Adding a parallel supplier to a laminated busbar programme looks like a purchasing decision and behaves like an engineering change. The drawing stays the same, so the quality file stays the same, so nothing appears to need re-testing. Then the first shipment from the new line arrives with a different film, a slightly different stud torque strategy, and a partial-discharge inception voltage nobody measured. The fix is not to re-run everything; it is to decide, attribute by attribute, which differences are load-bearing.
Adding a parallel supplier to a laminated busbar programme looks like a purchasing decision and behaves like an engineering change. The drawing stays the same, so the quality file stays the same, so nothing appears to need re-testing. Then the first shipment from the new line arrives with a different film, a slightly different stud torque strategy, and a partial-discharge inception voltage nobody measured. The fix is not to re-run everything; it is to decide, attribute by attribute, which differences are load-bearing.
Two plants can build the same drawing and produce different electrical hardware. The conductor geometry is fixed by the drawing; what is not fixed is the film grade, the bond pressure, the stud attachment method, the plating thickness distribution, the edge trim and the cleanliness of the press. Those variables do not appear in a dimension check, and several of them dominate the dielectric behaviour of the finished stack.
This is why second-sourcing decisions tend to fail quietly. A first-article inspection passes, the paperwork matches, and the failure surfaces months later as a field return or, worse, as a partial-discharge reading taken by a customer's own incoming test that your file cannot explain.
The useful question is not "is the new supplier good enough" but "which attributes of this design carry the qualification evidence, and does the new supplier reproduce them". That reframing turns an open-ended audit into a finite list, and a finite list can be argued about and signed off.
If the design itself is still open, start with the laminated busbar design guide; requalification is a separate exercise that assumes a frozen drawing.
This page is the technical half of the question. The commercial framework — how to score candidate suppliers, structure a dual-source programme and audit them — sits in the dual-sourcing qualification framework; what follows assumes the supplier has already been selected and the question is now evidence.
In practice every attribute of a busbar falls into one of three tiers, and the tier decides the response.
Most requalification programmes over-test Tier C and under-test Tier B, because Tier C is easy to specify and Tier B requires somebody to understand the process. The matrix in the next section is an attempt to correct that imbalance rather than add another checklist.
Two practical notes on the tier system. First, tiers are a property of the design, not of the supplier: a rail assembly with friction-welded studs and a tight creepage budget has more Tier A surface than a datacenter stack with generous clearances, so the number of attributes requiring re-test differs project by project. Second, a Tier B change that repeats three times stops being Tier B. If a supplier has swapped film vendor twice in a year, the accumulated variation is a design-level risk and should be treated as one.
Read this as a decision table. For each attribute, the middle column is what you compare, the right column is what you may have to run again.
Twelve rows is short enough to sign. That is the point: the matrix exists so that a decision can be attributed to a person rather than to a mood.
Of all the repeat tests on the list, partial discharge is the one where skipping costs the most and rerunning costs the least. A PD measurement at the working voltage, using the method described in our partial discharge primer, takes minutes per sample and detects exactly what a supplier change tends to alter: voids, bond quality at the film-copper interface, and edge field crowding.
Hi-pot does not substitute. A stack can pass an over-voltage test and still have an inception voltage well below its working level, which means it discharges continuously in service and degrades from there. The two tests measure different quantities at different sensitivities, and a supplier who offers hi-pot as PD evidence is either misunderstood or hoping you are.
The practical protocol: three samples from the first production lot, PDIV and PDEV at the declared working voltage, with the ambient and the measurement circuit described. Compare against the incumbent's numbers, which you should already have. If the incumbent never measured PD, that is the more important finding, and it changes the framing from "qualify the new supplier" to "qualify both".
Temperature rise is usually where buyers expect a full re-test and where a comparison is often sufficient. The reason is that rise is dominated by conductor cross-section, layer arrangement and the thermal environment — all Tier A attributes that the drawing fixes — with bond quality and interface resistance as second-order effects.
So the sequence is: confirm the cross-section and layer geometry match, confirm the terminal attachment method matches, then run one instrumented comparison at rated current on a sample from each source. Two identical stacks in the same rig, same ambient, same thermocouple placement. A difference beyond a few kelvin means an interface problem, most often at the studs or at the bond between layers, and it is worth chasing before the lot is released.
What you should not do is accept a calculation instead of a measurement at this stage, or accept a measurement without the ambient and the mounting condition written down. A rise figure without its test context cannot be compared to anything, including to itself a year later. The measurement conventions we use are described in thermal management for busbars.
The terminal interface is the highest-variance part of a busbar across suppliers, and the least documented. Three things differ even when the drawing is identical: how the stud is attached, what the plating thickness is at the contact face, and whether the supplier controls torque at assembly or only states a value on the drawing.
