Views: 0 Author: Elsa Long Publish Time: 2026-11-19 Origin: Site
Readiness guide · Published by I.C.T · For process engineers, plant engineers and equipment buyers
Most installation delays are not equipment failures. They are site conditions that were assumed rather than measured: a door that is 20 mm too narrow for the crate, a transformer that was sized from a nameplate that did not include the chiller, an air drop that reads 6 bar at the compressor and 4.6 bar at the tool, an exhaust connection with no make-up air behind it. Each of these costs a day or more, and each of them is visible weeks in advance to anyone who walks the route with a tape measure and a checklist instead of a drawing.
This PCBA factory readiness checklist before installation is organized the way the site is inspected rather than the way a contract is written. It covers the floor and the access route, electrical capacity and earthing, compressed air and exhaust handover points, network and ESD, consumables and sample boards, and the people and safety conditions that decide whether the line can legally run on day one. Every numeric example is a stated assumption for illustration, not a measurement from a named project. The equipment those sections are preparing for is described under the production line configurations that this checklist is written for, and the site conditions remain the variable part of every project.
Readiness is not a feeling about the site. It is a set of conditions that somebody with authority has inspected, measured where measurement is possible, and signed against a date. The distinction matters because the failure mode is almost never a contractor refusing to do the work; it is everyone believing the work is done. A floor is poured, so the floor is ready — except the overlay has not cured and a pallet jack will mark it. A panel has a spare breaker, so power is available — except the breaker feeds a circuit that also carries the compressor. Compressed air reaches the wall, so air is ready — except the pipe is 12 mm and the tool needs 400 liters a minute. Turning those beliefs into evidence is the whole job of a readiness review, and it has to happen while there is still time to fix what it finds. A PCBA factory readiness checklist before installation is worth exactly what the evidence behind each line is worth, which is why the sections below ask for readings, dates and photographs rather than for confirmations.
Every line on a readiness checklist should resolve to one of three states rather than a yes or no: closed, conditionally closed, or open. Closed means the condition exists and there is evidence dated within the review window — a photograph with a tape measure in frame, a meter reading, a signed handover sheet. Conditionally closed means the condition will exist by a stated date, the person who will make it exist has agreed in writing, and there is a fallback if that date slips. Open means nobody has committed, and it should be treated as a scheduling problem rather than a site problem, because it will move the installation date.
The reason to insist on three states is that binary checklists produce dishonest ticks. Faced with a yes/no column and a truck already booked, most site teams will tick yes and add a note in the margin. A three-state column lets the same team record the truth without feeling that they are cancelling the project, and it gives the project manager something actionable: a conditional item with a date and an owner can be chased, while a marginal tick cannot. In practice, a first-pass review of a hall that is about to receive a line tends to come back with roughly half the items closed, a third conditional, and the rest open.
Someone has to own the gate. On a project run by a single equipment supplier the gate owner is usually the customer's plant or process engineering lead, because the supplier can only confirm what is inside its own scope of supply. When several parties are on site — a building contractor, a utility contractor, an MES integrator and one or more equipment suppliers — the gate needs an interface owner as well, and that role is worth writing down before the first review rather than after the first argument. The interface rows that generate most of the disputes are predictable, and they are laid out in who owns each readiness item when several parties share the site.
Evidence has a quality scale, and auditors should know it. A verbal confirmation is not evidence. A photograph of a machine in position is not evidence that it is powered. A meter reading with no date, no instrument and no load condition is close to useless, because voltage at rest tells you nothing about voltage at startup. The evidence worth filing is narrow and boring: a reading, taken at a named point, under a stated load, on a date, by a named person. Once that standard is applied, the number of genuinely closed items usually drops by half in the first review, which is uncomfortable and considerably cheaper than discovering the same gap on the day the crew arrives.
A readiness review that ends in a discussion ends in optimism. The alternative is to weight the items before the review, score them during it, and let the arithmetic decide. Weights should reflect consequence rather than effort: an item that stops energisation is worth more than an item that costs a day of rework. A workable weighting for a single-line project is 30 percent for anything that gates power-on, 30 percent for building and utility items with long lead times, 20 percent for access and material, and 20 percent for people, documentation and safety. Items are then scored 1 for closed, 0.5 for conditionally closed with a date, and 0 for open.
