Audio Features — Tab5 Real-Time Decoding (Beta)¶
Audio decoding is a BETA feature. These implementations are first-generation and subject to further development. Real-world feedback shapes the roadmap.
Getting audio — start here if you hear nothing¶
Silence is usually one of five things, and they are listed here in the order they actually catch people out. Work down the list; each step says where the control is.
1. Check the mode¶
Only CW, CW-R, USB, LSB and AM produce audio. DiGi and FM produce none, by design — see Which modes produce audio.
This catches more people than anything else, because FT8, FT4, JS8 and WSPR all run in DiGi. If you are on a digital mode and expecting to hear the band, you will not. Switch to USB to listen.
2. RX audio is off by default, and enabling it needs a restart¶
Open the settings drawer → Audio Settings and tick RX Audio (speaker/headphone). Then restart the Tab5.
The restart is not optional and it is not a bug: the audio codec is opened at boot, before the USB host claims the DMA memory, so it cannot be opened safely part-way through a session. From v1.20.1 the Tab5 tells you this with a toast when you tick the box. Before that it said so only in the serial log, which is how three operators arrived at silent audio in one week.
The same switch and volume are on the web API from v1.20.1, in the rxaudio group of
/api/settings.
3. In the browser, press the play button¶
The browser does not start playing on its own. Click the ▶ button at the top right of the page, just to the right of F. It turns into ■ while it plays.
This is the single most-missed step — two users reported "no web audio" in one week and in both cases the stream was working and the button had never been pressed. See Listen in the browser.
4. Know which volume control does what¶
There are three, and they act in different places:
| Control | Where | What it changes |
|---|---|---|
| QMX volume | Drawer → Radio | The radio's own AF gain, over CAT — the same number the QMX shows when you click its volume knob. Affects everything downstream, including the browser stream |
| RX Volume | Drawer → Audio Settings | The Tab5's speaker and headphone level only. Does not affect the browser stream |
| Your browser's own volume | The browser or the operating system | The browser stream only |
A fourth control sets the level before any of these: QMX RF gain (Drawer → Radio), per band, 0–99 dB, QMX default 54. If signals look weak on the waterfall as well as sounding weak, this is the one to look at — not the volume.
!!! note "Full volume and still quiet?"
The level is governed by an **automatic gain control**, not by the volume slider.
If a signal is quieter than you expect at full volume, the AGC ceiling is holding
it back — see [The AGC controls](#the-agc-controls).
5. The speaker and the browser are exclusive¶
While a browser is listening, the Tab5 mutes its own speaker, and hands the audio back when you stop listening or close the page. One listener at a time is deliberate. So if the Tab5 went quiet, check whether a browser somewhere on your network is holding the stream.
Panoramic CW Split¶
Listen to two CW stations on different frequencies, panned left and right across the stereo stage. The CW receiver demodulates each sideband independently, allowing separation by frequency.
How It Works¶
The RF passband is split into two channels: - Left: Lower-frequency station (panned fully left) - Right: Upper-frequency station (panned fully right) - Center blend: Dry/wet mix between the two
Three live controls tune the effect mid-QSO without leaving the exchange, grouped under Panoramic split — adjust while listening:
- Width — Frequency spread between the two stations. Wider = greater separation.
- Blend — Dry/wet audio mix. Left is pure lower channel; right is pure upper channel.
- Overlap — Frequency range for each station. Wider ranges allow more signal capture.
Enabling Panoramic CW¶
- Open the settings drawer and tap Audio Settings (just below SelfSpotter)
- Tick RX Audio (speaker/headphone), then Binaural CW (stereo separation)
!!! warning "Switching RX audio on needs a restart of the Tab5"
The audio hardware claims its memory during start-up, before the USB host
takes that pool, so it is started at boot or not at all. If you boot with
**RX Audio** off and tick it later, the setting is saved but you will hear
nothing until the Tab5 is restarted — it tells you so on screen when you
tick it. Your QMX will need a power cycle after that restart, as it does
after any Tab5 restart.
!!! note "Which modes produce audio"
**CW, CW-R, USB, LSB and AM.** **FM is still silent** — there is no FM
demodulator, so an FM signal has to be listened to in USB or not at all.
