Blog · Field Notes

PoE Injector vs Splitter: Two Ends of the Same Idea

Injectors add power to a link; splitters take it back out. How the pair works, the classic combined design, and the mistakes that come from confusing the two directions.

The injector and the splitter are the most confused pair in the PoE accessory world, and the confusion is understandable: they are mirror images. An injector takes a data-only Ethernet link and adds power to it. A splitter takes a powered Ethernet link and separates power back out, handing a device clean data on one port and DC power on another. Same cable plant, opposite directions. Most real designs use them together, and most ordering mistakes come from asking for one when the design needed the other — or from not knowing the pair pattern exists at all.

The injector's job, precisely

An injector (a PSE, in the standard's language) sits in a link and performs a negotiation: it detects whether the device at the far end is a legitimate powered device, classifies how much power it may draw, and only then energises the cable, at the class agreed. Its input side is ordinary Ethernet from any switch — PoE or not — plus a power feed (mains-derived in commercial units, wide-range DC in the industrial families). Its output is one powered link, indistinguishable to the device from a PoE switch port. Injectors are how a single camera, access point or phone gets standards power where the switch has none to give, at classes from 30W 802.3at up to the 60–100W 802.3bt shelf.

The splitter's job, precisely

A splitter serves the device that cannot negotiate at all: the legacy camera, the single-board computer, the access controller with a barrel jack and an Ethernet port. It presents itself to the PSE as a proper powered device, draws its class, then converts: data passes through untouched, power comes out as a clean DC rail at the device's voltage — the documented splitter families print their outputs in their names, from a 5V 2A Micro USB model for the small-computer world to 12V-class units for cameras, and an industrial 802.3bt splitter whose outputs span 12/19/24/36V for cabinet devices. The non-negotiable discipline of the splitter is voltage: its output must match the device's requirement exactly, because the device has no negotiation with which to protect itself.

The pair pattern: injector and splitter as one design

Put the two together and a useful pattern appears: power can be transported over the network plant to a device that speaks neither PoE nor the local power language. An injector at the rack energises a long link; a splitter at the far end converts that link into the device's native diet. The link between them is standard, monitored, backed-up network power — while the device sees only its familiar barrel jack. This pattern rescues devices stranded by every partial upgrade: the switch went PoE but the device predates it, or the device is fine but its wall-wart power was the failure point everyone was tired of.

The mistakes, enumerated

  • Ordering a splitter to 'add PoE' to a link: a splitter only separates what is already there. No injector or PoE port upstream means no power for the splitter to split.
  • Ordering an injector to feed a non-PoE device directly: the injector will (correctly) refuse to energise a device that does not negotiate. The device needs a splitter at its end.
  • Voltage guessing at the splitter output: 12V into a 5V device is destruction, and 5V into a 12V device is a haunting — intermittent, temperature-dependent, blamed on the network. The device's required voltage and connector belong in the inquiry, verbatim from its label.
  • Forgetting the negotiation at the far end of a long design: bt-class transport needs a bt-class PSE. An at injector cannot feed a bt splitter's high-power outputs, however impressive the splitter is.

Two neighbours that are neither

Extenders are the third device in this family portrait: they sit mid-link, consume a little of the passing PoE, and re-present data and power for a further channel — the answer to distance, not to direction. And media converters solve a different axis entirely (copper to fiber), though the families interbreed usefully: an industrial PoE media converter powers a local camera while converting its uplink, which is injector logic wearing converter clothing. Knowing which axis your problem lives on — power direction, distance, or medium — is the whole taxonomy.

The summary fits on one line: injectors put power in, splitters take it out, and the interesting designs usually need both ends of the idea. Specify the device, its voltage, its class and its distance, and the pair selects itself.

Worked micro-designs

Three sketches make the taxonomy stick. First: a warehouse adds one Wi-Fi access point above a racking aisle; the nearest switch is not PoE. One injector at the switch, the AP's own negotiation does the rest — injector alone. Second: a school corridor camera from 2011, healthy, 12V barrel input, in a building whose switches are now all PoE. One splitter at the camera, set to its documented 12V output, fed from the switch's PoE port — splitter alone, and the camera never learns the network changed around it. Third: a gate intercom at 90 metres, 5V USB-powered, no local supply worth trusting. A bt-class injector at the head end energises the run; an active gigabit splitter at the gate delivers data plus 5V on Micro USB — the pair pattern, each device doing the half the other cannot. Every real inquiry is one of these three sketches with the nouns changed, which is why naming the device, its voltage and its distance in the first message collapses the whole correspondence into a confirmation.

Safety and the standards, briefly

It is worth understanding why the negotiation theatre exists at all. A PoE link can carry up to roughly 100 watts at voltages that deserve respect, down cabling installed by many hands over many years, to devices of every provenance. The detection and classification ritual — the injector's refusal to energise a non-negotiating device, the splitter's proper presentation of itself as a powered device — is what makes that power safe to leave in walls. Passive workarounds that bypass negotiation defeat the ritual for a small saving and inherit the liability the ritual was designed to retire. Professional practice in this corner is short: active, standards-based devices at both ends (the catalogue's splitters and injectors are specified on that basis), voltages matched exactly, classes confirmed in writing. The theatre is the safety case; attend it.

Procurement shorthand, to close: if the sentence describing your problem contains 'the switch has no PoE', the answer involves an injector. If it contains 'the device has no PoE', the answer involves a splitter. If it contains both — a non-PoE switch and a non-PoE device, with a long powered road between them — it contains the pair, and the middle of that road is where extenders join if distance demands. Three sentences that between them specify most of the accessory orders ever placed.

λ.02

Keep reading

Blog

Video over Fiber: A Working Guide to Putting Pictures on Light

What video over fiber is, which signals it carries, how a link is built from transmitter, fiber and receiver, and the specification questions that decide whether a link works first time.

Blog

Choosing a PoE Injector for a PTZ Camera: Power Class, Peaks and Placement

PTZ cameras are the hardest common load on PoE: motors, heaters and IR push them past PoE+ into 802.3bt. How to size an injector by worst-case draw, input power and rate.

Blog

Single-Mode to Multimode Conversion: Bridging the Two Fiber Worlds

Why buildings are multimode and campuses are single-mode, what an OEO mode converter actually does, and how to specify the bridge between a riser plant and a spine plant.

Confirm the exact configuration before payment

Send the model name, the distance, the fiber type and the quantity. We will confirm the matching transmitter/receiver pairing, the optical module and the full specification of the exact model in writing before you pay.