Camera drops off stream intermittently

Likely causes

• Marginal Ethernet link: split pair, damaged pair or poor termination

  most common

  Switch port reports 10 Mbps, or 100 Mbps half duplex, where an identical camera on the same switch shows 100/1000 full. Port error counters climb while the stream is still up.

• PoE voltage sag under load on a long run

  common

  Drop coincides with something that raises draw: IR engaging, heater cycling, PTZ movement. Port draw sits at or near the class limit immediately before the camera disappears.

• IP conflict with a device that is only online part of the day

  less common

  Check ARP repeatedly: the entry for the camera's address flaps between two MACs, or the switch logs a MAC move between ports, or you get a duplicate-address warning. Usually a staged laptop, a printer, or a second camera left on the same address.

• Energy Efficient Ethernet negotiated on the camera port

  Link flaps with no PoE event logged and no error-counter growth, and it stops the moment EEE is disabled on that port. Some camera PHYs handle the low-power idle transition badly.

• Loose field plug, untensioned RJ45, or moisture in an outdoor connector

  Flex the drop at the camera and the link LED blinks or ping drops. Verdigris on the pins or water in the pigtail boot confirms it. Work it as water ingress from there.

• Network path: multicast flap, STP re-convergence, or a saturated uplink

  Several cameras behind the same switch or uplink drop at the same instant. The camera's own web UI streams perfectly while the VMS tile is black.

• Camera firmware defect or memory leak

  Uptime resets with no PoE cycle logged on the switch. The camera's own log records a software restart, not a loss of power.

What to bring

• Certification tester (qualification tester at minimum)
• Phone or laptop with access to the camera web UI
• Known-good stranded factory patch cords
• Known-good camera head
• PoE injector
• Switch management or CLI access
• Ladder or lift

Steps

1. 1

   Step 1: Pull up the camera's own web UI on the same VLAN

   Live view direct from the camera while the VMS shows black means the camera and the drop are both fine and the fault is VMS or network side. Black in both narrows it to camera, power or cable.

   If that doesn’t do it

   Camera unreachable in both: go straight to the switch port.

2. 2

   Step 2: Read the switch port: speed, duplex, error counters, PoE draw

   You want 100/1000 full duplex, flat error counters, and PoE draw with real headroom under both the port limit and the switch total. 10 Mbps, half duplex or climbing CRC/FCS counters means cable. Draw pinned at the class limit means power.

   If that doesn’t do it

   Counters clean and power healthy: clear them and set up a soak.

3. 3

   Step 3: Log a soak that spans a real failure window

   Continuous timestamped ping at a 1400-byte payload written to a file, or poll the port error counters on an interval, running across the hour the customer reports, overnight if that is what it takes. A 10-minute clean soak proves nothing on a fault with a daily period; only a soak covering a known failure window is evidence. ICMP is often rate-limited or deprioritised on cameras and switches, so treat ping loss as a lead and confirm it against the port counters and the stream before you call the link bad.

   If that doesn’t do it

   Soak spans a failure and stays clean: export the VMS event log and check whether other cameras drop at the same moments.

4. 4

   Step 4: Patch the camera to a known-good port with a known-good stranded cord

   Free swap, nothing gets re-terminated. Fault follows the port or the cord, it is the switch or the cord. Fault stays with the drop, it is the permanent link or the camera.

   If that doesn’t do it

   Fault stays with the drop: try the free switch-side settings next.

5. 5

   Step 5: Disable EEE on that port and retest

   One command, instantly reversible, and it costs nothing to rule out. Link flap with no PoE event and no error-counter growth is the signature.

   If that doesn’t do it

   No change with EEE off: check for an address conflict.

6. 6

   Step 6: Re-address the camera to a clean static outside the DHCP scope

   Conflicts stop the instant the camera is on a unique address. With switch access, check ARP repeatedly and look for a MAC move to confirm before you change anything.

   If that doesn’t do it

   Still dropping on a clean address: swap the head.

7. 7

   Step 7: Swap in a known-good camera on the same drop

   Same model and firmware if you have one. Good camera stays up means the original head is faulty: RMA it. New camera drops too means it is the cable path.

   If that doesn’t do it

   New camera also drops: certify the link.

8. 8

   Step 8: Certify the permanent link, do not just wiremap it

   Wiremap and continuity will happily pass a split pair and an over-length run. Certify to the installed category and read the NEXT and return loss margins plus length. A permanent link over 90 m is out of spec even when the tester passes everything else.

   If that doesn’t do it

   Link passes certification: update firmware and open a manufacturer case with the uptime log and port counters attached.

9. 9

   Step 9: Where the run is genuinely over-length, fix the topology

   The answer is a fibre leg with a media converter, or an intermediate switch in a rated enclosure with power and a bonded ground. PoE extenders work, but they consume budget, add latency and cascading them is outside what the standard covers. Price the fibre and put the recommendation in writing.

Check it with a calculator

• Channel & PoE Check length and power

More Video Surveillance guides

• Camera drops or whites out when IR engages
• Camera cycles power in a boot loop
• Motion recording triggering on nothing
• Customer can't view the system remotely
• Water or fog inside the camera
• Faces black against a bright doorway

All 54 guides