With that scheme, a project sitting at 0.85 with all power-on items closed is in a very different position from a project at 0.85 with one power-on item open, and the weighted score shows it while an unweighted count does not. As an illustration, assume a review of 24 items where six are power-on: five closed and one open gives 0.30 × (5 ÷ 6) = 0.25 from that block alone, and the same overall average cannot hide the missing fifth. The score is not a substitute for judgment, but it forces the judgment to be stated in a form that can be argued with, which is the most a checklist can do.
State | What it means in practice | What closes it | Evidence an auditor accepts |
|---|---|---|---|
Closed | Condition exists now and has been observed | Nothing; record and date it | Dated reading, photo with scale, signed handover |
Conditionally closed | Will exist by a stated date, with an owner and a fallback | Delivery of the named action on the named date | Written commitment plus a revisit date in the plan |
Open | No owner, no date, or both | Assign an owner and a date, or move the installation | Held as a scheduling risk until it becomes conditional |
The route into the building is the first thing to walk and the last thing anyone photographs, and it decides whether the installation takes two days or six. A loaded crate is not a machine: it is heavier, wider, taller, and it cannot be lifted by the corner because the center of gravity has moved. The check therefore has four parts — what the slab will carry, what the openings will admit, what the lifting equipment can do, and where the crate can sit while it is opened. All four can be verified with a tape measure, a load calculation and one walk-through, long before the shipping date is confirmed, and all four are cheap to fix at that point and expensive to fix afterwards. On any PCBA factory readiness checklist before installation this is the block to walk first, because a route that fails invalidates the plan for every item behind it.
Floor checks usually begin and end with the slab rating in kilograms per square meter, which is the least useful number on its own. A machine rated at 2,200 kg over a 1.2 by 1.0 meter footprint produces a distributed load of about 1,830 kg/m², but the crate it travels in weighs more, sits on bearers that concentrate the load into a much smaller area, and is moved on a pallet truck or forklift whose wheels apply the load to two narrow strips. As an illustration, assume a crate at 2,800 kg standing on three bearers each 0.15 m wide and 1.2 m long: the contact area is 0.54 m², so the average pressure under the bearers is roughly 5,200 kg/m², nearly three times the distributed figure for the machine itself.
The practical remedy is to spread the load rather than argue about the number. Steel plates or thick plywood pads under the bearers bring the pressure down to something the slab and the floor finish will survive, and they cost a fraction of a repair. Two other floor items belong on the same line of the checklist: the finish, because a freshly applied epoxy coating needs its full cure before a pallet truck turns on it, and levelness, because a line assembled on a floor with a visible fall will be shimmed into a shape it was not designed for. If the building is being prepared around existing production rather than standing empty, the sequence changes and the constraints come from what is still running — the distinction is set out in the difference between preparing an empty hall and preparing a running plant.
Measure the opening, not the door. A nominal 1.8 meter double door loses 100 to 150 mm to frame, hinges and a closing mechanism, and the crate needs clearance on both sides to avoid damaging the frame it is being pushed through. The same applies vertically: a crate that clears the header by 50 mm will not clear it once the pallet truck lifts it, and a dock leveler in the raised position can eat 200 mm of headroom without anyone noticing. The useful discipline is to measure the narrowest point of the whole route rather than each opening in isolation, because the route width is set by its worst constraint, not its average.
Corners are where the arithmetic surprises people. A crate 2.4 m long and 1.1 m wide on a pallet truck needs a corridor width roughly equal to its diagonal swept path, which for a right-angle turn is closer to 3.4 m than to the 2.4 m length. If the corridor is narrower, the options are to open the crate outside and carry the machine in pieces, to use a different entry, or to remove a partition temporarily — and all three need to be decided before the truck arrives, not while it is waiting at the gate. Photographing the route at each corner, with the tape visible, takes twenty minutes and settles the argument permanently.
The last ten meters are usually the most expensive. A ramp steeper than about 5 percent becomes unsafe for a manual pallet truck with a load over a tonne, and above 8 percent it is a controlled risk that most site safety rules will not accept. If the route uses a goods lift, the lift needs a rated capacity above the crate weight plus the handling equipment, internal dimensions that admit the crate with clearance, and a floor that will take the point load of a laden truck. Capacity plates are often unreadable after a few years, in which case the lift certificate is the document to ask for rather than the plate.
