Power Supplies for Telecom and Datacom Equipment

Telecom power is built around an assumption that other sectors treat as a goal: the equipment does not stop. That produces a distinctive architecture — a DC bus permanently floated across a battery string, rectifiers feeding it in parallel, and equipment designed to keep running while any single element is replaced.

The conventions are old, well documented and worth following rather than reinventing, because the installed base an engineer is connecting to was built to them.

The −48 V bus, and the range that actually matters

ETSI EN 300 132-2 defines the DC power interface at the input of information and communication technology equipment. The nominal voltage is −48 V DC with the positive conductor connected to earth, a convention that dates to telephone practice and persists because it reduces galvanic corrosion on the exposed conductors of an outside plant.

The figure that governs design is not the nominal one. The standard sets the normal service voltage range at −40.5 V to −57.0 V, and equipment must operate throughout it with no degradation of service. The top of that range follows from a charging battery — the standard derives it from 2.35 V per cell across 24 lead-acid cells, plus regulation allowance, minus distribution drop. The bottom reflects a battery that has been discharging for some time.

Beyond that, the standard defines an abnormal range: equipment has to survive 0 to −40.5 V and −57.0 to −60.0 V without damage, and recover automatically once the supply returns to normal. Some older installations run a −60 V system instead, for which the normal range is −50 V to −72 V.

The practical consequence is a specification question with a clear answer. A converter that stops regulating at −42 V is inside the nominal figure and outside the standard, and it will drop out partway through a battery discharge — during the outage the battery exists to cover.

Rack power systems and hot-swap

The rack power category covers the shelf-and-module architecture this application uses. A rack subsystem — the RCP, RKP, RHP and DHP rack systems, with RCP and DRP modules populating them — holds several supply modules in a common shelf sharing a backplane, an output bus and a monitoring interface.

Two properties follow from that arrangement. Modules can be inserted and removed while the system is live, so a failed unit is replaced without taking the load down. And capacity is added by fitting another module rather than by replacing the system, which matters when a site grows past the load it was built for.

Monitoring is what makes redundancy real rather than theoretical. Controllers such as the CMU units in the same category report module status, and without something in that role a redundant system quietly degrades to a non-redundant one the first time a module fails — with no outward symptom until the second failure takes the site down. A redundant system with no alarm path is a system that has bought itself one silent failure.

N+1, and what redundancy is actually buying

N+1 means the load needs N modules and the system has one more, so any single failure leaves capacity intact. It is the common arrangement and the right default for most sites.

Two details decide whether it works. Current sharing has to be handled properly, so modules divide the load rather than one carrying most of it and ageing faster than the rest. And the shared elements have to be honest: a shelf whose modules all feed one output bus through a single connector, or draw from a single AC feed, has a component whose failure defeats the redundancy entirely.

How far to take it is a question about what the site is worth. 2N — a complete duplicate system, ideally on a separate AC feed — costs roughly twice as much and removes whole classes of common-mode failure that N+1 does not. Most access sites are N+1; core sites are frequently not.

Airflow, and the direction it has to run

Racks are cooled front to back, and equipment rooms are laid out on that assumption, with cold aisles feeding intakes and hot aisles collecting exhaust. A unit that draws air from the side or discharges toward the front breaks the pattern locally, recirculating hot exhaust back into its own intake and into whatever sits above it.

Airflow direction is a specification to check rather than assume, particularly when mixing equipment from several vendors in one rack.

The thermal budget itself is arithmetic that gets skipped. Every watt the equipment consumes leaves as heat, and the power supplies' own losses are added on top of the load's. A rack's usable capacity is usually set by what the room can remove rather than by what the distribution can deliver, and derating curves have to be read against the temperature at the intake — which, high in a rack in a busy row, is not the room's nominal figure.

Getting board voltages from the bus

Equipment connected to a −48 V bus rarely uses −48 V internally. Isolated DC/DC conversion produces the rails the boards need, and the isolation is not incidental: it separates the equipment's internal grounds from a bus that is shared across the whole site and earthed at one point.

The enclosed DC/DC category — the SD family, and the RSD and RSDH families built for railway and ITE duty — covers this at the chassis level, with the wide input ranges the bus range demands. For conversion on the board itself, the embedded and board-level page covers the module options.

Size these against the low end of the bus range. A converter drawing its rated output power pulls its highest input current when the bus is at −40.5 V, and that is also the moment the upstream distribution and its protection are under most stress.