Repository navigation
Releases
astar 0.1.22-beta
A release for the node. astar-server now holds a link to an AllStarLink node
the way a stock node does, passes speech through without damaging it, and
keeps a record of every link, so a drop is no longer a mystery.
Added
- A link log on the node. astar-server writes one line when a link
starts, when the codec is agreed, when the link is up and when it ends. The
last line says who hung up (us, them or a time-out), the reason given, and
how long the link lasted. A link that ends within ten seconds of coming up
also gets a warning line, because that is what a node "dropping off" looks
like. See
Reading the link log.
Changed
- The parrot keeps a linked AllStar node connected. In parrot mode
astar-server used to hang up after every echo, which suits someone calling
in for a signal report but made a linked node drop a few seconds after each
over. A node that links in now stays linked and can talk again. Other
callers are still hung up after their report.
Fixed
-
Unfinished calls no longer lock everyone out of the node. A caller that
started a call and never finished the opening handshake held one of the
node's call slots until the server was restarted. On a node open to the
internet, stray packets could fill every slot, after which every real node
was turned away as busy. Such a call now ends by itself, and the node no
longer sets anything up for a caller until that caller has proved its
address by returning a signed token, the way Asterisk does it. The node
also used to accept a token it had never issued; it now checks them. -
Speech passes through the node's bridge undamaged. The bridge read each
member's audio once every 20 ms with no slack, so a packet a few
milliseconds late missed its turn and about half the overs picked up a
20 ms stall or a 40 ms repeat. Each member now has a small buffer that
absorbs late packets. It adds about 40 to 60 ms of delay. -
The parrot plays back what you said, pauses included. When a node keyed
by voice level, the parrot dropped the quiet between words, so a five-second
over came back as under four seconds with the words run together. A late
packet could also leave a hole in the recording. Both are fixed.
Full changelog: https://rcludwick.github.io/astar/changelog/
astar 0.1.21-beta
Bluetooth headsets that work for a whole call, a push-to-talk button you can
hold in your hand, and a directory search that finds AllStar nodes and
EchoLink stations alongside the reflectors.
Added
-
A Bluetooth push-to-talk button. The Inrico B02PTT, a Bluetooth
speaker-mic with a push-to-talk button, keys astar while you hold the
button, the same as the Space bar. Choose it as the PTT in Quick settings →
Station, press it once to prove it is yours, and astar finds it by itself
from then on. It goes through the same transmit time-out as everything
else, and if the link drops while you are transmitting, astar stops within
about 2.5 seconds and waits for you to let go before it will key again.
See
Bluetooth PTT. -
AllStar nodes in the directory search. The search pane has an AllStar
filter, and All networks includes AllStar nodes as soon as you type. Search
by node number, callsign, town or frequency. Node numbers now show with
their callsign and location in recents, favorites and the call header. The
list ships with the app, so it works on first launch with no network, and
refreshes itself weekly from hamcall-db. -
EchoLink stations in the directory search. While you are logged in to
EchoLink, the filter in the top right of the search pane offers EchoLink,
and All networks includes the stations too. If the login drops, the filter
goes back to All networks, and returns to EchoLink when you log back in.
See Searching the reflector directory. -
Net recordings you can browse. The net recorder from 0.1.20-beta now
has pages to go with it: a list of recorded nets, each session's overs
with a waveform and who spoke, and a player that jumps to any over.
astar-server serves them under/nets/behind sign-in, and
astar-netrecserves them on 127.0.0.1 on the Mac it runs on. See
Net recordings.
Changed
-
VOX ignores clicks. VOX used to key on the single loudest sample of
the last few milliseconds, so a tap on the mic, a breath pop or a Bluetooth
glitch keyed you for the whole hang time. It now listens to how loud the
mic has been over the last 50 ms and keys only once that has stayed above
the threshold for 120 ms; nothing shorter than about 70 ms can key it.
To keep the first word whole, astar sends the 250 ms before VOX keyed
along with the over. Because loudness reads about 10 dB lower than peaks,
your threshold is converted once (−40 becomes −50), and the slider now
stops at −58, because a reading at the very bottom of the meter is
silence. While VOX is on, the other ways of keying send that 250 ms too. -
Settings is easier to scan. Audio configs (renamed from "Saved audio
configs") and Mic Profiles are at the top, and Transmit folds away like
every other section.
