English
Views: 0 Author: Vinci Zhang Publish Time: 2026-11-10 Origin: Site
A PCB handling machine is deceptively simple to watch and deceptively easy to accept on hope, because it moves boards smoothly on the demo video and then, three weeks into production, drops a board once a shift and nobody can say why. The defense against that gap between looks-fine and actually-works is a real factory acceptance test and a real site acceptance test, written before the machine ships and ticked on the floor with evidence. This guide is that checklist, built around the things a loader or unloader can hide until they cost you yield, so you can sign off on proof instead of on a friendly demo.
Factory acceptance and site acceptance are two different moments in the life of the same machine, and confusing them is how a unit that passed a polite bench test arrives and fails the line it was meant to join. Each one exists to catch a different class of problem.
Factory acceptance is the test the builder runs before the machine leaves the plant, and its job is to prove the unit was built and configured to the order, not just that it powers on. You verify the as-built height, width range, fixed edge, and travel direction against the order form, and you watch a representative set of boards move through without jams, because a builder who runs the unit on their own bench has every incentive to show you the one configuration that works. The FAT is also where you confirm the SMEMA connector is wired and labeled as specified, so the wiring you receive is the wiring you approved, and the documentation you get matches the machine you get, which is the foundation the site test later builds on.
Site acceptance is the test you run after the machine is placed, leveled, and wired into your real line, and its job is to prove the unit works in your environment, not in the builder's. The neighbor machine, the floor, the power, and the real boards all differ from the factory, and the SAT is where those differences are either confirmed harmless or exposed as faults, because a loader that passed FAT on a clean bench can still fight a real printer over a one-millimeter height step. The SAT is the last chance to reject or condition the machine before warranty and production start, so it is worth more of your attention than the FAT, not less, and it is worth writing down as firmly as the factory test was.
Handling machines are often accepted as an afterthought bundled into the line acceptance, and that is a mistake, because a loader or unloader that fails subtly degrades every downstream station instead of failing loudly on its own. A printer fault stops the printer, but a loader fault starves the printer and the fault looks like a printer problem, so the handling machine needs its own test points spelled out so its behavior is judged directly rather than inferred from line symptoms. This guide gives those points as a list you can copy, and it pairs naturally with the SMEMA compatibility checklist, because the compatibility items are exactly the items the acceptance test must prove on the floor.
Before any motion test, you confirm the machine actually accepts the boards and magazines you will feed it, because a loader specified for one board family can silently mishandle another even if it never jams on the demo board.
You test the minimum and maximum board length, width, and thickness you actually run, including the extremes, because a loader that handles the median board can still skew or drop the smallest or flex the largest if the clamp or lift was not verified at the edges of the range. The acceptance point is a written record that each extreme board transferred cleanly a set number of times, not a verbal it-handles-everything, because the board you will regret is the one outside the demo set. For lines running small or thin products, the mini loader and unloader should be accepted against the same board extremes, since mini formats are exactly where edge cases hide.
Magazine loaders and unloaders live or die on magazine compatibility, and the acceptance test confirms the machine indexes the exact magazine you use, with the exact pitch and the exact side-wall style, because a magazine that is a few millimeters off in pitch will eventually mis-stack a board. You bring your real magazines to the test rather than accepting on the builder's generic one, and you record that the lift steps and the push-in align across a full magazine of boards, since a unit that works on an empty magazine and drifts on a full one is a unit that fails in week two. This is a quiet failure mode, and the acceptance record is what turns it from a mystery into a measurable pass or fail.
Flex and very thin boards bend, and a loader that transports a rigid board flat can still crease a flex one or let it sag into the rail, so the acceptance test includes a representative fragile board run with inspection of the board afterward for marks or stress. You define what acceptable handling looks like before the run, not after, because a mark you notice after sign-off is a mark you agreed to accept, and the handling machine should be held to the same board-care standard as the placement equipment it feeds. The acceptance point is a before-and-after board check, simple to write and powerful to enforce, and it is one of the items most often skipped and most often regretted.
The transfer is the heart of the machine, and it is where jams, double boards, and registration errors live, so the acceptance test spends real time here rather than a single pass.
The acceptance record should also note the transport speed used for fragile boards, because a handler that whisks a flex panel through at full line speed can induce vibration the inspection later reads as a defect, and the gentler setting is part of what you are accepting. You confirm the machine offers a separate transfer profile for delicate products and that switching to it is a recipe choice rather than a manual override, because a profile only an engineer can invoke will not be used by the night shift. The test is a short run on the fragile board at the gentle profile with a board check afterward, recorded as a pass only if no mark or stress is visible, and this is one of the items where a written criterion prevents the quiet acceptance of damage that shows up as scrap three stations later.
