Why 100 metres is the wall
The 100-metre channel (90m of fixed cable plus patch leads) is written into the Ethernet standards for twisted pair: beyond it, signal attenuation and crosstalk erode the margin that gigabit and 100M signalling need, and DC resistance in the pairs drops PoE voltage along the run. It is not that a 130m cable never works — it is that it works unguaranteed, temperature-dependent and one re-termination away from failure. Professional designs treat 100m as a hard limit and engineer beyond it deliberately.
Power has its own distance physics. The standard's voltage-drop budget assumes the channel resistance of a compliant run; stretch the run and a 12W camera may still light up while browning out at night when its IR LEDs and heater add load. Distance faults are therefore intermittent, seasonal and maddening — the worst class of fault — which is why the fixes below are products, not improvisations.
Fix one: long-distance PoE (250m at 10M)
Drop the signalling rate and the physics relaxes. ONV's long-distance models run selected ports at 10M and drive power and data to 250 metres over Cat5 — stated in the datasheets ('10M/250 meters transmission') and enabled per model or port group. Ten megabits is useless for a workstation and ample for a camera (a 4K H.265 stream is typically 8–16 Mbps at maximum settings, and most cameras run well under). The fix costs nothing extra — it is a mode of the switch you already bought — and it preserves gigabit uplinks for the aggregated traffic. Its limits are equally clear: 250m ceiling, camera-class bandwidth only.
Fix two: extenders, and fix three: fiber
A PoE extender is an active repeater placed mid-run, powered by the PoE itself: it regenerates data and re-drives power for a further 100m hop, and multi-port models feed a small far-end cluster. Chains of two or three hops are practical; each hop consumes budget and adds a device, so the arithmetic must be done (see the power budget guide). Beyond a few hundred metres, or where the far end needs full bandwidth, fiber is the honest fix: a media converter pair (or an SFP port) makes distance a non-issue — single-mode links in the ONV datasheets run to 40 km at 1310nm — at the price of solving power separately at the far end, via an injector or local supply.
The decision in one line: under 100m, anything; to 250m with camera-class loads, long-distance mode; to ~300m with full rates, extenders; beyond that, or between buildings (lightning alone argues for glass), fiber. Anyone who quotes you a fourth option is selling you a future maintenance contract.
Measuring a run before you believe it
Every distance fix in this guide presumes the run length is known — and in real buildings it usually is not. 'About 120 metres' from a drawing that excluded the riser, the patch-panel detours and the service loops can be 150 metres of actual cable. The reliable methods are unglamorous: a cable certifier or TDR-capable tester reports length electrically; failing that, read the metre markings on the cable jacket at both ends and subtract — installers print a running meter count on the sheath for exactly this reason.
Do the same honesty check on cable material. Copper-clad aluminium (CCA) masquerading as Cat5e has roughly 60% higher resistance than solid copper: it steals signal margin and PoE voltage simultaneously, and it is the hidden culprit behind a large share of 'impossible' long-run faults. Solid copper, certified, is the substrate every figure in this guide assumes. With length and material proven, the three fixes stop being gambles and become what they should be: interchangeable engineering options with known envelopes.
Specifying distance on a real camera schedule
A camera schedule is where distance stops being abstract. Take a logistics yard: 26 cameras, of which 19 sit within 85 metres of the cabinet, five at 130–210 metres along the fence, and two on a gatehouse at 340 metres. The schedule-driven design: standard ports for the 19 (any gigabit PoE switch), a long-distance model or mode for the five fence cameras (10M is ample for their 6 Mbps streams, and the same cabinet switch can carry both groups where the model supports mixed modes), and a fiber pair with far-end power for the gatehouse pair — 340m is past extender-chain prudence and the gatehouse needs a cabinet of its own anyway.
Note what did not happen: no switch was bought 'for distance' as a headline feature, and no run was left to luck. Each run got its fix from its measured length and its device's bitrate, and the schedule — camera, metres, fix — went into the project file where the next engineer will find it. Distance engineering is mostly this: unglamorous measurement, matched to three well-understood fixes, written down.
The decision, compressed
Measure the run, price the three fixes against its true length, and never let a drawing's optimism do the measuring. Standard PoE to 100 metres on certified copper. Long-distance mode to 250 metres where the device is camera-class. Extenders where full rate must survive a few hundred metres and the power arithmetic closes. Fiber everywhere beyond — between buildings by default — with power solved deliberately at the far end instead of hopefully. And whichever fix is chosen, the run, its length and its fix go into the site record: distance faults are only mysterious to estates that did not write their distances down.
One last practical note on commissioning: a long run is not proven by a link light. Commission each extended link under load — camera streaming at full bitrate with IR on, or a load tester drawing the device's rated watts — and leave it soaking while the rest of the install proceeds. Marginal runs pass idle tests and fail at duty; the soak is what separates them, and it costs nothing but the patience to run it before the scaffolding comes down. Record the measured length and the soak result beside the camera schedule entry; the engineer who inherits the site will start from your evidence instead of from a ladder and a suspicion.
There is also a budgeting corollary worth stating once, plainly: distance fixes are cheapest at design time and most expensive at retrofit, in strict proportion to how finished the building is. A fiber pair pulled during construction costs cable and minutes; the same pair pulled after handover costs containment, access equipment, permits and a night shift. If a route might ever need to exceed 100 metres — a future gate, a planned extension, a camera position marked 'phase two' — the time to solve its distance is while the trays are open and the scaffolding is up, not when the phase-two camera is already on order.