Aetherial Audio is AetherSDR's client-side audio DSP suite. Since v0.9.0 it runs on both TX and RX with fully independent state on each path: shape your microphone audio on the way out, and shape what comes out of your headphones on the way in. Every stage runs on your computer at the radio's native 24 kHz sample rate, independent of the radio's onboard processing. The intent is to give FLEX users the same caliber of broadcast-style audio shaping that studio engineers have in DAWs, without leaving the SDR.

TX chain (microphone shaping, before encoding to the radio):

[MIC] → [GATE] → [EQ] → [DESS] → [COMP] → [TUBE] → [PUDU] → [VERB] → [TX]

RX chain (received-audio shaping, before your headphones):

[RADIO] → [DSP] → [EQ] → [GATE] → [COMP] → [TUBE] → [PUDU] → [SPEAK]

Both chains use the same DSP cores (Parametric EQ, AGC-T expander/gate, AGC-C compressor, Tube saturator, PUDU exciter) but with separate parameter sets so you can dial in transmit and receive independently. The applet titles read Aetherial Parametric EQ, Aetherial AGC-T, Aetherial AGC-C, Aetherial Tube, Aetherial PUDU, with the side (TX or RX) appended.

Aetherial Audio TX processes PC microphone audio. To route your mic through the chain you need:

1. Mic source = PC (set in the P/CW applet)
2. DAX TX = OFF (DAX bypasses client-side processing)

When both are correct the CHAIN widget's [MIC] endpoint turns green, mirroring the SQL button's colour in the RX Controls applet. If [MIC] is grey, your mic audio is not going through the chain — double-check the P/CW applet.

The [TX] endpoint pulses red while you are transmitting on your own slice (MultiFlex transmissions from other clients don't trigger the pulse).

The RX chain processes the audio the radio is already streaming to you, so there's no source-routing setup. The bookend tiles report status:

• [RADIO] turns green when PC audio is flowing from the radio (something to process).
• [DSP] is the client-side noise-reducer indicator. When NR2, NR4, or BNR is engaged it shows that name (e.g. NR4) with the same blue-ring / green-dot styling as an active stage; when no client NR is active it reads DSP and dims out.
• [SPEAK] turns green when your output is unmuted.

If [RADIO] is grey there's nothing to process — usually a disconnected radio or a stalled stream. If [SPEAK] is grey your output is muted at the panel level.

The CHAIN applet — top of the Aetherial Audio container — is the central interaction surface, and drives both TX and RX depending on which side is selected at the top.

Reordering: drag [COMP] in front of [EQ] and the chain runs compressor-before-EQ. The applet-stack panel below the CHAIN re-orders its tiles to match so your vertical scroll matches the signal flow. TX and RX each have their own chain order — drag on TX changes only TX, drag on RX changes only RX.

• [TX] / [RX] — side toggle. TX shows the microphone chain, RX shows the received-audio chain. The applet stack below CHAIN filters to match the selected side.
• [⏺] / [▶] — PUDU monitor (record/playback — see below). TX-only; hidden when the RX side is selected because there's nothing to pre-audition (you're already hearing the chain output live).
• [BYPASS] — one-click disable of every currently-enabled stage on the selected side. Click again to re-enable only the stages that were on before. Useful for A/B comparing processed vs dry signal. Each side has its own BYPASS state.

The RX chain is brand-new in v0.9.0 and tends to surprise operators the first time they enable it. A few notes to set expectations.

[RADIO] → [DSP] → Aetherial RX stages → [SPEAK]

The radio's onboard AGC, RX EQ, NB, NR, etc. all run before the audio gets to your computer. The Aetherial RX chain processes that already-decoded audio — it doesn't replace the radio's RX DSP, it adds finishing on top. This is by design: the radio's noise reduction and AGC are well-tuned for SSB band conditions; Aetherial RX is for the listening polish that's hard to do at the radio level.

Each Aetherial stage holds two complete parameter sets — one for TX and one for RX. Switching the CHAIN's side toggle from TX to RX flips every applet tile and editor to the RX values. There is no "copy from TX" button: settings that work on transmit (compression designed to fight intermodulation, EQ tuned to your microphone) rarely sound right on receive.

• Aetherial Parametric EQ — RX: pull harshness out of bright stations, add chest to thin ones, tame a hot 800 Hz bump that fatigues over a long contest. Set the LPF to whatever your slice filter is (3 kHz for SSB) — boost above that is wasted.
• Aetherial AGC-T — RX: the gate, used as an expander, can knock down inter-syllable hiss between transmissions on a noisy band without adding the pumping you'd get from a hard gate. Use Expander mode (green button), not Gate mode.
• Aetherial AGC-C — RX: low-ratio compression (1.5:1 or 2:1) on RX evens out signal-strength variation across operators, so a pile-up doesn't make you ride the AF gain. Aim for 2-4 dB of reduction, not the heavy hand you'd use on TX.
• Aetherial Tube — RX and Aetherial PUDU — RX: useful in small doses for adding warmth or presence to thin-sounding stations. Easy to overcook — your ears will tell you when.

