The phone call usually starts the same way: “The Wi-Fi was down again last night. The kids woke up and the cameras went offline around 3 AM. By the time I got out of bed it was already back.” Sometimes the homeowner has already power-cycled the router; sometimes the controller dashboard shows two APs that disconnected and reconnected within a few seconds of each other. Almost every time, the cause lives somewhere in the chain between the switch port and the access point — not in the AP itself, and not in the cloud.
This is the diagnostic post we wish more homeowners had read before replacing perfectly good APs. It covers PoE budget math, midspan injectors, the 802.3af / at / bt distinctions that actually matter, and the loose-pair problem that quietly takes down Wasatch Front installs every winter.
What “the AP rebooted” usually means
Access points don’t spontaneously reboot for fun. A modern UniFi, Aruba, or Ruckus AP that resets in the middle of the night is almost always doing one of three things:
- Brown-out. The PoE source momentarily can’t deliver enough power. The AP under-volts, drops the radios, and reboots when the supply recovers. This is the most common cause we see, and it almost always traces back to PoE budget or a marginal injector.
- Link renegotiation. A bad pair in the run, a flaky keystone, or a slightly oxidized RJ45 contact causes the link to drop from 1 Gbps to 100 Mbps to nothing and back. The AP looks like it’s “rebooting” on the dashboard, but it’s actually the link layer flapping. Same symptom, totally different fix.
- Firmware watchdog. A specific firmware bug causes the AP to panic and reset when an obscure code path fires — often during a 2 AM cron-driven channel scan or an overnight client-list cleanup. Less common, and usually fixed by a firmware update once you figure out it’s the cause.
Telling these three apart is most of the battle. The rest of this post is the field checklist we walk through on every “Wi-Fi was down again” ticket.
PoE budget math: the part everybody skips
Every PoE switch has two numbers on its spec sheet that matter: per-port budget and total switch budget. A UniFi USW-Pro-24-PoE, for example, gives you up to 30 W on any individual 802.3at port, but the whole switch only delivers 400 W total across all 24 PoE ports. If you’ve got 16 cameras, eight APs, and a handful of PoE doorbells and intercoms, you can easily exceed that 400 W ceiling under load even though no individual port is over budget.
The standards stack:
- 802.3af (PoE). Up to 15.4 W at the source, ~12.95 W at the device. Fine for older APs and most basic IP cameras. Not enough for a Wi-Fi 7 AP under load.
- 802.3at (PoE+). Up to 30 W at the source, ~25.5 W at the device. The current minimum for a UniFi U6-Pro or U7-Pro. Still tight for U7 Pro Max.
- 802.3bt Type 3 (PoE++). Up to 60 W at the source. Required for high-end Wi-Fi 7 APs with multi-gig Ethernet, PTZ cameras, and any AP doing simultaneous tri-band radios at full power.
- 802.3bt Type 4. Up to 90 W at the source. You see this on managed PoE++ switches feeding outdoor PTZ cameras with built- in heaters, or large displays.
The trap: a U7 Pro Max plugged into an 802.3at port will boot, associate clients, and look fine on the dashboard most of the time. Then, the moment a handful of clients all hit the AP at once and all three radios are pushing traffic, the power draw spikes past what 802.3at can deliver, and the AP browns out and reboots. The dashboard reports it as “disconnected,” not “under- powered,” so the homeowner blames Wi-Fi.
Our PoE explained primer covers the standards in more depth. The short version: when in doubt on a Wi-Fi 7 install, plan on PoE++ at the switch, not PoE+.
Midspan injectors: where chains go to die
A midspan injector is the little brick-shaped power supply that sits between a non-PoE switch and a PoE device. They look harmless. They are not.
Common things we see go wrong with injectors on Utah installs:
- Wrong standard, right voltage. A bargain-bin “PoE injector” from Amazon outputs 48 V at 0.5 A — which is 24 W, fine for a U6-Lite, dead for a U7. The AP boots, runs at idle, and reboots the moment a client hits it. The injector spec sheet says “PoE” but doesn’t commit to af vs at vs bt.
- Passive injectors on the wrong gear. Some old UniFi APs shipped with 24 V passive injectors. Plugging a modern 802.3af AP into a 24 V passive injector damages the AP. Plugging an old 24 V AP into a real 802.3af switch port just doesn’t work. Mix-ups happen during parts swaps and we’ve replaced more than one fried U6-IW because of it.
- Heat. An injector tucked behind a TV cabinet or stuffed in a media box will run 50 °C+ in the summer. Capacitors age fast at that temperature. The injector starts sagging voltage under load after a year or two, and the AP downstream starts rebooting at peak times. The same injector tested cold on a bench reads fine, which makes the failure infuriating to diagnose.
- Daisy-chained injectors on a long run. We’ve seen homeowners run an injector at the rack, then another in the attic, then a third near the AP, all 24 V passive, all different brands, in a desperate attempt to feed a backyard pavilion AP on a 280 ft cable run. The voltage drop is brutal, the impedance mismatches cause weird reflections, and the AP reboots every few hours.
Our recommendation: if you have a midspan injector anywhere in your install, replace it with a real PoE switch port. A managed PoE switch pays for itself in injectors-not-bought after about six APs and gives you per-port power monitoring you cannot get from a brick.
The loose-pair problem
Ethernet has eight wires arranged as four twisted pairs. 1 Gbps Ethernet uses all four pairs. Multi-gig Ethernet (2.5 Gbps, 5 Gbps, 10 Gbps) uses all four pairs with extremely tight signal-integrity requirements. PoE is delivered on all four pairs in the modern (4PPoE) implementations used for 802.3bt.
