How to Plan 4G / 5G / 6G RF Coverage Online — A Free Atoll Alternative

Every mobile network you have ever used started life as a prediction. Long before a single antenna was bolted to a rooftop, an RF engineer sat in front of a map and answered three questions: where should the cells go, which bands and antennas should they use, and how far will each one usefully reach. Get those answers right and the network is fast, reliable and cheap to build. Get them wrong and you either leave holes in coverage or waste millions on sites you never needed.

That work — turning geography, physics and a traffic forecast into a buildable network — is RF planning. For decades it has lived inside expensive desktop tools like Atoll and Planet. This guide teaches the whole discipline from first principles for 4G LTE, 5G NR and 6G, derives every number it uses, and then shows you how to do the same thing for free, in your browser, with CellScope Pro. By the end you will be able to predict a cell's coverage by hand and reproduce it on a real map.

What RF planning really is

At its core, RF planning is a single repeated calculation: given a transmitter and a point on the ground, how strong is the signal there, and is it strong enough? Do that for one point and you have a link budget. Do it for a few million points on a map and you have a coverage prediction. Do it for every cell in a city, accounting for how they interfere with each other, and you have a network plan.

Three quantities drive everything:

• Received power — in 4G/5G this is reported as RSRP (Reference Signal Received Power), the power of one resource element in dBm. It decides whether you have coverage at all.
• Quality — SINR (Signal to Interference plus Noise Ratio) and RSRQ. A strong signal drowning in interference is useless; SINR decides how fast the link can run.
• Capacity — how many users and how much traffic a cell can carry, which follows from SINR through the modulation-and-coding the channel can sustain.

What makes it an engineering discipline rather than guesswork is that all three are computable. The signal weakens with distance, frequency, terrain and buildings in ways that are described by well-validated mathematical models. Plug the geometry into those models and you get a number you can trust to within a few decibels — which, on a calibrated network, is good enough to decide where to spend the money.

The RF planning workflow — eight steps

Whether you use a €30,000 desktop licence or a free browser tool, professional RF planning follows the same eight steps. Everything in the rest of this article is one of these steps explained in depth.

Propagation models — the physics engine

A propagation model answers one question: how many decibels does the signal lose travelling distance d at frequency f, through this kind of environment? That loss is the path loss, and it is the single most important number in the whole tool. Everything starts from the simplest possible case — free space.

Free space is the best case — nothing in the way. The real world adds rooftops, foliage, walls and the curve of the earth, all of which take more decibels. Different models capture different environments.

4G era — Okumura–Hata and COST−231

For decades the workhorses were the empirical Hata models, fitted to measurement campaigns. They are still excellent for sub−3 GHz macro coverage and remain the default for many 4G plans.

5G / 6G era — 3GPP TR 38.901

5G needed models that work from 0.5 GHz all the way to 100 GHz, on the same footing, with explicit line-of-sight and non-line-of-sight behaviour. The answer is 3GPP TR 38.901, today's industry standard. It defines three environments — each with separate LoS and NLoS equations:

The clever part — LoS probability

A real user is sometimes in line of sight of the tower and sometimes shadowed by a building. 38.901 captures this statistically: it gives the probability that any point at distance d has line of sight, and the predicted path loss is a blend of the LoS and NLoS values weighted by that probability.

Add penetration loss for going outdoor-to-indoor (the 38.901 O2I model), a log-normal shadow-fading margin for the clutter the model can't see in detail, and you have the complete picture: PL = base-model + diffraction + penetration + shadow margin. That sum, subtracted from the transmitted power, is the received signal.

The link budget — term by term

The link budget is the accountancy of the radio link: start with the power leaving the antenna, subtract every loss, add every gain, and see what arrives. It tells you the maximum allowed path loss (MAPL) — the largest path loss the link can survive — which, fed back through the propagation model, becomes the cell radius.

Each term has a story:

• EIRP — transmit power plus antenna gain minus feeder loss. A 5G massive-MIMO panel reaches a high EIRP not by brute power but by beamforming gain: focusing energy toward the user.
• Receiver sensitivity — the weakest signal the UE can still decode, set by the thermal noise floor (−174 dBm/Hz), the receiver's noise figure (~7 dB), and the SINR the target modulation needs.
• Shadow-fade margin — the model gives the median loss; real loss varies log-normally (σ ≈ 6–9 dB). To cover, say, 95% of locations you must hold back a margin of several dB.
• Interference margin — in a loaded network, neighbours raise the noise floor; you reserve 2–4 dB for it.
• Body / penetration loss — the hand, the body, the car, the building wall — each costs dB depending on the scenario.

