Fabricating busbars for EV charging equipment

A switchgear shop that takes on EV charging work usually finds the parts easy and the process hard. The bars are small, the sections are thin, the tonnage requirement is trivial. What catches people out is that the tolerance band is tighter than switchgear work, the batch sizes are smaller, the drawings change between production runs, and a bad edge comes back as an insulation failure eighteen months into service rather than as a cosmetic complaint at goods-in.

What the parts look like

EV charging equipment splits into two quite different busbar populations.

The AC and distribution side of a charger cabinet looks familiar: rectangular copper, 6 to 10 mm thick, punched and bent, bolted joints, much like a small distribution board. Nothing about it is unusual.

The DC side is where the work changes. Charging architectures have converged on operating ranges of roughly 200 to 920 V DC for CCS installations, with the specification allowing as wide as 50 to 1000 V DC. The Megawatt Charging System now being deployed for heavy-duty vehicles is rated at 1,250 V and 3,000 A DC. Those numbers mean short, wide, thin conductors carrying very high current between power modules, and they mean the conductors are packed close enough together that air alone is not doing the insulating.

Typical DC-side and module-interconnect work runs 1.5 to 6 mm thick, often under 300 mm long, in quantities of tens to low hundreds per configuration. The parts are frequently laminated or film-insulated rather than bare. And the drawing revision changes, because the power electronics inside a charger cabinet are on a two-to-three-year product cycle, not a twenty-year one.

Laminated and insulated interfaces

Three insulation approaches dominate, and each puts a different demand on the metal.

Film lamination bonds polyester or polyimide to the conductor, sometimes with two or more conductors built into one stack to cut loop inductance. The film is thin, often well under a millimetre, and it has to conform to the conductor's edge without thinning or bridging.

Powder coating applies epoxy powder electrostatically and cures it in an oven. It is cheap and it survives handling. It is also unforgiving of sharp geometry, because during cure the coating flows and surface tension pulls it away from convex edges, leaving the film at its thinnest exactly where the field is highest.

Overmoulding and encapsulation put the conductor in a tool and mould resin around it, or pot the whole assembly. That gives the best mechanical result and the least tolerance for entrapped air, because a void formed during moulding is sealed inside and cannot be inspected without a dielectric test that may pass anyway.

Published guidance on laminated busbar edge conditioning identifies four approaches: open edge, pinch sealed, FR4 and epoxy edge-filled. Pinch sealing requires an insulation overlap of at least twice the conductor thickness beyond the copper edge, and that overlap only works if the copper edge is where the drawing says it is. It is a fabrication constraint dressed as a design choice.

The edge is where the insulation fails

Chamfer quality dominates this sector for a physical reason that most specifications state as a bare dimension without saying where it comes from.

Start with the geometry. A sheared or punched edge is not square. It has a rollover on the die-entry side, a burnished band, a fracture zone and a burr on the exit side. The "corner" is effectively a radius of a few tens of micrometres, and in places it is a burr tip with a radius smaller than that.

Now put insulation on it. Whatever the process, the material has to make a right-angle turn around that corner. Film bridges rather than conforms. Powder pulls back. Moulding resin has to fill a re-entrant corner against its own surface tension and against whatever air is already sitting there. In each case the likely result is a small pocket of air trapped between the conductor and the insulation, right at the corner.

That pocket is the problem, for two reasons that compound.

First, the field is highest there. Electric field concentrates at a convex conductor feature, and for small radii the local enhancement scales roughly with the inverse of the radius. A 20 µm burr tip in a field that is comfortable across a flat face is not comfortable at the tip.

Second, air is the weakest material in the assembly and it is carrying more than its share of the field. In a void surrounded by a solid dielectric, the field in the void is higher than the field in the solid by roughly the ratio of their relative permittivities. Epoxy and polyester sit around 3 to 4; air is 1. So the void sees three to four times the field of the surrounding resin, while having roughly a tenth of its dielectric strength.

The void ionises, and once it ionises it does not stop. Partial discharge erodes the insulation from the inside, the eroded cavity grows, the discharge activity grows with it, and the part eventually fails at a voltage it withstood on the day it was tested. The research literature on laminated busbars is consistent on this: PD inception in embedded defects occurs at a few kilovolts, which is inside the operating range of an 800 V or 1000 V DC architecture once transient overvoltages are counted.