Fixed
-
A Bluetooth headset no longer goes silent when the call starts. Opening
a Bluetooth headset's microphone switches it into its hands-free mode,
which changes its sample rate and used to kill the speaker for the rest
of the call: astar said Connected and you heard nothing. astar now opens
the mic first and rebuilds a stream whose device changes under it. If a
rebuild ever fails, the window says "Speaker stopped" instead of staying
quietly deaf. -
Turning VOX off no longer forgets its settings. The threshold and hang
time went back to their defaults until the next launch.
Full changelog: https://rcludwick.github.io/astar/changelog/
astar 0.1.20-beta
A way to capture exactly what a call receives, so a bad moment on the air
can be studied afterwards, and a net recorder that joins a reflector on a
schedule and keeps what it hears.
Added
-
Packet capture, for chasing audio problems. Some problems, like the
hollow echo heard on a long EchoLink call, only show up on the air and are
gone by the time you reconnect. Packet capture keeps the evidence. Turn on
Debug mode in Settings → Recording → Advanced, and a capture button
with a ladybug appears next to Record (⇧⌘R). Press it during a call and
astar writes every packet that call receives to a.pcapfile, which
Wireshark opens, with a small.jitter.jsonlfile beside it recording how
late each packet arrived. Press it again, or hang up, and it stops. The
packets are saved as they arrived, after any decryption, and they are
written on their own thread, so a capture can't disturb what you hear.
The folder keeps itself under a size limit you choose, deleting the oldest
files first. These files hold other people's voices, so keep them to
yourself. See
Packet capture. -
The node can capture packets too. astar-server has the same switch
([packet_log] debug_modeinnode.toml, or Debug mode on its Settings
page) and a capture button on its Recordings page, where the captures can
also be downloaded. See
Packet capture on the node. -
A net recorder.
astar-netrecis a small receive-only program for a
Linux machine or a Mac. You list your nets (which M17 or D-Star reflector,
which day, what time) and shortly before each one it joins, records the
whole evening plus one file per over, logs who talked when, and leaves. It
has no microphone or push-to-talk line, and three separate layers keep it
from ever keying. It is built from source for now. See
Net recorder.
Changed
- The Account section of Settings is now called AllStarLink, which is
what it holds: your callsign, node number and allstarlink.org password.
Fixed
- Settings → Recording → Advanced opens. Clicking it turned the
triangle, but nothing appeared below it, because the whole group sat
inside one row of the Settings list and the row never grew. Each setting
under it is now a row of its own.
Full changelog: https://rcludwick.github.io/astar/changelog/
astar 0.1.19-beta
EchoLink that stays in time on a busy conference, a built-in D-Star vocoder,
recordings without a beat, and a Settings pane you can fold.
Added
-
D-Star can run without a vocoder stick. astar now has a built-in
software vocoder for D-Star: Bruce Perens' AMBE 3600×2400 implementation.
Choose it in Settings → Vocoders, or with[vocoder] dstar = "software"innode.toml; the dongle stays the default, and nothing
switches by itself. With no stick added, the Vocoders section and a
refused D-Star connect offer it in one click. It is D-Star only: System
Fusion, NXDN and DMR still need a dongle, because their AMBE+2 codec is
patented until 2028-05-20. The vocoder is LGPL-3.0; see
LICENSE-EXCEPTIONS.md§9. See
Vocoders. -
Settings folds. Every Settings section sits behind a disclosure
triangle, with Expand All / Collapse All beside the title and
Option-click to open or fold them all. astar remembers which sections you
left open; a fresh install opens just Operator and Account. Buttons that
jump to a section, such as "Open Vocoder Settings…", open it on the way. -
The node warns that an M17 reflector it monitors goes out on your
links. The monitor never sends voice to the reflector, but what it hears
is relayed to every AllStar node and EchoLink station on the node, and a
linked radio transmits it. The admin page, the settings page's M17 card,
thePOST /m17reply (relays_to,warning),/status(m17_relay),
Discord's/dial m17reply and the log all say so, with how many stations
are connected at that moment. The 0.1.17-beta notes and the Discord page
understated this and are corrected.
