Joggle, flare and fish plate: how a busduct joint is made

Almost everything published about busway describes the finished system: ratings, tap-off units, IP class, installation sequence. Very little describes how the joint is actually formed, which is odd, because the joint is the part a fabricator has to make and the part that decides whether the run passes a temperature-rise test. Three forming operations turn plain conductor stock into busduct ends: the joggle, the flare and the fish plate. Here is what each one does and what it demands of the machine that makes it.

The stack has one pitch in the run and a different pitch at the joint

Sandwich busway packs its conductors face to face with a thin insulating film between them. That is the whole point of the construction: minimum spacing means minimum loop area, which means low reactance and a low voltage drop even at poor power factor. In a four-pole run with 6 mm bars and 0.5 mm insulation, conductor centres sit about 6.5 mm apart.

At the joint that spacing cannot hold. Each conductor pair has to be clamped between two conductive plates, and each phase's plate pair has to be separated from the next phase by an insulator thick enough to satisfy the clearance and creepage the design claims. The conductor pitch inside a joint block is typically three to four times the run pitch. The conductors therefore have to be pushed apart as they approach the joint end, each by a different amount, and each has to arrive at the exact height the joint block expects.

That displacement is the joggle. It is a pair of coupled bends in the same direction that translate the bar from its run plane to its joint plane while leaving the two faces parallel. Bar A moves 9 mm up, bar B moves 3 mm up, bar C moves 3 mm down, bar D moves 9 mm down. The figures depend entirely on the joint block, but the pattern is symmetric about the stack centreline and the end faces all finish on the datum the block defines. Get one bar's offset wrong by half a millimetre and its fish plates go in skewed, the Belleville stack loads unevenly, and that phase runs hotter than the other three for the life of the run.

Joggle forming force

The joggle looks small on a drawing and is not small on a press. Two things drive the force. The first is section: force in bending scales with width and with the square of thickness. The second gets underestimated more often. The two bends of a joggle sit close together, so the effective die opening is short, and a short span needs a high force for a given deflection in the same way a short beam needs more load than a long one.

An order-of-magnitude check makes the point. Take 300 mm × 6 mm Cu-ETP in the half-hard temper, R240/H065 to EN 13601, tensile strength 240–300 N/mm². The full-section tensile capacity of that bar is around 480 kN. Local plastic bending needs a fraction of that, but not a small fraction once the die opening is short and the material is work-hardened by the first bend before the second one starts. A busduct forming machine rated at 300 kN for the joggling station, like the EMAC-BDM busduct line, is sized for exactly this: repeated offsets in wide, hard copper without the ram deflecting enough to change the offset from one bar to the next.

Springback matters here as much as it does in ordinary busbar bending, and it is harder to compensate. A single bend has one angle to correct. A joggle has two bends whose springback partially cancels, and the residual is expressed as an offset error rather than an angle error, so the operator sees a bar that fits badly rather than a bar that is visibly at the wrong angle. This is why joggle offsets are set as a programmed dimension against a stop rather than trimmed by eye.

The flare is a lead-in, not a second joggle

The joggle sets where the bar sits. The flare is a shallow angled bend at the very tip of the conductor, opening away from the joint centreline, and its job is to let the bar find its slot. Site assembly of a busway joint means sliding one section's bare conductor ends into the block already fitted to the neighbouring section, overhead, on a lift, often at an awkward angle. Without a lead-in taper the bar catches on the edge of the fish plate or on the insulator and the fitter forces it, which scores the plating and drives copper swarf into the joint.

Terminology varies between builders. Some call the whole end-forming operation flaring, some reserve flaring for the tip taper and joggling for the offset, and some machine catalogues use both words for the same station. What matters on the shop floor is that the two features have different tolerances. The offset is a fitting dimension and needs to hold to a few tenths of a millimetre. The flare angle is a handling feature and a degree either way changes nothing, which is why machines that form it usually do so with a programmable angle rather than a dedicated tool. On the EMAC-BDM the flaring stroke is 30 mm with the angle set in the program, so a change of joint block does not mean a tooling change.

One consequence for production planning: because the conductor is cut to length and formed in the same clamping, the length datum and the offset datum are the same. If the cut and the form happen on separate machines with separate fixtures, every transfer adds a stack-up error to a dimension that has no slack in it.

Fish plates and the punch, stretch, cut sequence

The fish plate (splice plate, connector plate or joint bar, depending on whose drawing you are reading) is the conductive strip that bridges the gap between two abutting conductors and carries the current across the joint. In a butted joint it carries the full phase current for the length of the overlap, which makes its section, its plating and its flatness electrical parameters rather than mechanical ones.

Material is normally the same alloy as the conductor, Cu-ETP for copper systems, in a thinner gauge: 3 to 4.5 mm is the usual band, against 5 to 10 mm for the conductor itself. Aluminium systems use aluminium plates for the same reason. A copper plate on an aluminium bar puts a dissimilar-metal couple in the hottest, least accessible part of the run.

The forming sequence on a dedicated fish-plate machine is punch, stretch, cut. The middle operation carries more design intent than its name suggests.

Punch puts the bolt clearance hole through the strip. Ordinary punching work, except that the hole position is the joint's only self-locating feature, so its tolerance is the tolerance of the whole assembly.