A radiused edge reduces the enhancement factor, and more usefully it gives the insulating material a geometry it can wet and conform to. The field never becomes uniform; it becomes survivable. That is the whole argument for chamfering in this application, and it rests on physics rather than on a clause.

What the standards actually say about it

IEC 60664-1 is the insulation coordination standard, and it is the right reference for the clearance and creepage numbers in a charger design. Clearance comes from rated impulse withstand voltage combined with pollution degree and altitude; creepage comes from working voltage combined with pollution degree and the comparative tracking index of the insulating material. Those are the parameters an EV charging designer works to.

What IEC 60664-1 does not do is tell you to chamfer. Neither does IEC 61439. That needs saying, because a good deal of vendor material claims IEC 61439 mandates a busbar edge radius, and no such clause exists. Edge radius is an engineering means of meeting dielectric, clearance and creepage provisions, and of making an insulation process producible. If a customer's specification calls for a particular radius, that number came out of their insulation process, so ask which process. The answer changes what "good" means: a radius sized for pinch sealing is not the radius sized for powder coating.

What a good edge looks like on the shop floor

Two separate operations, often confused.

Deburring removes the burr and the oxide. It is a surface condition operation across the whole face, and it is what you do before plating or before any adhesive-backed film goes on. A two-stage deburring machine handling 0.5 to 50 mm thickness at 5 to 30 m/min gets the whole batch to a uniform condition, which matters more for coating adhesion than the average finish does.

Chamfering is a controlled removal of the corner to a defined profile. On thin EV parts this is where the difficulty sits, because a 2 mm bar has very little corner to remove and a heavy-handed chamfer eats into a section that was sized for current. A CNC chamfering machine working 3 to 15 mm thick stock at ±0.05 mm positioning and repeatability, with a BT40 spindle to 6,000 rpm, produces a profile that is the same on part one and part four hundred. A bench grinder does not, and on a laminated part where the film overlap is specified against the copper edge, a chamfer that wanders is a chamfer that moves the edge.

Sequence matters as much as the operations do. Chamfer before deburr, so the deburring pass cleans the secondary burr the chamfering cutter raises; deburr before any plating or coating, because oxide under a coating is a bond failure waiting for a thermal cycle; and do not bend after coating unless the coating specification says you may, because the outer fibre of a bend stretches and a coating that was continuous at 3 mm radius may not be at 2 mm. On laminated assemblies the forming has to be finished before lamination, full stop.

Two practical checks that catch most problems before they reach the insulation line. Look at the corner under magnification rather than running a thumb along it, because a thumb finds a burr and misses a re-entrant fold. And check the chamfer on the last part of a batch, not the first, because tool wear on copper is slow and cumulative and shows up as a gradually shrinking radius.

Changeover is the production constraint

Fifty parts of one configuration, then a revision, then eighty of the next. That is the rhythm of EV charging work, and it means the machine spends more of its life being set up than being run. Three things decide how much of the shift you lose to it.

Turret capacity comes first. A high-mix part family needs a lot of hole sizes, and every tool that is not in the turret is a manual tool change mid-job. A processing centre with 24 tool stations, extendable to 36, will run a mixed job list without an operator touching the tooling. The same job on a machine with eight stations becomes a sequencing exercise where you group parts by tool instead of by delivery date, and the scheduling cost of that stays invisible until someone measures it.

Then program-to-part time, which is the single largest hidden cost in low-volume work. If a revision takes an hour of manual programming, a product family on a two-year cycle burns weeks of engineering time that was never quoted. Software that imports the geometry and produces the program without re-keying dimensions removes both the hour and the transcription errors that come with re-keying.

Nesting is the third. Small parts from thin stock generate a lot of offcut, and whole-project nesting with an oddment library lets the next job consume the remnants of the last one. On a shop running fifty part numbers a month that shows up in the material yield figure, not just in a tidier rack.

For the lighter end of the work, a compact all-in-one machine at 300 kN covering 160 × 15 mm is normally more than enough tonnage. The tonnage was never the constraint. Setup time was.

If you are checking whether a thin section will take a bend without cracking at the radius your drawing specifies, the bending force calculator will give you the force and let you sanity check the die opening before you cut metal.