How Power Surges Damage ICT Systems

Nobody watches a surge happen. There is no smoke at first, no alarm — just a router that stops negotiating gigabit and quietly falls back to 100 Mbps, a CCTV channel that goes black next Tuesday, a server that reboots once a month for no logged reason. Six months later three devices are dead, the pattern looks like bad luck, and the actual culprit — months of electrical violence arriving over wires nobody protected — goes uninvestigated. This guide explains exactly how surges damage ICT systems, why Uganda is one of the harshest places on earth for electronics, and what real protection looks like. For the device-selection half of this story, see our power infrastructure services.

What a Surge Actually Is

Mains power should be a clean 240V sine wave. A surge is any brief excursion far above it — hundreds to thousands of volts lasting microseconds to milliseconds. Duration is the trap: a surge too short to trip a breaker or even flicker a bulb still carries enough energy to punch through semiconductor junctions rated for tens of volts. Your switch's Ethernet PHY chips operate at 2.5V internally while sitting centimetres from a cable that just carried 600V. That ratio is the entire story.

Where Surges Come From in Uganda

Load-shedding restoration. The single biggest killer. When grid power returns, transformers energise, capacitor banks switch, and every motor on the feeder starts simultaneously — the first seconds of restored power routinely carry the day's worst transients. Equipment that survived the outage dies at the recovery. If your site loses power weekly, it absorbs dozens of these events yearly.

Lost or floating neutrals. Common in ageing installations and informal extensions: the neutral conductor corrodes or loosens, phases unbalance, and 230V sockets silently receive anything up to 400V. This cooks power supplies slowly over weeks — the classic "everything keeps dying and the electrician says the power is fine" (measured at noon, under no load, it is fine).

Lightning, direct and induced. Kampala sits in one of Africa's highest lightning-density zones, with violent convective storms most afternoons half the year. A direct strike destroys everything conductive; far more common is the induced surge — a strike hundreds of metres away couples thousands of volts into overhead lines, gate wiring, CCTV cables, and long Ethernet runs. Buried cable helps; unprotected terminations still die.

Generator changeover and switching. Transfer switches (especially manual ones thrown under load), generator voltage excursions during warm-up, and UPS-to-bypass transitions all inject transients. Sites proud of their generator sometimes have worse surge exposure than grid-only sites, because nobody protected the changeover path.

Internal switching. Large motors, welders, lifts, and pump starters inside your own building generate daily switching transients that propagate through shared distribution boards into the server room. The enemy is often already inside the building.

How Each System Dies

Network switches and routers: Ethernet magnetics and PHY chips are the most surge-sensitive components in common IT. Symptoms progress: intermittent port flaps → ports stuck at 100 Mbps → dead ports → dead switch. PoE switches add a cruel twist — the same cable carrying data carries 48V power, giving surges a second conductor into the device, and a surge-killed PoE port often takes the connected camera or access point with it.

CCTV systems: cameras at gates and perimeters sit at the end of long outdoor cable runs — perfect lightning antennas. Recorders die through HDMI ports (a surge path everyone forgets: camera → recorder → monitor/TV), through poorly earthed coax shields, and through power. The pattern we see repeatedly: one storm, four dead channels, recorder alive but blind.

Servers and storage: redundant PSUs do not protect against common-mode surges that ride both feeds. Enterprise servers tolerate a lot — then fail all at once: RAID controllers corrupt arrays mid-write (data loss, not just hardware loss), BMCs die silently, and cumulative PSU capacitor damage shortens life by years. A server that "just needs rebooting monthly" is often a server being slowly electrocuted.

PABX and access control: legacy phone lines are copper pairs running kilometres back to the exchange — superb surge collectors. Modern IP phone systems move the risk to the network, which is why voice networks need the same protection discipline as data.

End-user equipment and POS: tills, printers, and desktops die in clusters after restoration events — the classic Monday-morning-helpdesk pattern following a weekend of load shedding.

The Two Deaths: Catastrophic vs Cumulative

Catastrophic failure gets the attention, but cumulative degradation costs more: every sub-lethal transient ages capacitors, weakens insulation, and pits contacts. Equipment doesn't fail on the day of the big storm — it fails six months later, cause unattributable, warranty expired, pattern invisible. This is why "we've never had surge damage" from a site without protection usually means "we've never diagnosed surge damage." When we audit such sites, the failed-PSU drawer tells the real story.

Why Uganda Is Exceptional (in a Bad Way)

Combine the continent's most lightning-active skies with daily load shedding, widespread ageing wiring, floating neutrals, generator dependence, and long overhead runs to gates and outbuildings — no European or American protection guide assumes this combination. Designs copied from temperate-grid countries under-protect here as a rule. Our specifications assume Kampala conditions: restoration transients as the design case, lightning-induced coupling on every outdoor run, and earthing verified by measurement, not by looking at a rod.

Real Protection Is Layered (and a Power Strip Is Not It)

A consumer "surge protector" power strip is a single undersized MOV that sacrifices itself once, silently, then conducts unprotected forever — most have no failure indicator the owner checks. Professional protection layers:

1. Service entrance (Type 1): heavy surge arresters at the incomer, sized for lightning-induced energy, bonded to a measured earth (target single-digit ohms — a rod driven into dry murram without testing is decoration).
2. Distribution (Type 2): SPDs on distribution boards, especially sub-boards feeding racks, CCTV, and gate systems.
3. Point of use (Type 3 + UPS): line-interactive UPS minimum; double-conversion online UPS where uptime matters — it rebuilds the sine wave entirely, erasing incoming transients rather than merely clamping them. See how we size this in power infrastructure design.
4. Data-line protection: PoE surge protectors at building entries, shielded cabling bonded at one end per standards, fibre (non-conductive!) between buildings instead of copper — fibre cannot carry surges, which is reason enough to run it.
5. Earthing and bonding: the foundation everything above stands on. No earth, no protection — SPDs need somewhere to dump kiloamps. Test it; write down the number; retest yearly as soils dry.

Then maintain it: SPDs have status windows/flags and finite lives — inspect after every storm season, replace spent modules, and re-torque the neutral connections that start this whole story. Protection is a system with a maintenance schedule, not a product.

What to Do This Week

Walk your site with this article: find the UPS (is it actually online/double-conversion, or standby?), the earth (tested value?), the gate cable route (overhead copper?), the recorder's HDMI path, and the distribution board feeding the rack (SPD present? indicator green?). Photograph everything. Then book a free assessment — we will measure the earth, map the surge paths, and specify layered protection before the next storm season does the mapping for you. Or talk to an engineer about the failures you are already seeing; the pattern in your dead-equipment drawer almost certainly has a name.