PoE in Surveillance and NVR Systems

A PoE NVR is a network video recorder with a built-in Power over Ethernet switch, so each IP camera plugs straight into the recorder and takes its power, its network connection and its recording path from one cable. It removes the separate PoE switch, the power supply at every camera position and much of the network set-up that a standard NVR needs.

PoE is the backbone of modern IP surveillance. Running a separate power cable to each camera is often the most awkward part of an install, and PoE removes it. The footage is aggregated and recorded centrally, either on a dedicated network video recorder or on an industrial PoE computer that powers the cameras, records them and analyses the video in the same box.

This guide compares a PoE NVR with a standard NVR, a DVR and a PoE edge computer, shows how to size the PoE budget, the bandwidth and the storage with worked examples, and covers the extra demands of outdoor, vehicle and rail sites. It ends with a checklist, examples from our range and answers to the questions buyers ask most.

The recorder choice comes down to what the cameras are, how they are powered and what you want the recorder to do with the video. Our FAQs on what an NVR is and NVR vs DVR cover the basics; the table puts the four options side by side.

A PoE NVR suits a self-contained system of a few cameras. Once the system has to survive a hot cabinet or a vehicle, run analytics, or fit into a managed site network, an industrial recorder or a PoE edge computer is the better fit for the installation.

A regular NVR has one or two network ports. The cameras connect to a separate PoE switch, or each takes power from a PoE injector, and the NVR records their streams across the network. A PoE NVR builds the switch into the recorder: the cameras plug into its rear ports, which supply power and often put the cameras on a private subnet that the recorder manages, so they are found and added with little configuration.

In industrial systems the second pattern is more common, because cameras are spread across a site and the network is shared. A managed switch then keeps camera traffic in its own VLAN and uses IGMP snooping so multicast streams reach only the recorders and viewers that want them. Our guide to managed vs unmanaged switches covers that choice.

The cameras are the powered devices and the recorder, switch or card is the power source; our FAQ on PoE PD vs PSE explains the two roles. The source is either a PoE switch feeding a recorder, or a PoE computer with built-in PoE ports that powers the cameras directly and records them, removing the separate switch. Which you choose depends on camera count, the power each camera needs, and whether you also want analytics running on the same box. Where a computer has too few PoE ports, a PoE card in a PCIe slot adds four more; our Ethernet and PoE frame grabbers include cards with PoE+ and PoE++ ports.

As many as its PoE budget allows, which is often fewer than it has ports. Each port has a per-port limit set by the PoE standard, and the recorder or switch has a total budget shared across all of them. Fixed cameras draw little; PTZ cameras, cameras with infrared illuminators and cameras with heaters draw much more, and their peak draw is what counts.

Our FAQ on PoE vs PoE+ vs PoE++ explains the standards, and how much power PoE delivers covers the voltages. Before settling on a recorder, work out the PoE power budget for every camera at peak load, since a source’s total budget is often lower than its per-port figure suggests.

Worked example: sizing the PoE budget

A depot yard has six fixed cameras with IR illuminators and two PTZ cameras with heaters. The figures are typical maximums from camera datasheets; always use the actual cameras’ peak figures.

A recorder with eight 802.3at ports and a 120 W budget would fail this site twice: the PTZ ports cannot supply 60 W, and the total is short. The PTZ cameras would either move to a PoE++ switch or card, or the recorder would be chosen for an 802.3bt budget. Cable length matters too: PoE and Ethernet both reach 100 m per run, and the power at the far end allows for the loss in that length of cable.

The number of channels, their resolution and frame rate, the codec and how long you keep footage drive the storage and the recorder. H.265 typically needs about half the bitrate of H.264 for similar quality, which halves both the network load and the disk; our FAQ on H.264 vs H.265 explains the trade-off, and H.266 is the next step on. Bitrates below are typical for continuous recording of a busy scene; quiet scenes and motion-only recording use much less.

Worked example: 16 cameras for 30 days

Sixteen 4 MP cameras record continuously in H.265 at an average of 6 Mbit/s each, which allows for busy periods. The total stream into the recorder is 96 Mbit/s. Each camera writes about 65 GB a day, so the system writes about 1.04 TB a day and about 31 TB over 30 days. With RAID 1 mirroring, that becomes about 62 TB of raw disk; with RAID 5 across four drives, about 41 TB.

Size the storage for the retention you actually need, not the maximum the cameras can produce, and check the recorder’s incoming bandwidth limit as well as its disk capacity. A surveillance computer often carries several drives, sometimes in RAID for redundancy, and removable or front-access trays make swapping a drive or collecting footage easier. Retention periods are set by your own policy and the regulations that apply to the site.

Where the equipment lives shapes the build. An indoor recorder in a comms room has an easy life; a roadside cabinet, a depot or a vehicle does not. For those, a fanless, wide-temperature computer survives the environment, and the cameras themselves need the right ingress rating. The recorder can sit centrally or out at the edge near the cameras, depending on bandwidth and resilience.

On vehicles, M12 connectors replace RJ45 so PoE links stay connected under vibration. Our guides to EN 50155 and E-mark certification set out the approvals, and our railway computers and in-vehicle computers include PoE models.

Recording is only half of what modern surveillance does. A PoE-powered computer with an NPU or GPU can run analytics on the streams it records, for example licence-plate recognition, people counting or intrusion detection, without sending video to the cloud. These are edge AI computers: doing the inference on site cuts bandwidth and keeps the footage local. Analytics also change the recording: many systems keep full-rate video only around detected events and store metadata for the rest.

Interoperability matters when the recorder and cameras come from different makers. ONVIF profiles define how cameras stream video (Profile S, and Profile T for H.265 and advanced streaming), record at the edge (Profile G) and share analytics metadata (Profile M). PTZ cameras add control commands that the recording software must support. Video management software then sits on top, and we check the chosen software against the computer and operating system when we build the system.

Our case studies show both sides: AI-enabled NVRs monitoring station activity, and PoE computers running ALPR for automated barrier control.

For straightforward recording, an industrial NVR captures and stores the streams. Where you want analytics as well, a PoE computer can power the cameras, record and run the inference in one box. See our guide to industrial Power over Ethernet for the power side, and our industrial PoE switches where the cameras are spread across a site.

We configure, image and test the computers and recorders we supply, with the PoE budget checked against the cameras. Antaira designs and builds its switches; we supply and specify them and pre-configure IP addressing and VLANs where you ask.