How do I calculate antenna clearance for a small PCB?

The short answer is that you need to treat the metal casing as an RF sink. It will absorb and reflect energy unless there is a large enough air gap or the casing is designed to be a parasitic element. When working with 4G in a tiny space, the rules for clearance are less about a fixed number of millimeters and more about ensuring the antenna has a clear path to the outside in the primary operational frequencies. You really want to avoid having any conductive material directly parallel to the antenna trace or the radiation path.

For a compact PCB like yours, the general rule of thumb is to keep the antenna trace at least two body heights away from any continuous piece of metal enclosure. If the PCB is mounted directly to the metal chassis, you need to consider using a stand-off or placing the trace path at a point where the metal is interrupted, like near a seam or corner. The specific height depends heavily on the type of antenna you are using—are you using a dipol trace, a microstrip, or a chip antenna? They all have very different requirements for ground plane separation.

If you are running a simple PCB trace antenna, you're essentially building a piece of wire that needs space to radiate. A metal shell turns that space into a short circuit or a reflector, which usually kills the signal strength faster than you can measure. Start by physically modeling the air gap. Try placing the PCB in the enclosure with dummy components and see how much of the trace path is actually exposed to open air versus shielded by the housing. If the trace is mostly trapped inside the metal box, you need to raise the board.

If you are designing the PCB, ensure the ground plane isn't restricted by internal components or the edge of the PCB itself, as that will compromise the entire antenna design, regardless of the casing clearances.

Back when I was messing with these handheld units, we didn't have these tiny surface traces that run along the edge. If you are trying to get good signal, you really need to be looking at the ground plane design, not just the clearance around the wire itself. If the chassis clamps down on the ground too much, the whole system is going to choke, even with all that air separation you're calculating. A simple way to fix this is to lift the board up a bit on some standoffs just to give the whole unit room to breathe. Try to keep the ground far from those metal corners.

I agree with the point about physically modeling the air gap first; trying to calculate that without seeing the whole setup is a waste of time. When I went through a similar project fitting a trace into a metal case, the biggest killer wasn't the trace proximity to the metal, but the ground connection. Even if the antenna trace was clear, if the ground part of the board was making contact with the chassis, the signal just died instantly. You have to isolate the entire ground area from any conductive surface.

One thing that saved me on a similar build was testing with the shell held togather by tape instead of screws. Signal looked okay with the case apart, then dropped hard once I snapped it shut, turned out a screw barrel was sitting right over the active part of the trace. Moved the board two holes over and it was fine. Also try walking around outside with it while someone watches a speed test, handhelds behave wierd near your hand anyway. Where in the shell are you planning to put the antenna, top or bottom?

On mine the fix ended up being a strip of adhesive foam between the board and the shell. The board kept flexing just enough to kiss the case in one spot whenever I pressed on the housing, and that killed reception in a way that made no sense until I slid a bit of paper in there as a shim. Even a couple mm of standoff foam changed everything. Worth checking yours with the shell assembled and a finger pressing on different spots while watching the signal readout. If it jumps when you press near the antenna, that's your gap closing up, not your trace layout.

Update: got it working. Turns out the replies about the ground plane were spot on, the issue wasn't the air gap at all. I also took the tip about taping the shell together instead of screwing it, and sure enough signal dropped the moment I snapped it shut. Shifted the board over a bit and reworked the grounding points, now it holds a solid connection with the case fully closed.