hamradiotools

Nexus 1.7.0

Released 18 August 2026. The current release is 1.24.0. Download it · Every version

Added

  • OmniRig rig control (Windows).

    The Connection dropdown in Settings ▸ Radio ▸ Rig & CAT gains OmniRig — VE3NEA's rig-control server, the one much of the Windows logging and contest world already runs. Pick it and your radio is configured once, in OmniRig, and every program shares it instead of fighting over the COM port. Because OmniRig owns the rig type, the port and the baud, Nexus stops asking for them: the only thing it needs is which radio — RIG 1 or RIG 2, matching OmniRig's own two tabs, defaulting to RIG 1, and per radio, so a two-radio station can put one on each. Dial and mode follow both ways, exactly as on a serial rig. PTT: leaving PTT Method on CAT keys through OmniRig with nothing else to set up, and the other choices are unchanged and deliberately independent of it — pick RTS/DTR with a PTT Serial Port and Nexus asserts that hardware line itself, which is how many operators run. It is Windows only (OmniRig is a Windows program), so on macOS and Linux the choice is shown greyed out with that reason rather than hidden. If OmniRig isn't installed Nexus says so plainly, and if it is running but the rig is off, on a busy port, or not answering, the CAT status shows OmniRig's own words for it instead of a bare "no reply". See the rig setup guide.

  • QPSK31.

    The PSK screen's mode picker now offers QPSK31 beside PSK31 — same speed, same feel, but with error correction: the four-phase signal carries a convolutional code, so the flutter or static crash that prints a wrong character in PSK31 is simply corrected. Receive and transmit both work — pick QPSK31, click a trace, and the same macros, continuous TX and stop controls apply. One thing to know on the air: QPSK31 cares which sideband the other station transmits on. Nexus runs the standard (USB) polarity; if a station warbles but prints garbage, click the Rev toggle beside the mode picker. Mode and polarity switch only between overs — never mid-transmission.

  • PSK31 transmit.

    The PSK screen now works both ways: type a line and press Enter for a one-shot over, fire the F1–F4 macros (CQ / Answer / Exchange / 73), or click TX for continuous transmit — stay keyed, idle on the classic PSK31 reversals, and what you type goes out as you type it, mixed case and all. TX keys at the same spot you netted the decoder (click the trace, answer on frequency), the rig is put in its data mode automatically (USB-side on every band, the PSK31 convention), and a band picker offers the standard watering holes — 14.070, 3.580, 7.070 and friends — showing only the bands your license can key. Stop TX, the Esc/Stop button and Esc all cut a transmission instantly, leaving the section or QSYing out of your privileges unkeys within a tick, and a hard 10-minute ceiling bounds a continuous over no matter what. One deliberate courtesy: Nexus transmits at a modest drive and the dock reminds you to keep the rig's ALC near zero — an overdriven PSK31 signal splatters into the neighbors (IMD).

  • Export a date range from the logbook.

    Compact from/to date pickers now sit beside Export ADIF — set either or both and the ADIF/CSV export carries only the QSOs in that UTC date range, which is what a POTA activation or one weekend's contest needs. Leave them empty and the export is the whole log, exactly as before. The per-operator export is deliberately untouched: it is the submission path, and a leftover date silently truncating an uploaded log would be the worse surprise. (#98)

  • PSK31 receive.

    A new PSK screen on the Digital rail decodes the classic narrow-band keyboard mode — open it, tune a watering hole (14.070 is the busy one), click a warble trace on the waterfall and the text prints, with faint characters marking doubtful copy. The receiver starts by itself when the screen opens (turn that off in Settings ▸ Digital ▸ PSK; stopping it by hand is remembered for the session), the click nets the decoder rather than the rig, and a gentle AFC (never more than ±25 Hz) rides small drift. Transmit ships in this release too (see the PSK31 transmit entry above), and so does QPSK31 (its own entry above).

  • Every Mac release is now checked harder before it ships.

    The release pipeline proves the signed app still carries the microphone entitlement and usage prompt (the exact defect that silently shipped a dead mic in 1.5.0–1.6.1), and it unpacks the self-update bundle, matches its version against the DMG and verifies its signature with the same key every installed copy uses — a broken or missing update bundle now fails the build instead of quietly publishing a release Macs could never update to.

