Publish Time: 2026-10-29 Origin: Site
Two factories can own exactly the same SMT inspection equipment and get completely different results, and the difference is usually where the machines stand. Inspection only pays back when it sits at the point in the process where the defect it looks for has just become visible and is still cheap to correct. Place it a station too late and you are paying for rework instead of prevention; place it without a rework path and you are generating data nobody can act on. This article walks through the SMT process in order and explains how to decide where solder paste inspection, AOI, X-ray, first article inspection, and electrical test should physically sit, including the handling, buffering, utilities, and data connections each position needs.
Layout is usually treated as a facilities question, decided late and decided by whoever owns the tape measure. It should be a quality question, decided early and decided by whoever owns the yield number.
Each defect becomes visible at a specific moment and stays visible for a limited time. Solder paste volume is visible the instant the board leaves the printer and becomes unreachable once components are placed on top. Missing parts are visible right after placement and become expensive once they have been reflowed. Hidden joints exist only after reflow and are visible only to X-ray. A machine placed outside its defect's window is a machine that produces reports instead of preventing escapes.
Where a machine sits also decides what happens to a failing board. A fail immediately after the printer means clean and reprint. A fail after reflow means a technician, a soldering iron, and a re-inspection step. A fail at the end of the line means a board that has already absorbed all its value. When you choose a position, you are also choosing the repair method that follows it.
The shorter the distance in time between the process step that creates a defect and the station that detects it, the easier it is to find the cause. Inspection placed close to the step it monitors produces actionable feedback; inspection placed three stations downstream produces statistics. This is why tightly coupled positions are worth fighting for even when floor space is tight.
By the end you should be able to take a process flow for your product and mark the physical position for each inspection method, specifying what happens to a failing board, what buffer it needs, and what data it must write. The methods themselves and the defects each one detects are covered in our guide to matching SMT inspection methods to defect risks, and the checkpoint sequence is set out in our SMT inspection plan from solder paste to final test.
Product mix changes, throughput requirements change, and equipment gets added. A layout drawn once and never revisited gradually stops matching how the line actually runs. Treat it as a living document with a version and an owner, reviewed whenever volume or mix moves materially.
Before discussing positions, write the process flow down in physical order. The sequence, not the equipment list, dictates where inspection belongs.
Most lines are not purely automatic. There are manual insertions, touch-up stations, board flips for second-side processing, coating stages, and offline checks in between. Draw every step, including the ones that are not machines, because defects and delays concentrate there. A step that is invisible on the equipment layout is often the step that explains your escapes.
Walk the flow and mark the step that creates each of your significant defect classes: print, placement, reflow, second-side processing, through-hole assembly, coating, and handling. Every mark is a candidate inspection position. The goal is not to inspect at every step but to know which steps create risk and which do not.
Detection and creation are not the same point. Voids are created during reflow and become detectable only after the joint solidifies. A wrong component is created at feeder setup and becomes detectable as soon as the board leaves the placement machine. The gap between creation and detection is what costs money, and layout determines how wide that gap becomes.
An inspection station inserted into the line consumes cycle time. Before adding one, know which machine currently sets the takt and how much headroom exists. Adding inspection to a line with no slack either transfers the bottleneck or reduces output, and either outcome gets blamed on the new machine rather than on the missing capacity analysis.
Some positions are naturally inline because the method is fast and the coverage requirement is high. Others fit better offline because the method is slow or the product does not justify full coverage. That trade-off is examined position by position in our article on inline versus offline SMT inspection equipment, and the choice should be made per position rather than once for the whole line.
Draft the layout with operators, technicians, and maintenance present. They know where boards already jam, which corners are too tight for a cart, which station always runs warm, and where an operator currently has to walk twice to do one job. Those details never appear on a drawing, and they decide whether the layout works on a Monday morning rather than only in the plan review.
The first inspection position has the clearest answer of any in the line: immediately after the stencil printer, before the first placement machine.
