TL;DR
- Private networks are designed against a use case, not against a coverage target — throughput, latency and determinism come first.
- Local spectrum changes the planning problem: you can trade coverage for capacity more freely than a public operator can.
- Industrial environments have propagation characteristics that general clutter models handle badly.
What is different about planning a private network?
The design driver is the application, not the population. A public network is planned to cover people; a private network is planned to serve a defined set of devices doing defined tasks — AGVs on a factory floor, cameras on a perimeter, sensors across a port. That inverts the usual priorities: a small coverage gap that would be a minor complaint in a public network can stop a production line.
It also changes the acceptance criteria. Throughput at the cell edge, latency distribution and determinism matter more than aggregate coverage probability, because the applications either meet their timing requirement or they do not.
| Requirement | Public network priority | Private network priority |
|---|---|---|
| Coverage | Population coverage probability | Coverage of defined operational zones |
| Capacity | Aggregate throughput per area | Guaranteed throughput per device or zone |
| Latency | Best effort | Bounded, and often deterministic |
| Interference | Coexistence with other operators | Internal resource partitioning |
| Change | Expansion planning | Reconfiguration tolerance built in |
How does local spectrum change the design?
When you hold the spectrum, you can use it without the reuse constraints a public operator faces. That allows a design to trade coverage for capacity — more sites at lower power, tighter reuse — without worrying about interference into a neighbouring operator's network.
It also means interference is largely self-inflicted and therefore controllable. The planning question shifts from 'how do we coexist' to 'how do we partition our own resources across the site', which is a much easier problem and often allows a simpler, cheaper design than teams expect.
Why do industrial environments break general propagation models?
Because the clutter is not the clutter the model was calibrated on. Metal racking, stacked containers, machinery in motion and reflective floors produce multipath that behaves nothing like an urban clutter class. Predictions are typically optimistic, sometimes by a wide margin, and the error is not uniform across the site.
The practical response is measurement-led design: model the site with a rough geometry, then validate with measurements in the actual environment before committing to antenna placement. For a single site this is faster than refining a model that was never built for the space.
What should a private network design deliver?
A site plan, an antenna placement plan, a capacity and throughput estimate per zone, and a validation plan. The validation plan is the part most often missing: it should specify where measurements will be taken, what will be measured, and what result counts as passing.
For industrial deployments it is also worth designing for change. Factories reconfigure, ports add cranes, warehouses re-rack. A design that has no margin and no documented assumptions becomes a rebuild the first time the layout changes.