Fixed

  • FLEX-8400 and FLEX-8600 owners can find their radio in the list.

    The rig picker, the detected-radio row and the Flex page all said "FLEX-6xxx", so the current flagship line appeared to be unsupported. It always worked — SmartSDR CAT presents the same commands for a 6000 and an 8000, and Nexus drives both through one profile — but nothing said so. The entry is now "FlexRadio FLEX-6xxx / 8xxx (SmartSDR CAT)".

  • Nexus now notices when the radio is keyed and no RF is coming out.

    Every check Nexus had was on the unkey side: it made sure the transmitter came down. Nothing ever asked whether anything went out in the first place. A transmit inhibit, an antenna port that only receives, an amplifier interlock, a transmit profile pointed at an audio source that is not running — each of them looks like a completely normal over: PTT is accepted, the meters move or don't, the QSO is logged, PSK Reporter is told, and the first evidence is nobody coming back to you. If your radio reports forward power over CAT, Nexus now watches that reading during an FT8, RTTY, PSK or SSTV over, and after a couple of seconds of zero watts it says so in the radio status line. It is a message and nothing else — it never stops or blocks a transmission. It stays quiet on radios that do not report forward power (most do not), during a tune-up, through the first moments of key-down, on phone and CW overs (silence between words and between elements is normal), and it never repeats inside one over or fires once the radio has been seen making power.

  • Phone tells a FlexRadio operator when native DAX has taken their microphone.

    Switching on Flex native DAX audio — the early-access toggle you would turn on for FT8 — tells the radio to take transmit audio from DAX. That is a radio-wide setting, not a Nexus one: while it is on, your Flex's microphone is disconnected on every slice, in every program, SmartSDR's own MOX included. Turn it on for digital, pick up the mic later for an SSB contact, and you transmit silence with nothing anywhere saying why. The Phone screen now shows a mic off (DAX) marker beside the frequency whenever native DAX audio is actually running, and says what to do about it: turn Flex native DAX audio off in Settings ▸ Radio ▸ Rig & CAT and the mic comes straight back. It appears only while the native audio really is live — not merely because the toggle is on with no radio address, or after the audio path has already fallen back to your sound card. The Flex setup guide gained a matching Phone (SSB) section.

  • CW on 160, 80 and 40 m now actually puts an SDR radio into CW.

    Below 10 MHz Nexus asks the radio for CW-L, which is the right thing to ask for and what most radios call it. Four rig profiles do not offer it under that name — FlexRadio SmartSDR CAT, SmartSDR native, PowerSDR and Thetis — and they refused the request outright. Nexus then gave up and left the radio in whatever mode it happened to be in while the keyer went right on sending, so a CW macro could go out of a radio still sitting in USB or a data mode, on the three lowest CW bands, with nothing on screen naming the cause. Nexus now notices the refusal and sets plain CW instead. It is the same signal on the same frequency — CW-reverse only changes which side of the carrier you listen on — so nothing about your transmission changes. The Flex setup guide gained a CW section covering the keyer choices on that radio.

  • Dragging text on an SSTV picture makes it stay where you put it.

    Text you dropped snapped straight back to the middle. The drag itself was fine — the picture is redrawn from scratch many times a second, and each redraw quietly restored the text to where it had last been saved, which for a fresh caption is the centre. Letting go then saved that restored position, so the snap-back stuck. Nudging with the arrow keys was never affected and still isn't.

  • Rotators: the baud now comes from the model, so five of them work for the first time.

    If you own a SPID Rot2Prog or Rot1Prog, an Idiom Press Rotor-EZ, a Hy-Gain DCU-1/DCU-1X or a Green Heron RT-21, your rotator has never answered Nexus — and this is why. Nexus gave every rotator model the same 9,600 baud and told you in the tooltip that was right, then forced it on the control daemon over the top of the rate Hamlib knows your controller actually uses: 600 for the Rot2Prog, 1,200 for the Rot1Prog, 4,800 for the Rotor-EZ, the DCU-1 and the RT-21. At the wrong line rate a controller simply ignores you, which looks exactly like a dead cable or broken hardware. Picking your model now fills in the rate its own backend declares, read out of the Hamlib that ships in the installer rather than typed from a manual; where a model genuinely accepts a range (the GS-232 family, EasyComm, SPID MD-01/02) your own setting is left alone, as it should be. If your rotator has never worked, re-pick your model in Settings ▸ Radio ▸ Rotator — and if the saved number cannot work, the hint under the box now says so in words, with the number to use.