Where none of these work, the fallback is a crane or a telehandler through a temporary opening, which needs a lift plan, a permitted area and a day when the weather allows it. That plan is worth writing even if it is never used, because the existence of a documented fallback changes the decision about the delivery date: a project with a fallback can accept a marginal route, while a project without one is betting the schedule on a door frame. The alternative equipment route — installing a line in more, smaller sections — is one of the reasons short lines are easier to site than long ones, and it is visible in something as simple as a short DIP line where the readiness set is smaller.
Staging is the item that turns a two-day installation into a four-day one. Crates have to be opened somewhere, the packaging has to go somewhere, and machines have to wait somewhere while the ones behind them are placed. A line installed from the far end toward the door needs a staging area at the near end; a line installed from the door inward needs the staging area outside the building or in a side aisle. Choosing the wrong direction means moving machines twice, and moving a machine twice on a finished floor is where most floor damage happens.
A practical rule is to reserve a staging strip at least as wide as the widest crate and as long as the two largest crates, plus a walkway that keeps people out of the forklift path. In a hall that is also going to run production afterwards, that strip usually has to become the main aisle, which is why aisle width should be fixed at the layout stage rather than during installation. The minimum for a line with inline equipment is commonly quoted as 1.2 m of clear walkway plus the service space the machines need at their own access panels, and the two should be added rather than compared. The conveyor handoff between adjacent units follows the SMEMA mechanical interface convention, which fixes the transfer height and the point where responsibility passes from one machine to the next; the readiness item is therefore the floor level and the access space under that handoff, not the interface itself.
Check | What to confirm | How to verify | Typical failure symptom |
|---|---|---|---|
Slab capacity | Rated load against crate pressure, not machine weight | Load calculation plus plate or pad sizing | Floor finish cracking under bearers |
Narrowest opening | Clear width after frame, hinges and door hardware | Tape measure at the tightest point of the route | Crate clears the door but not the frame |
Right-angle corner | Swept path of the longest crate | Diagonal estimate plus a walk-through | Machine unloaded in the yard instead of the hall |
Ramp or lift | Gradient, rated capacity, internal dimensions | Certificate, not the capacity plate | Load refused by the site safety officer on the day |
Staging strip | Space for two unopened crates plus a walkway | Marked on the floor before delivery | Machines moved twice; packaging blocks the aisle |
Installation direction | Sequence agreed with the layout and the doors | Walk the sequence in reverse | Last machine cannot be reached without dismantling |
Electrical readiness is the item that most often looks complete and is not, because the evidence people collect is the wrong evidence. A voltage reading taken with the line idle says that the transformer exists; it says nothing about what happens when three servo axes accelerate at the same moment. Capacity is the first question and power quality is the second, and both are answered by measurement under load rather than by reading a drawing. The good news is that the measurements take an hour, need one meter and one person who knows how to use it, and produce numbers that a contractor cannot argue with.
Start with the utility sheet for every unit rather than with the machine nameplate, because the nameplate often describes the largest single load while the utility sheet describes the connected load including heaters, chillers and pumps. Sum the connected loads, apply a diversity factor that reflects how many units actually run at full power simultaneously, and compare the result with the available capacity at the panel — not with the capacity of the building. As an illustration, assume a printer at 3 kVA, a mounter at 6 kVA, a reflow oven at 45 kVA, an AOI unit at 2 kVA and a dedicated compressor at 15 kVA: the arithmetic total is 71 kVA, and a diversity factor of 0.75 gives about 53 kVA to reserve, before any chiller or upstream equipment is added.
Then measure. Record line-to-line voltage with the line idle, again with every heater on, and again at the moment the largest motor starts. A reading that falls from 400 V to 386 V under steady load is a 3.5 percent drop and is usually acceptable; a transient dip to 372 V at startup is close to 7 percent and, combined with an unbalanced phase, may put a drive into undervoltage trip on the first production shift. Phase imbalance is worth measuring at the same time: a difference of more than about 2 percent between phases indicates single-phase loads distributed badly across the panel, and rebalancing them is a morning of work compared with a year of intermittent faults.
