Key takeaways
- Size each run for the rack it serves, not for a room average that hides the high-density rows.
- Higher rack power raises current, heat and voltage drop at the same time, so the section and the route have to be checked together.
- Tap points, joints and plug-in connections are the parts of an overhead run that decide reliability.
- Fault level and protection clearing time belong in the design, even when the continuous current looks comfortable.
Why rack density changes the bar
A data hall used to be designed around a power density that was fairly uniform. A row of racks drew roughly the same current, the cooling was sized for the row, and the distribution bar could be sized from a room total. High-density computing changed that. A single rack can now draw several times the power of its neighbour, and the row average is no longer a safe design input.
The busbar sees the consequence first. More power per rack means more current in the dropper, more heat in the aisle and a larger voltage drop at the server input. If the bar is sized from a room average, the high-density row runs hot and the racks at the far end see a lower voltage than the ones near the source. The design has to start at the rack and work back.
Current per rack, not current per room
The current a rack draws depends on its power, the supply voltage and the power factor. For a three-phase supply, the line current is the power divided by the product of the square root of three, the voltage and the power factor. The figures in the infographic are illustrative and assume a 415 V three-phase supply at unity power factor. They show the shape of the problem, not a design value for a particular hall.
A rack at 10 kW draws about 14 A per phase. At 30 kW it draws about 42 A. At 60 kW it draws about 84 A, and at 120 kW it draws about 167 A. The current rises in proportion to the power, but the heat in the bar rises with the square of the current. Doubling the rack power quadruples the heating in the same section, which is why a bar that was comfortable at the old density can be marginal at the new one.
The practical response is to design the distribution in short runs with known rack loads, rather than one long bar sized on an average. A run that serves four high-density racks can be sized for those four, with a tap for each, and the next run can be sized separately. That keeps the copper where the current is and avoids over-sizing the whole aisle for one dense row.
Space, bends and cable entry
An overhead busway has to fit above the racks and below the ceiling services. That space is shared with cable tray, containment, fire suppression and lighting. A bar that is deep enough for the current may not fit the available height, and a bar that is wide enough may not leave room for a bend radius or a tap box.
Bends are the first thing to check. A flatwise bend needs a radius that suits the bar thickness, and an edgewise bend needs a different radius and a different support. A bend that is squeezed into a ceiling void can crack, thin the section or reduce the clearance to the tray beside it. Draw the bend in the section, not just on the plan.
Heat, ambient and derating
A data hall is not a cool place at the bar. The inlet air may be controlled, but the space above the racks is warmer, and the bar is often close to hot exhaust or to other current-carrying conductors. The ambient used for the current rating has to be the temperature in that space, not the set point of the cold aisle.
Grouping matters too. Several bars running side by side heat each other, and a bar in a closed containment or a riser has less air movement than one in open ceiling space. A rating from a free-air table has to be adjusted for the real mounting. If the bar is painted or insulated, the surface treatment also changes the heat transfer a little.
The thermal check should include the joints and the taps, because they run hotter than the bar when they are poorly made. A thermal camera survey after commissioning is a useful check, provided it is done at a known load and the results are compared with the design expectation rather than with an absolute limit alone.
Measure at the load you know
A thermal survey at half load tells you less than one at a stated current. Record the current with the image so the result can be compared later.
Taps, joints and plug-in points
An overhead run has more connections than a panel bar. Each tap box, each joint between lengths and each end connection is a place where resistance, heat and human error can collect. A plug-in tap is convenient, but the contact spring, the alignment and the rating of the plug have to suit the load. A tap that is rated for the run but not for the rack is a hidden limit.
The joint between two lengths of busway needs the same discipline as a panel joint: flat faces, correct plating, controlled torque and a contact resistance check. In a ceiling, the joint is harder to reach and harder to inspect, so the design should reduce the number of joints or place them where a person can work.
Label every tap with its circuit and its rating. In a dense hall, a tap that was installed for a 10 kW rack can be reused for a 30 kW rack by mistake. A label and a load schedule are cheaper than a thermal incident.
Fault level and protection
A high-current run has a high prospective fault current, and the bar and its supports have to survive it until the protection clears. The thermal duty depends on the current squared and the clearing time, and the mechanical duty depends on the peak current and the support spacing. A busway with a large continuous rating can still be weak in the fault case if the supports are far apart or the joints are not braced.
Protection settings and the busbar rating have to be read together. A bar rated for a short-time current assumes a clearing time. If the upstream device is set slower than that, the bar sees more heat than the rating allows. The fault study should state both the current and the time, and the bar selection should use the same pair.
For a run that feeds several racks, the fault contribution from the downstream equipment also matters. The design should consider the fault level at each tap, not only at the source, because the impedance of the run changes the current along its length.
What to specify
Send the rack power schedule, the supply voltage, the power factor, the ambient above the racks, the run length, the tap positions and the fault level. Add the enclosure or containment details, the required ingress protection and the monitoring or metering points. A busway drawing without the rack schedule is a shape without a duty.
Ask for the rating basis: the ambient, the mounting, the allowable temperature rise and the joint rating. A current number without those conditions cannot be compared with another supplier. Ask for the fault rating with its duration, and for the tap rating as well as the run rating.
If the hall will be built in phases, say so. A run that is sized for the final fit-out may be lightly loaded at first, and a run that is sized for the first phase may not take the later racks. The phasing plan is part of the electrical design, not a commercial afterthought.
| Rack class | Power per rack | Current at 415 V | Design focus |
|---|---|---|---|
| Low density | 5 to 10 kW | About 7 to 14 A | Standard dropper, shared tap |
| Medium density | 20 to 40 kW | About 28 to 56 A | Dedicated dropper, voltage drop |
| High density | 60 to 100 kW | About 84 to 139 A | Section, tap rating and heat |
| Very high density | Above 120 kW | Above 167 A | Short runs, bracing and cooling |
The power bands and currents are illustrative and assume a 415 V three-phase supply at unity power factor. Real values depend on the rack, the power supply and the hall.
Frequently asked questions
Can one busway serve a whole data hall?
It can, but a single long run sized on a room average hides the high-density rows. Short runs with known rack loads keep the copper where the current is and make the voltage drop and tap ratings easier to check.
How does rack density affect busbar temperature?
Heating rises with the square of the current, so doubling the power in the same section roughly quadruples the heat. The ambient above the racks is also warmer than the cold aisle. Both effects have to be included in the current rating.
Are plug-in taps a weak point?
They can be. The contact spring, alignment and rating of the plug decide the joint resistance, and a tap rated for the run may not be rated for the rack. Label each tap, check its rating against the load schedule and inspect it at commissioning.
What fault level should a data centre busway be rated for?
The prospective fault current and the clearing time from the protection study. The thermal and mechanical checks use both, and the rating should state the duration. A continuous current rating alone does not show whether the run will survive a fault.
Sources and further reading
- IEC 61439-1, Low-voltage switchgear and controlgear assemblies
- IEC 60364-5-52, Selection and erection of electrical equipment, wiring systems
- Qumek data centre busbar and busway reviews
Qumek Engineering Desk
Engineering team
The engineering team behind Qumek drawings, DFM notes, tolerances and inspection standards. Every guide is reviewed against the drawings and processes used on live programmes.
Sizing a run for a high-density row?
Send the rack power schedule, supply voltage, ambient and fault level. The engineering desk will review the section, taps and route before quoting.
Send the rack schedule