  • A rotator command that got no answer was reported as success.

    Nexus treated silence from the rotator daemon as "done", so the compass slew, the ↗ point-at-call and a whole satellite pass could report that the antenna had moved while it sat still — and the "the rotator stopped answering, point it yourself" warning could never fire, because every unanswered command was counted as a good one. Silence, a timeout and a hang-up are all failures now, and the daemon's own refusal reaches you with its reason. The wait before giving up is matched to what Hamlib itself allows your model, instead of being shorter than it for almost every rotator in the list.

  • A rotator daemon that died at launch was logged as "launched".

    The commonest rotator failure — a port that is not there, a port another program already has, a model number the bundled Hamlib does not carry — kills the daemon within milliseconds, and Nexus reported success and then went quiet forever; re-saving Settings did not bring it back. It is noticed now, at launch and afterwards, restarted, and reported with Hamlib's own words for what went wrong in the Connections log, where before there was nothing to read at all.

  • A Green Heron RT-21 refused the last tenth of a degree of the compass.

    Bearings from 359.95° up were sent as 360.0, which that controller's Hamlib backend rejects outright (it declares a 359.9° ceiling). They are sent as 0.0° now — the same bearing, and one every rotator accepts.

  • "Allow flip" did nothing.

    The setting promises to take a high satellite pass by running elevation past 90° instead of swinging the mast 180° at the top of the pass, and the condition it was gated on could never be true — so the mast raced round on every high pass with the box ticked. It works now: above 85° peak elevation, a mount whose owner has said it can go over the top keeps the antenna on the bearing the bird rose on and runs elevation up through 90° and out the far side.

  • A failed park at the end of a satellite pass drove the antenna FLAT.

    If the park command was refused, the fallback commanded elevation zero — laying a dish or a long boom into the wind, which is the thing a park position exists to avoid — and said nothing about it. The fallback is now azimuth-only for rotators that genuinely have no elevation axis, and for everything else a park that did not complete is reported and the antenna is left where the pass ended.

  • A stuck or jammed mast was invisible during a pass.

    Nexus never asked the rotator where it actually was, so a controller that accepted every command and then stopped moving — a jam, a hit stop, a slipped belt, a box left in local — tracked "perfectly" for the whole pass. The position is read back now, and a mast that is neither on target nor moving is called out in the Connections log. A rotator that stopped answering mid-pass is logged there too, instead of only appearing on screen for whoever was watching.

  • A rotator that cannot report its position no longer loses its controls.

    Some models genuinely have no read-back — the Hy-Gain DCU-1 is one — and the Connect rotor pane used to delete itself entirely when the readout failed, taking the compass rose, the typed bearing and the STOP button with it. It stays, with "—" where the needle would be.

  • Rotator settings live in the Rotator section.

    The model, port and baud were filed under Rig & CAT, so the "Rotator not answering" chip — the one affordance the app has for this failure — opened a section containing neither, and searching Settings for "rotator" found only the pointing manners. Clicking the chip now lands on the model and the port.

  • The rotator list: a wrong entry out, six models in.

    "EA4TX ARS (az)" was Hamlib's parallel-port backend offered with a serial port and a baud box — it could not work as presented, and the brand name steered ARS-USB owners away from the GS-232 generic entry that does (which now says so). Added, all of them models operators actually own: SPID MD-01/02, Prosistel Combi-Track az+el, Hy-Gain DCU2/DCU3/YRC-1, DF9GR ERC, AMSAT LVB Tracker, and the Kenpro GS-23/GS-232. Entries now say (az) or (az/el) where the backend declares it, so an az/el owner is not steered onto the azimuth-only variant of their own manufacturer. And a saved model that is not in the list shows its number instead of an empty box.