Fixed
-
EchoLink calls no longer fall further and further behind on a busy
conference or repeater that never pauses, which could leave the audio
sounding hollow. Received EchoLink audio now plays through the same
adaptive jitter buffer as AllStarLink (the Jitter buffer switch and
its Min/Max apply), and the call-quality line under the meters now shows
on EchoLink calls too. -
The jitter buffer gives back an over-deep cushion and recovers after a
stall, on AllStarLink as well as EchoLink. It now drops the odd frame to
shrink while audio keeps arriving, as Asterisk's does, and after a network
stall it skips the frames that are badly overdue instead of staying behind
for the rest of the call. -
Recordings no longer have a regular beat. An app recording of an
EchoLink call could carry a flutter about once a second: the recorder
wrote short stretches of silence the speaker never played, then dropped
audio to catch up. Recordings of M17, D-Star, DMR, YSF and NXDN calls on a
44.1 or 48 kHz output had the same flaw. -
dstar-listenand the other listen tools build again with the
software vocoder in, and the YSF, NXDN and DMR listen tools build on their
own.
Full changelog: https://rcludwick.github.io/astar/changelog/
astar 0.1.18-beta
Favorites shared across every network, EchoLink on the same dial row as the
rest, an EchoLink section in Settings, sharp station text and even call
buttons.
Updating from 0.1.16-beta? 0.1.17-beta wasn't released on its own, so its
notes apply to this update too (see the full changelog): EchoLink, recording, and the server's
public page, sign-in and Discord bridge. One of them needs you to do something.
If you use D-Star, System Fusion, NXDN or DMR, add your DVstick or ThumbDV
once in Settings → Vocoders, because astar now opens only the vocoder
sticks you've added.
Added
-
Favorites shared across every network, and EchoLink on the same dial
row. The saved-nodes menu beside the dial field now lists favorites and
recents from every network, whichever one is selected, each marked with its
network: pick a DMR talkgroup while on AllStar and the picker switches to
DMR with the network, talkgroup and timeslot filled in, ready for
Connect. Saved nodes are keyed by network and target, so EchoLink node
1234 and AllStar node 1234 are separate favorites, and Settings → Favorites
lists them all with a network tag. EchoLink no longer swaps the dial row
for a form of its own: type a callsign or node number in the same field,
or click the magnifying glass to browse the station list in its own pane
and choose a station into the field. One line under the row logs in or out
with what Settings → EchoLink saved, or points you there. Stars in the
station browser, on the dial row and on a call are the same favorite, and
the browser's Favorites filter shows them with their live free/busy
dot, still listing the ones that aren't online right now. They travel in
an exported config with the node directory. -
Settings → EchoLink. The EchoLink callsign, password and Log in
automatically now have a section in Settings, beside the AllStarLink
account: the one place they are edited now that the dial card has no login
form. The eye button in the password field shows the saved
password. The section also holds the location text other stations see
beside your callsign, which had no control before, plus Log in / Log out
and a Clear that asks first. None of it goes into an exported
.astarconfig.
Fixed
- The EchoLink station list's text was blurry. Switching the network to
EchoLink slid the station list in from the top. The list is an AppKit table,
and the slide left its scroll position on a fraction of a pixel, so every
row's text sat between pixels and came out soft, most visibly on a non-Retina
(1x) display. The list now fades in instead, and lands on whole pixels. - The call controls are one height. Hold to Talk and Record grew from
their text and symbol and stood 35 pt tall, while Disconnect, whose phone
symbol is short, stood 28 pt. All of them, including the VOX, listen-only and
receive-only blocks that take the push-to-talk button's place, now share
Disconnect's 28 pt, and the recording clock no longer changes it.
Full changelog: https://rcludwick.github.io/astar/changelog/
astar 0.1.16-beta
The macOS 27 release: the main window laid out again, and the frosted look
reaches the top edge.
Changed
-
The title bar is part of the frosted window. The blurred material now
runs from the top edge down, with the traffic lights and title sitting on
it, instead of a separate strip above the content. -
The network picker folds instead of clipping. Regular where it fits,
small where that fits, and a popup menu below that — every network stays
reachable at every window width. A fresh call window opens at 470 pt,
wide enough for the full control on macOS 27; a window you have already
sized stays the size you left it. -
The Discord invite is on the front page and in the site's title bar,
and the README no longer carries version numbers that can go stale.
Fixed
-
The title bar was completely transparent on macOS 27. A titled window
with a clear background gets no title bar backdrop at all on 27 — the
traffic lights and title floated on the bare desktop. The window keeps its
stock background now; the blur never needed a clear one. -
The window's minimum width cut both edges off every row on macOS 27.
Its roomier segmented controls pushed the six-network picker past the
310 pt minimum, and the scrolling pane centred and clipped the overflow.
See the folding picker above. -
The jitter buffer sliders drew a dotted smear in a narrow window. Those
were tick marks, seventeen per slider. The sliders still land on 10 ms
steps; they just no longer draw them.