Paste deposit volume drives a large share of downstream joint defects, and this is the last moment when a print fault can be fixed by wiping the stencil and reprinting. Move the inspection to after placement and the same defect now requires removing components. Move it after reflow and it requires desoldering. The cost curve is steep enough that this position almost always justifies itself.
Inline 3D SPI needs a short conveyor section between the printer and the first placement machine, stable board support, and enough clearance for its camera gantry and lighting head. Because the printer already handles board transport, the mechanical interface is usually straightforward. Platforms such as the S400, the S400D, and the S1200 are built for this spot, with differences driven by board size, line configuration, and the detail of the measurement required.
Decide the failure path before installation. The cheapest loop is a short reverse return into the printer so an operator can clean the stencil and reprint. Where that is not mechanically possible, the loop becomes a manual remove-and-reprint station, which costs more and introduces handling risk. The mechanical return path is part of the layout drawing, not an afterthought.
A small buffer upstream of the SPI keeps the printer running while a measurement completes, and a short buffer downstream prevents a single inspection cycle from blocking placement. These buffers should be sized deliberately, because a large buffer delays the feedback that makes this position valuable: if fifty boards sit between the printer and the detection point, the printer has already produced fifty suspect boards before anybody knows.
SPI should write per-board results with board identifier, job, and program version, and it should surface volume trends to whoever owns the printer. The physical position gives the data its meaning: because SPI sits directly behind the printer, a trend can be attributed to that specific printer rather than argued about. Configure the reporting before the machine ships rather than months later.
The recurring error is placing SPI where space happens to be available rather than immediately after the printer, for example after the first placement machine, because the line was easier to modify there. That move gives up the cheapest correction point in the whole process in exchange for installation convenience.
The second position covers everything that can go wrong during placement, while parts can still be moved.
Pre-reflow AOI verifies presence, orientation, polarity, and position of components on the board immediately after placement. Its value is timing: a missing or rotated part found here is corrected by placing it correctly, whereas the same part found after reflow requires desoldering, cleanup, re-inspection, and possibly scrapping the pad.
The station goes after the last placement machine in the sequence and before the reflow oven entrance. This matters more than it sounds: if there are three placement machines and the AOI sits after the second, everything the third machine did is unsupervised. Always inspect after the last machine that can introduce the defect you care about.
Double-sided boards pass through printing, placement, and reflow twice. The pre-reflow AOI position is used on each pass, inspecting the side just placed. Plan the board flip so that the flip happens after inspection of the first side, otherwise an operator or a conveyor turns the board over before anyone has checked it, and any handling damage is attributed to the wrong step.
This position must keep up with the placement line or it becomes the bottleneck. Where cycle time is tight, verify actual inspection time with your own boards before committing to the position, because AOI program complexity and board density affect per-board time more than any datasheet suggests. Inline platforms such as the AI-5146 are designed to sit at line rate in these positions within an SMT line.
A pre-reflow fail should route to a repair position where parts can be corrected with tweezers or a small rework tool, then return to the line before reflow. Because the board has not been through the oven, the repair is quick and leaves no trace. Keeping that return path short is one of the highest-value details in the whole layout.
Changeover risk belongs at this position too. A first article inspection machine such as the FAI1095 verifies the first board of a job against the bill of materials, the placement program, and the reference image before volume production begins, catching the setup errors that affect every board in a run rather than individual random faults.
Reflow creates a new family of defects and reveals others, so the oven exit is the third natural inspection position.
Placing AOI immediately after the oven means joint defects are detected while the board is still in the flow and before any further value is added. Every additional step after the oven, cooling stations, second-side handling, coating, or assembly, increases the effort required to repair what a hot joint reveals. Short delay, cheap repair: that is the whole argument for this position.
This position grades joint shape, fillet formation, component position after reflow, bridging, tombstoning, and surface contamination. It cannot see inside a hidden joint and it cannot confirm continuity. Those limits are physical, not programmatic, and they are why this position always hands off to something downstream. Guidance from IPC such as IPC-A-610 for assembled board acceptability provides the reference for what an acceptable joint looks like, without inventing limits of your own.
