New industry Technology regarding to Bussmann fuse, ABB breakers, Amphenol connectors, HPS transformers, etc.
Data Centre · Selective Protection
A data centre has four protection levels between the utility supply and a server rack, and each one uses a different device for a different reason. Get the levels confused and the result is either a full-hall outage from a single rack fault, or a rack that cannot be isolated for maintenance. This guide sets out what each level protects and how the devices are chosen to remain selective.
Data centre electrical design is dominated by one requirement that ordinary industrial design does not share: a fault at the lowest level must not escalate. A short circuit in a single server rack should blow one branch fuse. If instead it trips the upstream breaker, the whole hall goes dark — and the cost of that event is measured in hours of downtime, not in the price of a fuse.
Selectivity, not protection, is therefore the governing criterion. Every device must be able to protect its own zone, and no device may operate for a fault in a zone below it.
| Level | Location | Typical device | Protects |
| 1 | Utility incomer / main switchboard | ACB or MCCB, high breaking capacity | The installation against the utility fault level |
| 2 | UPS input / output distribution | MCCB, sometimes fused switch | UPS module, bypass path, output bus |
| 3 | PDU / busway tap-off | MCCB or high-speed fuse in a fused switch | Branch circuit to the rack row |
| 4 | Rack PDU / server branch | Class J, Class T or Class CC fuse; MCB | The branch cord and the IT load |
The fuse families appear mainly at levels 3 and 4, because that is where current-limiting performance buys the most: a fast fuse at the rack limits the let-through energy enough that the upstream device does not see the fault as a short circuit at all.
| Criterion | Fuse | Breaker |
| Current limiting | Sub-cycle, limits peak let-through | Depends on the frame; slower at the same rating |
| Selectivity with upstream | Easy to achieve with a ratio of ratings | Requires coordination tables or electronic trip units |
| Recovery after fault | Replace the fuse; no settings to verify | Reset; settings must be checked for drift |
| Operator safety | No live parts exposed if the holder is correct | Reset may be at the source of the fault |
| Monitoring | Blown-fuse indication is an accessory | Trip signal available |
The fourth row is underrated. A rack PDU that must be reset at the rack is a different operational risk from one where a fuse can be replaced by a technician at the row PDU. In a live hall, the fuse is often the safer choice for exactly this reason.
The classic approach is a ratio: if the upstream device is rated at least 1.6 to 2 times the downstream device, selectivity is assumed. The rule works for fuses in the current-limiting region and fails elsewhere.
Three cases where the ratio rule fails:
| Level | Family | Why |
| Main / large distribution | Class L (bolt-in), very high current | Highest current ratings with 300 kA class breaking capacity |
| UPS output / busway | Class J or Class T | Compact, current limiting, wide rating range |
| Row / rack feed | Class J or Class CC | Small footprint, rejection features prevent wrong rating |
| Semiconductor loads (UPS rectifier) | aR high-speed square body | Protects the thyristors, not just the cable |
Class CC and Class J carry rejection features: a fuse of the wrong class cannot be inserted into the holder. In a hall with hundreds of holders, that feature is worth more than it costs, because it prevents the most common maintenance error — fitting a 60 A fuse where a 30 A one belongs.
Data centres usually run TN-S with a full neutral. Two practical points:
Why not use breakers everywhere?
Current limiting and selectivity. A fuse limits fault energy faster and achieves selectivity with a simple rating ratio, which matters in dense distribution with high fault levels.
How do I know a fuse has operated in a remote PDU?
Use a fuse holder with a blown-fuse indicator or a microswitch, wired to monitoring. Without it, a blown fuse in a parallel path may go unnoticed until redundancy is lost.
Is a 300 kA fuse necessary?
Only where the fault level approaches it. Most row and rack positions see far less; the requirement comes from the main board.
Does a higher breaking capacity improve selectivity?
Not directly. Selectivity comes from the time-current and let-through characteristics, not from the breaking capacity alone.
Data centre protection is a selectivity problem with a safety margin attached. Each of the four levels protects its own zone, and the devices are chosen so that a fault at the bottom never reaches the top. Fuses earn their place at the row and rack levels because they limit fault energy and make selectivity a matter of rating ratios rather than coordination studies.
The failure that matters is not a fuse blowing. It is a fuse not blowing, and a breaker upstream doing the job instead — taking an entire hall offline for a fault confined to one rack.
New industry Technology regarding to Bussmann fuse, ABB breakers, Amphenol connectors, HPS transformers, etc.