A camera covers a band, not a distance
Ask how far a camera sees and you get an answer in feet, which is the wrong shape of answer. Pick a pixel density you are working to and there are two distances: the far one, where the frame has spread so wide that the density drops through your target, and the near one, where the ground has fallen out of the bottom of the frame because the camera is up on a post looking down a line. The band between them is what the camera actually contributes to a perimeter, and that band is what sets the spacing.
On the defaults — 1080p, a 90 degree lens, 12 ft of mount, 40 px/ft — the far edge lands at 23 ft and the near edge at 3.3 ft. Just under twenty feet of usable band, from a camera that a spec sheet will happily describe as having a 100 ft range. Both statements are true. They are answering different questions.
Why the count moves so fast with the density target
Double the pixels per foot you demand and you halve the line of sight over which you get them, exactly. The near edge does not halve with it, because it is set by the mount height and the vertical angle rather than by the pixel count. So the band is squeezed from one end while the other largely stays put, and the camera count on a fixed run usually goes up by rather more than the factor you moved the target by.
Swept across pixel counts from 1280 to 3840, lens angles from 30 to 130 degrees, mounts from 8 to 25 ft and targets from 10 to 120 px/ft, doubling the target more than doubled the count in about 96 percent of combinations. The exceptions are the cases where the near edge was already almost at the post — a wide lens on a low mount — so there was nothing there to hold the band open, and the count went up by 1.8 to 2 instead. The table at the bottom of the results does this for your own numbers rather than asking you to trust the rule. If the count looks absurd at your target, the target is doing that, and the honest options are a longer lens, more recorded pixels, or admitting that the whole run does not need the same treatment as the door.
Along the line or broadside to it
Looking down a fence line is the cheap layout: one camera covers a long thin band and the pixels are concentrated where the fence is. It also means the subject is walking towards or away from the lens, which is the worst geometry for a face and the best for a long run of coverage.
Set the camera back and look at the line broadside and something useful happens: for a rectilinear lens, the pixel density measured along a straight line parallel to the sensor is the same at the edges of the frame as at the centre. The extra distance to the edge of the view is exactly cancelled by the way the projection stretches off axis. So a broadside camera has one density figure for its whole stretch instead of a value that decays with distance, and the calculator prints it as one number. What it costs is that the stretch is finite and short — the lens angle and the set-back fix it, and stepping back to see more line drops the density everywhere at once.
The near edge is where the trouble happens
A camera on a 12 ft eave with a 90 degree lens has a few feet of ground it cannot see at the base of its own post. Raise the mount and the far edge always comes in, because the extra height is height above the subject and it is spent on the line of sight rather than on the ground. That part has no exceptions: going from a 10 ft mount to a 20 ft one pulled the far edge in on every combination swept.
What it does at the near end is not one-directional, and this is where the received wisdom is wrong. Aiming higher forces a steeper tilt, and a steeper tilt can bring the bottom of the frame closer to the post rather than further from it. Across the same sweep the near edge grew about five times in six and shrank in the rest, mostly on narrow lenses at high density targets. The band itself shrank in all but three combinations out of two hundred, and those three were narrow lenses at 60 to 80 px/ft where the steeper aim bought back more at the near end than the height cost at the far one. So height costs you reach every time and blind zone most of the time, which is worth knowing before somebody mounts a camera out of reach for good reasons.
The fix on a run is not a wider lens. It is pointing every camera the same way so that each one covers its neighbour approach, which is also why facing two cameras at each other across a gate is a worse layout than it looks: they cover the same middle twice and leave both near zones to nobody.
What this does not tell you
It does not tell you the layout is enough. Pixel density is one input into whether footage is any use, and lighting, motion blur, compression, weather and lens quality move the answer as far as the pixel count does. The way that question gets settled is by putting a camera up, having somebody stand on the spot at night, and looking at the recording. Everything here is for deciding what to buy and where to dig before that test is possible.
Questions people ask
How many cameras do I need for a 400 ft fence?
That depends entirely on the pixel density you are working to, and the calculator refuses to pick it for you. On the defaults — 1080p, a 90 degree lens, 12 ft mount, 40 px/ft, 15 percent overlap — a 400 ft run takes over twenty cameras looking along the line. Drop the target to 15 px/ft, which is a movement record rather than a description, and the same run takes a fraction of that.
How much overlap should there be between security cameras?
The page takes whatever figure you give it and defaults to 15 percent as a placeholder. With zero overlap the bands meet at a single point, so any error in post position, aim or tilt opens a real gap exactly where somebody crosses the seam. Overlap is bought in feet of coverage rather than in equipment, which is why it is usually the cheapest thing to add.
Does mounting a camera higher cover more ground?
It covers more area and less useful line. Height is measured against the subject rather than against the ground, so raising the camera lengthens the line of sight to a face and pulls the far edge in every time. The blind patch at the base of the post usually grows too, though not always: the steeper aim that height forces can bring the bottom of the frame back toward the post. Height keeps a camera out of reach and gives a wider sweep, and those are the reasons to do it.
Is it better to point cameras along a fence or across it?
Along it gives a long thin band from one camera and puts the subject face-on to nothing. Across it gives a short stretch at a single density that does not fall off toward the edges of the frame, because a rectilinear projection stretches off-axis by exactly as much as the extra distance costs. Most perimeters end up with along-the-line cameras for coverage and one or two broadside at the places that matter.
What pixels per foot do I need for identification?
This page will not tell you, and be careful with anything that does. Different bodies publish different figures for detection, recognition and identification, they do not agree with each other, and none of them is a rule anyone can hold you to. Put in the number you are prepared to defend, then record a person standing on the spot in the dark and look at the file.