What makes a bolted busbar joint reliable

A busbar run is mostly a solved problem. The bar is sized from a current, the temperature rise is verified, and the metal does what metal does. The joints are where the engineering actually is, and they are where switchboards burn. Much of what decides a joint's service life is settled in the fabrication shop before the bar reaches site: how flat the mating faces are, whether the burr came off, how the holes were made, and what the surface looks like where the two bars will touch.

The apparent contact area is a fiction

Two 50 mm bars overlapped by 70 mm give a nominal contact area of 3500 mm². Almost none of it carries current.

Real surfaces are not flat. Pressed together, two bars touch only at asperity peaks, and current crosses only where those peaks have penetrated the surface films into metal-to-metal contact. These are the conducting spots, universally called a-spots after Holm's treatment of them. The Copper Development Association's busbar guidance puts them at the order of 1% of the overlap area.

Measurements are less flattering. Braunovic's work on high-power overlapping bolted joints, published in IEEE Transactions on Components and Packaging Technologies in 2002, calculated the ratio of real conducting area to expected contact area for four- and six-bolt connector-to-busbar joints. With no surface preparation the ratio came out at 0.0003% to 0.0023%. Machined and lubricated, the best case reached about 4%. Even in the best-prepared joint tested, 96% of the contact area was doing nothing.

The load-bearing relationship behind this is straightforward: contact force F relates to apparent area A and hardness H as F = ξHA, and Holm showed hardness relates to yield stress as H ≈ 3σy. Softer material yields at lower pressure and gives more real contact area for the same bolt load, which is why aluminium makes intimate contact easily and why aluminium then has a creep problem.

Constriction resistance

Because current can only cross at the a-spots, current lines in the bulk metal converge into each spot and diverge out of it. That convergence is a resistance in its own right, independent of any surface film, and it is called constriction resistance. For a single circular spot of radius a between two identical metals of resistivity ρ:

Rc = ρ / 2a

Between dissimilar metals it becomes (ρ₁ + ρ₂) / 4a. Total contact resistance adds film resistance Rf = σ/πa², where σ is the resistance per unit area of any oxide or contaminant film the current has to tunnel or fritt through.

The denominator is the important part. Constriction resistance scales with spot radius, not spot area, so doubling a spot's area only cuts its constriction resistance by about a factor of 1.4, while doubling the number of spots halves it. This is why surface preparation beats bolt torque as a lever: torque grows existing spots, preparation creates new ones.

Put a number on it. Copper at 1.724 × 10⁻⁸ Ω·m with one a-spot of 10 µm radius gives Rc = 862 µΩ for that spot alone. A thousand in parallel would give 0.86 µΩ. Real joints sit above that, because closely spaced spots interfere with each other's current fields and the cluster behaves partly as one larger constriction, but the order of magnitude explains why a good joint measures in single-digit microhms.

Contact pressure targets

Contact resistance falls steeply with contact pressure and then stops falling. The CDA guidance is specific: above roughly 30 N/mm² there is little further improvement, below 7 N/mm² is not advisable in most cases, and pressures above 10 N/mm² are preferred. Independent work on copper busbar connections tested 6 to 36 MPa and found contact resistance falling from about 16 µΩ at 6 MPa to about 11 µΩ at 30 MPa, with no significant gain beyond.

So the design target is a band, not a maximum: somewhere between 10 and 30 N/mm² averaged over the overlap. It is a band rather than a number because the pressure is nowhere near uniform across the overlap.

Torque is a proxy for preload, and a bad one

Contact pressure cannot be measured in the field. It is inferred from bolt torque through:

T = K F D

where T is torque in Nm, F is bolt force in N, D is nominal bolt diameter in m, and K is the nut factor, which depends on thread friction, surface finish, plating and lubrication state. The CDA quotes 0.20 to 0.22 dry, 0.19 to 0.21 with a contact aid compound, and 0.15 to 0.16 with a molybdenum disulphide boundary lubricant. Work that through on the CDA's own tabulated arrangement for 50 mm bar: 70 mm overlap, 3500 mm² joint area, two M12 bolts at 45 Nm.

Same torque wrench, same setting, same fitter. A 47% spread in preload, and a joint that lands either comfortably inside the preferred band or slightly below it depending on whether somebody wiped the threads. That is before wrench calibration or bolt sequence. Torque control is not useless, but on its own it gives you a joint you cannot claim a contact pressure for, which is much of why the recommended band is so wide.

