Almost every modern camera system in a commercial building runs on Power over Ethernet. One cable to each camera carries both the video and the electricity, which is why a camera can sit on a soffit well away from the nearest outlet with no electrician involved. The device supplying that electricity is the network switch in your communications room, and when someone specifies one, the number they say out loud is the port count. Sixteen ports. Twenty-four. Forty-eight.
That number tells you how many cables you can physically plug in. It tells you almost nothing about how many cameras the switch can actually run.
The constraint that decides that is a different number, printed in smaller type on the same datasheet, and it is a total — one pool of power shared across every port at once. Ports are individual. Power is communal. Buy on the first number and you will meet the second one eventually, and probably not on installation day, which is what makes this worth an owner's attention.
Two numbers, and only one of them is on the box
A PoE switch has two independent limits.
The first is the per-port maximum: the most power the switch is willing to send down any single cable. This is set by which PoE type the port supports, and it exists because the cable and the connector have physical limits.
The second is the total power budget: the most power the switch's internal supply can deliver across all ports simultaneously. This is set by the size of the power supply inside the box.
The two are not related by multiplication. This is the whole article in one sentence. A switch that supports a high per-port maximum on twenty-four ports does not carry a budget equal to twenty-four times that maximum — it carries whatever its power supply provides, which is typically a good deal less. The manufacturer is not hiding anything. The assumption baked into the design is that a real installation is a mix: most devices drawing modestly, a few drawing hard, and the total landing comfortably under the ceiling. That assumption is correct for a phone system. It is less correct for a camera system, and it stops being correct entirely once someone adds heaters.
So the honest way to read a datasheet is to stop reading the port count as a capacity figure. The capacity figure is the budget, and the question that matters is what your devices will draw at the worst moment, not the average one.
PoE types, without the wattage table
PoE is standardized, and the standard has gone through successive revisions, each one raising the ceiling on how much power a single port may deliver. The names you will see on a quote — PoE, PoE+, and the higher-power tiers usually written as PoE++ — are marketing shorthand for those revisions.
We are not going to print the per-tier wattage figures here. They are published in a standards document that costs money to obtain, and this company does not publish numbers it has not read in the primary source. What matters to an owner is the relationship rather than the values, and the relationship is simple:
Each step up in PoE tier raises the per-port ceiling substantially, and every watt a device draws comes out of the same shared budget. A switch is not more capable because its ports are higher-tier. A higher tier means each hungry device is permitted to take more — which means fewer of them fit inside the same pool.
There is a second mechanism worth knowing, because it is the one that produces the confusing symptoms later. PoE is negotiated, not simply supplied. When a camera is plugged in, it and the switch have a brief conversation: the camera signals what class of power it needs, and the switch decides whether it has that much left to give. If it does, the port comes up. If it does not, the port stays down, or comes up and is later dropped.
That negotiation happens at power-up, and again whenever a device's demand changes. Which is why a system can be stable for eleven months.
Why a 24-port switch cannot power 24 cameras
Work through it structurally, without any invented numbers.
Take a fixed, indoor, unheated camera with no motorized lens. It sits near the bottom of the PoE range. Twenty-four of those on a mid-range 24-port switch is usually fine, and this is the installation the switch was designed for.
Now change the site to a real one.
Add infrared illuminators. Most exterior cameras have IR LEDs that switch on in darkness. They are off during the daytime survey and on every night. A camera's draw at 2 p.m. in July and its draw at 2 a.m. in December are different numbers, and only one of them appears on a spec sheet's "typical consumption" line.
Add heaters. This is the British Columbia part. Exterior cameras in the Interior, in the Fraser Canyon, on the Coquihalla corridor, and on any exposed Lower Mainland rooftop are commonly specified with integrated heaters and blowers to keep the housing above freezing and the dome free of condensation and ice. A heater is a resistive element. It is, by a wide margin, the largest single consumer in a camera housing, and it is thermostatically controlled — meaning it draws nothing at all for most of the year and a great deal on a cold morning, all at once, across every exterior camera on the site, because they are all experiencing the same weather.
Add PTZ cameras. A pan-tilt-zoom camera has motors. Motors draw a surge when they start moving, above their steady-state draw. A PTZ on a guard tour or responding to an operator drag is asking for its peak repeatedly. PTZ housings are also large, which means their heaters are large.
