RF simulation: the two models

Everything about a run that involves the air passes through one layer, and that layer can decide reception two different ways. Which one is active changes what a result means, so it is stamped into every saved run.

A concrete case: two repeaters relay the same flood within a millisecond of each other, and a third node hears both at similar strength. Calculated mode sums each transmission into the other's noise and both fail. A real SX1262 usually recovers the stronger one, and waveform mode shows that, because the chirps are synthesised, summed, and handed to a real demodulator that locks onto the stronger frame. When "did it collide" is the question, the mode is part of the answer.

Shared by both: transmit power, antenna gain in the true direction, free space + terrain diffraction + buildings, the receiver's noise floor then the verdict is reached two ways Calculated the default signal, in dBm overlaps summed into the noise SNR vs the demodulator floor above it decodes, below it does not no capture effect Waveform 20 to 50 times the cost IQ samples, chirp by chirp overlaps summed coherently the real receive chain preamble, header, FEC, CRC capture emerges, it is not a rule The channel never decides anything in either mode. It produces a signal; what reads it differs.
Both modes price the same path. They differ only in what happens at the receiver: arithmetic against a floor, or a demodulator given samples.

Which to use#

Calculated is the default and the right choice for anything large - a national network, a coverage raster, a sweep across seeds. It is the mode most questions should be asked in.

Waveform answers what arithmetic cannot: whether two overlapping transmissions capture or collide, what a partial overlap actually costs, and what a receiver would hear if an SDR were attached to it.

CalculatedWaveform
verdict fromlink-budget SNRa decoded frame
concurrent transmissionssummed into the noise, in dBmsummed coherently, as samples
capture effectnoneemerges from the physics
partial overlapcounted as full-length interferencecosts what it actually costs
cost, 300-node burst46 ms1.0 to 2.3 s

What calculated gets wrong, and in which direction#

Two known biases, and they pull opposite ways:

Measured against waveform mode on a dense 60-node burst with six simultaneous senders: roughly half the collision-affected pairs decode under the waveform that calculated calls lost. In a contended scenario, calculated understates delivery.

That is a reason to compare two runs in the same mode rather than to distrust the mode. It is also why the mode is stamped into every saved run - two runs under different models are not comparable.

Switching#

Configuration → RF Simulation, or the rf.mode verb. It applies on a whole-transmission boundary, so a run can be switched while it plays without cutting a frame in half. The status bar names the active mode.

Where to go next#

MeshBench documentation. Built from the running application, not from mock-ups. Screenshots are window-only captures; see CLAUDE.md for the rule that keeps them current. Edit this page.