A loader or unloader must reliably know when a board is present and when the neighbor has taken it, and the acceptance test confirms the sensors read correctly across board types and across a sustained run, because a sensor that misses one board in a thousand is a sensor that causes one unexplained jam a shift. You verify the sensing on the demo board and on your extreme boards, and you confirm the machine reacts correctly when a board is missing or stuck, since the safe failure is a stop, not a push into an empty or occupied space. This is a small test point with large consequences, and it is worth ticking with evidence rather than assumption.
When the board leaves the loader and enters the neighbor, it must arrive centered and at the right height, and the acceptance test measures the registration at the handoff rather than trusting it, because a board that is off by a couple of millimeters at the loader becomes a misregistration at the printer. You place a marked board and measure where it lands downstream, and you repeat across widths, because alignment that holds at one width can drift at another, and the test is there to catch that drift before production does. The companion guide on how to choose a loader and unloader frames the selection, but the SAT is where the choice is proven against your real alignment tolerance.
A good handling machine detects a jam or a double-board feed and stops instead of crushing or stacking, and the acceptance test deliberately provokes those conditions to confirm the detection works, because a machine that only fails safely when nothing goes wrong is not safe. You simulate a stalled downstream and a double release and confirm the unit halts and alarms rather than forcing the board through, and you record that the recovery is clean, since a detector that stops the line but cannot be cleared is its own kind of failure. This is the test that separates a robust handler from a fragile one, and it is worth the few minutes it takes on the floor.
The machine must speak to the line and to the safety system, and the acceptance test proves the conversation actually happens, because signals that look wired on the drawing often fail in practice.
The acceptance test should also confirm that recovery from a detected jam does not itself cause a second fault, because a handler that stops safely but then double-feeds when restarted is only halfway safe, and the restart behavior is part of what you are signing for. You clear the simulated jam and confirm the board count reconciles, with no extra board released and no board lost, since a unit that loses track of boards during recovery will eventually desync the magazine and the line. The record notes the recovery sequence and who can perform it, because a recovery only the builder understands is a recovery the line will bypass, and bypass of a safety interlock is the start of the next incident rather than the end of the test.
You test the SMEMA request, ready, and stop signals against the real neighbor, confirming the loader releases only when the downstream is ready and halts when the line stops, because a handshake that works on a loopback and fails on the line is the most common hidden defect. The test is a sequence: stop the neighbor, confirm the loader waits, release the neighbor, confirm the loader sends, and fault the line, confirm the loader halts, recorded as a checklist rather than a glance. This directly confirms on your actual equipment the same handoff risks the loader and unloader placement guide flags at the line boundary, and it is the moment the paper checklist becomes a floor fact.
Every meaningful alarm on the loader or unloader should map to the line stop so a fault upstream of the inspection never lets a bad board ride into the good bin, and the acceptance test confirms the mapping by triggering each alarm and watching the line respond. For an NG/OK unloader after inspection this is critical, because a divert fault that does not stop the line can misroute a rejected board into the good magazine, and the SAT is where that risk is caught or accepted consciously. You record which alarms stop the line and which are local only, because an alarm no one mapped is an alarm no one will trust.
The E-stop test runs the stop signal through the loader and unloader as part of the cell safety chain, confirming a stop anywhere halts the whole cell and that the handling machine cannot be restarted out of sequence, because a handler that ignores the cell E-stop is a hazard, not a convenience. You trip the stop at the handler and at a neighbor and confirm both halt the cell, and you confirm the reset order is safe, since a machine that restarts while a human is inside is the failure mode this test exists to prevent. The acceptance record is what the safety audit asks for, so writing it down turns a likely finding into a confirmed control.
These mechanical items were specified at order and checked for compatibility, and the acceptance test confirms they arrived as specified, because a machine built to the wrong fixed edge looks correct until the first board is offset.
The acceptance test also confirms that the handler cannot be left in a state where the E-stop is defeated, because a machine that passes the stop test but can be silently overridden is a machine that will be overridden on a bad day. You verify the reset requires a deliberate, sequenced action and that the handler will not move until the whole cell is cleared, and you record that the behavior is the same from every E-stop point, since a stop that works from one panel and not another is a stop that fails exactly where the hazard is. The signed safety record is what the audit expects, and writing the test down turns a probable finding into a confirmed control that protects both the line and the people beside it.