Boards leaving the oven are warm, and shrink varies with board design and material. The layout must allow enough conveyor distance for boards to stabilise before measurement, otherwise programs drift with board temperature. Provide that distance in the plan rather than discovering during commissioning that every first board after a pause measures differently.
This is the position where a real repair station becomes mandatory. Every reject needs a defined destination with tools, lighting, trained technicians, a defined acceptance standard, and a mandatory re-inspection before the board returns. If your layout has no room for that station, the AOI will generate fails that operators quietly push back onto the line, which is worse than having no inspection at all.
For double-sided boards, the post-reflow position inspects the side just soldered before the board flips. In some products the more valuable arrangement is to inspect side one after reflow and again after the second-side reflow, accepting the cycle time cost for the coverage. Whether that is justified depends on whether defects on the first side are accessible for repair after the second reflow, which is usually worse than doing it earlier.
Most high-volume lines carry this position inline because it feeds directly into the repair loop and the data system. Lower-volume operations sometimes use offline AOI at this stage and accept sampled coverage, which is a reasonable trade when the product tolerates it and the detection delay is understood.
X-ray is the position that most often goes missing from layouts, usually because it is slower and more expensive than optical inspection and because nobody is certain where it belongs.
Hidden joints only exist once reflow has formed them, so X-ray always sits after the oven. Placing it earlier inspects joints that have not yet formed. In practice this means either inline automated X-ray at the oven exit or an offline station fed from the end of the line or from a post-reflow buffer.
Inline automated X-ray keeps the coverage and the data flow continuous, at the cost of cycle time and significant floor space. Offline X-ray, such as the X-7100, suits sampled audit, new product validation, and failure analysis, taking the position off the critical path entirely. Higher coverage or higher resolution requirements point toward platforms like the X-7900 or the X-9200. The driver here is whether the product demands full coverage, not whether the line has room.
X-ray needs more than floor space. Plan for radiation safety interlocks, restricted access envelopes, shielding requirements, maintenance clearance, and any local licensing or periodic inspection obligations. These are site-level items that belong in the layout drawing early, because discovering them after installation usually means moving the machine.
An offline station needs a defined physical flow: where boards queue, who carries them, how results get recorded against the right identifiers, and how repaired boards return. A rack, a cart route, and a short walk are legitimate parts of the layout. Skipping them means operators invent their own route, and traceability quietly disappears.
For sampled X-ray, write the sampling rate, the trigger conditions for going to full coverage, and the record requirements next to the position on the drawing. A layout without that note invites a later argument about who decided a board was not inspected.
The final inspection family is electrical, and it sits after every assembly step that could change the netlist.
In-circuit and functional test verify the assembled board as an electrical object. Any placement performed after the test — a manual insertion, a shield, a connector — changes what the test proved. The rule is simple: electrical test goes after the last operation that can alter connectivity or component values.
Fixture-based ICT is fast per board and precise at locating faults, and it is naturally placed inline near the end of the line where the board can be loaded onto the fixture automatically. Online ICT fits this arrangement; offline ICT suits sampled or batch testing where the line cannot absorb another station. Fixture storage needs its own space, which layouts routinely forget.
Flying probe test is fixtureless and slower per board, which usually places it offline, fed by magazine racks from the line. It suits high-mix production, prototypes, and engineering changes. Because there is no fixture to store, the space requirement is modest, but operator time per batch is higher and belongs in the layout as a labour allocation rather than as floor space.
Functional test is usually the last station before packing, and it is product specific. Its layout needs are practical: power, fixturing, any environmental chamber or load, failure labelling, and a defined path for units that fail and need diagnosis. Treat it as a production station with its own takt rather than as a laboratory bench that happens to be on the floor.
Where the product has a through-hole stage after reflow, that stage has its own defects and deserves its own position. Inline solutions such as a DIP online inspection system like the AI-4026, the inverted camera AI-4540, or the dual-side AI-5146W can be placed directly after the soldering or insertion operation, keeping the same logic of inspecting where the defect is created.
