Camera cycles power in a boot loop
very common
“Camera comes up, gives me video for about thirty seconds, then the light goes out and it starts over. Again and again. It's not the camera. I already swapped it.”
Likely causes
-
Switch PoE budget exhausted
most commonMore than one camera is cycling and the ones added most recently suffer worst. The switch reports total PoE consumed at or above budget, or logs power-denied / over-budget events. Unplug two other cameras and the loop stops. That is the diagnosis.
-
PoE type mismatch: camera wants Type 3/4, PSE offers Type 1/2
commonThe camera boots far enough to serve a web page, then dies as soon as the whole board loads. Switch reports a lower class than the data sheet requires. Put it on an 802.3bt injector and it runs clean.
-
Power negotiation not completing: LLDP disabled on the port, or the vendor's high-power mode not enabled
commonLooks almost identical to a type mismatch and has a completely different fix. The switch is bt-capable and the camera is bt, yet the port reports an allocation well below the data sheet requirement and the camera loops under load. Several vendors only allocate above the hardware-classification result after a successful LLDP power exchange. The distinguishing test: allocated power rises the moment LLDP is enabled with nothing else changed. Feature naming, and whether LLDP is mandatory for high power at all, vary by vendor. Check the switch documentation.
-
Excessive DC loop resistance: channel too long or too many couplers
Only the longest drops on the job do it; short drops on the same switch are rock solid. Measure length: permanent link over 90 m, or a drop plus long patch cords blowing the 100 m channel.
-
Bundle heat derating: PoE-loaded cable confined in a sleeve, tray or conduit
The failing runs are in the middle of the bundle; loose drops on the same switch are fine. It worsens as the space warms and again under full night IR load. Elevated conductor temperature raises resistance and degrades transmission margin at the same time.
-
Faulty PD power stage in the camera, or a wet connector shorting a pair
The loop follows the head to a different port, or the switch logs a short or overcurrent fault. Water in an outdoor RJ45 usually presents as the port shutting down with an overcurrent error rather than a clean reboot.
-
Non-standard injector, passive midspan or PoE splitter in the path
A no-name injector or a passive 24 V brick somewhere in the ceiling or the pole base. Passive gear performs no negotiation at all, so an 802.3at or bt camera browns out under load. Look in the pathway, not the rack.
-
Switch firmware bug or a locked-up PoE controller on one port
The loop stops on a different port of the same switch, and the original port stays dead for every device you try in it.
-
Ground potential difference on a pole or rooftop camera
Measurable AC potential between camera chassis and cable shield, or the loop only occurs with the shield landed. Common where the pole is fed from a separate service.
What to bring
- Switch CLI or management with PoE event logging
- Switch documentation for LLDP and high-power configuration
- 802.3at and 802.3bt compliant injectors
- Certification tester with DC loop resistance test
- Multimeter
- Thermal imager or temperature probe for bundle checks
- Camera data sheets by exact part number
- Ladder or lift
Safety
Bonding checks on a pole or rooftop camera can put you across a genuine potential difference between two services. Meter before you touch the shield, and do not lift a bond to clear a ground loop. That is a code violation and a shock hazard for whoever comes next, not a fix. A bundle that has been carrying full load can be hot enough to be unpleasant, and tray work is at height alongside power.
Steps
-
Step 1: Read the switch PoE totals and the per-port event log
You want total consumed with real headroom and no power-denied or over-budget entries. Add up the night-load figures from the data sheets rather than the idle draw: a switch that balances at noon will shed ports at dusk.
If that doesn’t do it
Budget clean: check negotiated type and class.
-
Step 2: Compare the negotiated PoE type and class against the data sheet
Pull the exact part number and look up its maximum draw and required PoE type. These figures differ significantly between brands, and between the IR and non-IR variants of one model number, so look it up rather than working from the last job.
If that doesn’t do it
Type and class are correct: check the power negotiation itself.
-
Step 3: Enable LLDP, or the vendor's high-power mode, on that port
Free, instantly reversible, and it is the fix that gets missed because the symptom looks like a hardware mismatch. If allocated power rises with nothing else changed, you have your answer and the switch was never the problem.
If that doesn’t do it
Allocation unchanged or already correct: go to a local injector.
-
Step 4: Run the camera from a standards-compliant injector at the camera
Disconnect the drop from the switch port, or disable PoE on that port, BEFORE connecting the injector. Two PSEs on one channel risks the injector, the switch PoE controller and the camera's PD front end, and it invalidates the test because the camera may still be drawing from the switch. Short cord, injector rated for that camera's type, drop out of circuit. Stable means the delivery path. Still looping means the head.
If that doesn’t do it
Still looping on a good injector: swap the head.
-
Step 5: Swap the head
Known-good, same model. A clean run means RMA the original. Still looping means the path, and you have now proven it.
If that doesn’t do it
Path confirmed: walk the run.
-
Step 6: Physically walk the run for unauthorised injectors, splitters and couplers
Above ceiling, in the soffit, in the pole base. Every passive injector, PoE splitter, no-name midspan and inline coupler comes out. Couplers eat margin and are a routine cause of failures that only appear under night load.
If that doesn’t do it
Run is clean: count the bundle while you are up there.
-
Step 7: Count the loaded cables in the bundle and check the ampacity requirement
Large bundles of PoE-loaded cable in a sleeve, tray or conduit run hotter than the same cable run loose. Conductor gauge, bundle count and ambient temperature all bear on permitted current, and the temperature rise degrades transmission margin as well as being a code matter. Check the counts against NEC 725.144 for the edition your AHJ enforces. It was renumbered under the new Article 722 in the 2023 edition.
If that doesn’t do it
Bundles within limits: certify the link.
-
Step 8: Certify the permanent link, including DC resistance if your tester does it
90 m permanent link, 100 m channel. Check DC loop resistance and resistance unbalance where the tester supports it: PoE faults show up there well before they show up in NEXT.
If that doesn’t do it
Link passes: check bonding and for wet connections.
-
Step 9: Meter for ground potential and open the housing
Check chassis-to-shield potential before you touch anything. Anything measurable gets resolved properly at the bonding, never by lifting a ground. Then open the housing and look for water staining, corroded pins, or a field plug packed with dirt.
If that doesn’t do it
All clean: move the camera to a different switch and log a manufacturer case with the port logs and certification results.
References
- IEEE 802.3-2022 Clause 33
- IEEE 802.3-2022 Clause 145
- IEEE 802.3-2022 Clause 79 (Power via MDI TLV)
- IEEE 802.1AB (LLDP)
- ANSI/TIA-568.0-D: generic premises cabling, horizontal distance limits (confirm current revision)
- NFPA 70 (NEC) Section 725.144: ampacity of conductors carrying power and data; renumbered under Article 722 in the 2023 edition, confirm the edition your AHJ enforces
- ANSI/TIA-607-D: bonding and grounding (confirm current revision)