v1.89

installer: the AppImage carries its license texts

LICENSE and ATTRIBUTIONS.md ride in every AppImage, staged into
usr/bin/wwwroot beside the operator manual so UseStaticFiles serves
them at /LICENSE.txt and /ATTRIBUTIONS.md — readable on the radio,
offline, and both AppDir variants inherit the copies. Release notes
now link LICENSE in this repository.

v1.88

What's Changed

• ci: a tag publishes again, and the freshness check guards that path too by @rampa069 in #30

Full Changelog: v1.87...v1.88

v1.87

What's Changed

• build: ship the libzeus_wspr that does not blow the stack by @rampa069 in #28
• ci: assemble releases as drafts, and refuse to ship a stale binary by @rampa069 in #29

Full Changelog: v1.86...v1.87

v1.86

What's Changed

• wspr: decode off the heap — 1.25 MB of tables blew the thread stack by @rampa069 in #22
• logbook: publish each QSO to QRZ.com as it is logged, when the operator asks by @rampa069 in #23
• spotting: resolve the station the way the rest of Zeus already does by @rampa069 in #24
• reporters: one software name for every network we report to by @rampa069 in #25
• ci: the arm64 zeus_wspr build needs the FFTW headers, not just the .so by @rampa069 in #26
• build: ship the libzeus_ft8 that actually has the measured SNR by @rampa069 in #27

Full Changelog: v1.85...v1.86

v1.85

What's Changed

• spotting: decodes reach PSK Reporter and WSPRnet — the uploaders behind the panel by @rampa069 in #12

Full Changelog: v1.84...v1.85

v1.84

What's Changed

• ci: move off the deprecated Node 20 actions by @rampa069 in #20
• feat: implement the logbook in core (ILogbookPlugin over LiteDB) by @rampa069 in #17
• digital: the keyer logs arm, disarm and halt by @rampa069 in #16
• wspr: enabling receive twice no longer throws away the slot being captured by @rampa069 in #15
• fix: FT4 actually decodes — the core receiver was hard-wired to FT8 by @rampa069 in #21
• ft8: report a measured SNR, not ft8_lib's demo score — decodes read ~28 dB high by @rampa069 in #14
• ft8: reply in the slot that just opened, not a cycle later by @rampa069 in #13
• feat: clicking a decode joins the QSO where it is, and arms TX by @rampa069 in #18

Full Changelog: v1.83...v1.84

v1.83

Full Changelog: v1.82...v1.83

v1.82

Full Changelog: v1.81...v1.82

v1.81

pan3d: the palette mapping stops wasting itself on the noise floor

Answering 'why is only blue visible even on strong AM signals' by
reading the maths instead of guessing at it.

rawLevel maps dB linearly into the display window, and I then mapped
that with 0.55 + level*0.45. Two consequences, and together they are the
whole complaint:

The noise floor is most of the screen, and 0.55 lifted it to BRIGHT
blue. A typical floor lands near level 0.2-0.3, which came out at
0.64-0.69 — solid bright blue before a single signal is drawn.

Signals then had only the top 45% of the ramp. An S8 carrier near level
0.5 reached 0.775, barely cyan; yellow needed level 0.7 and red 0.87,
which across a 120 dB window means a signal 25-35 dB stronger than it
actually was. The upper palette was effectively unreachable.

Now: floor 0.46, just above the palette's black shelf, so a quiet band
is DEEP blue; and the full remaining ramp carries the signal range. The
spread is deep blue floor → blue quiet band → cyan mid → green strong →
yellow very strong → red at full scale.

The gamma constant stays at 1.0. My first attempt set it to 0.62, which
is the wrong way round — expanding the low end brightens the floor, the
exact fault being fixed. Checking the mapping against the palette
anchors caught it before it shipped; both constants are named and easy
to move if a band wants a different bias.

v1.80

Full Changelog: v1.79...v1.80