AM is demodulated on the Tab5 itself, from the I/Q the radio sends, so the
bandwidth is set by the Tab5 and not by the filter in the radio. Tap **BW**
in the top bar to choose it: **2.0, 3.0, 4.0 or 6.0 kHz**,
quoted as the width of the whole signal, the way amateur AM
figures always are. 6.0 kHz is the default and the standard — 3 kHz per
sideband, what most commercial rigs hard-code. The same list is on the
web page's **BW** button.
Go **narrower** when someone is sitting on top of you: 2.0 and 3.0 kHz
sound like a telephone line and that is the point, they keep a signal
copyable in heavy QRM. 6.0 kHz is the widest on offer and the one to use
on a clear channel. Broadcast channels are spaced 5 kHz, so 6.0 already
reaches into the neighbour's carrier on a crowded shortwave evening.
**6.0 kHz is the widest the Tab5 offers, from v1.20.1.** Four wider
settings were available earlier and have been withdrawn, for two
different reasons.
**10.0 and 20.0 kHz** ran the receive chain at half the decimation the
others use. Measured on the bench, that leaves the second CPU core with
no idle time at all and the audio stutters.
**8.0 and 9.6 kHz cost nothing extra** — matched 60-second windows
measured AM at 4.0 and at 6.0 kHz at the same processor load, so width is
free up to that point and those two would have been free as well. They
are withdrawn because they were never run end to end: they entered the
list on a filter measurement and a listening report, and nothing since
has confirmed them. The list now holds only widths that have actually
been measured playing.
CW and SSB are untouched at any setting.
The three pan controls sit directly under the Binaural CW row in the same window. They stay greyed out until both boxes above are ticked, because they shape a split that is not being produced otherwise.
On the Air¶
All three controls are live-tunable during an exchange: - Slide them while the QSO is running - No reboot, no mode change required - Changes take effect immediately
Binaural CW — First Implementation¶
Frequency-based station separation in CW. Each station gets a distinct audio tone, allowing the ear to distinguish them without losing either signal.
Bandwidth Requirements¶
Binaural separation works best with bandwidths of 300 Hz and above.
Do not expect good separation below 200 Hz — the audio bandwidth is the limiting factor. At 200 Hz or narrower, the two stations' spectra start to overlap, and separation breaks down. This is a fundamental constraint of the audio bandwidth, not the algorithm.
Recommended Bandwidth Settings¶
| Bandwidth | Clarity | Use Case |
|---|---|---|
| 300 Hz+ | Excellent | General CW operation; two stations easily distinguished |
| 200–300 Hz | Fair | Tight pileup; separation possible but marginal |
| <200 Hz | Poor | Not recommended; too much spectral overlap |
Audio Settings¶
Audio decoding shares the Tab5's compute and memory budget with the spectrum, waterfall, and FT8/FT4 decode paths. Not all features can run simultaneously.
!!! note "Renamed in v1.16.4"
This window used to be called **Resource Management**. It is **Audio
Settings** now, because that is what it is — volume, AGC, binaural and the
panoramic split. The old name described how it works rather than what it does.
Opening Audio Settings¶
Open the settings drawer and tap Audio Settings, directly below SelfSpotter.
!!! note "Top-bar shortcut removed in v1.16.8"
A long-press on the top bar used to open Audio Settings too. It shared its
area with the Band, Mode, BW and Frequency pickers, so it was easy to open
the wrong thing by accident. The drawer button is now the only way in.
The window lists RX audio and the background network feeds that compete with it for the same scarce memory, each with a tick box:
- RX Audio (speaker/headphone) — with RX Volume and the three AGC controls described below
- Binaural CW (stereo separation) — with the three panoramic-split controls
- SelfSpotter, RBN, DX cluster, PSK Reporter — who's hearing me — these grey out and are held off while RX audio is on
- POTA / SOTA spots, PSK Reporter — report my decodes — these keep running
The list scrolls; the Close button stays put at the bottom.
Toggle audio on and off as needed — the resource footprint is immediate, and disabling a feature instantly frees that headroom.
Typical Constraints¶
- Audio ON, Waterfall ON, Decode ON → Some margin left; FT8/FT4 decode still runs steadily
- Audio ON, Waterfall ON, Decode ON, Web streaming ON → Tight; any spike can cause temporary frame drops
- Audio ON, full WiFi + web load → Audio is the first to suffer if the load spikes; consider disabling audio or the waterfall if you need reliable web streaming
Headphones¶
Plug headphones into the Tab5's 3.5 mm jack and the internal speaker goes quiet on its own. Unplug them and it comes back. Nothing to switch.
