📶 5G Coverage Lab: Link Budget, mmWave Blockage and Beamforming

2D interactive simulation · educational model

Drag the phone or a tower on the city map. Switch between low, mid and mmWave bands, change the antenna array and rain, and see where the speed collapses.

The science

Link budget. Received power in dB: Prx = Ptx + Garray − PL − M, where M = 25 dB is a fixed margin for clutter, body/indoor loss and antenna inefficiency. Free-space path loss is FSPL = 20 log10(4πd f / c) (about 32.45 + 20 log dkm + 20 log fMHz): every doubling of distance or frequency costs 6 dB, which is the main reason higher bands reach less far. Buildings on the straight path add a penetration loss per building (about 6 dB at 700 MHz rising to 20+ dB at 28 GHz), capped at three buildings to stand in for diffraction and reflection. Rain adds γ = k Rα dB/km, important only at mmWave.

Noise, SINR and speed. Thermal noise is N = −174 dBm/Hz + 10 log10B + NF (7 dB receiver noise figure). The phone connects to the strongest tower; the other towers on the same band are interference: SINR = S/(I + N). Speed follows Shannon, C = B log2(1+SINR), limited to 7.4 bit/s/Hz per layer (256-QAM), multiplied by 1-4 spatial MIMO layers when the signal is strong and by 0.7 for protocol overhead. Below about −6 dB there is no connection.

Beamforming. N antenna elements phased to add coherently on the user give an array gain of 10 log10N dB (3 dB per doubling). Interference from other towers is counted without this gain (their beams point at their own users), which is a simplification. This is why mid- and mmWave 5G rely on many small elements: they win back the extra path loss of higher frequencies.

Honest simplifications. Flat 2D city of boxes with a fixed cap on blockage, no ground reflections, fading, handover, load or traffic sharing, equal power on every tower, a single user, one antenna height, no latency model. Real throughput depends on the operator's configuration, spectrum and number of users; use this to see the physics trends (frequency vs reach, bandwidth vs noise, beams vs blockage), not to predict a real network.