Earthing is checked in two ways and both are needed. The first is the resistance of the earth electrode system, for which local code sets the target — 4 ohms or lower is a common figure, but the applicable value should come from the electrical contractor who signed the installation rather than from a checklist. The second is continuity from each machine's protective earth terminal back to the building earth bar, measured with a bond tester rather than inferred from the fact that the cable has a green-and-yellow core. A daisy-chained earth that relies on the mechanical joint of a cable tray is the single most common finding in an ESD and electrical audit, and it is invisible until something goes wrong.
Residual current protection needs a design decision rather than a default. Machines with variable frequency drives and three-phase rectifiers leak current to earth by design, and a Type A device may trip on normal operation while a Type B device will not. Where the board carries both drive loads and office circuits, splitting them onto separate devices prevents the classic nuisance trip that shuts a line down for no diagnosable reason. The neutral deserves the same attention: switch-mode power supplies in modern equipment inject harmonic current, the neutral conductor can carry more than its nominal rating, and a neutral sized on the assumption of linear loads is a slow-burning problem rather than an immediate one.
If the site runs on a generator, on a shared transformer, or on a supply that is already close to its limit, the readiness question changes from capacity to stiffness. A generator rated at 100 kVA does not deliver 100 kW to a load with a poor power factor, and it will not absorb the starting current of a large motor without a voltage dip that drives will interpret as a fault. For a line that must run through commissioning on temporary power, the two numbers worth confirming are the generator's continuous rating at the site's power factor and its total harmonic distortion under the actual load, both measured rather than quoted from a brochure.
A shared transformer introduces the opposite problem: the capacity exists, but it is shared with loads nobody on the project controls. A welding bay on the same transformer will produce dips that a reflow oven controller may see as a supply fault, and the fix is either a dedicated feed or a conversation with the neighbor about scheduling — both of which are possible before installation and awkward afterwards. The general rule is that any supply condition which depends on somebody else's behavior belongs on the checklist as a conditionally closed item with an owner, not as a closed item with a hope attached.
Air and exhaust are the two utilities where the number in the document and the number at the machine diverge most often, because both are measured at the wrong end. Compressed air is quoted at the compressor and needed at the tool; exhaust is designed at the fan and needed at the flange. Between those two points sit pipe diameters, filter regulators, flexible hoses, duct lengths, bends and dampers, each of which takes a share of the margin. A handover that specifies the value at the machine connection point, names the instrument, and records the reading on the day is worth more than a page of design intent, and it is the only form of handover that survives a change of contractor. It is also the block of a PCBA factory readiness checklist before installation where numbers are most often quoted from the wrong end of the system, which is why the measurement point is part of the specification rather than an implementation detail.
Pneumatic readiness has three parts: pressure, flow and quality. Pressure should be confirmed at the machine's inlet regulator with the machine running, not at the compressor gauge, and a drop of more than about 0.3 bar between the main header and the tool usually points to undersized pipe or a clogged filter rather than to the compressor. As an illustration, assume a header at 7.0 bar and a machine inlet reading 6.2 bar after 40 meters of 15 mm pipe with four elbows: the 0.8 bar loss is more than twice the acceptable figure, and the usual fix is one size up in pipe diameter rather than a bigger compressor.
Flow is the figure most often omitted from the RFQ. A machine may need 300 to 600 normal liters per minute when it is cycling, and several machines on one drop will starve each other if the pipe was sized on average rather than peak demand. Quality is the third part and the one that causes the strangest faults: moisture in a line shows up as intermittent sensor failures and sticky cylinders, oil carryover shows up as contamination on boards, and both are prevented by a dryer and filter set sized for the actual flow rather than for the compressor. The quality class should be stated in the handover document using the ISO 8573-1 notation, so that it can be verified rather than assumed.
Exhaust is verified at the machine flange with an anemometer and a duct cross-section, and the result should be compared with the figure in the equipment's utility sheet rather than with the fan curve. A reflow oven with three heated zones, for example, may call for roughly 500 CFM per zone in the supplier's documentation, which for three zones is about 1,500 CFM or 42 m³/min in total; the exact figure for a specific model belongs in that model's utility sheet, and it should be confirmed there rather than inferred. The measurement point matters because a reading taken at the roof fan includes every leak in the duct, and the machine sees none of that margin.