Belleville washers earn their place thermally

The second reason preload does not stay where you put it is temperature.

IEC 61439-1 permits a temperature rise of 105 K on bare copper busbars, so a bar in a 35 °C ambient can legitimately sit at 140 °C. The bolt through it is neither at the same temperature nor of the same expansion coefficient. Copper runs at about 16.5 × 10⁻⁶ per °C, high-tensile steel at 11.1 × 10⁻⁶, aluminium bronze CW307G at 16.2 × 10⁻⁶.

With a steel bolt through a copper joint the copper grows faster than the bolt, so bolt tension rises as the joint heats. That sounds helpful and is not. The CDA requires maximum bolt tension anywhere in the working range to stay below 95% of proof stress, because a bolt taken past yield does not come back, and on cooling the joint is looser than it started. Cycle that daily for ten years and the joint relaxes to nothing.

Two ways out. Use a bolt whose expansion matches copper, which is the argument for CW307G at 16.2 × 10⁻⁶ against copper's 16.5, a differential small enough that tension barely moves. Or keep the steel bolt and put a disc-spring washer in the stack, sized on height and spring rate so thermal growth compresses the spring instead of stretching the bolt.

Belleville washers are usually sold as an anti-vibration measure. That is not what they do here. They are a compliance element that turns a stiff, low-travel bolted joint into a softer one, so a few hundredths of a millimetre of differential expansion does not become a large change in load. Braunovic's joints used thick 3 to 4 mm flat washers, 1.4 times the disc-spring diameter, to stop the spring edge biting in.

What surface preparation is actually worth

Braunovic measured joint contact resistance on identical hardware with nothing changed but the surface treatment. Four- and six-bolt joints, 60 Nm, disc-spring washers, 150 A DC, mean of readings over an hour.

Those are aluminium connector pads against aluminium busbar, so do not transplant the absolute values onto a copper joint. The ratios are the point. Between an as-received surface and a machined and lubricated one on the same hardware, contact resistance fell by a factor of 24 on the four-bolt configuration and 70 on the six-bolt. Nothing you can do with a torque wrench comes close to that.

Machined surfaces also outperformed wire-brushed ones, attributed to the uniform height and spacing of the machining serrations giving a more controllable load-bearing area. Note the direction of that: a deliberately roughened surface beats a smoothly polished one, because sharp asperities penetrate oxide films at much lower load than blunt ones. Polishing a joint face to a mirror is counterproductive.

You will also see the figures "bright copper 3 to 5 µΩ, light tarnish 15 to 25 µΩ, heavy oxide 50 to 200 µΩ" quoted in several places. We could not trace them to a primary measurement, and a contact resistance figure means nothing without the joint geometry, bolt count and test current alongside it. Treat them as an illustration of the ordering, not as design data.

Plating: four options, four different failure modes

Bare copper is the default, and a better default than most people assume. Copper oxidises slowly, and the film that forms is semiconducting rather than insulating, because copper ions diffuse into it and conduction proceeds in both directions. The CDA does not recommend plating copper-to-copper joints unless the environment demands it, for a reason that surprises people: plating materials are soft, and a soft layer can flow when hot and reduce contact pressure. Bare copper does need a contact aid compound to fill the voids between the a-spots and keep oxygen and moisture out. Petroleum jelly works, and silicone vacuum grease works hotter.

Tin is cheap and its oxide is easily disrupted under pressure, but it is soft, so it creeps under sustained pressure and damages readily under fretting. Pure tin grows whiskers that can cause transient short circuits, so the CDA advises against it, and the tin-lead alloys that historically solved that are now off the table environmentally. Copper-tin intermetallics grow at the interface with temperature and are more resistive than either parent metal, giving tin the lowest service ceiling of the four.

Silver is the best electrical performer. Its oxide is conductive and thin enough to be displaced under contact pressure, and it holds up at the highest temperatures. It is expensive, and it is wrong in sulphur-bearing atmospheres where silver sulphide forms readily: paper mills, sewage plants, anywhere near a refinery.

Nickel is the CDA's preferred general-purpose coating, being cheap, durable and hard enough to survive installation handling. The trade-off is electrical. Nickel oxide is tough and needs high contact pressure to break through, so a nickel-plated joint assembled at the low end of the band will not perform, and nickel does poorly at high humidity.