Coverage prediction — pixel by pixel

A coverage prediction is the link budget run on a grid. The map is divided into pixels (say 10–50 m each), and for every pixel the tool computes the same thing:

1. The 3-D distance and terrain profile from each cell to the pixel.
2. The path loss from the chosen model, plus diffraction, plus penetration, blended by LoS probability.
3. The received RSRP = EIRP-per-RE − path loss, after applying the antenna's horizontal and vertical pattern (azimuth and downtilt matter enormously here).
4. The best server — which cell gives the strongest RSRP at that pixel.
5. SINR = serving RSRP over the sum of all other cells' RSRP (the interference) plus noise; from SINR comes predicted throughput.

A city-scale prediction can be tens of millions of these calculations, which is why CellScope Pro runs them in a background Web Worker so the map stays responsive and you watch a live progress bar instead of a frozen tab. The output is the familiar coverage heatmap — green where RSRP is strong, fading to red at the cell edge — plus a best-server map showing each cell's footprint and the overlap zones where handovers happen.

4G vs 5G vs 6G — what changes

The workflow is the same across generations; the inputs change. Understanding what shifts is the difference between a 4G planner and an engineer who can plan the next decade.

The headline tension is simple: higher bands carry more capacity but lose more signal. A 3.5 GHz 5G cell at the same power as an 1800 MHz 4G cell loses about 6 dB more just from frequency, plus more diffraction and penetration. 5G claws this back with beamforming — a 64-TRX panel can add 20+ dB of array gain toward a user. 6G's FR3 and sub-THz bands push the loss even higher, which is why 6G planning leans on density, cell-free coordination and reconfigurable intelligent surfaces (RIS) to paint signal into shadows that a single tower can't reach. CellScope Pro models all three generations, including FR3, sub-THz molecular-absorption terms and RIS-assisted paths that commercial 4G-era tools simply don't have.

PCI, PRACH & frequency — ACP / AFP

Coverage planning gets the signal there. Interference planning stops the cells from sabotaging each other. Two automated disciplines do this work:

ACP — Automatic Cell Planning

ACP optimises the physical configuration: which candidate sites to keep, and each cell's azimuth, downtilt, height and power, so coverage is maximised while overlap (and therefore interference) is minimised. It is a constrained optimisation — too much downtilt and you create holes, too little and neighbouring cells bleed into each other and crush SINR.

AFP — Automatic Frequency / Identity Planning

AFP assigns the scarce identities and channels so neighbours never clash. In 5G the key resource is the Physical Cell ID (504 values), and it is not enough to make neighbours simply different:

• Collision — two neighbours with the same PCI; the UE cannot tell them apart. Forbidden.
• Confusion — one cell has two neighbours sharing a PCI; handover targets become ambiguous. Forbidden.
• Mod-3 — PCI mod 3 sets the PSS and the PSS/SSS subcarrier offset; same mod-3 neighbours interfere on the sync signals.
• Mod-30 — PCI mod 30 sets the cell-specific DMRS sequence; clashing it hurts demodulation.
• PRACH RSI — the root-sequence index must differ between neighbours or random-access preambles collide.

This is a graph-colouring problem: cells are nodes, neighbour relations are edges, and the planner must colour every node so no edge joins a forbidden pair while keeping mod-3/mod-30 spread out. CellScope Pro's planner detects collision, confusion and mod-3/mod-4/mod-30 conflicts and assigns a clean, interference-aware set — the same job AFP does inside Atoll, done in the browser.

Capacity planning — sites from traffic

A cell can have perfect coverage and still fail if too many people use it at once. Capacity planning turns a traffic forecast into a site count. The chain is:

The crucial coupling is that capacity depends on SINR, which depends on the plan. Push sites closer together for capacity and you raise interference, which lowers SINR and per-cell efficiency — so doubling sites does not double capacity. A good tool lets you see the SINR map and the throughput map together, so you add capacity where the geometry actually supports it. CellScope Pro maps DL/UL throughput per pixel directly from the predicted SINR, so a coverage layer and a capacity layer come from one run.

Small cells & in-building (IBS)

Macro planning is only half the network. Two other layers carry a growing share of traffic and have their own planning rules.

Small cells

Small cells are low-power, low-height nodes (micro, pico) that densify hotspots. They use the UMi / street-canyon model with antenna heights around 5–10 m and inter-site distances of 50–200 m. Planning them is about infill — finding the SINR-limited pockets between macros and dropping a small cell exactly there, without creating a new interference source for the macro layer.

