When a buyer moves from quoting a semiconductor process equipment supplier to signing a purchase order, most of the negotiation is already over. What remains is the part that decides whether the machine earns its cost back: how the two sides agree, in writing, on what "arrives working" means. A lamination line that bonds samples in the supplier's workshop can still fall short on a production floor, and the difference is almost never visible in a brochure. The checks below are the ones that hold up during commissioning, because each one produces evidence rather than an opinion.
Before the order: agree on a machine-readout specification
The single most useful document in an equipment purchase is not the datasheet but the acceptance sheet that defines measurable pass criteria. A tolerance that a supplier prints on a listing becomes a contract term only when both sides state how it is measured and on what material. For bonding and laminating work, the practical items are:
- Alignment tolerance, and the reference marks used to read it — CCD vision alignment on a UVW platform, for example, is only meaningful if the mark pattern and measurement point are fixed in the test protocol.
- Cycle time, stated as a range per size and per material combination, with the substrate size that the figure refers to.
- Defect criteria for the process itself, such as acceptable bubble size and count after bonding, and the debubbling step that follows it.
- Utility and consumable requirements: compressed air, vacuum level, power supply, chamber atmosphere and nitrogen use where the process calls for it.
- Spare-parts and service terms, including lead times for wear items that a line cannot run without.
Write these criteria before the purchase order, not during commissioning. A buyer who asks for an acceptance sheet at the FAT stage usually gets one; a buyer who asks at installation has already given away the negotiating position.
FAT: run the process, not the machine
A factory acceptance test should be run with the buyer's own material, or the closest available substitute, and with the recipe the line will actually use. Watching a machine cycle empty tells you about mechanisms; running your film, glass or substrate through it tells you about yield. Three things are worth insisting on during FAT:
First, test the difficult size at both ends of the range. A medium-sized automatic laminating machine specified for 10–32 inch products can behave differently at each extreme, and the largest size is where roller bonding pressure distribution usually shows its limits. Second, run the machine long enough to expose heat drift and consumable behaviour, which a short demonstration never surfaces. Third, capture the readouts: machine logs, alignment records, defect counts. Those records become the baseline you compare against when the same machine runs at your site, and without them a commissioning dispute turns into two people arguing from memory.
Site acceptance: repeat the test under your conditions
Site acceptance is where environment stops being a variable you described and becomes a variable you measure. Power quality, floor space, operator access, ambient temperature and dust all change results, and none of them appear on a datasheet. The pattern that works is simple: rerun the FAT protocol unchanged, then compare line by line against the workshop baseline. Where a result differs, ask for the cause before accepting a workaround, because the two most common causes — utility mismatch and fixture or handling differences — are cheap to fix early and expensive to fix a year later.
Post-process steps deserve the same treatment as the bonding tool itself. A high-pressure debubbling machine, for instance, is sized by chamber volume and vessel construction; if it is treated as an accessory rather than as part of the acceptance scope, defects judged good at bonding can still fail downstream. Same for cleaning, curing and handling between stations, which is where many lines quietly lose yield.
Customisation and integration: the questions that predict pain
Most semiconductor and display process equipment is bought as a customised cell rather than a catalogue item, and this is where commissioning problems are usually created — months earlier, during specification. Four questions separate a well-scoped build from a difficult one:
- Which change is in the machine and which is in the recipe? A supplier that says both are easy is worth pinning down with a written change procedure and a training session.
- How does this cell connect to upstream and downstream stations mechanically and in data terms? Handoff and signalling are where integration schedules slip.
- Who owns the software licence, the recipe storage, and the parameter backup after handover?
- What happens the first time your process changes? Equipment specified for today's product stack should be checked against the product you expect next year.
Documentation and people: the check most buyers skip
Commissioning is not finished when the line passes its test; it is finished when your own engineers can run it, adjust it and diagnose it without waiting on someone else. Ask for the parameter list with its normal operating ranges, the alarm and fault list with responses, the wear-parts schedule, and the wiring and pneumatic drawings. Then have your operators run the acceptance shift themselves while the supplier's engineer observes. A machine that only the vendor's engineer can commission is a machine that stops whenever that engineer is unavailable.
Choosing a supplier who can pass these checks
The value of a supplier is not measured by the number of models in a catalogue but by whether they can produce the evidence above without being pushed for it. Shenzhen PowerDe Automation Precision Equipment, established in Shenzhen in 2007, builds vacuum lamination, bonding, debubbling and related process equipment, along with customised automation cells and engineering services that cover development, hardware testing and manufacturing automation. That scope matters for commissioning in a concrete way: when the same supplier handles the process tool, the fixture and the integration, responsibility for a failed acceptance test cannot be split between three vendors.
For a buyer comparing offers, the practical test is to ask each supplier to fill in the same acceptance sheet — tolerance and how it is measured, cycle time per size, defect criteria, utilities, spares, service response. Suppliers who already work this way answer in a day. The gap between those answers is usually a better indicator of a smooth commissioning than any headline specification in the quotation.
A short checklist to carry into the meeting
- Acceptance criteria written and signed before the purchase order.
- FAT run on your material, at the extremes of the size range, with logs captured.
- Site acceptance repeats the same protocol and compares results item by item.
- Post-process steps — debubbling, cleaning, curing — inside the acceptance scope.
- Change procedure, recipe ownership, parameter backup and licence terms settled on paper.
- Fault list, wear-parts schedule and drawings delivered, and your team runs an acceptance shift.