Before v1.16.4 you heard both at once, which made headphones close to useless in a quiet shack (Roy KI0ER).
!!! tip "The internal speaker is small and trebly"
Several operators have said so, and they are right — it is a tablet
speaker. Headphones or a powered external speaker on the jack are a large
improvement, and now that the internal one mutes itself either is
practical.
Which modes produce audio¶
Audio and the panoramic split are not the same thing, and it is worth being clear about which you get where.
| Mode | Audio | Panoramic / binaural split |
|---|---|---|
| CW, CW-R | Yes | Yes |
| USB, LSB | Yes | No — single channel, both ears the same |
| AM | Yes, to about 3 kHz | No — AM has no stereo content to split |
| DiGi | No | No |
| FM | No | No |
So you do get audio on SSB and AM; what you do not get there is the left/right split, because that is a CW feature. In DiGi and FM there is no audio at all — nothing is broken if you hear silence in those.
Known Limitations¶
- Audio breaks up while a web page is open. Known and unfixed as of v1.16.3. With a browser connected to the panadapter page the audio task misses some of its deadlines and you hear a regular crackle; closing the page clears it. Measured, but the cause is not yet found — if you need clean audio, close the web page.
- Stereo rendering requires a headset or speaker with stereo output. Mono output collapses both channels; panning does not work.
- The split is CW-only (see the table above). On SSB you hear the signal, but it is not placed left or right.
- Limited to the audio passband. If the audio filter is 300 Hz wide, the maximum usable separation is about 300 Hz. A wider filter in the radio settings increases the available range.
- Level is set by an automatic gain control, not by the volume slider. If a signal is quieter than you expect even at full volume, the AGC's ceiling is what is holding it back. All three AGC controls are adjustable live — see below.
The AGC controls¶
All four audio level controls live in Audio Settings (settings drawer → Audio Settings), on their own rows under RX Audio:
| Control | Range | Default | What it does |
|---|---|---|---|
| RX Volume | 0 – 100 | 60 | Output level to the speaker or headphones. |
| AGC Ceiling | 50 – 1500 | 200 | How far the AGC may lift a weak signal. Raise it if quiet stations stay quiet. |
| AGC Attack | 1 – 50 ms | 3 ms | How fast the gain comes down when a signal arrives. Shorter catches sharp CW edges; too short can sound clipped. |
| AGC Release | 10 – 500 ms | 150 ms | How fast the gain comes back up afterwards. Longer is steadier on CW; shorter follows speech better. |
All four are remembered across restarts and travel in a config backup.
AGC presets. Four buttons under the sliders set the timing in one tap:
| Preset | Attack | Release | For |
|---|---|---|---|
| Fast | 1 ms | 30 ms | CW |
| Med | 3 ms | 150 ms | general, and the shipped default |
| Slow | 5 ms | 400 ms | SSB, where a fast release pumps on speech pauses |
| Off | — | — | no automatic gain at all |
Move either slider and the caption reads Custom — the sliders are the custom setting, so there is no fifth button.
Off is a real bypass, not a very slow AGC. The gain follows the signal at every attack and release value, so no combination of the sliders switches it off. With Off selected the gain is pinned at AGC Ceiling, and that slider becomes a plain manual gain control.
⚠ What to expect when you try them. The AGC only starts regulating once a signal is strong enough; below that it already sits at the ceiling, so the gain is the same number whether the AGC is on or off. On a quiet band every preset correctly sounds identical — the difference is heard in how fast the noise returns between signals when the band is busy. The attack values (1–5 ms) are all far quicker than the ear resolves; Release is the audible one.
Volume and Ceiling are not the same thing. Volume sets how loud the output is; Ceiling decides how hard a weak signal is lifted before it gets there. Running Volume to 100 does not mean you have run out of level — if it is still too quiet, raise AGC Ceiling.
What changed in v1.16.3¶
- AGC Ceiling was called "Gain". The name was wrong: it never was a plain gain, it is the ceiling on the AGC's own weak-signal boost, and calling it Gain led people to look for an AGC that was already there. Its range also went up from 800 to 1500, because the Tab5's internal speaker wants more headroom than the old ceiling allowed.
- Attack and Release are new — the AGC's timing was fixed before, and is now yours to set.