  • FlexRadio, the ORDINARY setup: your Flex settings stay put, the radio's address is kept, and a switched-on feature says what it needs.

    Unlike the two entries below — which are about the opt-in "early access" toggles nobody is required to turn on — this one is the everyday, field-verified FlexRadio configuration: SmartSDR CAT, DAX audio, the way the guide sets you up. The worst of it was silent: the Flex radio address and the two native toggles were stored once for the whole station instead of per radio, so two Flexes could never both be configured, and — the expensive part — going into Settings to configure a different radio and pressing Save wiped the Flex settings of the radio you were not even looking at, while Save reported success. They now travel with the radio they belong to, and an existing address is carried over on first launch. Alongside that: the setup wizard threw away the Flex address it had just discovered, so a Flex set up through the wizard arrived in Settings with both native features offered and no address for them to use; switching a native feature on before filling the address in left it permanently dead — typing the address afterwards did nothing at all, and neither did re-picking the radio, so only restarting Nexus (or toggling the feature off, saving, on, saving) brought it back; and turning one on with no address set did nothing and said nothing, which now reads as a plain message telling you which field to fill in. Two more: the one-click "found a Flex" button applied a Windows-only SmartSDR CAT address on macOS and Linux, where that program cannot be installed — it now applies only what is true on your platform and tells you what to enter; and two radios pointed at one network CAT address now warn, exactly as two radios on one COM port already did. Settings' radio list also labels its addresses now (CAT, Flex radio, CAT helper port) — a network Flex has three, and they were bare numbers.

  • A radio moved out from under you is finally visible in the FT8 screen.

    Change the mode at the radio — in SmartSDR, from a Maestro, or with the front-panel knob — and Nexus keeps its own idea of the mode, by design: the read-back is a display value and never overrides what you commanded. Every screen that could show the disagreement did, except the one that transmits unattended for hours. The Operate strip now shows a "rig: …" chip when the radio disagrees, and flags a receive filter far too narrow for an FT8 window — a CW filter left in, or a slice narrowed at the radio, throws away most of the band with nothing on screen to explain the quiet.

  • A CAT level the radio refuses is no longer re-sent fifty times a second.

    RF power, mic gain and noise reduction each kept re-issuing a setting the radio had rejected, on every 20-millisecond tick, for the rest of the session — a continuous stream of CAT traffic on the one thread that also draws the waterfall, times the FT8 slot and holds PTT. That is the "waterfall hangs for a moment, then it's fine, then it lags again" report, at many times the rate of the case already fixed for the DSP controls. Nexus now tries once, gives up, and says the radio would not take it; moving the slider tries again, as does any CAT reconnection. The same discipline reached two more places: the 750 ms rig read-back now has a time budget, so a radio at the end of a slow or remote link degrades how often it is read instead of stalling the loop that has to stay responsive to Stop TX, and the S-meter re-check backs off on a radio that has no CAT S-meter at all instead of costing three blocked reads every 30 seconds, forever. Every CAT radio benefits, not only a Flex.

  • The CAT helper is watched now, and split is put back the way it was found.

    If rigctld — the helper process Nexus talks to your radio through — died, nothing noticed: CAT stayed dead until you re-saved Settings, and because the unkey command travels the same path, a crash mid-transmission took away the ability to stop transmitting. Nexus now spots it, restarts it, unkeys through the fresh connection and tells you it happened. Separately, ending an over with Split Operation set to "Rig" wrote split OFF and left VFO B where Nexus had put it — cancelling a split you had set yourself at the radio (or that another program sharing the radio had set) and clobbering its transmit frequency. Both are restored to what they were. Also: when the radio reports it is already transmitting under another program's control while Nexus is armed, that is now said out loud instead of passing unmentioned.

  • FlexRadio native audio and panadapter (early access): Nexus now shares the radio properly, listens to the right slice, and comes back after a network blip.

