802.3af, 802.3at and 802.3bt are usually presented as a power ladder — 15 W, 30 W, 60 W, 90 W — and that framing is accurate but incomplete. The three amendments differ in voltage window, current limit, number of pairs carrying power, and classification scheme. The last two are what decide whether a camera works reliably at the far end of a long cable.
For a system integrator the useful question is not which standard is larger, but what the camera actually sees when the link is negotiated at each level.
What actually changed between the three amendments?
802.3af, published in 2003, defined Type 1: 15.4 W at the PSE, 12.95 W at the PD, 350 mA maximum over two pairs, PSE output between 44 V and 57 V, and a PD that had to keep working down to 37 V. Detection was a resistive signature plus a single classification event.
802.3at, in 2009, added Type 2: 30 W at the PSE, 25.5 W at the PD, 600 mA, still two pairs, and a tighter window — PSE between 50 V and 57 V, PD input minimum 42.5 V. It also added two-event classification, so a Type 2 PSE can tell a Class 4 PD from a Class 0 one, and LLDP-based negotiation for finer granularity during operation.
802.3bt, in 2018, added Type 3 and Type 4 on four pairs. Type 3 delivers 60 W at the PSE and 51 W at the PD with 600 mA per pair; Type 4 delivers 90 W and 71.3 W with 960 mA per pair. Classes 5 through 8 were added (45/40 W, 60/51 W, 75/62 W, 90/71.3 W), the PSE minimum output rose to 50 V for Type 3 and 52 V for Type 4, and Autoclass was introduced so the PSE can measure real consumption and release the unused allocation.
| 802.3af | 802.3at | 802.3bt Type 3 | 802.3bt Type 4 | |
|---|---|---|---|---|
| PSE / PD power | 15.4 / 12.95 W | 30 / 25.5 W | 60 / 51 W | 90 / 71.3 W |
| Max current | 350 mA | 600 mA | 600 mA per pair | 960 mA per pair |
| Pairs carrying power | 2 | 2 | 4 | 4 |
| PSE voltage | 44–57 V | 50–57 V | 50–57 V | 52–57 V |
| PD minimum input | 37 V | 42.5 V | 42.5 V | 41.1 V |
| Classes | 0–3 | 0–4 | 0–6 | 0–8 |
What does that mean at the camera end?
Datasheets quote watts. Volts decide whether the link survives. Every PD has a minimum input voltage, and the higher the class the higher that minimum, because the standard assumes a PD drawing more current is paired with a PSE and a cable able to hold the voltage up.
Long runs are where this bites. On 100 m of Cat5e at 600 mA, the pair set drops about 5.6 V; at 300 m, about 17 V. An 802.3at camera fed from a PSE regulating at 50 V over 300 m sees roughly 33 V, which is below the 42.5 V minimum and below even the 37 V af floor. The camera may not refuse power outright — many will boot, run for a few seconds, then reset when the zoom motor engages. The symptom reads as a firmware fault; it is a voltage budget fault.
A camera running under-powered does one of four things: it refuses to start, it boots with IR disabled, it reboots cyclically at dusk, or it browns out only during PTZ movement. All four get reported as "camera problems", and all four are resolved by measuring voltage at the camera rather than at the switch.
Why do Type 3 and Type 4 need all four pairs?
Because the limit a single pair can carry is current, not power. A twisted pair terminated in a standard RJ45 plug and drawn through Category cable sits around 600 mA continuous with acceptable heating at normal ambient; Type 4 pushes that to 960 mA and accepts the thermal consequences. To move 90 W through a connector limited that way, every conductor has to be used: two pairs for the positive rail, two for the negative.
The cable then carries more loss than most people expect. On 100 m of Cat5e 24AWG the four-pair path has about 4.7 Ω of loop resistance; at 1.92 A total that is close to 17 W dissipated in the cable — inside the standard's 18.7 W allowance, and enough to warm a bundled conduit noticeably. On Type 4 links, Cat6 with 23AWG conductors or Cat6A is the sensible choice — not for bandwidth, but for copper cross-section — and bundled runs should be derated for ambient temperature.
What happens when a PSE and a PD do not match?
Most combinations work, because the standard is deliberately backward compatible: a Type 4 PSE powers an af camera at af levels, and a Type 2 PSE powers an af camera at af levels. The failures cluster in three places.
An af PSE with an at camera is the classic. A 350 mA limit and a 15.4 W ceiling cannot feed a 25.5 W PD, and an af PSE running one-event classification reads the camera's Class 4 request as Class 0. The result is a camera that powers up and then drops when it draws its real load — typically at dusk when the illuminator engages.
An at PSE with a Type 4 PD is subtler. A single-signature bt device may run at reduced power, which on a PTZ means no heater or no wiper. A dual-signature device, effectively two PDs sharing one cable and one per pair set, sees power on only two of its four pairs and may not start at all.
A passive injector with a non-PoE device is the damaging case. There is no detection and no classification: 48 V is applied to the pairs continuously, and a device with an unprotected PHY or a non-isolated design can be destroyed. Passive injectors belong on devices known to be PoE-compatible, with the correct polarity documented.
How do you confirm which type a link is running?
On a managed switch, the port detail page reports the negotiated class, the allocated power and the delivered power; on many units it also reports the detected type. LLDP Power via MDI carries the same information to the endpoint. Where the switch is unmanaged, an inline PoE tester between switch and camera shows voltage, current and class, and it is the only way to see what the camera receives rather than what the switch claims to send.
Measure at night with the illuminator on, and during a pan cycle for PTZ. A camera that reads 9 W at noon and 13 W at midnight passes any daytime check and still fails in service. Record voltage at the camera as well as wattage: 44 V and 12 W is a healthy af link, 36 V and 11 W is a link that will not survive a cold morning.
FAQ
Can an 802.3bt PSE power an 802.3af camera?
Yes. A Type 3 or Type 4 PSE detects the af signature, classifies the camera at its own class and supplies power within af limits. The camera never receives more than it asked for, and the unused PSE capacity stays available to other ports under dynamic budgeting.
Is Cat5e acceptable for 802.3bt?
For Type 3 at 600 mA per pair, Cat5e is workable at full length, though the voltage drop should be checked on long runs. For Type 4 at 960 mA per pair, Cat6 or Cat6A is the sound choice, and long runs in hot bundles should be evaluated for both loss and heating.
Does 802.3at guarantee 25.5 W at any cable length?
No. 25.5 W is the figure at the PD when channel loss stays inside the standard's allowance. Beyond 100 m, or with cable whose resistance is higher than assumed, the PD receives less and may fall below its minimum input voltage.
Why does the same camera reboot every evening?
The usual cause is that total draw rises when the IR illuminator engages and pushes the port or the supply past its limit. On a long run it can also be voltage: the extra IR current increases the drop enough to cross the PD's minimum input threshold.
Do splitters work with 802.3bt?
Splitters exist for Type 3 and Type 4 inputs and output 12 V, 24 V or 48 V at higher current than af/at types. They suit fixed installations where the device has no PoE front end, but the conversion loss, typically 10–12%, comes out of the port budget before the device sees anything.