The half of the problem nobody writes down is make-up air. Every cubic meter extracted has to be replaced, and in a closed hall the replacement arrives through the door gaps, which produces drafts, dust and a negative pressure that makes doors hard to open and pushes the building's own ventilation out of balance. Make-up air should be at least equal to the extracted volume, delivered where it does not blow onto the line, and tempered so that the hall does not drop below the temperature the process needs. Where the building has no make-up air unit, the practical options are a dedicated inlet with a filter, a louvered opening sized to the extract volume, or a reduction in extract — and the third option usually conflicts with the oven's own process requirements.
A utility handover needs a physical point, a value and a signature. The physical point is the flange, valve or terminal where the supplier's responsibility stops and the customer's begins; the value is the measured quantity at that point; the signature is the person who will be called if it is not there on the day. Without a named point, handovers drift: the electrician believes the connection is the supplier's, the supplier believes it is the electrician's, and the gap is discovered when a commissioning engineer arrives without the right fitting.
The handover document should also record the condition, not only the value. Compressed air at the right pressure but with visible water in the filter bowl is not a passed item. Exhaust at the right flow but with a flexible connection that collapses under negative pressure is not a passed item either. Photographing each handover point on the day it is signed takes ten minutes, and it converts a disputed recollection into a dated record — which is the same standard the site will apply to the equipment acceptance two weeks later.
Utility | Value to confirm in writing | Where to measure it | Common failure |
|---|---|---|---|
Compressed air pressure | Dynamic pressure at the machine inlet regulator | Machine connection, with the tool cycling | Header reading accepted instead of inlet reading |
Compressed air flow | Peak normal liters per minute for the whole line | Flow meter at the drop, or summation of utility sheets | Pipe sized on average demand |
Air quality | ISO 8573-1 class for particles, water and oil | Downstream of the dryer and filter set | No dryer; moisture in cylinders and sensors |
Exhaust flow | Figure from the equipment utility sheet, per connection | Machine flange, duct cross-section measured | Roof fan reading used as proof |
Make-up air | At least equal to extract volume, tempered | Inlet louver or AHU supply | Negative pressure, drafts, doors that will not close |
Handover record | Point, value, date, instrument and signature | On the drawing, marked and photographed | Verbal agreement between two contractors |
Material readiness is what separates an installed line from a producing line, and it has a longer lead time than most project plans assume. The machine arrives with the ability to place parts; it does not arrive with the parts, the feeders to carry them, the nozzles to pick them, the stencil to print them, or the boards to prove it. Each of those has its own clock, several of them run on the supplier's schedule rather than the customer's, and the ones that depend on a customer-supplied item are the ones that slip. The section below treats them in the order they bite: paste and stencil, tooling, boards, and the money that has to be spent before the first good unit comes off the end of the line.
Solder paste is a dated material and it starts aging the day it is manufactured, not the day it is opened. Refrigerated storage typically extends the working life to around six months from the date of manufacture, and ambient storage shortens it to weeks, so the first decision is whether the site has refrigerated space with a logged temperature before the paste is ordered. Thawing is a second clock: a jar taken from the refrigerator needs two to four hours at room temperature before use, and a site that thaws paste in a warm cabinet or a water bath is creating a printing problem that will be blamed on the printer.
Stencils carry their own dependencies. The aperture design follows the process route and the component mix, so the stencil cannot be finalized until the route is decided — the equipment and process sequence that the stencil is designed for is described in how the process route fixes the equipment list before anything can be checked. Tension, frame type, thickness and the number of spares all need to be agreed with the printer supplier rather than assumed, and a stencil ordered after installation adds days to the first-article date. Ordering one spare with the first set costs little and removes a single point of failure from the first week.
Feeder count comes from the bill of materials, not from the machine's slot count. The number of unique part numbers on a run, the number of those that need more than one lane for continuous operation, and the width mix across 8, 12, 16 and 24 mm reels together determine how many feeders are actually needed, and the arithmetic is usually larger than the number of slots the machine has. Where the count exceeds the slots, the practical answers are a second setup, a pre-staged feeder cart, or a sequence that runs the high-mix part numbers in batches; all three are planning decisions that get made badly under time pressure on the day of installation.
