Rig control and CW keying

A small Python script on the operator's own machine bridges Hamlib's rigctld to the browser over a local WebSocket. Band and mode then follow the VFO, and if the rig supports CAT keying, a CW macro window becomes available.

The EzFD server is never involved. The bridge runs entirely on the operator's computer and talks to the browser over ws://localhost:4575. Nothing about the radio reaches the server; only logged QSOs do.

This is entirely optional. Operators who don't use it lose nothing but the convenience.

Two ways to connect

There are two transports. The bridge is the default and works almost everywhere; the browser can also open the CAT port itself on some setups, which needs nothing installed.

Bridge Directly from the browser
Needs installing Python, and Hamlib nothing
Kept running a second process, per radio nothing
Browsers all of them, except Safari when the server uses https:// Chrome, Edge and other Chromium browsers only — not Firefox, not Safari, and no browser on iOS
Served over plain HTTP works not available — the browser only offers serial ports on a secure page
Radios around 200, through Hamlib Kenwood and Elecraft commands, which also covers FlexRadio SmartCAT
Band and mode follow the VFO yes yes
CW keying yes no — use the bridge
Two radios a second bridge on another port nothing extra; each window grants its own port

The bridge is not going away, and if you already have one running it keeps working exactly as before — it is the better-informed of the two, since Hamlib knows each radio's quirks. Use the direct path when you want rig control at a site where installing Python is the thing standing in the way.

Connecting directly

Open the Rig button in the Operators panel and choose Choose the radio's port. The browser asks which serial port to use — pick the radio's — and band and mode start following the VFO.

Each browser window asks once, and the permission does not carry between windows, so a second radio in a second window simply grants its own port.

If the panel says Web Serial is unavailable, it names the reason: either the browser has no support, or the page is served over plain HTTP. Both are cases for the bridge.

Reading only, for now. CW keying goes through Hamlib's \send_morse, which has no direct equivalent here, so the macro window still needs the bridge.

Setup

Download ezfd-rig-bridge.py from the logger — the Rig button in the header opens the panel it is linked from — connect the radio by USB or serial, and run:

$ python3 ezfd-rig-bridge.py

It needs Python 3 and nothing else installed by hand. The first time it runs it fetches the one library it uses, websockets. Some Pythons refuse to install anything system-wide, notably Homebrew's on a Mac and the system Python on recent Debian, Ubuntu and Raspberry Pi OS, which say externally-managed-environment. On those the bridge sets up a private environment of its own in ~/.ezfd/rig-bridge-venv and runs from it, so your system Python is left alone and later starts go straight there. Deleting that folder is safe; the bridge rebuilds it on the next start.

The script will:

  1. Find rigctld, offering to install Hamlib if it's missing.
  2. Start rigctld if it isn't already running, prompting for radio model and serial port the first time and remembering the answer.
  3. Probe whether the rig supports CAT CW keying.
  4. Open the WebSocket the browser connects to.

Open EzFD and the header's Rig button turns green and shows the live frequency. It is there whether or not a radio is attached: with none, it reads Rig off and opens the panel you connect from.

Options

Flag Default Purpose
--port 4575 WebSocket port the browser connects to
--rigctld-port 4532 TCP port rigctld listens on
--rigctld-host localhost Host running rigctld

--rigctld-host covers the case where rigctld already runs elsewhere on your network — a shack PC, say — and you want the browser on a laptop.

What it does

Band and mode are polled about four times a second and pushed to the browser, so the QSY drawer tracks the radio without anyone touching it. Logging a QSO records whatever the VFO was on.

Frequency is displayed in the header. On a special event you can also record a planned frequency on your checkout, which is free text for other operators to read.

CW keying

If rigctld reports CAT keying support, a ⚡ CW button appears. It opens a separate window (window.open, so it can go on a second monitor) with:

  • F1–F12 macros, editable, with separate sets for Run and S&P
  • Speed control in WPM
  • Auto-CQ with an adjustable repeat interval
  • ESM (Enter Sends Message), N1MM-style: Enter in the callsign field sends the appropriate macro rather than only moving focus
  • Space or Escape stops transmission instantly

The CW window is a full logger in its own right — it has the same entry form and dupe checking, and shares the log live with the main tab.

Macro placeholders

Placeholder Expands to
{call} The callsign currently in the entry field
{class} Their received class
{section} Their received section
{rst} The RST you're sending
{name} Their name
{mycall} The event's callsign
{myclass} Your class
{mysection} Your section
{exch} Your full exchange

On a special event, {exch} expands to the signal report rather than a class and section, since there is no contest exchange to send.

Radio-specific notes

These are the ones that have actually caused trouble:

FlexRadio via SmartCAT emulates a Kenwood TS-2000. Use Hamlib model 2014, not any FlexRadio-specific model — those are TCP-only and cannot open a COM port at all.

FlexRadio via AetherSDR on a Mac or Linux works the same way: turn on its TS-2000 CAT emulation, pick Hamlib model 2014, and give 127.0.0.1: and the port AetherSDR shows (for example 127.0.0.1:5001) as the serial port. A FlexRadio model number here sends commands the emulation doesn't answer, and every reading arrives late.

Virtual and software CAT ports often don't implement RTS/CTS hardware handshaking, which causes silent write failures. The bridge defaults serial connections to serial_handshake=None for this reason. It's a safe default for real hardware too.

Capability detection is slow. Asking a rig what it can do exercises real CAT round-trips and can take several seconds with irregular gaps. The bridge does this on a separate throwaway connection so it can't desynchronise the polling connection — a lesson learned the hard way, because a misread byte there permanently offsets every later frequency reading by one field.

Troubleshooting

The header's Rig button still reads off. Check the bridge is running and shows no errors. The browser connects to localhost, so the bridge must be on the same machine as the browser, not the server.

The bridge shows the frequency but Safari never connects. Safari blocks a page served over https:// from talking to the bridge's plain ws://localhost link, and says nothing on screen. Chrome, Edge and Firefox allow it. Use one of those for rig control, or a server on plain HTTP.

"Rig not responding (no reply within 2s)" keeps repeating. rigctld is answering slowly or not at all, most often because the Hamlib model doesn't match what the radio, or the program emulating it, actually speaks. Check the model number first; see Radio-specific notes. The bridge reopens its connection after each late reply, so a late answer can no longer be mistaken for the next one.

CW button never appears. The rig didn't report CAT keying support. Not all radios have it, and some need it enabled in a menu. The rest of rig control still works.

CW sends "VFO" as text. A Hamlib invocation is passing a VFO argument to the keying command. The bridge deliberately never does this, because rigctld only accepts one when started with --vfo — and without it the word "VFO" gets sent as Morse, which sounds confusingly like "4FO" on the air.

"Could not create ~/.ezfd/rig-bridge-venv". Your Python can't install packages system-wide and is also missing its venv module. On Debian, Ubuntu or Raspberry Pi OS run sudo apt install python3-venv and start the bridge again. See Setup for why the bridge needs it.

Serial port permission denied on Linux. Add yourself to the dialout group and log back in.

More in Troubleshooting.