• Heavy compression on RX: makes weak signals leap up alongside strong ones — exactly what the radio's AGC was already trying to prevent. If you find yourself fighting AGC pumping, that's the parameter to fix, not RX compression.
• High Tube drive on RX: you can't fix a noisy band with saturation. The tube character is musical on TX where you control the source; on RX it just adds harmonic mud to whatever the band is doing.
• De-essing on RX: there's no DESS stage in the RX chain — that was an intentional omission. Inter-station sibilance varies dramatically by operator; tracking it with a single sidechain produces audible artefacts on every voice.

Each Aetherial stage's floating editor has a frameless title bar that reads, e.g. "Aetherial Parametric EQ — RX" or "Aetherial AGC-C — TX". Double-click the title bar to maximise, drag it to move, use the trio at the right edge to minimise / maximise / close.

Every stage has:

• An applet tile — compact dashboard view, 4–5 most-used knobs.
• A floating editor — full parameter set, larger viz, tuning depth.

Knob values stay synchronised between tile and editor at 30 Hz, so tweaking one surface updates the other live.

Purpose: kill room noise, HVAC rumble, and mic hiss during your pauses so the rest of the chain isn't feeding on garbage.

Modes

The Flip button at the bottom of the editor toggles between the two preset pairs. All other parameters (attack/release/hold/return) survive the toggle, so you can fine-tune from either starting point.

Key parameters

Viewing the action: the big Ableton-style level view in the editor shows input envelope (white line) on top of an amber "audible" band; reduction appears as dark gray above the amber. Two cyan lines mark Thresh and (Thresh − Return).

Example: SSB voice with room noise (-50 dB hiss, voice peaks -15 dB)

1. Start in Expander mode (green button).
2. Set Thresh to -35 dB — just above your noise floor, below your quietest voice syllables.
3. Ratio at 2.5:1 — gentle attenuation on quiet stuff.
4. Attack 1 ms, Release 80 ms — fast enough to catch the opening of a word, slow enough not to chatter on breath noise.
5. Hold 20 ms — prevents stutter on sibilants.
6. Floor -15 dB — caps attenuation so the gate never fully slams; preserves room tone.
7. Return 3 dB — prevents flutter when voice hovers near Thresh.

Watch the viewer: the amber band should stay full during speech and shrink between words. If it's pumping audibly (fast open/close on syllables), lengthen Release.

Purpose: shape tonality. Cut the mud, add presence, tame harshness.

The EQ has fixed frequency anchors (HPF / LF shelf / 7 peaks / HF shelf / LPF), four filter families (Clean / Warm / Vintage / Modern), and a per-band gain + Q (bandwidth). It replaces the radio's onboard TX EQ — leave the radio EQ flat when Aetherial is active.

Editor layout

• Icon row — one button per band, click to select for editing.
• Response curve — live FFT analyzer overlay (post-EQ signal) with band handles you can drag.
• Parameter row — Freq / Gain / Q for the selected band.
• Output fader — overall EQ output level.

Example: male baritone voice, stock dynamic mic

A common "broadcast" curve:

Always set the LPF to match your voice filter (3 kHz is the SSB default). Boosting above the filter is wasted energy.

Example: bright condenser mic

Flip the LF/HF relationship — add a little low-mid warmth, pull back 2-4 kHz to tame sharpness, roll off above 4 kHz.

Purpose: suppress sibilant energy ("sss", "shh") that sounds harsh in compressed SSB, without dulling the rest of the signal.

De-essing is sidechain dynamics: a bandpass filter watches the sibilant frequency band, and when its energy crosses threshold, broadband attenuation (up to Amount) is applied.

Key parameters

Example: harsh "esses" on a bright mic at 5 kHz

1. Set Freq = 5500 Hz, Q = 2.5 — tight-enough band to target sibilants without catching "t" consonants.
2. Speak a test phrase with plenty of "sss" — the curve widget shows the bandpass response with a live ball at the current centre. Watch the meter: sibilants should peak above -20 dB.
3. Set Thresh = -25 dB — below the sibilant peaks, above your normal vowels.
4. Amount = -6 dB — gentle, musical. Start conservative; go to -10 or -12 dB only if you still hear harshness.
5. Attack = 1 ms, Release = 80 ms — fast enough to catch the onset, relaxed enough to not pump.