Nozzles are the item most often discovered missing, because they are small, ordered by family rather than by part number, and easy to leave out of a purchase order that was written around machines. The reliable method is to list the component families on the BOM, map each family to the nozzle types the placement head supports, and count how many of each are needed for a complete setup plus a spare set. A missing nozzle type stops a whole product family rather than a single part, so the cost of the spare set is almost always smaller than a day of lost commissioning time.
The first-article plan should exist before the line does. It defines how many boards are needed, which ones are representative, what is measured on each, and what the acceptance criteria are — and it is the document that decides whether commissioning is finished. A typical first run uses three to five boards: one to establish the process, one to adjust it, and one to keep as a reference for the next time the product runs. The boards should be the real product or a faithful equivalent, because a test coupon that prints differently from the product proves nothing about the product.
Large-format work changes the arithmetic, and it is worth checking the board size against the carrier and conveyor capability before the first-article kit is assembled. On a line built for large panels, such as an end-to-end line for large smart-panel boards, the sample boards are themselves a logistics item: they need transport frames, they need a place to be stored flat, and they need to be counted before the run rather than after it. Whatever the size, the acceptance data from the first article — print volume, placement offset, reflow profile, inspection results — becomes the baseline that later production is measured against, so it has to be recorded rather than remembered. The reflow profile in particular should be captured against the process windows described in IPC-7530 and filed with the run, because a profile that exists only in a machine's memory is lost the first time the machine is serviced.
Consumables are usually budgeted as a monthly cost and needed as a first-week cost, which is how sites end up with a fully installed line and no paste. The first-month list is short and predictable: paste in the quantity the first builds require plus one spare jar, stencils for the first products plus a spare, wipes and cleaning solvent, nozzle spares, feeder spares for the most-used widths, ESD-safe bags and trays, and the labels or carriers the traceability system needs. Buying these before installation is not over-ordering; it is the difference between commissioning and waiting.
A useful discipline is to price the list against the cost of one idle day. As an illustration, assume four people on site at USD 180 per person-day, which is USD 720 a day, plus the unrecovered share of a commissioning engineer's visit: a single lost day is comparable to the entire consumable list for a small line. Framing the purchase that way tends to end the discussion about whether the spares are necessary, and it also produces a defensible number for the project's contingency line.
These three items share a property that makes them dangerous: nothing visibly fails when they are wrong. A line with no network drop still runs, right up to the moment somebody wants traceability data, remote support or a program change. A line on a floor with the wrong surface resistance still runs, right up to the moment a batch fails a customer audit or a board arrives at test with a damaged input. Because the failure is delayed, the work is deferred, and because the work is deferred it is usually done in a hurry after the first problem. Checking them before installation costs a day; retrofitting them costs a shutdown and a conversation with the customer. That is why the third block of a PCBA factory readiness checklist before installation is devoted to the items that nobody can see, and why they are scored with the same weight as the machines themselves.
Count the drops from the equipment list and add one spare per group, then confirm that each drop is where the machine's cable actually reaches rather than where the drawing puts it. A drop that is one meter short is a trailing cable across an aisle, which is both a trip hazard and an interference source. Cabling should be at least Category 6 for new work, terminated and tested rather than merely connected, and labeled at both ends with the same identifier the IP plan uses — the labeling is what makes a fault findable in ten minutes instead of an afternoon.
The test that gets skipped is the one that matters for machine-to-MES communication: a short sustained transfer between the machine's own port and the server, measuring round-trip latency and packet loss under production-like traffic. A line that answers a ping is not proven; a line that completes ten thousand small transactions without loss is. Segmenting machine traffic onto its own VLAN keeps office downloads from competing with program transfers, and putting the switch on the same UPS as the line prevents the situation where machines hold position through a power dip while the network does not.
ESD readiness is measured, not declared. The floor or mat surface resistance is checked with a resistance meter against the range the site's ESD control plan specifies, wrist straps are tested with a tester that reports both a low and a high failure, and the earth bond of every workstation is verified to the same earth reference the machines use. Humidity belongs in the same check: a dry hall in winter can push surface resistance out of range even when the flooring is correct, which is why the ESD plan usually states a humidity band alongside the resistance band.
