IEC 61439 temperature-rise verification, route by route

Temperature-rise verification is where busbar fabrication quality becomes an audit finding. Everything else in an assembly's design verification can be settled on paper by somebody who has never touched the copper. This one cannot, because the number being verified is produced by joints that a fitter made with a torque wrench.

Most published guidance stops at naming the three routes. The parts that decide whether a project succeeds are the boundary conditions on each route, what the rated diversity factor actually multiplies, and which fabrication variables move the result. Clause references below are to IEC 61439-1

The three routes

Clause 10.10.1 lists them plainly. Verification shall be made by one or more of:

a) Testing, clause 10.10.2. A physical assembly, loaded, in an accredited laboratory.

b) Derivation of ratings for similar variants from a tested design, clause 10.10.3. The comparison route: you already have a tested arrangement and you are assigning ratings to variants of it under stated rules.

c) Calculation, by two separate methods with two different current ceilings. Clause 10.10.4.2 covers a single-compartment assembly not exceeding 630 A; clause 10.10.4.3 covers an assembly not exceeding 1600 A using the method of IEC TR 60890.

Two constraints sit above all three. Every route is limited to rated frequencies up to and including 60 Hz for the calculation methods, and for assemblies rated above 60 Hz, clause 10.10.1 requires verification by test or by derivation from a design tested at the same frequency. Combination is permitted and normal: a multi-section assembly can have different sections verified by different routes.

Route c) in detail, because it is the one with the fine print

The calculation route is the one most manufacturers reach for, because it needs no laboratory. It also carries the conditions that most often disqualify a project after the design is frozen.

The 630 A method (10.10.4.2)

A power-loss balance. Total power loss from components and conductors inside the enclosure is compared against the enclosure's ability to dissipate it without the internal air exceeding the maximum operating temperature of any device. Quick, conservative, and narrow.

Conditions, all of which must be met:

• Power-loss data for all built-in components is available from the component manufacturer.
• Power losses are approximately evenly distributed inside the enclosure.
• The rated current of each circuit does not exceed 80 % of the rated conventional free-air thermal current (Ith), or the rated current (In), of the switching devices and components in that circuit.
• Mechanical parts and installed equipment do not significantly impede air circulation.
• Conductors carrying more than 200 A, and adjacent structural parts, are arranged to minimise eddy-current and hysteresis losses.
• All conductors have a minimum cross-section based on 125 % of the permitted current rating of the associated circuit.
• The enclosure's temperature rise as a function of installed power loss is known, whether from the enclosure manufacturer, from test per 10.10.4.2.2, or from the cooling-equipment manufacturer where active cooling is fitted.

The 1600 A method (10.10.4.3)

Same 80 % derating and 125 % conductor rule, with the temperature rise itself determined by the IEC TR 60890 method rather than a simple power-loss balance. It adds two geometric conditions that catch multi-compartment designs:

• There are no more than three horizontal partitions in the assembly or in the section being assessed.
• For enclosures with compartments and natural ventilation, the cross-section of the ventilating openings in each horizontal partition is at least 50 % of the horizontal cross-section of the compartment.

Above 1600 A, both calculation routes close. You are on testing or derivation from a tested design, and that is a decision to take before you build, not after.

The standard is candid about why the margins exist: clause 10.10.4.1 states that because the actual local temperatures of current-carrying parts cannot be calculated by these methods, some limits and safety margins are necessary and are included. The 80 % derating and the 125 % conductor sizing are those margins. They are not conservatism you can argue away with better analysis.

The 80 % derating in practice

The derating is not a note in the file. It changes the bill of materials.

A 200 A circuit cannot use a 200 A device. It needs one whose free-air rating is at least 250 A, set or fixed at 200 A, or a larger frame set lower. The power loss used in the calculation is the loss at 200 A, but the device itself must be a 250 A frame or above. The conductor rule compounds it: a 200 A main busbar circuit is sized on 1.25 × 200 = 250 A, and that applies to every internal conductor in the assembly.

Then the iteration. Conductor cross-sections are first chosen for an assumed internal air temperature. Once the calculation produces actual air temperatures, components and conductors may need changing and the calculation rerun. Two or three passes is normal, and each one can move the bar size, which moves the hole pattern, which moves the tooling. Settle the verification route before the copper is detailed, not after.

The rated diversity factor, and what it actually does

RDF is the most misunderstood term in the standard, and most of the misunderstanding comes from assuming it is a discount applied inside the calculation. It is not.

Clause 5.4 defines it: the per-unit value of the rated current, assigned by the assembly manufacturer, to which outgoing circuits can be continuously and simultaneously loaded, taking into account mutual thermal influences. It can be stated for groups of circuits or for the whole assembly. Two clauses in the definition do the work. RDF multiplied by the rated current of the circuits shall be equal to or higher than the assumed loading of the outgoing circuits, and RDF applies with the assembly operating at its rated current InA.