Stretch, sometimes called bulging or embossing, draws a raised boss out of the plate around or beside the hole. That boss is a spacer. When two plates face each other with their bosses touching, the boss height sets the gap between them, and that gap is the slot the conductor enters. It also acts as a hard stop against over-insertion, so a fitter cannot push a bar so far in that it crushes the phase insulator behind the plate. The technique is old and well documented: US patent 5760339 describes spacing tabs formed from the material displaced when the hole is made, maintaining plate separation and limiting insertion depth in exactly this way. Forming the spacer out of the plate itself removes a separate part from the bill of materials and removes a separate stack-up from the tolerance chain.

Cut shears the finished plate off the strip. Because the strip is processed continuously and sheared last, there is no offcut between plates.

Why hole pitch accuracy compounds

A dedicated fish-plate machine is specified on hole pitch accuracy rather than on force, and the reason is cumulative error.

Take the within-joint case first. Every plate in a joint block shares one bolt. If one plate's hole sits 0.3 mm off position, that plate cannot centre on the bolt, so the Belleville washers load it eccentrically. Contact resistance at a bolted joint is set by the true contact area at the asperity tips, which is a small fraction of the apparent overlap area, and that true area is set by how pressure is distributed. Skew the pressure and you lose contact on one side of the plate faster than you gain it on the other. The joint still assembles, still passes a continuity check, and runs a few kelvin hotter than its neighbours. Over a design life measured in decades of load cycling, that is how joints degrade.

Now the along-run case. Busway is set out against building geometry: the run starts at a switchboard and finishes at a fixed point, and the sum of every section length plus every joint gap has to land on that point. A strip processed at ±0.1 mm/m accumulates at most 0.6 mm end to end over a 6 m length. A machine holding ±0.5 mm/m accumulates 3 mm over the same strip. Across a 60 m run built from 3 m sections, that is roughly twenty joints. If the error is systematic rather than random, the run drifts by twenty times the per-section error in the same direction, and machine error usually is systematic, because it comes from a lead screw or an encoder rather than from noise. Six millimetres of drift is absorbed by joint clearance. Thirty is not, and the last section does not fit.

So the reason to hold better than ±0.1 mm/m is arithmetic: it keeps the accumulated error inside the joint clearance the block was designed with.

Assembly: the single-bolt joint

Modern busway joints are single-bolt designs. One high-tensile bolt passes through the whole stack, insulator to plate to conductor to plate to insulator, phase after phase, and one tightening operation compresses every contact interface in the joint. Published figures from large sandwich-busway ranges put the clamping force of that one bolt above 4,000 lbf, roughly 18 kN, spread across the contact faces by large-diameter cupped conical spring washers. Above about 2,000 A the same ranges move to two or three bolts, one per conductor group, because a single bolt can no longer distribute pressure evenly across the wider conductor.

The bolt is usually a double-headed shear nut. The outer head is machined to snap off at a defined torque, leaving the inner head behind, so a walk-down inspection can confirm every joint in a run was tightened correctly without a torque wrench and without a calibration record for that wrench. Some designs add a coloured indicator disc that falls away when the head shears.

None of that helps if the parts feeding the block are out of tolerance. The joint block assumes a conductor at a known height with a known plating finish, a fish plate with a hole in a known position and a boss of known height, and faces flat enough that the pressure from one bolt reaches all of them.

Where the standards sit

Two documents frame busduct in most tenders, and they do not overlap the way people assume.

UL 857 is the North American busway product standard. Its scope runs to 600 V and 6000 A, covering service-entrance, feeder and branch-circuit busway. Conformity is by NRTL listing with follow-up factory inspection, which means the construction that was tested is the construction you are obliged to keep building. NEMA BU 1 sits alongside it as the product standard and, with BU 1.1, covers handling and installation for busway of 100 A and above.

IEC 61439-6

is the busbar trunking part of the IEC 61439 series. Check its scope before quoting it. It applies to busbar trunking systems up to 1000 V AC or 1500 V DC and, unlike UL 857, it states no current ceiling. It also cannot be used alone: it is written as a delta against IEC 61439-1 and has to be read with it, the same way IEC 61439-2 is. It replaced IEC 60439-2 and brought busway onto the verification structure the rest of the series uses: test, calculation, or comparison with a verified reference design. For a fabricator, the practical consequence of that third route is that your process has to be repeatable enough for the units you ship to be comparable to the one that was verified. The obligation lands on manufacturing, not on the document controller.

Neither standard tells you how to form a joggle. Both of them make the consequences of forming it badly your problem.

What the data-centre load is doing to the joint

Busway joints used to be a low-current afterthought in a distribution design. That has changed. As racks move from a 5–15 kW baseline to AI-optimised cabinets drawing 60, 80 and 100+ kW, tap-off and joint ratings of 400 A continuous and above have become ordinary rather than exceptional. A joint that spends its life at 400 A instead of 100 A dissipates sixteen times the power for the same contact resistance, in the same enclosed volume, with the same insulator sitting next to it.

That changes what a fabricator has to control. Plating on the contact faces stops being a line item that gets value-engineered out. Flatness of the fish plate stops being cosmetic. And the temperature-rise margin that used to absorb a sloppy joint is no longer there. If you want to see how quickly conductor temperature moves with current for a given section, the busbar ampacity calculator makes the I²R relationship concrete.