You command the width to your minimum, median, and maximum and measure the result each time, and you repeat the cycle to confirm it returns to the same place, because a width that drifts changes where the board sits at every downstream station. The acceptance point is a measurement, not a readout, since the PLC value and the physical rail can disagree after transport, and the floor is where that disagreement shows. A 90 degree loader or unloader in an L-shaped line is especially sensitive to width repeatability at the corner, so the test there is worth the extra cycles.
You confirm the travel direction and the fixed edge match the line as built, by running a marked board and seeing which edge it references and which way it leaves, because a machine that is wrong here will offset every board and the error compounds downstream. This is a quick test but an easy one to skip, and it is the one most often discovered during the first real product, so ticking it at acceptance saves a confusing first shift. The direction and edge are mechanical facts, and the acceptance test treats them as facts to be proven, not promises to be trusted.
If the line drives width by recipe, the acceptance test sends a changeover and confirms the loader or unloader reaches the new width and only then signals ready, because a handler that releases a board before the rail is set is a handler that jams the new product immediately. You run at least one full changeover during SAT and confirm the handshake waits for width, since this is where standard versus mini loaders and unloaders differ in how they behave under auto-change. The test turns a specification into a demonstrated behavior, and it is the last of the width-family checks before the machine is signed off.
The machine must keep up with the line without colliding boards at the boundary, and the acceptance test proves the overlap rather than assuming it from the datasheet.
The acceptance test should also confirm that a width change interrupted mid-cycle completes or aborts safely rather than leaving the rail half-set, because a handler caught between widths is a handler that jams the very next board, and the interrupted case is the one most often left untested. You start a changeover and halt it partway, then confirm the machine either finishes to the commanded width or returns to the prior safe width before signaling ready, and you record which behavior it chose, since both are acceptable but only if deliberate. This is also where the magazine versus vacuum loader choice matters, because the two mechanisms recover from an interrupted width change differently and the SAT is where that difference is proven on your floor.
You run the loader, the line, and the unloader at the line tact speed and confirm all three share a comfortable overlap, and you push to the extremes to confirm the handler never starves or floods the neighbor, because a speed number on paper is not a speed proven on the floor. The test records the minimum and maximum at which the handoff stays clean, so the operating window is known rather than guessed, and the line can be run inside it with confidence. This is a throughput item, but it is also a reliability item, since the speed mismatch is what causes the boundary jams described earlier.
You measure the board pitch and the gap at the handoff during the run and confirm they stay within the neighbor tolerance across speeds, because pitch that drifts with speed is pitch that eventually collides boards or gaps the line. The acceptance point is a measurement at several speeds, not a single observation, since the drift is exactly what shows up only under variation, and the test is where variation is intentionally introduced. A magazine buffer or vertical buffer at the boundary should be included in the pitch test, because the buffer handoff is another place pitch can drift unnoticed.
The most revealing acceptance test is a sustained run that lets small errors accumulate, so you run the handler for a meaningful period with a realistic board mix and watch for the slow drift, the occasional jam, and the creeping misregistration that a five-minute demo never reveals. You also test accumulation, filling and emptying the buffer, to confirm the handler respects the buffer state instead of overrunning it, because a handler that passes a short run and fails a long one is the classic almost-works machine. The sustained run is the test that earns its time, and it is the one most often cut for schedule, which is exactly why it should be protected.
A machine the team cannot run or maintain is not accepted no matter how well it moved boards, and the acceptance test includes the human and paper handover as firm items.
The acceptance record should state the board mix and the approximate duration in terms of cycles rather than clock time, because the drift you are hunting appears after a number of transfers, not after a number of minutes, and a cycle count is what makes the test repeatable on the next machine. You also note the ambient and the shift conditions if the run spans a break, since a handler that behaves differently after a temperature swing is a handler that will confuse the first cold-start shift, and the SAT is the place to surface that rather than production. The sustained run is the single most valuable test in this guide and the one most often shortened for schedule, so protecting it in the plan is protecting the reliability of the whole cell, not just the handler.
You require a documented training session with the actual operators and maintenance staff, not just the engineer who ordered the machine, because the people who run it daily are the ones who will catch the first anomaly, and they need to know what normal looks like. The acceptance point is a signed training record and a short demonstration by the operator, since a handler the operator cannot run safely is a handler that will be bypassed, and bypass is where accidents and yield loss begin. Training is part of SAT, not a courtesy after it, and writing it into the checklist makes it unavoidable.