So RDF is a declared rating, not a calculation input. Clause 10.10.1 puts it exactly: the current-carrying capability of the circuits to be verified is determined by the rated current and the RDF. It defines what you are verifying.

Where RDF enters the test route

In the test methods, RDF is explicit and physical.

• Clause 10.10.2.3.5, verification of the complete assembly: incoming and outgoing circuits are loaded at their rated currents, which the standard states is equivalent to a rated diversity factor of 1. Quick, conservative, and it proves nothing about diversity because it assumed none.
• Clauses 10.10.2.3.6 and 10.10.2.3.7 d): the assembly is verified by loading the incoming circuit to its rated current and the outgoing functional units to their rated current multiplied by the diversity factor. This is where a declared RDF below 1 is actually earned.

Annex O sets out the trade: method a) needs the fewest tests but the test arrangement is more onerous than necessary and the result applies only to that arrangement.

Where RDF does not enter: the calculation route

Both calculation clauses say the same thing: the effective power losses of all circuits including interconnecting conductors shall be calculated based on the rated current of the circuits. There is no RDF multiplier. Diversity enters by a different door. The total power loss is computed taking into account that the total load current is limited to the rated current of the assembly.

The standard's own worked note makes it concrete. A single-compartment assembly rated 100 A, limited by the distribution bars, is fitted with 20 outgoing circuits each assumed to be 8 A. Total effective power loss is calculated for 12 outgoing circuits at 8 A each, because 12 × 8 = 96 A is what the 100 A incoming rating will actually distribute. Twelve circuits at full load, not twenty circuits at a diversity factor.

That distinction is the practical one. Guidance published by BEAMA on non-test verification puts it as plainly as anyone has: rated diversity factor is not considered when verifying temperature rise by calculation, and the most onerous power-loss combination of circuits operating at their rated current is used to distribute the assembly's rated current. Anyone who has been applying an RDF discount inside a TR 60890 spreadsheet has been calculating a lower temperature rise than the standard allows them to claim.

The tabulated values

Where the manufacturer and user have not agreed actual load currents, the product parts give fallback values of assumed loading, indexed to the number of main or outgoing circuits. In Parts 1 and 2 the value starts at 1,0 for an incoming circuit and steps down as the number of outgoing ways rises, reaching 0,6 at ten or more circuits. In Part 3, for distribution boards operated by ordinary persons, the ladder sits one step lower throughout and bottoms at 0,5.

A figure of 0,4 circulates widely online as the bottom of this range. It does not appear in the tabulated assumed-loading values of Parts 1, 2 or 3. A manufacturer may of course declare an RDF lower than the tabulated fallback, since RDF is assigned by the manufacturer and the table only applies in the absence of agreement. But if a specification quotes 0,4 as coming from the standard, ask where.

Annex E of Part 1 works examples at RDF values such as 0,68. The number is not a round one from a table; it falls out of the loading arithmetic.

The limits you are verifying against

Table 6 of IEC 61439-1 says more than the two numbers usually quoted from it.

Terminals for external insulated conductors are limited to 70 K. The footnote qualifies it twice. The 70 K is a value based on the conventional test of clause 10.10, and an assembly used under real installation conditions may have connections whose type and disposition differ from the test, producing a different terminal temperature rise that may be required or accepted. And where the terminals of a built-in component are also the terminals for external insulated conductors, the lower of the two applicable limits governs: the component manufacturer's limit and 70 K, taking the smaller.

Busbars and conductors get no single number in the table. The row states that the limit is set by the mechanical strength of the conducting material, possible effects on adjacent equipment, the permissible temperature limit of insulating materials in contact with the conductor, the effect of conductor temperature on the apparatus connected to it, and, for plug-in contacts, the nature and surface treatment of the contact material.

105 K appears in note g, as a ceiling on top of those criteria: assuming all the listed criteria are met, a maximum temperature rise of 105 K for bare copper busbars and conductors shall not be exceeded. Note 1 gives the reason: 105 K relates to the temperature above which annealing of copper is likely to occur. It is a metallurgical limit. Reach it and the bar softens, which is a mechanical failure mode, not an electrical one.

The ambient assumptions sit underneath all of it. Table 6's limits apply for a mean ambient air temperature up to 35 °C in service. Annex O adds the peak: daily average and peak ambient temperatures not exceeding 35 °C and 40 °C respectively. During the test itself the ambient must be between 10 °C and 40 °C, measured by at least two sensors around the assembly at approximately half its height and about 1 m away, shielded from air currents and radiation.

