Write requirements as statements a model can contradict
Most requirement lists for an alignment and packaging cell are phrased so that nothing could disprove them. A line such as "alignment must be precise" has no outcome that would make it fail. Rewrite every line so that it names an observation and the person who makes it: where the datum marks are read, which direction the offset is measured in, what counts as a bubble worth recording, which side of the stack may stretch during the bond, what the station does when a plate arrives late from the station before it. A model can answer all of those. It cannot answer an adjective.
Sort the lines into three groups before the model is built. Some are yours alone to decide, such as the part families you will actually feed and the batch you intend to run. Some belong to the supplier's design, such as how a plate is held and which sensing closes the loop. Some are shared, and the shared lines are the ones the model exists to settle. Writing down the shared lines is the whole of the requirement plan; the rest of the model stage is paperwork.
Use the published description as a test list, not as a claim
The published wording for this class of equipment names selectable recipes used to reduce alignment bubbles. Treat that as two questions you must answer on your own parts, not as a sentence you may copy into a specification. The first question is what a recipe is allowed to contain inside your plant: which settings move together when a recipe changes, who may change one, and what is recorded afterwards. The second is what a bubble means on your stack: its size, its position, whether one bubble at an edge behaves differently from several across the field, and whether the definition changes between the film families you intend to run.
A model run that never changes a recipe has tested nothing about the recipe capability. Ask for the run in which the same pair of parts is processed under more than one setting and the results are placed side by side. That is the demonstration worth attending, and it is the one a delivered machine can no longer give you cheaply, because by then the settings are locked by production.
Test how the stack is handled before testing the alignment
A glass plate carrying through-glass vias and a polymer film are handled as one object, and they do not behave as one object. Films creep, curl, and keep a memory of how they were stored. Glass does not forgive an edge chip, and a chip introduced while handling is a handling finding, not a bond finding. Record which of the two materials caused each stop while the model is still in front of everyone.
Make the model run your parts, in the orientation and on the carrier you plan to use, with the same hand-offs you plan for the real cell. Bringing a prepared sample from a laboratory is the usual shortcut, and it hides precisely the requirement lines that matter most: how a plate is presented, how it is located before the datum marks are read, how the film is lifted from its liner, and where the film waits while the plate is being located.
Ask for the failure behaviour, not only the good cycle
Every demonstration ends with a good part. The requirement lines that decide whether you can live with the machine concern the bad ones. Make the model answer these: what happens when a plate is missing a via, when the film is laid down at the wrong angle, when the vision system cannot find a datum mark, when the upstream station sends a part too early, when a cycle is interrupted halfway and a half-bonded stack is left inside. Write down the behaviour you saw, in the same words each time, and compare that list afterwards with the fault handling the design promises for the delivered machine.
A half-finished stack with nowhere to go says something about the cell, not about the machine. If the model has no route for scrap, rework, or quarantine, put that gap into your requirement list before the quotation, because adding a reject path later is a mechanical change and it arrives billed as a change order.
Decide who observes, and with what instrument, before the run
Alignment findings are only as strong as the measurement that produced them. Agree before the run which instrument is authoritative, who operates it, whether it belongs to you or to the supplier, and what it can resolve on a stack of glass and film rather than on a calibration artefact. Two parties arguing afterwards about which measurement counts have wasted the model stage; a written agreement about the instrument is what turns a number into evidence.
Agree the record in the same way. For each run note the parts used, the setting in force, the observed result, the departures from the planned run, and the signature of the person who observed it rather than the person who set the machine up. A model stage that produced a good feeling and no record cannot be used in the acceptance clauses of an order, and acceptance clauses written from memory are the clauses that fail at handover.
Treat every unanswered question as a cost, and settle it on the model
The published listing for this class of equipment carries no parameter rows, so questions about footprint, services, room conditions and service access stay open until somebody answers them in writing. That is not a flaw to complain about; it is the normal condition at the model stage, and it is why the model stage is worth paying for. Put each open question to the supplier as a sentence they must complete: what the machine consumes at the point of use, what the room must hold, how consumables arrive, where used material goes, and who can reach the inside of the machine without dismantling a neighbour's work.
Keep those answers on one sheet, dated, with the name of the person who gave each one. Bidders who receive the same sheet quote the same scope, and that is the first moment your comparison stops being a guess about what each supplier assumed.
Let the model stage end in a document, not in an opinion
A finished model stage produces a small and boring set of artefacts: the requirement lines that were confirmed, the lines that were rejected and how, the lines that were never reached because time ran out, the handling rules for the stack, the fault behaviours that were observed, the instrument and record agreement, and the open questions that must be answered in writing before any order. If the people standing in front of the model cannot produce that set, they have watched a machine and have not tested a requirement.
The supplier's own order of work follows the same logic. The engineering route recorded on this site runs from the requirement through a proposal, mechanical and control design, a built model, and testing and verification, before any whole line is delivered; the same sequence is presented for automated production line solutions, where a cell has to be planned together with the stations around it rather than as a machine standing alone. A model stage that hands you confirmed requirement lines shortens that sequence. A model stage that hands you enthusiasm lengthens it, and the extra length is billed.
Before you agree to a model stage, ask these questions
- Which requirement lines is this model meant to answer, and how would we notice if one of them failed?
- Are our own plates and films being used, or a supplier sample, and who decided that?
- How many settings will be run on our parts, and will the results be shown side by side?
- Which fault cases will be provoked on purpose, and who writes the behaviour down?
- Which instrument settles an alignment dispute, and whose instrument is it?
- Is there a route for a rejected or half-finished stack, and was it used during the run?
- What document do we receive at the end, and does it carry the names of the observers?
The last question is the one worth insisting on. A model stage that ends in a signed record changes what you send in the next enquiry; a model stage that ends in a photograph changes nothing except the date of the decision. When the equipment class itself is still being identified, this site invites a written brief for engineered equipment work through the enquiry route used for customized steam cleaning equipment, a page that handles process-equipment enquiries alongside cleaning projects, so a sheet of confirmed requirement lines can go there instead of a description of a machine nobody has specified yet.