    These are the opt-in "Flex native DAX audio" and "Flex native panadapter" toggles in Settings — both off by default and still unverified on real hardware, so nothing here changes anything for an operator who has not turned them on. The theme is that the code assumed it was the only client on the radio, and a Flex almost never is: SmartSDR's own window is usually running, and a Maestro, N1MM or a second Nexus window can be too. What that cost: the native panadapter would latch onto any panadapter on the radio — including SmartSDR's own — then retune it, rescale it and delete it on the way out; native audio would adopt whatever DAX stream held channel 1, even another program's, and remove it on teardown. Nexus now creates and steers only its own, and removes nothing it did not create. Which slice you hear was the other half: audio followed whichever slice had focus while the dial, the waterfall and the log all came from the slice your CAT connection drives, so on a two-slice radio you could be decoding one slice and logging another with nothing saying so. Audio now follows the slice your CAT port names, re-binds when you switch at the front panel (it never could before), and the S-meter shows your slice's signal instead of whichever slice reported last. Also fixed: a dropped audio packet used to be spliced over silently, shifting the timing of every decode after it — the gap is now filled with its own length of silence and named in the log; ordinary network reordering could leave the native waterfall completely blank; a lost connection or a radio reboot ended native audio and the native panadapter for the rest of the session — both now re-dial with a backoff; native audio that died mid-session left you deaf with no warning, and now falls back to the sound card and says so, as it already did for audio that never started; and the RX Gain slider, which did nothing at all on native audio, now boosts a quiet slice exactly as it does on the sound card.

  • FlexRadio native audio (early access): the transmit half is fixed, and it no longer freezes the app when the radio is unreachable.

    This is the opt-in "Flex native DAX audio" toggle in Settings — off by default, and still unverified on real hardware, so nothing here changes anything for an operator who has not turned it on. With it on, both directions ride the network, transmit included (the toggle's old "RX-only" wording was simply wrong and has been rewritten). What was wrong with that transmit path: the Pwr slider did nothing — the radio got full-scale audio wherever Pwr sat, Pwr at 0 included, which is the drive-discipline problem the control exists to prevent; the same over was sent twice, over DAX and out the configured sound device, so operators using the one-click "Pair DAX audio" setup fed the radio two copies and anyone with speakers selected heard every over in the room; a 12-second over was flung at the radio in a few milliseconds instead of streamed in real time; and if the radio refused or was slow to answer the request that sets the transmit stream up, Nexus switched the rig's audio to DAX anyway and never switched it back — leaving the microphone dead even after Nexus had quit. All four are fixed: one route carries an over, at the drive you set, paced in real time, and the mic is always put back. Nexus also puts your slice's own DAX channel back the way it found it now, instead of leaving it moved to channel 1 for good. Stop TX cuts a native over the same way it cuts any other. Finally, a Flex address that does not answer — a stale home LAN IP used from away, a VPN down — no longer freezes the app for up to two minutes when the feature is switched off, and a Flex that reboots or drops the connection no longer pins a CPU core.

  • Test CAT stopped speaking Icom to Xiegu owners.

    When Test CAT walks the CI-V rates to find the one your radio answers on, its verdict used to quote an Icom rig menu (MENU ▸ SET ▸ Connectors ▸ CI-V), explain that your USB port was following a [REMOTE] jack, and finish by telling you to install Icom's USB driver. None of that exists on a G90, X6100, X6200, X5105 or X108G — those radios reached this diagnostic when it was broadened to every CI-V rig, and the advice was not broadened with it. A Xiegu now gets the one cure that is certain (set Baud here in Settings, and the found rate is named), the right driver pointer, and no invented menu path. Icom verdicts are unchanged.

  • Detect no longer badges both of an X6100/X6200's serial ports "use this one".

    Those radios present two USB serial ports and CAT answers on only one of them, but the disambiguation was written for Icom's dual-port cable and matched both — so Detect showed two identical-looking rows for one radio and half of them sent you to the port that opens cleanly and then returns nothing forever. Exactly one row is now badged, and Auto-test sweeps that port first instead of burning a full baud sweep on the dead twin.

  • An FT8/FT4 decode that overran the 15-second period could replay a transmission into the wrong slot.

