Device reboots when it draws peak power
very common
“Camera's fine all day and drops out every night. Comes back when the sun comes up. Or it reboots the second the PTZ moves.”
Likely causes
-
PD peak draw exceeds what the port allocated: IR illuminators, heaters, PTZ motors, speakers
35%Failures line up with the load, not the clock: dusk for IR, cold mornings for the heater, a preset tour for PTZ, a page for speakers. Switch shows the draw climbing to the allocated ceiling in the seconds before the drop, or logs an overload on that port.
-
Switch out of budget and shedding low-priority ports when total draw peaks
20%Several PDs drop together and always the same ones. Used power sits near the switch's budget at the moment of the drop, and unplugging an unrelated PD elsewhere keeps them all up.
-
Voltage drop on a long or high-resistance run, so the PD undervolts only at peak current
20%Voltage at the PD sags under load and recovers at idle; that idle-recovers behaviour is the tell, not any particular number. Whether the sag matters depends on the class the port detected: a Class 4 PD is out of spec below about 42.5 V, Class 5-8 below about 41 V, Class 0-3 below about 37 V. Long run, 28 AWG cords, or both. The same PD on a short cord at the switch never misbehaves.
-
Allocation capped by class or a missed LLDP negotiation
10%The PD reports low power mode before it reboots, and the allocated figure is well below its datasheet peak in PD watts. This is the class mismatch fault wearing a different hat.
-
PD firmware watchdog or a genuinely faulty PD
10%It reboots the same way on a lab injector with plenty of headroom, and the PD's own log shows a firmware or watchdog event rather than a loss of power.
-
PSE overcurrent trip on inrush, or a PoE surge protector or extender in the path
5%The port logs overcurrent at the exact instant of the load step. Pull the extender or SPD out of the path and it stops. Check whether that hardware is on any drawing; often it is not.
What to bring
- Switch CLI with inline power logging
- PoE loaded tester
- Inline power meter or multimeter at the PD
- PD datasheet peak figures in PD watts
- Known-good compliant injector
- Ladder or lift for the PD location
Safety
Most of these faults only show at dusk or in the cold, which is exactly when you should not be improvising at height. Set the ladder or lift properly for night work, light the base, and do not work off a roof edge or near mast-mounted antennas without confirming RF transmitters are down and the fall protection is rated. Cold, wet steel and a camera that needs two hands is how people come off ladders.
Steps
-
Step 1: Ask what is different when it fails
Dusk means IR. Cold mornings mean the heater. Patrols mean PTZ. A page means the speaker. Nothing changing at all points at the switch's total budget or at firmware. This one question saves an hour of metering.
If that doesn’t do it
Log the port's power draw across a failure
-
Step 2: Log the port's inline power draw through a full fail cycle
Draw rising to the allocated ceiling and then going to zero means the allocation is too small. Draw dropping away with no ceiling hit means the link or the PD. Force the load if you can rather than waiting for dusk: cover the sensor, trigger the tour, run the heater.
If that doesn’t do it
Compare switch budget and port priority
-
Step 3: Compare total PoE used against the switch's budget and note the port's priority
If used power is within one port's worth of the budget, the switch will shed the lowest priority ports first and it will look like a device fault. Budget, shedding behaviour and any thermal derating of the budget vary by model and PSU; check the datasheet for that exact SKU.
If that doesn’t do it
Go to Switch out of PoE budget
-
Step 4: Note the class the port detected, then measure voltage at the PD while it is drawing peak
The minimum is class-dependent and the measurement means nothing without it: about 42.5 V for Class 4, about 41 V for Class 5-8, roughly 37 V for Class 0-3; confirm against the clause. These symptoms are mostly PTZ cameras, heaters and IR, which are Class 4 and above, so do not read 42 V as comfortable headroom. It is the cliff edge. Read the PSE's own reported output and subtract rather than assuming the source voltage.
If that doesn’t do it
Feed the PD from a known-good injector
-
Step 5: Feed the PD from a known-good compliant injector on a short cord and exercise the load again
Rock steady means the run or the switch. Still rebooting means the PD or its firmware: check its version and its own event log before you condemn the cabling.
If that doesn’t do it
Fix the allocation or the plant
-
Step 6: Fix it at the source: raise the port Type, enable LLDP, re-prioritise the port, or add PSU or midspan capacity
Give the PD headroom above its datasheet peak, not a number equal to it: a camera allocated exactly its peak will reboot the first cold night. If you add a midspan, disable PoE on the upstream switch port first and verify it reads administratively off, not just not powering: two PSEs on one channel is outside 802.3 and is a known way to kill a midspan, a port or the PD. If a midspan ever comes back out, re-enable the switch port or you have built the next callback.
If that doesn’t do it
Heavier gauge run, a closer power source, or fibre plus local power
References
- IEEE 802.3-2022 Clause 33
- IEEE 802.3-2022 Clause 145
- ANSI/TIA TSB-184-A