Requalification actions are modest and concrete. Pull-test or torque-test the stud attachment on samples from each source. Measure plating thickness on a coupon rather than accepting a bath datasheet range. Confirm that a calibrated torque tool is used and that the applied torque is recorded, not merely specified. Then run a thermal cycle and re-check torque retention, because the difference between a good and a mediocre interface often only appears after the first fifty cycles.
Stud attachment deserves a note of its own. A friction-welded stud and a drawn-arc stud can look identical from the outside and behave differently under vibration, which matters for rail and EV programmes. If the incumbent uses one method and the second source uses another, that is a Tier A change even though the drawing's stud callout is unchanged. The relevant standards context sits in the IEC 61439-1 verification guide, where terminal accessibility and torque documentation are treated as verification items rather than workmanship details.
"Equivalent film" is the most abused phrase in a second-source conversation. It usually means "same class of material, similar thickness, available and cheaper". It rarely means the same comparative tracking index, the same bond activation treatment, or the same behaviour under thermal aging.
For requalification purposes, ask three questions in writing. What is the exact grade designation and who manufactures it? What is the measured thickness tolerance on the incoming roll, not the nominal? What evidence exists that the bond strength to copper matches the incumbent's, measured rather than assumed?
If the answers point to a different grade, the design's material group assumption may no longer hold, and the creepage numbers on the drawing were derived from that assumption. That chain — grade to tracking index to material group to creepage — is set out in IEC 60664-1 clearances and creepage. A second source that quietly changes the film grade can invalidate a distance calculation that nobody thought to re-check.
Ask for this with the RFQ and again before first article. Two of the items are usually missing and both matter.
The last line is the one that prevents next year's surprise. Without it, a supplier can change a film supplier in March and you learn about it in an incoming inspection in June, with no contractual basis to reject the lot.
One more item deserves its own line, because it is the one most often refused: the press logs. A supplier who cannot export three lots of press data has no bond process control worth the name, and a customer who accepts that has bought a certificate rather than a process.
A controlled parallel run is the cheapest form of assurance and the easiest to mismanage. Done well it produces a comparison dataset; done badly it produces two suppliers, two sets of habits and no comparison at all.
Four rules make it work. Split by part number rather than shipping both sources for the same part in the same window, so traceability stays clean. Keep the incumbent on the highest-volume or highest-risk part until the comparison data exists. Instrument the first three lots from the new source with the same tests the incumbent ran, not a lighter set. And set an exit criterion in advance — a numeric agreement on PD, torque retention and temperature rise — so the decision is a comparison rather than a negotiation.
Volume splitting deserves care too. If the new source gets a token 5 percent of volume, its process never reaches steady state and the comparison is meaningless. A workable split is large enough to exercise the line across several weeks and small enough that a failure is survivable.
There is a documentation shortcut that pays for itself: build one comparison sheet per part number with a row for every matrix attribute and two columns, incumbent and candidate. Every incoming answer lands in a cell. Six months later, when somebody asks why the second source was approved, the sheet is the answer, and when a candidate changes its film, the same sheet shows exactly which rows need reopening. It also keeps the exercise from drifting into a general audit, which is the usual way these programmes lose momentum and budget.
The business case for a second source is usually resilience, and the accounting often ignores the carrying cost of a parallel qualification.
Three cost lines belong in the model. The direct qualification cost — samples, tests, engineering time, and the PD and thermal work that cannot be skipped. The learning-curve cost, meaning the scrap and rework rate on the new line for the first few lots, which is real and usually underestimated. And the administrative cost of maintaining two documentation sets, two audit cycles and two change-notification streams for the life of the programme.
Against those, the benefit is not only price. It is the ability to keep shipping when one plant stops, and the negotiating position that comes from a genuine alternative rather than a theoretical one. Both are worth paying for; neither is free. Putting the three cost lines beside the two benefits is usually enough to decide whether the second source is a strategic move or an expensive hedge.
For programmes where the volume is low and the design is stable, the honest answer is sometimes that a well-audited single source with a documented change clause beats a thin second source with a weak file.
Four clauses cover the recurring disputes.
If a supplier resists clause two, that is information. Suppliers with disciplined change control usually welcome it, because it protects them from being blamed for a change they never made.
Before the second source ships its first production lot, confirm each line and name the owner.
The matrix is not a bureaucracy; it is the shortest defensible answer to "why is this supplier approved". If you want a second opinion on a stack before it enters a parallel run, the technical review we run during quoting covers the same attributes — start with a quote request, or compare procurement routes in sample to series procurement. Teams also weighing material changes will find the copper versus aluminium comparison and the quality documentation route useful in the same review.
Buying a laminated busbar is not like buying a standard connector. The part is custom, the
Read article
Nadia Rahman is a quality and compliance engineer with 9 years in electrical-component quality systems. She manages SHANYUAN's ISO 9001 and IATF 16949 programmes, incoming and outgoing inspection, and UL/IEC compliance documentation. She holds a B.Sc. in Industrial Engineering from the Universi
Send the current profile, voltage class and terminal types — our engineers reply within one business day.