If even one of those eight wires has a marginal contact — a strand cut by a punch-down tool, a half-seated keystone, an RJ45 pin that didn’t quite reach the conductor — the link will work, mostly. It will negotiate at 100 Mbps when it should be at 1 Gbps. Or it will auto-negotiate up and down, up and down, all night. Or PoE current on the affected pair will spike and dip as the contact makes and breaks under thermal expansion.
This is a particularly Utah-flavored failure, because cable runs in attics and crawlspaces here see a 90+ °F summer-to-winter temperature swing. A marginal connector that’s fine at 70 °F in July creeps loose by November, and you get 3 AM disconnects starting around Thanksgiving. This is the same physics that causes outdoor APs to struggle in Wasatch Front winters — we covered it from the AP side in the outdoor Wi-Fi winter post.
How to find a loose pair without a $5,000 cable certifier:
- On the controller, watch the link speed. A 1 Gbps AP that periodically appears as 100 Mbps is almost always a bad pair, not a bad AP.
- Check error counters on the switch port. RX CRC errors that climb steadily over time indicate a physical problem on the wire.
- Re-terminate both ends. On a residential install, one fresh RJ45 punch-down at each end fixes 70%+ of intermittent link issues. It’s 10 minutes of work and removes a whole category of failure from the suspect list.
- Pop the keystone off the wall plate and check the punch-down. Half-seated jacks and untwisted pairs that exceed the half-inch limit are the most common installer mistakes we re-do on inherited installs.
How to read your dashboard for this
A lot of this is visible in the controller if you know where to look. Our full UniFi dashboard guide walks through it screen by screen, but the high-yield signals are:
- Per-port PoE power draw. A U7 Pro Max should sit around 15–20 W idle and spike to 25–28 W under load. A port reading 9 W consistently means the AP isn’t getting PoE+ and is running in reduced-power mode.
- Total switch PoE consumption. Compare against the rated budget. If you’re above 80% of the spec, you’re in intermittent-failure territory under peak load.
- Link speed history. Look for ports that flap between 1G and 100M, especially overnight. That’s loose-pair fingerprint.
- Disconnect timestamps. If an AP disconnects at exactly 3:00 AM every night, that’s firmware. If it disconnects when the kids start streaming Netflix, it’s PoE budget. If it’s irregular and clusters during temperature swings, it’s a loose pair.
The diagnostic walk-through we actually use
On a service call for “the AP keeps rebooting at night,” here’s the order we work through it:
- Pull the controller event log for the last 7 days. Look at the disconnect pattern: regular cron-time, traffic-correlated, or thermal/random.
- Check the PoE source. Is it an injector, a switch, or a passthrough on a downstream device? What standard does it actually deliver, not just claim?
- Check the switch’s total PoE budget vs actual draw. Add up the per-port numbers and look for headroom.
- Check link speed history on the affected AP’s port. If it ever drops to 100 Mbps, that’s a smoking gun for a bad pair.
- Look at the AP’s firmware version. Compare against UniFi’s release notes for known overnight-reboot bugs at that version.
- Only if all four of the above check out, replace the AP. We replace AP hardware as a last resort, not a first resort.
Following this order saves homeowners $200–500 in unnecessary parts swaps on a typical service call.
What we install to avoid the problem
On a fresh install in Lehi, Draper, or Park City, here’s the stack that doesn’t see this class of failure:
- PoE++ (802.3bt) switches with at least 30% PoE budget headroom over the planned device count. On a 16-AP house with 24 cameras, that means a switch rated for 600 W PoE total, not 400 W.
- All APs and cameras fed from real switch ports. No injectors anywhere in the install except for a single retrofit case where running new cable isn’t feasible.
- Cat6A end-to-end on every AP run, terminated at keystones at both ends, certified to TIA-568 Cat6A specs at handoff. We covered why on the Cat5e vs Cat6 vs Cat6A post.
- Switch and gateway on a real rack UPS, so a brief utility-power sag doesn’t show up to the APs as a brown-out.
- Documented firmware version per device, with a quarterly review. New firmware doesn’t go out the same week it ships — we wait for the field reports.
When to actually replace the AP
Sometimes the AP really is the problem. Specifically:
- Anything > 6 years old. The capacitors have aged enough that even a properly delivered PoE feed produces unstable behavior.
- APs that have been physically exposed to water intrusion, even briefly. Indoor APs in attics after a roof leak, outdoor APs whose gaskets failed in a Park City winter. The corrosion is inside the casing and is not coming out.
- APs whose model has been EOL’d for two or more firmware generations. UniFi’s long- tail firmware support is excellent, but it isn’t infinite, and at some point security patches stop arriving.
Bottom line
“My AP keeps rebooting at night” is one of the most common tickets we get, and the AP itself is rarely the problem. The cause lives in the PoE feed, the cable, or a marginal injector somewhere upstream. Walk the diagnostic in order, and the actual culprit usually shows up in 15 minutes — no AP swap required.
The deeper lesson is that a network is a system. A $400 AP fed by a $19 injector and a hand-crimped RJ45 isn’t actually a $400 AP install — it’s a $19 install with a good radio. Build the underlying delivery right, and the AP just sits there and works for years.
Keystone Integration designs and installs properly powered, properly cabled Wi-Fi networks across Draper, Holladay, Lehi, Park City, and the rest of the Wasatch Front. If your APs keep rebooting and you’re tired of swapping hardware, see our full service list or get in touch for a service call — we walk the actual delivery chain, not just the AP.