TL;DR
- 5G NR coverage is beam-based, so a single coverage prediction is no longer sufficient — you need per-beam and per-SSB analysis.
- mmWave link budgets are dominated by penetration and blockage loss, not by distance.
- Massive MIMO modelling requires 3D building data; 2D clutter classes are not enough for beam-level predictions.
Why is 5G NR planning different from LTE planning?
LTE planning was largely about a single sector-level coverage prediction and a frequency reuse plan. 5G NR breaks that model in two places. First, the cell is no longer the unit of coverage — the beam is. A site can show acceptable sector-level RSRP while individual beams leave holes that only appear when you test with a real device.
Second, the antenna is no longer a fixed pattern. Massive MIMO with 3D beamforming means the radiation pattern depends on the traffic and the scheduler, not just on the mechanical tilt. A planning tool that treats the antenna as a static pattern will over-predict coverage in some directions and under-predict in others.
| Planning input | LTE practice | 5G NR requirement |
|---|---|---|
| Antenna model | Fixed pattern, mechanical tilt | 3D array model with per-beam weighting |
| Building data | 2D clutter classes | 3D geometry with material properties |
| Coverage unit | Sector-level RSRP | Per-beam and per-SSB coverage |
| Interference | Reuse-based estimation | Beam-based spatial separation |
| Validation | Drive test after launch | CW measurement before commit |
How do you build a 5G NR link budget?
Start from the same components as any link budget — transmit power, antenna gain, path loss, receiver sensitivity, margins — then change what dominates. At sub-6 GHz the budget still behaves broadly like LTE with a wider channel. At mmWave, free-space path loss rises sharply with frequency, but the practical limit is usually not distance: it is blockage.
Penetration loss through modern glazing, concrete and metal cladding can exceed the entire remaining margin. That is why mmWave planning depends on 3D building models and material properties rather than on a clutter class per pixel. If your building data is coarse, the link budget is optimistic in exactly the places that matter.
What does massive MIMO actually change in the planning tool?
Three things: the antenna model, the traffic assumption and the interference model. The antenna model has to be a 3D array with per-element weighting, not a horizontal and vertical pattern pasted together. The traffic assumption has to include a load factor, because beamforming gain depends on how many users are being served simultaneously.
The interference model has to account for beam-based spatial separation, which means neighbouring cells interfere less than a simple reuse calculation would suggest. Getting these three right is the difference between a plan that survives field testing and one that gets rewritten after the first cluster.
Which planning outputs should you actually trust?
Treat the coverage prediction as a hypothesis and the measurement as the verdict. The outputs worth trusting early are relative ones: which candidate site is better, which beam configuration covers a street canyon, where the coverage boundary falls. Absolute RSRP predictions carry model error and should be validated with CW measurements before capital is committed.
The single most useful early output is a beam-level coverage map overlaid on real building geometry for the densest part of the target area. If that map has holes the design cannot fill, the site plan needs changing, and finding that out in the tool is far cheaper than finding it out after installation.