    Field report, visually confirmed at the rig: with Tx 1st/even set, the station's own CQ keyed real RF a few seconds into an odd slot. On a machine that stalls long enough for a period's decode to outlive the period itself, the late result re-ran the transmit decision with its original slot number — which still passed the even/odd check, because that check trusts the slot it is handed. A decode result whose slot is no longer the one on the clock now folds its decodes and reports its period but can never key. Latent since 0.13 on every platform; it takes a badly stalled machine (a long AV scan, heavy swapping) to trigger.

  • The Setup-health RX-audio light can finally say "No RX audio".

    Its threshold was written for the wrong dB scale, so with a radio configured the light was green even on a stone-dead input — the one state it existed to catch. A silent capture now shows "No RX audio" and points at the audio device settings.

  • A capture that delivers only pure silence now says so — with the Mac's likely cause named.

    On macOS, denying the microphone permission doesn't error and doesn't stop the audio stream; it just delivers exact digital silence forever, which every health check read as "capture alive". After 15 seconds of that, a banner now explains it — and on a Mac adds: if the RX meter never moves, check System Settings ▸ Privacy & Security ▸ Microphone. It clears itself the moment real audio arrives.

  • "System default" audio input now notices its device vanishing on macOS.

    With the out-of-box device selection, unplugging the rig's USB audio (or a sleep/wake reshuffle) silently stopped the audio with no banner and no recovery — only explicitly named devices got disconnect detection. The default now resolves to the same device through the path that watches for disconnects, so the usual "Sound card stopped — reopening" self-heal covers it too.

  • Pounce desktop notifications actually appear now — on the Mac for the first time.

    The rare-DX alert used a browser notification API that Apple's webview simply does not have, and no platform ever asked the OS for permission, so the notification half of Pounce was silently dead everywhere. Alerts now go through the operating system's own notification center; macOS asks you once, at your first alert (never at launch), and if you decline, the alert sound and the in-app banner carry on exactly as before.

  • The waterfall's "move both markers" click works on the Mac.

    Ctrl+click there is the system right-click gesture — it silently moved only the TX marker — and ⌘-click (the chord a mac WSJT-X operator's muscle memory sends) did nothing. Both now set RX and TX together, the on-screen hint reads ⌘ on a Mac, and Ctrl+click keeps working everywhere else exactly as before.

  • Memory quick-recall answers to ⌘1–9 on the Mac.

    The recall chord was Ctrl-only — which macOS itself uses to switch desktops, so the press never even reached the app — and Cmd was explicitly rejected. Either modifier now recalls a favorite on every platform, and the chip tooltips advertise the platform's own chord.

  • F-key advice for Mac keyboards.

    The CW and RTTY macro buttons, the voice keyer, the Operate Decode button and the WSJT-X switchers' guide advertise F-keys that a default Mac keyboard treats as media keys; on a Mac their hints now say to hold Fn (or enable standard function keys in System Settings). Nothing is rebound — the keys themselves are unchanged.

  • Every "open in your browser" link in the app now actually opens it.

    The CAT-driver download link, the repeater directory credits, the contest-calendar rules links and APRS station pages were silently dead on every platform — the webview swallowed the click and nothing happened. They now route through the same mechanism the QRZ links always used, and ⌘-clicking a link works on the Mac too.

  • Field Day exports save a real file on macOS.

    All four export buttons (Cabrillo, ADIF, summary, dupe sheet) went through a browser download path that macOS discards outright; they now write straight into your Downloads folder like the Logbook exports, and the confirmation names the saved path.

  • The Journey share card no longer claims success while doing nothing on macOS.

    Copy-to-clipboard now uses the pattern Safari's engine accepts, and when the clipboard refuses, the card is saved as a PNG in Downloads with the message naming the file — a success message only ever reports a copy or a write that really happened.

  • A crash or force-quit can no longer strand rigctld/rotctld on macOS and Linux.

    Windows has always killed the CAT and rotator daemons with the app; on Mac and Linux a crash, a Force Quit, or a hung shutdown could leave one running — holding your serial port open and sometimes making the next launch's CAT land on the stale daemon. Nexus now tracks every daemon it starts, stops any stragglers on quit, and each launch first cleans up anything a dead instance left behind (a second running Nexus and its daemons are recognized and left alone).

