldering creates a defect profile that a reflow-era AOI was never designed to handle. Larger joints, uneven heating, flux residue, and silkscreen push rule-based systems into false-call territory. Once operators stop trusting the flags, the inspection station quietly becomes a bypass. An online AI AOI built for that station—like the T7 from IRIS ANT—approaches the problem differently: a 5 μm pixel size covers the joint geometry, deep-learning classification handles the color variation, adaptive learning trims false calls, and a nominal 5-minute CAD-free setup keeps changeover cost low. If you are comparing online and offline for your post-wave station, the next step is practical. Send your board stackup, defect list, and conveyor speed to an AI AOI machine manufacturer or AOI inspection equipment supplier, and ask for a trial run on your actual panels so the numbers you see come from your own boards.
A：Post-wave defects usually include solder bridges between adjacent pins, insufficient fill or cold joints, lifted leads, excess solder bulging past the pad, and missing through-hole parts. An AI AOI trained on joint appearance can also flag fillet shape deviations that suggest a marginal joint even when solder volume looks acceptable. The T7 classifies these within a nominal detection accuracy of ≥ 99.5%, as a process-control tool for post-wave inspection. Actual escape rates depend on board condition, lighting, and process stability.
A：An inline AOI sits directly in the conveyor path between the wave machine and the next process step. The connection happens at two levels. Mechanically, the entry and exit track must match the upstream and downstream conveyor height, speed, and board edge clearance. Electrically, the AOI exchanges board-present and board-leaving handshakes with the line controller; the T7 provides Ethernet, USB 3.0, and RS-232 for that exchange and accepts boards from 0.3 to 6 mm thick. The Hermes Standard provides protocol background for how inline SMT equipment usually coordinates, but the exact handshake is worth checking with your supplier and line integrator before installation.
A：Choose online when the defects you care about can appear between sampling points, when your customers require per-board traceability, or when you run several designs per shift and setup time is expensive. In those cases, an offline AOI only sees the boards you carry to it, and the wave process keeps drifting between checks. Offline AOI still fits first-article verification and low-volume runs, where sampling frequency is not the limiting factor. The T7's nominal 5-minute CAD-free setup, optional barcode scanning, and optional MES connection support the line-integrated scenario.