The result is a site whose total draw is not a constant. It is a curve with a pronounced winter-night peak, and the peak can be several times the summer-afternoon figure that everyone quietly used when the switch was chosen.
The failure mode, and why it gets misdiagnosed
Here is what happens when the peak crosses the budget.
The switch does not blow a fuse. It does not shut down. It behaves exactly as it was designed to: it protects itself by shedding load. When the demand on the shared pool exceeds what the supply can deliver, the switch drops power to one or more ports, usually according to a priority scheme that nobody configured — which in practice often means the highest port number, or the most recent device to ask.
So on a cold morning, some cameras go offline. Not all of them. Not the same ones each time. The recorder logs them as disconnected. Then the sun comes up, the heaters cycle off, demand falls back under the ceiling, and the dropped cameras negotiate power again and come back. By the time anyone checks, everything is online and healthy.
Consider how that presents to the people looking at it:
- It is intermittent, so it fails the most basic test of a real fault — you cannot reproduce it on demand.
- It is weather-correlated, which points a reasonable person straight at water ingress and corroded connectors, because that is the other thing in a camera system that fails when it is cold and wet. Corrosion is a genuinely common fault in this climate and it produces similar-looking dropouts. This is the single most likely wrong diagnosis.
- It moves between cameras, which argues against any single camera or cable being the problem, and so tends to get blamed on "the network".
- It is self-healing, so every site visit finds a working system.
The usual sequence is a camera gets replaced, then a cable gets re-terminated, then the recorder gets a firmware update, and the dropouts continue, because none of those things was ever the cause. The cause is arithmetic in the communications room.
The tell is correlation. If dropouts cluster on cold nights, involve exterior cameras, affect a rotating cast, and clear by themselves in the morning, the power budget is the first thing to check, not the last.
Distance is part of the same equation
One more mechanism, because it interacts with the budget and gets left out.
An Ethernet run has a published maximum length, and beyond it a link becomes unreliable. But a separate effect arrives before that limit: copper has resistance, and some of the power sent down a long cable is lost as heat along the way, so the device at the far end receives less than the switch sent. The standard accounts for this — a compliant switch is expected to supply enough that a compliant device still gets what it needs at full distance — but the allowance assumes cable that meets specification.
In practice, long runs, thin or copper-clad-aluminium cable, extra patch panels, joins, and couplers all increase the loss. The camera at the end of the longest, worst run fails first when the budget tightens, and it looks like a camera fault rather than a system-wide one.
That is why "it works if I plug it in at the rack" is not proof that a camera is healthy.
What to ask, and what to ask for
You do not need to become a network engineer. You need to make the budget a stated fact rather than an assumption. Ask for these in writing:
1. The switch's total power budget, and the calculated total draw of the cameras on it — at winter peak. Not typical consumption. The figure with every heater, every IR illuminator, and every PTZ at maximum, at once. If the answer is a shrug, the calculation was not done.
2. The headroom, stated as a number. A design that lands just under the ceiling has no room for the two cameras you will add next year, and no room for the day the temperature is lower than the design assumed. Ask what percentage of the budget is spare and whether that is deliberate.
3. Which cameras have heaters. Owners are rarely told this. It changes the arithmetic more than anything else on the site.
4. Whether exterior and PTZ cameras are split across multiple switches or fed from a switch sized for them. Concentrating every heated camera on one switch is how you build a site that fails in a block.
5. Run lengths for the longest cables, and what cable was used. Full-specification copper matters here; copper-clad aluminium is cheaper and loses more.
6. What the switch does when it runs out. Some switches let you set per-port priority, so that if the budget is exceeded, a low-value camera drops instead of the one covering the loading bay. Most installations never configure it. Asking is free.
An integrator who works with heated exterior cameras in this province will have these figures to hand. One who has never had to think about a January morning will not — and that answer is itself useful.
The short version
Port count is a socket count. The number that decides how many cameras a switch can genuinely run is the total power budget shared across all of them, and it is not the port count multiplied by the per-port maximum.
Heaters, infrared illuminators, and PTZ motors are what push a site past that budget, and all three peak in the same conditions at the same time — cold, dark, and usually months after the installer has gone. When the budget is exceeded, the switch sheds cameras rather than failing outright, so the fault presents as random, weather-linked, self-healing dropouts that survive every attempt to fix the camera, the cable, or the recorder.
If your system loses cameras on cold nights and looks perfect by lunchtime, get someone to do the arithmetic before they touch anything on the wall.