  • Installing an update on macOS and Linux now actually restarts Nexus.

    The banner said "Nexus will restart…" but on those platforms the updater only swapped the app on disk and left the old build running — the banner hung there forever and nothing restarted (Windows was fine; its installer restarts by itself). The restart now happens, and it goes through the normal shutdown: transmitter unkeyed, logs and window position saved, never a hard kill.

  • ⌘Q now quits Nexus properly on the Mac.

    The menu quit (and its ⌘Q shortcut) bypassed the entire shutdown path: no wait for the transmitter to unkey, no conversation or Field Day flush, and no window-geometry save — a Mac operator who always quit with ⌘Q reopened at the default window box every launch, with the last 15 seconds of chat at risk. Every quit route now runs the same cleanup closing the window does.

  • The docs caught up with macOS shipping.

    The FAQ, README, install guide, manual and wiki no longer say "macOS does not ship" — the Mac build (signed, notarized Apple Silicon DMG, self-updating) has been out since 1.5.0. The install guide gains real macOS sections: drag-to-Applications install, upgrade, uninstall, where your data lives (~/.config/tempo — same place as Linux, on purpose), a shasum checksum example, and the one Mac-only gotcha: on first launch macOS asks for microphone access, and declining it silently kills all decoding — Troubleshooting's "No decodes" checklist now starts with System Settings ▸ Privacy & Security ▸ Microphone. The old "put rigctld on PATH" advice for Mac is gone everywhere; the actual mechanism is brew install hamlib (Nexus searches the Homebrew/MacPorts prefixes itself — a Finder-launched app never sees your shell PATH).

  • CAT trouble on a Mac is now diagnosed in Mac terms — and a missing Hamlib is never blamed on your cable.

    Three verdicts stopped guessing: Test CAT's baud ladder no longer buries the correct "Hamlib's tools aren't installed — brew install hamlib" diagnosis under a "close other CAT software" port guess (it was the probe tool itself that failed to start, and no port was ever touched); Auto-test says the same instead of "check the cable and that the rig is on" when its throwaway daemon can't spawn at all; and a rotator's rotctld failing to launch names the per-platform install cure instead of a raw "No such file or directory (os error 2)". The CI-V no-answer walkthroughs and the Settings port hints also stop talking Windows at Mac operators: /dev/cu.* examples and "the port list" instead of COM16 and "Windows Device Manager", the dual-port Icom tie-break in cu.* name order instead of a driver label that only exists on Windows, and no more advice to install a USB driver macOS already ships in-kernel. Messages also stop claiming a "bundled" rigctld — nothing is bundled on macOS or in the AppImage.

  • A Mac serial port saved as /dev/tty.\* now heals itself to its /dev/cu.\* twin.

    Earlier Mac builds (1.5.0–1.6.1) listed every serial port twice and let you pick the /dev/tty.\* row — a node that hangs CAT on carrier detect instead of failing. The later picker fix only stopped NEW picks; a port already saved kept hanging after the upgrade. Nexus now substitutes the cu.\* twin wherever the stored name is consumed — the CAT daemon, native CI-V, PTT keying and Test CAT's baud ladder — and rewrites the saved setting once at launch when the twin is present, so the Settings screen shows the port that actually opens. A lone tty.\* with no cu.\* twin is left exactly as stored: it is the only node there is.

  • WSPR and FST4W beaconing works the way WSJT-X's does.

    Four fixes from one report (#101): the TX watchdog no longer halts a beacon a few intervals into a session — beaconing is unattended repeated transmission by design, and WSJT-X exempts exactly these modes (each transmission is still hard-bounded by the period clock, and your TX switch remains the stop); a Round-Robin rotation of one station no longer transmits every interval — one slot is no rotation, so the transmit-% schedule applies and Settings says so; your own beacon transmissions now appear in ALL.TXT as Tx lines like every other mode's; and a WSPR spot ("CALL GRID DBM") now files in the operating roster under its callsign with its grid, instead of the grid showing up as a phantom station.

  • N3FJP general logging now carries your reports, the operator's name, and power.