In-building (IBS / DAS)

Indoor systems — distributed antenna systems and indoor small cells — are planned per floor. Here the dominant physics is penetration loss: external signal fighting through walls (the 38.901 O2I model adds 20–30+ dB for concrete), and internal signal attenuating through floors and partitions. The planner places antennas per floor, applies wall/floor losses, and checks that every room clears the indoor RSRP threshold (often stricter, e.g. −95 dBm, because indoor users expect strong coverage).

CellScope Pro covers all of it from one tool: macro (UMa/RMa), urban small cell (UMi), low-height micro/pico, outdoor-to-indoor penetration and per-floor in-building layouts — so a dense-urban plan with a macro layer, a small-cell layer and an indoor DAS can be dimensioned in a single project.

A fully worked prediction example

Let's predict the coverage radius of one realistic 5G macro cell, by hand, with every number shown — then you can reproduce it in CellScope Pro and watch the map agree. We'll plan a single n78 (3.5 GHz) urban macro to a cell-edge target of RSRP = −110 dBm.

The inputs

Now the teaching point. At that same 823 m, a user who happens to be in line of sight down a street sees a very different number:

That 35 dB spread between a line-of-sight and a shadowed user at the same distance is exactly why coverage is a probability map, not a circle — and why the LoS-probability blend in 38.901 is the heart of a credible prediction. A tool that only draws circles is lying to you; one that blends LoS/NLoS over real terrain is telling the truth.

How CellScope Pro computes it

CellScope Pro is not a toy that draws coloured circles. Its prediction engine implements the same physics a commercial planner does, and every core formula has been validated against the published 3GPP and ITU references to a fraction of a decibel.

On top of the physics, it behaves like a real planner: real terrain (SRTM) and land-use clutter, antenna patterns with azimuth and electrical/mechanical downtilt, best-server and SINR layers, a background Web Worker for city-scale runs, automatic PCI/PRACH and frequency planning, capacity dimensioning, CSV/XLSX site import, PDF/image export — and it auto-saves your project locally in your browser, so you can close the tab and pick up exactly where you left off.

CellScope Pro vs Atoll / Planet

CellScope Pro will not replace a tier-1 operator's full Atoll workflow with proprietary ray-tracing and measurement databases — and it doesn't try to. What it does is put genuinely spec-correct, Atoll-grade prediction physics in the hands of anyone with a browser, for free: to learn RF planning, to sanity-check a design, to plan a campus, a small operator's footprint or a private 5G network, and to see the formulas in this article come alive on a real map.

Frequently asked questions

What is RF planning, in one sentence?

It is the engineering process of deciding where to put cell sites and how to configure them so a target area gets the required coverage, quality and capacity at the lowest cost — built on a propagation model, a link budget, coverage prediction and interference (PCI/ACP/AFP) planning.

Is there a free alternative to Atoll or Planet?

Yes — CellScope Pro by CafeTele. It runs in the browser, needs no install or licence, and implements 3GPP TR 38.901 propagation, ITU-R P.526 diffraction, real link-budget/SINR computation, ACP/AFP, capacity, small-cell and in-building planning, with exportable heatmaps.

Which propagation model should I use for 5G?

3GPP TR 38.901: UMa for rooftop macros, UMi for below-rooftop small cells, RMa for rural/fixed-wireless — each with LoS and NLoS equations blended by a LoS-probability function. For sub-3 GHz 4G, Okumura–Hata and COST−231 are still valid. CellScope Pro picks and blends these automatically and adds terrain diffraction.

What's the difference between ACP and AFP?

ACP (Automatic Cell Planning) optimises the physical/RF setup — site choice, azimuth, downtilt, height, power. AFP (Automatic Frequency Planning) assigns identities and channels — PCI (mod-3/mod-30), PRACH root sequence, carrier — so neighbours don't collide or confuse. ACP shapes the energy; AFP labels the cells so they don't clash.

How accurate is a free online tool?

Accuracy comes from the physics and the geodata, not the price tag. CellScope Pro's path-loss, diffraction and LoS-probability engines are validated against the 3GPP/ITU references to within a fraction of a dB. The remaining error, as with any planner, is set by terrain/clutter resolution and how well you calibrate against drive-test data — which the tool also supports.

Can I plan 4G, 5G and 6G in the same tool?

Yes. CellScope Pro covers 4G LTE bands with Hata models, 5G NR low/mid/mmWave with 38.901, and 6G FR3 and sub-THz with extended models including molecular absorption and RIS-assisted paths — all in one project, so you can plan a multi-layer, multi-generation network together.

You now know how RF coverage is predicted — the models, the link budget, the per-pixel pipeline, interference and capacity. The fastest way to make it stick is to plan a cell yourself. Open CellScope Pro →