Full changelog: https://rcludwick.github.io/astar/changelog/
astar 0.1.15-beta
Received AllStarLink audio gets a jitter buffer, and the "+" that adds a
config or a mic profile is one visible button everywhere it belongs.
Added
-
The jitter buffer is a setting, on both clients. Quick settings →
Speaker gains a Jitter buffer switch and the depth window it may adapt
between — Min (40 ms) and Max (200 ms), Asterisk's own numbers. On
by default, live mid-call, saved and exported with the rest of the audio
settings. No config-version bump: a reader that predates the keys reads the
defaults. -
A call-quality line under the meters. On AllStarLink:
jitter 12 ms · buffer 60 ms · lost 0 · late 0, with· underruns N
appended only when there are some — that one is the alarm, not a statistic.
Those figures are in the snapshot, the C header and both bindings too.
Changed
-
Adding a saved config is one obvious button, and the new config is
there. The label wedged above the list is now a + on the Saved configs
header. The config it makes arrives expanded, scrolled into view, with its
name selected so you type over it — and the config you are running stays
the one you are running. -
A + starts a new mic profile everywhere one can start. Mic Profiles,
Quick settings' mic row, a saved config's mic row and the analyzer's own
header all carry it, and it opens a blank sheet with the keyboard already
in the name field. "Open Mic Analyzer…" and both "Analyze…" links are gone:
three names for one action.
Fixed
- AllStarLink audio no longer stutters when the network does. Received
voice went straight from the wire to the speaker, so a packet more than
about 10 ms late played as a hole and a call that started choppy stayed
choppy. The receive path now runs an adaptive buffer on the playback clock,
absorbing late and reordered packets and interpolating the odd lost one.
Off by a switch if you want the old path.
Full changelog: https://rcludwick.github.io/astar/changelog/
astar 0.1.14-beta
One fix, reported on GitHub the day the public repo opened.
Fixed
-
The Keychain items are filed under
com.astar.app, not the author's
callsign. Keychain Access shows a password item by its service name, and
astar's wascom.aj7hr.astar— so what a new user saw first was somebody
else's callsign. Both items, the AllStarLink account and the DMR passwords,
now live undercom.astar.app. Anything saved under the old name is moved
on the next launch and the old item removed; nobody re-enters a password.
(Fixes #1.) -
The unit tests no longer touch the login Keychain. One test built a real
session through the live factory and read the Keychain on every run, which
raised a password prompt — and parked the whole suite when the screen was
locked. It uses in-memory stores now.
Full changelog: https://rcludwick.github.io/astar/changelog/
astar 0.1.13-beta
A microphone-setup release, mostly. The analyzer is a pane of the main window
rather than a window of its own, it opens on the microphone your profile
actually uses, and its threshold is now a level you can read straight off the
canvas instead of an estimate drawn a few dB too high. Last heard grew from one
name into a short history, the mic gains reach as far as the speaker's, and the
AllStarLink account asks for nothing but a password.
-
The analyzer's threshold is an absolute level in dBFS, and the background
noise has a line of its own. The slider used to read "Noise floor +N dB"
and the dashed line was tagged "(est.)" because it genuinely was one: it came
from the coarse display spectrum while the detector medianed a finer FFT, so
it sat several dB high and notches landed visibly beneath it. The slider is
now a threshold in dBFS (-100 to -20, default -60) drawn on the canvas's own
axis, and the background line is drawn and labelled separately. Nothing is
estimated any more: a peak above the orange line is a peak Analyze notches. -
A microphone with nothing above the threshold saves as a pass-through
profile. A clean mic used to come back with notches invented out of its
noise floor. Now the profile records that there was nothing to remove, and
says so. -
Mic Level and TX Gain reach 400%, matching the speaker.
-
**The AllStarLink account asks for a password only,
Full changelog: https://rcludwick.github.io/astar/changelog/
astar 0.1.12-beta
A System Fusion release, mostly. Your over now carries your callsign, so the reflector and the radios on it know who is talking, and the transmitted audio is no longer scrambled on the way out: every frame had been permuted a second time before it left, so 1 bit in 49 landed where a receiver expected it. Receive was never affected, and neither was D-Star.
Last heard now names whoever keyed up on M17 and System Fusion, as it already did on D-Star. The AllStarLink account asks for a password only; the portal never needed the node number.
Signed, notarized macOS build, arm64 only, macOS 13 or later.
Full changelog: https://rcludwick.github.io/astar/changelog/