    The ACLog push was sending only the contest fields, so RST sent/received, name and TX power logged in Nexus never reached the ACLog side. The Field Day push is unchanged byte for byte — a contest exchange carries none of these. (#106)

  • Changing bands mid-period no longer lets the old band's last seconds decode into the new one.

    Audio captured before a QSY was still decoded at the period boundary after it, so stations from the band you just left repopulated the roster the band change had rightly cleared. That period's decode is now dropped — the next full period on the new band decodes normally. Receive-side only. (#103)

  • The Log QSO button no longer claims success for a QSO it refused to log.

    The engine deliberately refuses a manual log when there is nothing to write — the contact already logged, no active QSO, no report exchanged yet — but the green "Logged QSO" toast showed anyway, so a double-click after auto-log looked like a second entry that later seemed to vanish. A refusal now says so honestly: nothing was logged, and why. (#100)

  • The docked Band activity strip tunes on scroll, like the pop-out band map.

    The strip is the same frequency scale the band map is, in the Phone and CW cockpits both, but a wheel over it did nothing while the map, the readout digits and the waterfall all tuned. It now rides the same wheel — your tuning step, your wheel sensitivity, and the same stop-at-the-band-edge behavior as every other dial. (#96)

  • macOS now says how to get Hamlib instead of quoting a Debian command.

    A Mac operator whose IC-7300/IC-9700 wouldn't connect was told sudo apt install libhamlib-utils — a cure for the wrong operating system. On a Mac the message now says brew install hamlib, the FAQ covers macOS, and the in-app guide no longer claims Hamlib ships inside the installer on platforms where it doesn't. (WSJT-X and the Mac loggers link the Hamlib library; Nexus drives CAT through the rigctld program, a separate package — those apps working is not evidence rigctld is installed.)

  • The rig port picker on macOS no longer lists the tty twin of every port.

    The #92 collapse (1.6.0) filtered rig detection but not the Settings picker, so /dev/tty.* rows still sat beside their /dev/cu.* twins — and a picked tty.* hangs on carrier detect instead of failing. The picker now gets the same collapse; a lone tty.* with no cu.* twin is still kept.

  • Removing a radio: the button is always there now, and a refusal says why.

    Nexus refuses to remove the active radio, but expressed that by not rendering the Remove button at all — from the operator's chair, "the radio won't delete." It now renders disabled with the rule spelled out: make another radio active first. A refused removal is a visible error instead of a silent success, and deleting the radio the rig form was editing no longer leaves later Saves silently writing to a profile that is gone.

  • The ATU button reaches the FT modes.

    Phone, CW and SSTV have carried the rig's built-in-tuner button since it was added; the FT cockpit's TX cluster was missing it. Same rules as everywhere: it appears only when the rig reports a tuner, sits beside Tune, and every refusal is shown with its reason.

  • The signed macOS bundle carries the microphone entitlement.

    Hardened-runtime apps need com.apple.security.device.audio-input before macOS will even offer the microphone permission prompt; without it, RX audio capture can open and deliver silence. The entitlement ships in the bundle now.

Changed

  • The Xiegu and FlexRadio setup pages say what the app actually does.

    The Xiegu page promised that Detect auto-matches an X6100/X6200 — it never could, because those radios report their USB bridge chip and not a model name — and never mentioned the two serial ports they present or the 19200 baud the family runs at. The FlexRadio page told you to launch a second copy of Nexus for a second slice, which is not how multi-radio works (add a second radio, tick "Run both radios at the same time", one window each), and named a PowerSDR model string the dropdown does not contain. Both are rewritten from the code, and the Flex page gains a section on the two early access native toggles: off by default, never run against a real Flex, same-LAN only (they cannot work over SmartLink or through NAT), and native DAX takes over the radio's transmit audio for every client while it is on.

  • macOS: the Local Network permission is now documented where it bites.

    A FlexRadio, or any network rig on the LAN, is reached through a permission macOS 15 gates and Nexus does not yet have a usage string for — so there may be no prompt, and a denial is silent rather than an error: "No radios found", or CAT reporting that nothing answered. The install, troubleshooting and FlexRadio pages now describe that and point at System Settings ▸ Privacy & Security ▸ Local Network.

Every version