Conformal coating introduces its own coverage and thickness risks. Where the product is coated, a dedicated inspection position after curing closes that loop, and platforms such as the AI-5146C coating AOI are built to sit inline at that stage.
The positions are only half the work. What connects them determines whether the layout functions under real production conditions.
Every inline station has to agree with its neighbours on transfer height, board width handling, direction, and the handshake signals governing release. These interfaces are mature and widely supported, but they still have to be commissioned station by station. A mismatch at one position produces jams that get blamed on the inspection machine rather than on the interface nobody checked.
Buffers between positions absorb short bursts and let upstream machines keep running during brief waits. They also increase the distance between cause and detection, which weakens the feedback loop. Size them for variation, not for hiding a cycle time mismatch, and prefer smaller buffers next to the positions whose feedback matters most.
Every transfer is a chance to bend a thin board, drop an assembly, or expose a board to an electrostatic event that later becomes an unexplained field failure. Inline positions remove transfers. Where manual handling is unavoidable, define the route, the carriers, and the grounding provisions, and keep the walking distance short.
Inspection machines need power, compressed air in some configurations, network connections, and often stable ambient conditions. X-ray needs its safety envelope. ICT needs pneumatic or vacuum connections for fixtures. Put these on the layout drawing as explicit requirements rather than assuming the electricians will work it out.
The whole layout should write to one record with one board identifier and one defect taxonomy. That requirement decides network topology, where servers or terminals sit, and which machines can export anything usable. Checking it before purchase saves a great deal of reconciliation later, because two machines that cannot agree on a defect vocabulary will never produce usable cross-stage analysis.
If you are adding inspection to a line that already runs, work backwards from the constraint rather than forwards from the ideal layout. Identify which machine currently sets the takt, find out how much line length can realistically be recovered, and decide which single position delivers the most improvement for the least disruption. Moving the highest-leverage position in first, usually solder paste inspection after the printer, then revisiting the rest once that has settled, produces far less disruption than attempting the whole layout in one shutdown.
Leave physical room and logical slack. A line drawn with no spare length cannot absorb a new station without a shutdown, and a network configured with no spare capacity cannot accept another data source. The best layouts assume that something will be added later and leave space, power, and connections for it.
Start from the process flow, place each inspection method inside the detection window of the defect it owns, give every fail a short and defined repair path, then specify the handling, buffering, utilities, and data connections each position requires. Get those four things right and the equipment will perform; get any of them wrong and no amount of machine capability compensates. If you are planning or reworking a line, the range of vision inspection equipment from I.C.T covers solder paste inspection, AOI, X-ray, first article, and electrical test in configurations for every position described here, and the AOI inspection solutions cover the optical positions in most layouts. Selection criteria for individual machines are set out in our guide on how to choose SMT inspection equipment.
Immediately after the stencil printer and before the first placement machine. This is the last point where a print fault can be corrected by wiping the stencil and reprinting, and paste volume predicts many of the defects that appear later.
Not always, but they cover different windows. Pre-reflow AOI catches missing, wrong, or rotated parts before soldering; post-reflow AOI verifies joint formation. Products with tight quality targets typically justify both positions, while simpler products may justify only one.
No. The joints X-ray inspects do not exist until reflow forms them. X-ray always sits after the oven, either inline as automated X-ray or offline as a sampled audit and failure analysis station.
After the last operation that can change connectivity or component values, including any through-hole assembly, manual insertion, or shield fitting. Anything performed after electrical test invalidates part of what it proved.
Enough to absorb normal variation and keep the printer running, but not so much that dozens of boards sit between the printer and the detection point. Large buffers delay the feedback that makes the position valuable.
Decide per position. Fast methods with high coverage requirements belong inline; slow methods, low volume, high mix, and shared resources fit offline. Most layouts end up as a mix, which is why the decision belongs on the layout drawing rather than being made once globally.
content is empty!