The audit question that catches most sites is not the reading but the record. A customer auditor will ask when the floor was last tested, with what instrument, by whom, and what the result was; a site that cannot answer has an ESD program on paper only. The same applies to garments, packaging and ionizers, each of which has its own test method and interval. For a line with fewer automated interfaces the exposure is smaller but the requirement does not disappear, which is one reason a semi-automatic line with fewer utility interfaces to sign off still needs the same ESD discipline, just with fewer points to test.
Traceability starts earlier than most projects expect. The barcode or laser mark on the board, the reader at the line entry, the station identifiers, the program names and the operator logins all have to exist in the system before the first board is marked, or the first day's production is unrecorded and the first-article evidence is incomplete. The practical test is to run one board through the whole chain — mark, read, associate with a program, record an inspection result, retrieve it by serial — before the line is considered ready. If that loop cannot be closed with a single board, it will not close with a thousand.
Two conditions make the loop harder and are worth checking in advance: whether the customer's system or the site's system is the master, and whether the data has to be retained for a period that outlives the hardware. Both are policy questions with engineering consequences, and both are easier to answer before the network is wired than after. The output of this check is a one-page data map — board identifier, read points, stored fields, retention period, owner — which is short enough to be read and specific enough to be audited.
Item | What to confirm | How to verify | Interval |
|---|---|---|---|
Network drops | One per machine plus spares, at the machine's cable reach | Physical test and label match to the IP plan | Before installation |
Data path | Sustained transfer with no packet loss | Round-trip test under production-like traffic | Before first board |
Floor resistance | Within the range in the site ESD plan | Resistance meter at marked points | Before installation, then periodic |
Wrist strap and bond | Pass on both low and high limits | Tester at each workstation, logged | Daily or per shift |
Humidity | Inside the band stated in the ESD plan | Logged sensor in the hall | Continuous |
Traceability loop | One board traced end to end by serial | Manual walk-through with a single board | Before first production run |
The last block of the checklist is the one that decides whether the installation ends with a running line or with a machine that is installed, powered and idle. People and safety are treated as soft items in most project plans, which is exactly backwards: they are the two items that can stop work immediately and without appeal. A missing lockout procedure stops an electrician. A missing safety induction stops a contractor at the gate. A missing permit stops a hot-work task for a day. None of these can be resolved by the supplier, and all of them can be resolved before the delivery date if somebody puts them on a list with a name next to them. They are also the block that a PCBA factory readiness checklist before installation tends to leave until last, which is why so many projects are technically complete and still unable to run.
The question is not whether the team has been trained, but who is authorized to do what, on which date. A commissioning engineer can jog an axis and run a program; an operator can load boards and start a recipe; a maintenance technician can open a panel and replace a part. Those three authorizations come from different places, and a site that has not written them down ends up with an operator pressing start on a machine that nobody has formally released. The authorization matrix is one page: role, permitted actions, prerequisite training, signatory, date.
The training itself has a sequencing problem that is easy to get wrong. Training delivered before the machines are in place is theory; training delivered during commissioning competes with the commissioning itself for the same hours; training delivered after handover is the most common and the weakest option, because the first production week is the worst possible classroom. The practical compromise is to have the future line lead present through installation and commissioning, absorbing the machine while it is being built, and to run the formal operator training immediately after the first successful first article, when the line can be run slowly without a delivery deadline attached.
Some safety conditions are prerequisites for energisation rather than follow-up actions, and they should be listed as such. Lockout and tagout procedures and locks for the people who will work on the line, a documented emergency stop layout with the positions verified, guarding in place on every moving axis, fire extinguishers of the correct class within reach of the equipment, and a safety induction record for every person who will be on site. Where the local authority requires a permit for hot work, pressure systems or electrical connection, the permit should exist before the work, not alongside it.
Two of these deserve specific attention because they are frequently assumed. The first is the emergency stop test: pressing the button and watching the machine stop is not a test unless the test also confirms that the energy is removed from the drives rather than only from the control circuit. The second is the fire risk inside a reflow oven, where residue accumulation is the usual fuel and a clean oven is a safety measure rather than a maintenance preference. Both items take minutes to verify and both appear in the acceptance documents that a customer will eventually ask to see.
A readiness review that runs for three hours produces a longer list of open items, not a decision. The format that works is short and unforgiving: every item is pre-scored by its owner before the meeting, the meeting only discusses items whose state changed or whose score is disputed, and the output is a decision rather than a discussion. Forty-five minutes is enough for twenty-five items if the evidence pack is circulated in advance, and the discipline of circulating it in advance is what forces the evidence to exist at all.