- Presets arrived in v1.16.5 (Samuel W7STF), along with a genuine Off.
- RX Volume moved here from the settings drawer, and the drawer's RX Audio section is gone entirely. That includes its on/off box: Audio Settings is now the only place to switch RX audio on or off.
What changed in v1.16.4¶
- The window is called Audio Settings, and it has a door of its own in the settings drawer. The top-bar long-press still works.
- Headphones mute the internal speaker automatically.
Strong signals¶
A peak limiter sits after the automatic gain, so a loud station cannot arrive at full scale and startle you in headphones. It holds the peaks down and leaves the average level alone, which means raising AGC Ceiling to hear weak stations does not make the strong ones painful.
This is the part of the beta I most want reports on: which mode, roughly how strong the signal was, and whether more ceiling fixed it.
Where the receive signal comes from¶
Until now the panadapter always listened to the radio's USB audio. There is now a receive source selector, on the Tab5 and in the browser, with a level meter and a clip indicator beside it.
| Source | State |
|---|---|
| QMX USB | The default, and what every earlier release did. |
| Line in (3.5 mm jack) | Declared beta - see the warning below. |
| Generic I/Q | Declared, not implemented. Selecting it is refused. |
| Internal mics | Declared, not implemented. Selecting it is refused. |
Line in (asked for by Tony, @tony1tf, in issue #17) feeds the panadapter from the Tab5's own 3.5 mm headphone jack instead of the radio's USB. Tony ran a test build on his own Tab5 to find where the jack lands — the mic ring arrives on the audio chip's third channel — which is the measurement the whole feature rests on. The jack is a single real audio channel, not I/Q, so the second channel is synthesised: a 255-tap filter builds the quadrature pair the decoders need. The jack wiring is drawn on screen so you can see which contact carries what.
⚠ No signal has been decoded through line in yet. The path runs, samples flow and the levels read correctly with nothing plugged in - that is the whole of what has been verified. Treat it as unfinished. Changing the source needs a restart, and the Tab5 does the restart itself when you press Save.
⛔ In line-in mode the USB radio connection is off by design. One or the other, not
both - so CAT and the S-meter go quiet while line in is selected. The Tab5 stops
pretending to have a dial with it: the frequency readout reads Line-in mode, the
band shows --- and the band-plan strip is hidden, because an audio spectrum has no
band plan and no RF frequency. The frequency axis under the display reads audio Hz.
The left half of the spectrum is empty and there is a step in the middle. That is correct and it is the honest picture: the jack carries one real channel, the second is synthesised, and a synthesised channel cannot recover what the missing one would have told you. The step sits at audio 0 Hz, which is the centre of the span.
Listen in the browser¶
From v1.20.0 the panadapter streams its own demodulated audio to any browser on your network. Open the web page and click the ▶ button at the top right, just to the right of F. It becomes a ■ while it is playing; click it again to stop.
What you hear is exactly what the Tab5's own speaker gets — the same demodulated audio, tapped after the filter and the AGC — so the two can never disagree about what the radio is doing. It works even if the Tab5 booted with its speaker turned off.
The speaker and the browser are exclusive. While a browser is listening the Tab5 mutes its own speaker, and it hands the audio back when you stop listening or close the page. There is no setting for this: one listener at a time was the point.
Only the modes the Tab5 demodulates produce audio: CW, CW-R, USB, LSB and AM. In any other mode the button lights up but nothing arrives, and hovering it says so.
Hover the ■ button for the health of the stream — it reports frame gaps, re-primes and chunks dropped to clock drift. On a quiet network all three stay at zero.
Notes on what this costs and what it does not:
- It runs over the web page's existing connection, so it needs no extra port and no extra setup.
- It costs about 9 percentage points of one CPU core — measured — and nothing measurable on the core that serves the web page.
- Audio is 12 kHz µ-law, which is a telephone-grade codec chosen because it survives a lossy Wi-Fi link without a jitter buffer deep enough to be annoying.
- Only one browser gets the page at a time. Opening the panadapter page in a second browser takes the connection from the first, which takes the audio with it.
Feedback¶
This feature is beta and in active development. Your feedback shapes the next iteration: - Does panoramic CW work in the field? - What bandwidth do you typically use? - What other audio features would you like to see? - Did you hit the resource limits?
Send feedback to the discussions page or open an issue.