Meshtastic: range vs. airtime

LoRa doesn’t send bytes, it sends symbols. The spreading factor sets both how many bits ride in one symbol and how long that symbol lasts: each step up adds a bit but doubles the duration, which is why a higher SF is always a net loss on speed.

Long Fast is SF11 at 250 kHz, so one symbol lasts 2048 / 250 000 = 8.2 ms and a 40-byte position packet fits into 68 symbols, or 0.56 s on air. Medium Fast is SF9, two steps down, so its symbols are four times shorter at 2.0 ms; the packet needs a few more of them (78, since each carries fewer bits) but still clears in 0.16 s.

The 16-symbol preamble and the error-correction overhead set by the coding rate are counted in that symbol total. Every other figure on this page is built on it.

Smart broadcast needs both enough elapsed time and enough movement, so the slower of the two sets the rate. The heartbeat is a ceiling on top of it: the firmware runs the two as one if/else sharing a single timestamp, so if the heartbeat is shorter it wins.

How far behind reality your own marker sits. No update can be fresher than the fix behind it, so a slow GPS interval adds to the gap directly rather than hiding inside it.

First how many seconds of air the whole mesh burns in an hour, then that as a share of the hour. An hour holds 3600 seconds, so one percent of it is 36 seconds, and dividing seconds by 36 turns them straight into percent.

Each group of nodes is its own stream: RBakens on the interval above, stock devices on the firmware’s 300 s, ATAK phones on their CoT rate. Relaying multiplies the total by the number of copies one packet actually produces, which is set by how the team is spread out rather than by how many nodes have relaying turned on.

The same 36 as above, but counting one node instead of the mesh. The node modelled is the worst-placed one: it originates on every stream and, if it relays, also carries its share of everything it did not originate. EU868 caps this at 10%, and the firmware starts holding telemetry and node names back above 5%.

The budget is how much signal the link can afford to lose on the way. Both ends of it sit on the same dBm scale, so it is really just the gap between them. The radio pushes out 27 dBm. The receiver on the far side still decodes a signal as faint as −131.5 dBm. Falling from +27 down to −131.5 is a drop of 158.5 dB, and that drop is what the air in between is allowed to eat. The antennas then help at both ends, 5 dBi each, worth another 10 dB, which brings the total to 168.5 dB. That is all the minus sign in the formula does: subtracting a negative sensitivity is how you measure the distance down to it.

The second line asks how far 168.5 dB buys. On 868 MHz the first metre alone costs 31.2 dB, fixed by the frequency; past that, loss climbs with distance at a rate set by n. Free space would be 2, but two radios at chest height do much worse, so this uses 3.3 in the open, 4.0 in forest and 4.3 in a built-up area, fitted to real walks. In forest that solves to 2.7 km.

n matters more than the budget does: the same 168.5 dB stretches to 14 km in the open and shrinks to 1.6 km in town. A relaying mesh chains up to three hops, so the figure shown is three shorter hops rather than one long one.

Average draw is idle current plus the GPS’s share of the hour, the screen, and the radio’s transmit time. GPS dominates: at 10 s or under it never sleeps, and above that every wake is a cold start of roughly 30 s, so a longer interval only pays off once it clears that comfortably.