A cellular base station does not run on mains alternating current. It runs on a nominal minus 48 volt direct current bus, produced by a shelf of rectifier modules and held up by a valve-regulated or lithium battery string wired permanently across that bus. Because the battery is always connected in parallel with the load, the site behaves as a continuously online DC uninterruptible supply with no transfer time at all: when the grid disappears the radios never notice, because nothing has to switch. Understanding that architecture is the difference between a network that holds five nines of availability and one that loses cells every time a feeder trips.
The choice predates mobile networks. Early telephone exchanges discovered that making the battery negative with respect to earth slows galvanic corrosion on buried copper, because the metalwork sits at the positive potential. Forty-eight volts then survived as a compromise between two competing pressures. Lower voltages force impractically thick busbars for a given power, while higher voltages cross the threshold where safety rules demand a different class of installation and a different grade of technician.
Open a typical macro site cabinet and you will find four functional blocks. The AC input section provides isolation, overcurrent protection and surge diversion, and increasingly a generator changeover contactor. The rectifier shelf holds between two and twelve hot-pluggable modules, commonly 1.5 kW to 3 kW each, sharing current actively so that no module carries more than its neighbours. The battery compartment holds one or more series strings sized for the target autonomy. The controller supervises everything, running float and boost charging regimes, temperature compensation, low-voltage disconnect and alarm reporting back to the network operations centre.
If the DC plant already provides uninterrupted power, why specify an AC UPS at all? Because a modern site is no longer only radios. Edge compute nodes, small server stacks, air conditioning for indoor shelters, security and access systems, transmission gear inherited from acquisitions and test equipment all expect ordinary mains. Feeding those from an inverter hung off the DC bus is possible but rarely economic above a few kilowatts.
The common arrangement at a hub or aggregation site is therefore an online double-conversion AC UPS in the range of 10 kVA to 60 kVA covering the alternating current auxiliaries, sitting alongside the DC plant that covers the radio payload. The two are sized independently because their autonomy requirements differ sharply: radios may need four hours, while an air handler needs only enough to bridge a generator start.
Rectifier capacity is calculated as the peak DC load plus the current needed to recharge a fully discharged battery inside the operator's recovery window, plus one spare module for redundancy. Recharge current is the term most often forgotten: a string that must refill within eight hours can demand a quarter of its rated capacity in charging current.
Battery sizing works backwards from autonomy. A site drawing 3.5 kW that must survive four hours needs roughly 14 kWh delivered at the battery terminals. Allowing for inverter-free DC efficiency, an end-of-discharge voltage of 43.2 volts and the fact that a lead-acid string should not be planned below about eighty percent depth of discharge, the practical requirement lands near 400 ampere-hours at the 8-hour rate. Lithium iron phosphate changes the arithmetic because it tolerates deeper discharge and far more cycles, which is why urban sites with frequent short outages have migrated to it fastest.
Across the field reports we see, four causes dominate lost cell hours. Battery capacity fade is first by a wide margin, and it is almost always thermal: every ten degrees Celsius above twenty-five roughly halves the service life of a valve-regulated lead-acid string, so an unventilated cabinet in a hot climate can destroy a five-year battery in eighteen months. Second is a rectifier module that failed silently months earlier, leaving no redundancy on the day it was needed. Third is a low-voltage disconnect set wrongly, either so high that available autonomy is thrown away or so low that the string is damaged. Fourth is corroded or under-torqued DC terminations, which raise resistance until the voltage at the load sags long before the battery is exhausted.
None of these are exotic. All of them are found by a quarterly discharge test and a thermal check, and all of them are missed by an inspection that only confirms the green lamp is lit.
Operators generally target 99.99 percent availability per site, which permits about 53 minutes of downtime a year, and 99.999 percent on aggregation nodes, which permits roughly five minutes. Grid supply in many emerging markets delivers nothing close to that on its own. A DC plant with N+1 rectifiers and a healthy four-hour string typically converts a supply averaging 99.5 percent availability into site availability above 99.98 percent, and adding an automatic generator start pushes it past 99.995 percent. The plant is not a peripheral: it is the single largest contributor to the availability figure the network is judged on.
If you are planning new sites, refreshing an ageing DC plant or trying to work out why one cluster keeps dropping while its neighbours hold, send us the load schedule, the autonomy target and the cabinet temperature history. Our telecom power configurations are set out at https://www.upsboss.com/telecom/, and rectifier, inverter and UPS ratings are listed across the range at https://www.upsboss.com/products/.
Key takeaway: the base station is already an uninterruptible system by virtue of a battery sitting permanently across a minus 48 volt bus, so availability is decided almost entirely by battery health and rectifier redundancy rather than by transfer speed. Manage cabinet temperature and test the string on a schedule, and the rest of the design largely takes care of itself.
Can a base station be powered from a conventional AC UPS instead of a DC plant?
Technically yes, but it adds a conversion stage that the radios do not want. Radio units are built to accept minus 48 volts directly, so inserting an inverter costs efficiency, adds a failure point and takes up space, all to deliver a voltage the equipment must then rectify again internally.
How often should the battery string be capacity tested?
Every six months for lead-acid in a hot climate, annually in a temperate one. A partial discharge to around thirty percent under a controlled load, with cell or block voltages logged, will reveal a weak block long before it takes down the site during a real outage.
Is lithium worth the premium at a remote site?
Usually, once the maintenance visit is priced in. Lithium tolerates heat far better, survives many more shallow cycles and weighs roughly a third of the equivalent lead-acid string. At sites where a technician visit costs more than the hardware, the longer replacement interval settles the argument.
Contact: Frank Zhang
Phone: +86-135 5688 8641
Email: frank@upsboss.com
Add: Jufeng Road, Guangming Street, Guangming District, Shenzhen City, Guangdong Province, China