Metal coatings are thin, typically 2 to 5 µm, and only protect if continuous. A pinhole in a corrosive environment is worse than no plating, because it concentrates the attack.

Pairing with aluminium

Copper and aluminium in direct contact with any electrolyte present form a galvanic cell, and the aluminium is the sacrificial half. Standard electrode potentials are +0.34 V for copper and −1.66 V for aluminium against the standard hydrogen electrode. In a dry sealed indoor cabinet you may get away with it for years; in anything humid or coastal you will not. The fixes, in ascending order of cost, are a jointing compound that seals the interface against moisture, tin plating so the mating surfaces are like-to-like, or a friction-welded or explosion-bonded bimetallic transition plate that moves the dissimilar-metal interface into the factory. The fabrication differences between the two metals go further than the joint.

Fretting and relaxation: how good joints go bad

Two slow mechanisms account for most joints that pass commissioning and fail years later.

Fretting is surface damage from very small oscillatory movement at the interface, and the amplitudes are startlingly small. Displacements under 100 nm are enough to initiate it, and because slip below roughly 125 µm cannot sweep debris out of the contact zone, oxidised wear particles stay put and build into a thick insulating third body. Contact resistance then rises sharply toward an open circuit. The oscillation comes from vibration, differential thermal expansion, or the joint simply heating and cooling as load cycles.

Creep and stress relaxation are the other. Creep is dimensional change under sustained load; stress relaxation is loss of contact pressure without dimensional change, as elastic strain converts to plastic. Both depend on time, temperature and stress, and both are markedly worse in aluminium. The CDA's comparison states the difference cleanly: annealed 1080 aluminium at 20 °C shows a minimum creep rate of 0.022% per 1000 hours at 26 N/mm², while annealed high-conductivity copper only reaches that rate at the same stress when heated to 150 °C.

That is the aluminium jointing problem in one line. At room temperature, aluminium under normal busbar contact pressure is already creeping at a rate copper does not reach until well above its service range. Bolted aluminium joints need disc springs as standard practice rather than as an upgrade, and they need retorquing schedules copper joints do not.

Where the fabrication shop comes in

Flatness first. Braunovic attributed the poor performance of unslotted joints to misalignment between rigid bars: some regions carry high stress and yield, others barely touch, and the load-bearing area ends up small. A bar with 1 mm of twist over its length will not lie flat against its neighbour whatever torque you apply.

Then burrs. A burr on a punched hole edge is raised metal standing proud of the mating face. It holds the bars apart around the hole, exactly where the bolt applies its pressure, and converts face contact into ring contact on the burr. The BND800-2 deburring machine handles bar 0.5 to 50 mm thick and 100 to 800 mm wide at 5 to 30 m/min, fast enough that deburring every bar is a scheduling decision rather than a bottleneck.

Hole quality follows from the same operation. Punching with correct clearance leaves a clean hole wall and a small, consistent exit burr; excessive clearance leaves a torn edge and a large one. Placement matters electrically too: holes should sit in-line along the joint, since offsetting them disturbs current flow further, and the dominant term in joint efficiency is the cross-section lost to the holes. DIN 43673-1 fixes hole locations and sizes and is a better starting point than a house standard.

On the mating faces you want uniform, controlled roughness rather than polish, and a controlled edge at the overlap boundary, where a sharp corner is both a stress raiser in the pressure distribution and a field concentration point. That second consequence is the subject of busbar chamfering and edge radius.

Overlap geometry is the last of them. The current distortion penalty at an overlapped joint, the streamline effect, falls rapidly as the overlap-to-thickness ratio rises to about 2 and then flattens. Very long overlaps buy nothing; you need enough length to fit the bolts that generate the contact pressure. Two geometry changes are free at the punching stage. Angling the bar ends at less than 45° cuts initial joint resistance by around 15% and slows the rise in resistance under current cycling by a factor of 1.3 to 1.5. A longitudinal slot cut in both bars reduces contact resistance by 30 to 40%, by evening out the pressure across each leg. Both are cheaper to make on the punching and shearing line than to explain afterwards.

A joint below 100% efficiency runs hotter than the bar around it, and in a cabinet full of joints that heat is cumulative. The ampacity calculator gives the bar's thermal baseline, which is the budget the joints have to live inside.