Antenna deployment planning over real topography for Estonia and Finland, from 136 kHz to 24 GHz. It answers the three questions an amateur actually asks before climbing a mast:
- "If I put this up here, how far will I be heard?" — Longley-Rice coverage prediction over Copernicus/LiDAR terrain, rendered as a heatmap.
- "Will this microwave link work?" — terrain profile, Fresnel clearance, diffraction, rain and gas loss, aiming angles, and the full budget both ways.
- "What does my 80 m dipole at 12 m actually do?" — modelled elevation pattern over real ground, matched against the takeoff angles each hop distance needs and against the current ionosphere.
Plus the data an antenna planner needs beside the map: public masts and towers, and the licensed-transmitter register — amateur repeaters and professional systems alike.
| Mode | Model | Confidence |
|---|---|---|
| VHF/UHF area coverage | ITM (Longley-Rice) v1.2.2, point-to-point mode down every radial | Good. This is the same engine as SPLAT!, ported faithfully. |
| Microwave links | Free space + ITU-R P.526 diffraction (Deygout + Bullington + smooth-sphere) + P.676 gas + P.838/P.530 rain, cross-checked against ITM | Good on clear paths, conservative on obstructed ones. |
| Fresnel & clearance | Exact geometry with a configurable k-factor | Exact, up to terrain accuracy. |
| Antenna patterns | Analytic: wires use the real Fresnel ground reflection with a power-balance normalisation; dishes use ITU-R F.699 | Within ~1 dB of NEC for wires. Not a substitute for modelling a real array. |
| HF ground wave | Norton flat-earth surface wave over the path's ground constants | Reasonable. Assumes homogeneous ground. |
| HF skywave | Hop geometry + secant-law MUF + ITU-R P.533 absorption and excess system loss | Indicative. This is not VOACAP. It gets the shape of the day and the skip zone right; do not plan a contest around the dB. |
Everything reports where its terrain came from and at what resolution, and the link report tells you when ITM has flagged a parameter out of range rather than quietly returning a number.
make on its own lists every target. The common ones appear below.
make sync # creates .venv from uv.lock
make serve # uvicorn with reload on :8088Open http://127.0.0.1:8088. Elevation tiles download on first use and cache
under ./data/cache. If you would rather not install uv, the container
gives you the same thing:
make dev # app only, on 127.0.0.1:8088make upTwo containers, both run by an ordinary user: the app on an internal network
only, and Apache with ModSecurity v3 and the OWASP Core Rule Set in front of
it, published on :8088. Nothing needs sudo, nothing is installed on the
host, and state lives in ~/.local/share/plotter. Put TLS in front of it
(Cloudflare, or deploy/nginx.conf) for anything public facing.
Then warm the caches and install the nightly refresh timer:
make warm # ~4 GB of DEM
make refresh # masts + register
make timer # nightly refresh at 03:20deploy/DEPLOY.md has the full flow, including how to migrate off the old
root install.
| Source | Resolution | Coverage | Key needed |
|---|---|---|---|
| Copernicus GLO-30 | 30 m | Global, used by default | No |
| Maa-amet LiDAR DTM | 1 m | Estonia | No, but bulk download |
| MML Korkeusmalli 2 m | 2 m | Finland | Free API key |
Copernicus is enough for coverage maps and HF. For microwave link planning,
point PLOTTER_LIDAR_EE_DIR / PLOTTER_LIDAR_FI_DIR at a directory of
downloaded national LiDAR tiles (or a .vrt over them) and the link tool will
prefer them automatically — a 30 m DEM will not tell you the truth about
whether a 5.8 GHz path clears a tree line.
OpenStreetMap, OpenTopoMap, Maa-amet base map / relief / orthophoto, the Finnish topographic map, and Esri satellite imagery. All keyless.
OpenStreetMap via Overpass (both countries, refreshed nightly) and the Estonian ETAK topographic database via the Land Board's WFS, which carries surveyed heights.
This is the one source that needs a caveat.
JVIS (jvis.ttja.ee) is the Estonian Consumer Protection and Technical
Regulatory Authority's information system. Its public side carries the
frequency-licence register — every registered emission in Estonia with
coordinates, frequency, power and antenna height. That is exactly what an
antenna planner wants, both to find repeaters and to avoid sitting on a
licensed user.
It is a server-rendered Java application, not an API, and its module paths and
form field names change between releases. So the scraper in
plotter/core/data/jvis.py discovers the search form rather than hard-coding
one layout: it maps form fields and result-table columns against a synonym
table (Estonian and English), prefers a CSV export if the portal offers one,
and reports what it found.
Check it before you trust it:
plotter-refresh discoveror click Check the JVIS portal in the Data tab. If the module has moved, set
PLOTTER_JVIS_MODULE_PATH to the right path. If the columns come back looking
nothing like a frequency register, the harvest will refuse to run rather than
silently collecting rubbish.