The decision itself has four outcomes rather than two: proceed as planned, proceed with named conditions, proceed on a reduced scope, or hold. Proceeding on a reduced scope is the option projects forget, and it is often the best one — installing the line but postponing the second oven, or commissioning the line without the traceability interface, keeps the critical path moving while the missing item is resolved. Whichever outcome is chosen, it should be recorded against the same list that will be used at the next review, so that progress is visible as a change in states rather than as a change in mood. The register that runs beside this list, and the scoring logic it uses, is described in the risk register that turns each open item into a dated exposure.
The first production shift is a test of the plan, not of the equipment, and it should be written down before installation starts. A workable shape is a short run at reduced speed with a single product, a full inspection of every board rather than a sample, and a stop condition that is agreed in advance — for example, stop if the first five boards show a defect that the process cannot explain. That structure produces usable data quickly and prevents the common failure of running a full shift to find out that the printer was set up wrongly at eight in the morning.
The plan should also name who is on call, who can authorize a change to a process parameter, and what happens to boards produced during the first shift. Boards made before the process is stable are evidence, not product, and treating them as saleable is how a customer receives units built on an unvalidated process. Marking them as first-article material, keeping them with the acceptance records, and disposing of them by a written rule costs nothing and removes an argument that is otherwise had months later, when nobody remembers which boards came from which shift.
Send the layout and the utility readings before the truck is booked
If a delivery date is already fixed, the most useful thing to send us is the layout with the machine positions marked, the utility readings you have taken at those positions, and the route the crates will travel from the gate to the hall. We will compare them against the conditions we see fail most often at this stage — crate pressure on a finished floor, a corner that is narrower than the swept path, air pressure measured at the header instead of at the tool, exhaust without make-up air, and a network drop that is one meter short — and tell you which items are still open. It is a desk review, not a site survey, and it usually takes a few days. What we ask for is deliberately small: one drawing, one equipment list, one set of readings, and a note of anything already ordered, so the review happens while the floor is still empty rather than after the crates are standing in the aisle.
Six to eight weeks before the delivery date is a practical target for the first full pass, with a second pass two weeks before. The first pass is early enough to fix lead-time items such as a transformer upgrade, a floor repair or a duct installation; the second pass is close enough to confirm that conditional items actually closed. A single review on the week of delivery is not a review, it is a notification.
Access, rather than equipment. In practice the recurring causes are a crate that cannot pass the narrowest opening, a floor that cannot take the crate's point load, exhaust with no make-up air behind it, and air pressure measured at the header rather than at the machine. All four are visible with a tape measure, a scale drawing and one walk-through, which is why they are worth checking before the shipping date is confirmed.
Usually not. Nameplates often describe the largest single load, while the utility sheet describes the connected load including heaters and auxiliary equipment. Sum the utility sheets, apply a realistic diversity factor, then measure voltage and phase balance under load and again at motor start. The measurement is what shows whether the supply is stiff enough, and it takes less time than the paperwork that follows a startup trip.
The figure belongs to the specific model and should come from that model's utility sheet, not from a general rule. What matters in practice is measuring at the machine flange with the duct cross-section recorded, and confirming that make-up air at least matches the extracted volume. A reading taken at the roof fan includes every leak in the duct and will overstate what the machine actually receives.
Physically yes, and that is exactly why it happens. The problem is that installation traffic damages fresh flooring, and resistance readings taken before the floor has settled are not the readings production will see. The better sequence is to finish and test the floor first, protect it during installation, and re-test at the marked points before the first production run, with the results logged against the site's ESD control plan.
Three to five representative boards, the stencil for that product, the paste in date, the feeder and nozzle set for the bill of materials, the program version, and the acceptance criteria written down before the run. The boards should be the real product or a faithful equivalent. The output of the run — print volume, placement offset, reflow profile, inspection results — becomes the baseline that later production is measured against.
The customer's plant or process engineering lead, because the supplier can only confirm what is inside the supplier's own scope. Where several contractors share the site, one interface owner should co-sign so that shared items have a single name against them. The signature should be against a scored list with dates and evidence, not against a summary statement, so that the next review can show what changed.