Should a data centre thermography survey cover the whole power path, from intake to rack?

Yes, as far as safe access allows. A fault anywhere between the incoming supply and the rack can affect the same IT load, so a live survey should follow the power path: LV intake and main switchboards, generator and ATS panels, UPS, busbar and tap-offs, PDUs and rack PDUs, plus the electrical panels for the cooling plant.

Why the whole path matters

A data centre is designed so that one failure does not drop the load. But a hot connection on one side of an A and B supply quietly removes that cushion: if it fails, the remaining path carries everything. Surveys that only cover the UPS room, or only the data halls, leave the rest of that chain unchecked. Following the power path end to end is how you find the weak connection before it matters, wherever it happens to be.

What we image along the path

LV intake and main switchboards: incoming terminations, busbar joints, ACBs and outgoing ways
Generator and ATS changeover panels, imaged on whichever supply is carrying the load
UPS input and output connections, bypass panels and battery isolators
Busbar trunking runs, joints and tap-off units
Floor-standing PDUs and the rack PDUs feeding the IT equipment
Mechanical panels for chillers, CRAHs and pumps

Where surveys usually stop short

The gaps we see most often are the ones that are awkward to reach. Tap-off units above hot or cold aisles need the right access equipment. Rack PDUs sit inside customer cages in a colocation site, so access has to be arranged with each customer. Mechanical plant panels get left out because they are seen as part of a different contract. Each gap is a section of the path nobody has looked at.

Planning the route

A whole-path survey is planned with the site team before the visit: the order of rooms, who opens each panel under the site’s permit system, which areas need escorts, and when the load is representative. Sites usually split the work over consecutive days so each area is covered at a normal operating load. Anything we could not image, because there was no safe access or no load, is listed in the report so the gap is visible rather than assumed to be a pass.

Reading results along the path

Readings are compared like with like: one phase against the others on the same circuit, and one connection against similar connections carrying similar load. The A and B sides of the same system are a useful comparison where they are loaded similarly. Each finding is located on its panel and circuit with the measured temperature, where it was taken, the reference it was compared against and the load at the time.

A finding from the busbar stage

On a London data centre build, commissioning thermography by our team found 16 busbar sections reading above 80 °C under test load, the highest 103.5 °C at a tap-off unit. All sat in one part of the power path that a survey limited to the UPS room would never have seen.

Frequently asked questions

Does the whole path have to be surveyed in one visit?

No. Large sites often split it over several days or phases. What matters is that every stage is covered within the survey cycle and at a normal operating load.

Can generator panels be imaged if the generator isn’t running?

Only parts carrying current show useful heat. Generator output connections are best imaged during a scheduled load test, if the site runs one; the ATS and mains side are imaged on normal supply.

Who arranges access to rack PDUs in customer cages?

In a colocation data centre it is usually the operator, with each customer’s agreement. See our page on colocation access.