You confirm the spare parts list, the consumables, and the manuals are delivered and match the machine, because a handler with no documented spare for its lift belt is a handler that will stop the line for a week when that belt eventually wears. The acceptance test checks that the manual covers the SMEMA wiring, the width mechanism, and the alarm list, and that the recommended spares are on hand or ordered, since a machine you cannot repair is a machine you cannot rely on. This is a quiet acceptance item that pays back the first time something wears, and it is worth ticking with the same seriousness as the motion tests.
You confirm the line recipe and the mode interlock are handed over working, so the handler enters auto with the rest of the line and honors manual when an operator steps in, because a handler that drops out of auto at changeover is a handler that stalls the new product. The acceptance test runs a changeover with the operator and confirms the mode and recipe behavior, and it records the expected behavior so future shifts know what normal is, since mode faults are the ones most often mistaken for board faults. The PCB handling category page lists the equipment families this applies to, and the handover should cover whichever ones are in your cell.
The sections above are the detail, and this section gathers them into a single list you can print and carry, so acceptance is a steady tick rather than a recollection after the fact.
The acceptance test also confirms that the handler reports its state back to the line so the line knows the handler is ready, because a handler that is ready but silent leaves the line guessing and sometimes waiting, and the handshake is incomplete without that report. You confirm the ready signal follows the width and mode settle and that a fault clears only with an explicit acknowledgment, since a handler that self-clears a fault and resumes can restart a board into a condition the operator has not reviewed. The handover record names the person trained and the date, so the warranty and the knowledge transfer are tied together, and the next shift inherits a working mode and a known contact rather than a mystery and a guess.
Before the machine leaves the builder, tick the as-built height, width range, fixed edge, and direction against the order, tick the SMEMA connector wiring and labeling, tick a clean transfer of a representative board set, and tick the documentation matching the machine. These are the FAT points, and ticking them with evidence means the site test starts from a confirmed build rather than a hope, because the factory is where the machine is most controllable and most honestly testable. Carrying this list to the factory visit turns a polite tour into a real acceptance, and it is the habit that prevents the most expensive surprises.
After install, tick the real-line SMEMA handshake, the alarm-to-stop mapping, the E-stop through the cell, the width repeatability, the direction and fixed edge, the speed overlap, the pitch consistency, the sustained run, and the training and spare-parts handover. These are the SAT points, and they are ticked on your floor with your boards and your neighbors, because that is the only environment that proves the machine works where it will actually live. The IPC acceptance guidance is a useful reference for what a serious acceptance record should contain, and the list here is built to match that seriousness without inventing numbers.
Acceptance ends with a signed record that states what was proven and what was conditionally accepted, because an unqualified sign-off hides the items you chose to live with and turns them into disputes later. You date the sign-off, you note the warranty start tied to a clean SAT rather than to shipment, and you keep the tick list as the baseline for future maintenance, since the acceptance record is also the first maintenance record. A clean sign-off is not paperwork, it is the moment the machine becomes yours with eyes open, and the checklist is what makes that moment real instead of ritual.
You need both, because FAT proves the machine was built to the order and SAT proves it works in your real line, and a unit that passes the builder's bench can still fight your floor over a height step or a signal polarity, so skipping SAT is how a machine passes on paper and fails in production.
The run should be long enough to let small errors accumulate rather than short enough to look good, and the right length is the one that covers a realistic board mix and a buffer fill-and-empty cycle, because the value is in exposing drift that a brief demo never shows, not in hitting a fixed clock time.
That is the classic hidden defect, and it is exactly why the acceptance test runs the handshake against the real neighbor rather than a loopback, because the loopback only proves the signals exist while the line proves they agree with the equipment the machine will actually talk to.
Yes, because the handler must respect the buffer full or empty state during both the pitch test and the accumulation test, and a buffer the handler overruns is a buffer that spills boards, so the buffer is treated as a normal station in the acceptance plan rather than an afterthought.
The actual operators and maintenance staff should be there, not only the engineer who ordered the machine, because the daily operators are the ones who will catch the first anomaly and they need to know what normal looks like before the line depends on them.
A clean sign-off records what was proven and what was conditionally accepted, dates the warranty start to a good SAT rather than shipment, and gives maintenance a baseline, so the moment the machine becomes yours is documented with evidence instead of assumed, which is what prevents later disputes and hidden liabilities.