The scraper honours robots.txt, identifies itself (set a real contact address
in PLOTTER_JVIS_USER_AGENT), and waits between requests. Do not drop
PLOTTER_JVIS_REQUEST_DELAY_S below 1 second, and do not run the harvest more
than nightly. If TTJA ever publishes a proper open-data export or an X-Road
service, point the importer at that instead — it will be faster, kinder and
more reliable than any HTML scrape.
I was not able to verify the live JVIS layout while building this — the portal disallows automated fetching from the environment I built in, and I did not work around that. The discovery step exists precisely so you can check it against the real site in one command on your own server, where you are the one making the request.
Live foF2 from the Sodankylä (or Juliusruh/Průhonice) ionosonde via the GIRO DIDBase service, and sunspot number / Kp from NOAA SWPC, with a diurnal and seasonal model as the fallback when the network is unavailable.
plotter/
core/
geodesy.py WGS-84 Vincenty, great circles, bearings
terrain/
providers.py Copernicus, WCS, local LiDAR, cache
profile.py path profiles, Fresnel, clearance, k-factor
propagation/
itm.py Longley-Rice v1.2.2, faithful port
diffraction.py ITU-R P.526 knife edge / Deygout / Bullington
atmosphere.py P.676 gas, P.838 rain, P.530 multipath
hf.py ground wave, NVIS, hop geometry, MUF
linkbudget.py assembles a point-to-point answer
coverage.py radial sweep + PNG/GeoJSON rendering
antennas/
patterns.py wires over real ground, yagis, dishes, sectors
library.py band plan and antenna presets
data/
db.py SQLite store for sites and transmitters
jvis.py TTJA JVIS scraper (see caveat above)
masts.py OSM Overpass + Estonian ETAK
ionosphere.py GIRO ionosonde + NOAA solar indices
api/ FastAPI routes and request models
static/ the web UI (Leaflet, no build step)
deploy/ deploy script, user timer, nginx reference
Makefile the commands above, self-documenting
tests/ reference-value tests
Everything the UI does is available over HTTP. GET /docs gives the full
OpenAPI browser. The interesting ones:
POST /api/link full point-to-point analysis with terrain profile
POST /api/link/best-heights clearance matrix over both antenna heights
POST /api/coverage area prediction; returns a PNG overlay + contours
POST /api/hf HF prediction with the antenna's elevation pattern
GET /api/horizon terrain horizon ring as GeoJSON
GET /api/profile terrain profile between two points
GET /api/elevation one point
GET /api/transmitters register query, bbox/band/callsign filters
GET /api/sites masts and towers
GET /api/antenna/pattern elevation and azimuth cuts for any preset
GET /api/registry/jvis/discover inspect the JVIS portal layout
Example — a 5.8 GHz link from the command line:
plotter-refresh link 59.4370 24.7536 59.3960 24.6650 \
--freq 5760 --h1 25 --h2 20 --antenna dish_600 --high-resAll settings are environment variables prefixed PLOTTER_, or a .env file.
See .env.example. The ones worth knowing:
| Variable | Default | Notes |
|---|---|---|
PLOTTER_LIDAR_EE_DIR |
— | Directory of Maa-amet 1 m tiles or a .vrt |
PLOTTER_LIDAR_FI_DIR |
— | Directory of MML 2 m tiles |
PLOTTER_MML_API_KEY |
— | For on-demand Finnish elevation |
PLOTTER_JVIS_MODULE_PATH |
/modules/side/raadiosagedus/avalik |
Set this if the portal moves |
PLOTTER_JVIS_USER_AGENT |
generic | Put a real contact address here |
PLOTTER_MAX_COVERAGE_RANGE_KM |
2000 | ITM's own documented ceiling |
PLOTTER_MAX_COVERAGE_POINTS |
500000 | ITM solves per sweep; the steps coarsen to fit |
PLOTTER_COVERAGE_WORKERS |
8 | Threads for the radial sweep |
make check # pytest plus pyflakesThe tests check against published reference values where they exist: free-space
loss at 1 GHz/1 km, the ITU-R P.526 knife-edge curve, P.838 rain coefficients,
the 4.12·√h horizon rule, dish gain against the aperture formula, and the
Vincenty round trip. The propagation models are also checked for the
monotonicity properties they must have — loss rising with distance, frequency
and terrain roughness — which catches most porting mistakes.
Transmitting requires a licence. The register data here is for planning and coordination; it is a snapshot of a public register and may be stale or incomplete. Check power and EIRP limits for your band and country before transmitting, and check whether a site needs aviation obstacle marking or planning permission before you erect anything on it. Plotter tells you what should work on the air; it does not tell you what you are allowed to build.