ThetisLink v2.0.4 — bandwidth toolkit, preventive TX-inhibit, power-cap, PstRotator

[2.0.4] — 2026-05 (bandwidth toolkit, preventive TX-inhibit, power-cap, PstRotator)

Backwards-compatible with 2.0.3. Wire-protocol additive only —
one new control ID (DxSpotsEnabled); older clients ignore it,
older servers default the new behaviour to ON. Mix v2.0.3 and
v2.0.4 freely while you roll out, but pair v2.0.4 with the
matching Thetis-fork build to unlock the full feature-set.

Added — Preventive RX-only TX-inhibit (Thetis-fork TL2-3 required)

A new chokepoint between ThetisLink and Thetis stops the radio from
transmitting on an antenna position marked as RX-only, before TX
can briefly come up. When the fork-side ThetisLink extensions are
enabled, ThetisLink drives Thetis' "Receive only" flag directly via
the new rx_only_ex TCI command — MOX, spacebar, hardware-PTT and
VOX are all refused at the source instead of being flipped back
reactively. The reactive ZZTX0 catch-all remains the safety floor
for stock Thetis (no fork extensions) and for any path the
preventive gate cannot reach.

• Server-side state machine handles takeover, level-maintain and
  release, including a bootstrap-stale clear so a leftover
  RXOnly=true from a previous session is wiped within ~1 ms after
  the cap is detected.
• The Thetis-fork RXOnly setter now broadcasts a TCI push-notify on
  every real transition, so external Setup → "Receive only" toggles
  are visible to ThetisLink in real time (was: only on TCI SET/GET
  echoes, which left ThetisLink with a stale cache).
• Server-side dedup on the rx_only_ex notification keeps the log
  clean when fork-broadcast and handler-echo arrive together.

Requires Thetis fork PA3GHM TL2-3 or newer for the full
preventive path. Stock Thetis falls back to the reactive ZZTX0
catch-all without any user action.

Added — Reactive RF power-cap per antenna position

Per-Amplitec-A position the server enforces a maximum forward-power
(amplitec_max_w) by sending the PA's own DriveDown button (SPE
Expert or RF2K-S) — not ZZPC, which the PA pushes back through the
TCI loop. Mode-multipliers are applied universally: SSB/CW × 1.0,
AM × 0.5, FM/DIG × 0.4. A counter remembers how many DriveDowns
were sent on the active position and restores them as DriveUps when
the user switches to a different position.

• New first-class GUI editor in the Amplitec tab (6 rows × max W
  • TX-blocked checkbox) replaces the previous file-edit workflow.
• Rate-limited to one DriveDown per second to let the PA-meter
  settle; brief CW-bursts under that interval may pass the cap
  (reactive only — preventive coverage exists on RX-only positions).
• Tuner first-config: the server UI now shows tuner slots without
  requiring an existing instance, breaking the catch-22 for new
  installs.

Added — PstRotator UDP/XML rotor backend

Native PstRotator support alongside the existing rotctl-TCP backend.
Per-installation choice via rotor_backend = pstrotator in the
server config. Integer-degree AZIMUTH commands; AZ/EL replies parsed
fallible; offline-timeout marks status false cleanly. Host field
is a numeric IP address — no DNS resolution. mDNS troubleshooting
notes added to the manuals.

Added — Editable WebSDR favorite names

Favorites in the WebSDR list now have an explicit Edit-toggle so a
rename commits on Done / loss of focus and survives reconnect.

Added — Server-tab bandwidth monitor + DX-spots opt-out

The desktop Server tab now shows the live UDP bandwidth in both
directions:

• Down (RX) and Up (TX) in Kbit/s, updated every 500 ms.
• Click on Down to expand a per-stream breakdown (audio, spectrum,
  S-meter, DX-spots, …) refreshed every 5 s.
• A DX spots ontvangen checkbox lets you opt out of the DX-cluster
  spot stream on metered links. The Android client has the same
  switch in Settings.

The monitor counts UDP application-payload bytes; the operating-
system network meter typically reads 1.5–2× higher because it
includes IP/UDP/Ethernet headers. The Android DX-spots toggle
resets to ON when the app restarts (no preference persistence).

Fixed — DX-cluster spot broadcast storm (~90 Kbit/s → ~6 Kbit/s)

The server used to re-send all cached DX-cluster spots to every
client on every equipment-tick (5 Hz). With ~100 spots in cache this
consumed ~90 Kbit/s steady-state on each client. The broadcast now
sends only new spots per tick and triggers a full age-refresh every
10 s — about 15× less data without a user-visible change.

Fixed — Server log spam

Two periodic log sources became state-change-driven:

• PowerCap state only logs when (pos, mode, pa_in_operate, cap)
  transitions — used to fire every 2 s regardless. PA-meter fluctuations
  are intentionally excluded from the snapshot so they don't reintroduce
  the spam.
• DX Cluster reconnect now emits one line per failure plus one
  line on recovery (reconnected after N failed attempts) instead of
  the previous three lines per backoff cycle.

Notes for upgraders

• A v2.0.3 client connects to a v2.0.4 server without problems; the
  new DxSpotsEnabled control is simply unused, default = ON.
• A v2.0.4 client connects to a v2.0.3 server without problems; the
  opt-out toggle is harmless (server ignores the unknown control)
  but you cannot turn off the spot stream on the older server.
• The new preventive TX-inhibit only activates when paired with a
  Thetis fork build PA3GHM TL2-3 or newer. Stock Thetis remains
  fully supported via the reactive ZZTX0 fallback that already
  existed in v2.0.3.

---