When selecting an online AI AOI machine for an SMT line, it is most effective to follow a logical sequence: first the inspection station, then the conveyor type, then the board envelope and pixel resolution, and finally any optional functions. This order works because each decision restricts the subsequent options.
The process of selecting an online AI AOI should begin with the inspection station, followed by the conveyor type, then the board envelope and pixel resolution, and only after that the optional features. Many are tempted to start with a speed figure or camera specification, but that approach frequently results in a machine unable to accommodate the largest panel or one that creates a bottleneck behind a high-speed placer. The T7 single-track online AI AOI and its T7D double-track version illustrate how each step limits the next, converting an extensive specification list into a concise, focused set of requirements you can review with a vendor.
How Inline AOI Fits Different SMT Inspection Stations and Conveyor Layouts
An online AI AOI is integrated directly into the line, meaning boards travel on the conveyor, get imaged and evaluated, and continue without manual transfer to a separate station. Consequently, the first choice is not a technical specification at all—it is the type of station the machine must support. A post-reflow station receives boards fresh from the oven, where solder joints are shiny, pads are covered with a thin flux residue, and typical defects include bridges, insufficient solder, tombstoning, and missing or misaligned components. A post-wave station presents a different surface: through-hole leads, taller fillets, and defects like pin bridges, icicles, and incomplete fill. Lighting angles and defect thresholds are not transferable between the two, so a machine set up for one station requires a separate recipe for the other.
1. Post-Reflow and Post-Wave Stations Need Different Image Conditions and Defect Criteria
Reflow solder joints reflect light inconsistently because each fillet has a distinct curvature, and factors like dark boards, white silkscreen, and gold or HASL finishes all alter the background. Deep learning inspection addresses this by learning what a good joint looks like on your specific boards rather than comparing pixel values to a fixed template, which keeps false calls low on mixed-finish production. Wave-soldered boards introduce additional complexity: leads protrude through the board, solder volume is greater, and the threshold between a full fillet and an acceptable one depends on the customer's workmanship class. First determine the station, then request the supplier to adjust the defect library accordingly.
2. Single-Track and Double-Track Conveyors Match Different Takt and Board Flow Conditions
A single-track machine processes one board at a time, so its inspection cycle must fit within the existing line takt. If the placer and oven deliver a panel every 20 seconds but the inspection takes 26 seconds, the AOI becomes the line's bottleneck regardless of algorithm quality. A double-track machine like the T7D runs two lanes simultaneously, matching the speed of fast reflow lines and allowing two different products to run side by side without reconfiguring the entire line. The key question to ask a supplier is not 'which is faster' but 'how many boards per hour do my placer and oven produce, and can a single lane handle that?' The T7 single-track model is suited for standard continuous inspection; the T7D double-track variant should be considered when takt pressure is the primary concern.
What Board Size, Thickness, and Pixel Resolution an Online AI AOI Must Match
Begin by compiling your actual board list: the smallest panel, the largest panel, the thickest and thinnest stack-ups, and any oversized boards planned for the coming year. The T7 supports PCB sizes from 50 × 50 mm to 460 × 460 mm, with a selection note for 500 × 325 mm and customization available beyond that. It accommodates board thicknesses from 0.3 mm to 6 mm. These dimensions determine whether the machine can handle your production, so verify them against your widest panel and thinnest flex-heavy assembly before discussing optics. Panels outside the standard size typically require a custom conveyor width or a different frame, an issue best addressed early rather than at installation. Pixel resolution then dictates what the camera can distinguish on those boards. A 5 μm pixel size provides clear edge definition on 0201 components and fine-pitch leads, and 0201 is standard on this platform. Pixel size is a camera characteristic that determines how much board area each pixel covers; smaller numbers yield finer detail but also a narrower field of view and more images to stitch. The EMVA 1288 standard provides a consistent way to describe these camera metrics. For most SMT assemblies, 5 μm is the practical baseline. Choose finer optics only when your product mix genuinely demands it, because higher magnification increases inspection time on every board, not just the challenging ones.
Which Optional Functions Change the Selection Decision for Your Product Mix
Optional features are where the selection becomes tailored to your specific production. Support for 01005 components is optional and should be added only if your roadmap includes that package size, as it requires a change in optical configuration rather than a simple software adjustment. 3D detection is also optional and becomes valuable when height information—such as coplanarity, lifted leads, or solder volume—is needed for parts that a 2D image cannot reliably classify. Barcode scanning and MES integration are optional modules that become important when traceability is required by customers or when inspection results must be sent to a shop-floor system. Each of these should be treated as a separate configuration decision. Programming effort is a related consideration because it affects the machine's behavior during changeovers. The T7 platform uses a single good board and a guided, CAD-free setup that takes about five minutes for standard boards; complex, dense, or unusual assemblies require more time and deserve a realistic engineering allowance. This is especially important on high-mix lines, where changeover frequency, not raw inspection speed, drives the actual cost. For example, a high-mix EMS line with frequent changeovers: every CAD import and manual tuning cycle consumes line time, so a machine that can be configured from one good board shifts the economics of small batches far more than a faster camera would. Inspection speed is up to 100+ mm/s, while defect detection accuracy is nominally ≥ 99.5% and false call reduction is nominally up to 80%; these figures support that workflow rather than replace it.
Conclusion
Following this sequential selection process keeps an online AI AOI project on track: decide on the inspection station first, then single or double track, then board envelope and 5 μm pixel resolution, and finally optional modules like 01005 support, 3D detection, barcode scanning, and MES integration. Working through it in order yields a concise, justifiable specification rather than a long list of desires. Present your board list, boards-per-hour target, and traceability requirements to a supplier, and request confirmation that the machine covers your largest and smallest panels, your changeover frequency, and the exact optional modules your product mix requires. IRIS ANT is an AI AOI machine manufacturer and AOI inspection equipment supplier, and builds the T7 single-track and T7D double-track online AI AOI for SMT lines. Pricing, MOQ, lead time, warranty terms, and certification documents are confirmed through an official inquiry.
FAQ
Q：How do I choose between a single-track and double-track online AI AOI for my SMT line?
A：Begin with the takt that your placer and reflow oven already provide. If a single inspection cycle per board comfortably fits within that takt, a single-track machine like the T7 is the simpler option; if inspection would create a bottleneck, or if you need to run two products simultaneously, then the T7D double-track variant should be evaluated. Verify your boards-per-hour target with the supplier so they can align the conveyor configuration to your actual throughput rather than a generic speed specification.
Q：What PCB size and pixel resolution should I match before selecting an online AI AOI machine?
A：First align the machine with your largest and smallest panels, since the board envelope is the most difficult specification to modify later. The T7 platform handles PCB sizes from 50 × 50 mm to 460 × 460 mm, with a selection note for 500 × 325 mm and customization options, and board thickness from 0.3 mm to 6 mm. A 5 μm pixel size offers good resolution for 0201 components and fine-pitch leads, covering most SMT assemblies; select finer optics only when your product mix truly demands it.
Q：Are 01005 component support and 3D detection standard features on an online AI AOI machine?
A：On the T7 platform, standard coverage includes 0201 components, whereas 01005 support and 3D detection are optional additions. 01005 support alters the optical setup, so include it only if that package size is truly part of your production plans. 3D detection is appropriate when height-based assessments like coplanarity, lifted leads, or solder volume are needed and cannot be determined from a 2D image. Specify each as a separate module during the configuration process.
No comments:
Post a Comment