Two practical readings. First, the 105 K ceiling is rarely the binding constraint. The insulating material in contact with the bar, or the device the bar lands on, usually bites first. Second, an assembly destined for an environment hotter than a 35 °C mean needs its limits adapted, and clause 9.2 notes that if the limits are changed to cover a different ambient, the rated currents of busbars and functional units may need changing with them.

What each route costs and how long it takes

Published price lists for accredited temperature-rise testing do not exist, because the cost is dominated by the number of test arrangements and by what you have to build. The cost structure is what you need for the decision anyway.

• Testing: you build one or more representative assemblies to the most onerous arrangement, ship them, and pay for laboratory time. The test runs until temperature rise reaches a constant value, which clause 10.10.2.3.1 defines in practice as variation at all measured points, including ambient, not exceeding 1 K/h. That is typically most of a working day per arrangement, before instrumentation and setup. Current is delivered through test conductors the standard specifies in detail. Above 800 A that means copper bars of stated sizes with minimum connection lengths of 3 m, reducible to 2 m only under a stated temperature condition. Above 4000 A the original manufacturer determines the arrangement and it forms part of the report. Cost is dominated by the sample build and the number of arrangements. Delay is dominated by laboratory queue time, months rather than weeks for high-current work. Do not book the slot before your joints are made the way they will be made in production.

• Derivation from a tested design: cheapest and fastest where you already own the tested design, which is why assembly-system suppliers sell it as part of the package. Stay inside its rules and most of the work is done. Step outside them and you are back to test.

• Calculation: no laboratory, so the cost is engineering time plus component data. The invisible cost is the data-gathering: condition a) requires power-loss data for all built-in components, and chasing that from every device supplier for a mixed bill of materials takes longer than the calculation does. Budget days rather than hours for a first assembly, less afterwards as the component data accumulates. The real risk is discovering at the end of the iteration that you needed a bigger frame or a bigger bar, so run a rough pass early.

Why fabrication quality decides the outcome

Every route ultimately measures the same physical thing, and clause 10.10.2.3.3 names it directly: particular attention shall be given to joints in conductors and terminals within the main circuits. Joints are named measurement points.

Joint resistance has two parts. Spreading resistance, from current diverting through the overlap, is a function of the overlap-to-thickness ratio and of the hole pattern: bolt holes across the width of the bar penalise it more than holes in line along its length. Contact resistance is a function of surface condition and pressure.

Contact area is not overlap area. Only about 1 % of the apparent overlap conducts, through discrete asperity contacts. Contact resistance falls steeply with pressure up to roughly 10 N/mm² and flattens out around 30 N/mm²; below about 7 N/mm² it is not advisable to operate at all. Since pressure cannot be measured in the field, it is inferred from torque through the nut factor, which varies from about 0.20–0.22 dry down to 0.15–0.16 with a boundary lubricant. The same wrench setting on greased threads gives roughly 30 % more preload than on dry ones. Torque values and lubrication instructions have to come from the same source, or the number in your procedure is fiction.

Surface preparation is the cheapest variable and the most often skipped. Faces should be flat and freshly abraded immediately before assembly, so that clamping pressure disrupts the oxide and forms metal-to-metal contact at as many points as possible. A burr left on the joint face does the opposite: it holds the bars apart, and the contact area you calculated is not the contact area you get. That is why deburring belongs in the process route rather than in the fitter's hands, and why a line like the BND800-2 handling 0.5 to 50 mm thick bar at 5 to 30 m/min exists at all.

Embossing is the other lever, and the underused one. Raising a controlled contact land concentrates the available bolt preload onto a defined, flat area instead of letting it spread across an overlap that may or may not be in contact. Where a joint is thermally marginal, a designed contact land is often a better answer than another bolt.

Hole pattern accuracy feeds straight into all of the above. Pitch error across a multi-bolt joint means bolts fighting each other in oversized holes and preload distributed unevenly. Punching that holds position tightly costs nothing extra per hole and removes a variable from a test you are paying a laboratory for. A processing centre such as the IMAC-CENTER 80 works to ±0.05 mm. The punching process page covers the tooling side, and the hole-pattern article covers edge distance and pitch.

Before you commit to a route

Size the bar against a ratings source with its correction factors intact rather than against a reference-condition table. DIN 43671 rates copper bar at 35 °C ambient and 65 °C bar temperature, with corrections for ambient from 0 to 60 °C, operating temperature up to 125 °C, and factors for bar arrangement and the number of bars per phase. Inside an enclosure at 55 °C internal air, those corrections are the difference between a bar that verifies and one that does not.

Our busbar ampacity calculator applies the corrections rather than assuming reference conditions. Where the assembly is going to a North American market, the limits change completely and so does the bar. That comparison is in UL 891 versus IEC 61439.