How to find your sibilant frequency: play with Freq while saying "sister silver cellophane" — you'll hear the bandpass sweep. Park on the frequency where sibilance is loudest.

Pitfall: too low a Thresh means the de-esser is always on, which dulls the whole voice. Too high and sibilants slip through. The viewer's GR bar should flash on peaks, not flatten.

Purpose: level the signal. Bring quiet syllables up, hold loud ones back. The result is consistent perceived loudness without clipping.

Key parameters

Limiter: the LIMIT button on the editor enables a brickwall peak limiter after the compressor. It catches transients the compressor missed and hard-stops the signal at Ceiling. The LIMIT button glows red when actively clamping — useful for confirming the limiter isn't constantly working (which would indicate over-aggressive Makeup).

Example: broadcast-style "always-on" compression for SSB

1. Thresh = -20 dB — compressor engages for everything above moderate voice level.
2. Ratio = 4:1 — firm but musical.
3. Attack = 20 ms — preserves consonant transients.
4. Release = 150 ms — smooth recovery, no audible pumping.
5. Knee = 6 dB — soft-knee for transparent "talk into it" engagement.
6. Makeup = +6 dB — pulls compressed signal back up to peak around -3 dB on the Out meter.
7. Ceiling = -1 dB, Limiter ON — safety net on peaks.

Watch the GR meter: 3-6 dB of reduction on voice peaks is musical. More than 10 dB is obvious compression — acceptable for contest or DX, fatiguing for rag-chew.

Threshold fader: the vertical fader on the left of the editor combines the input level meter with the threshold handle. Drag the amber handle to set threshold while watching your voice's peak level right next to it — very intuitive.

Purpose: add analog-style warmth and harmonic complexity. Replicates the character of a tube preamp without the hardware.

Models

Switch between models with the A/B/C buttons stacked vertically between the Tone and Bias knobs.

Key parameters

Example: subtle analog warmth

1. Model = A, Drive = 6 dB — gentle saturation.
2. Tone = +0.2 — hint of brightness to compensate for tube-style HF softening.
3. Bias = 15 % — tiny asymmetry, a whisper of even harmonics.
4. Mix = 40 % — blend with dry signal to preserve clarity.
5. Output = -1 dB — match unity gain.

Example: broadcast "tube mic" sheen

1. Model = C, Drive = 12 dB.
2. Bias = 35 % — pronounced even-order colour.
3. Mix = 65 % — saturated signal dominates.
4. Envelope = +40 %, Attack = 10 ms, Release = 120 ms — drive swells slightly on voice peaks, adding dynamic warmth.

The curve widget shows the current transfer function — you'll see it bend visibly as you raise Drive, and shift asymmetrically as you raise Bias.

Purpose: add sparkle (high-frequency exciter) and body (low- frequency enhancement). The finishing stage.

Two algorithmic lineages, selectable via Even / Odd:

The six-knob layout splits into two groups:

Poo (LF section):

Doo (HF section):

Example: intelligibility boost for DX

1. Mode = Odd — the brighter, edgier lineage for cutting through noise.
2. Poo · Drive = 6 dB, Tune = 120 Hz, Mix = 20 % — subtle chest presence, don't blow out your speaker.
3. Doo · Tune = 4 kHz, Air = 9 dB, Mix = 30 % — adds "crispness" to consonants. Listen for sibilance — if it's harsh, the DESS stage catches it upstream.

Example: warm, rag-chew voice

1. Mode = Even — warmer character.
2. Poo · Drive = 9 dB, Tune = 90 Hz, Mix = 35 % — fuller low-end.
3. Doo · Tune = 5 kHz, Air = 6 dB, Mix = 15 % — understated top-end enhancement; don't fight the EQ.

Pitfall: layering PUDU on top of aggressive EQ high-shelf and TUBE high-drive produces a harsh, hyped sound. Pick one character surface — usually PUDU for broadcast, EQ for personal preference.

The PUDU wordmark on the editor glows in response to the wet RMS level — a visual confirmation that the exciter is actually adding energy, not just filtering.

Purpose: add a sense of space. The finishing stage of the chain, positioned after PUDU so the exciter harmonics get their own tail.

Important: reverb on TX voice is unusual for ham radio. Most operators want dry, punchy, intelligible audio — reverb trades intelligibility for "bigness". It ships disabled by default and is best used for broadcast-style rag-chewing, AM, or stylistic choices. Contest and DX operators should leave it off.

DSP: Freeverb (Jezar's canonical 2000 implementation) — 8 parallel lowpass-feedback comb filters summed through 4 series allpass filters, stereo-spread by 23 samples between channels. CPU-cheap (under 1 % of one core at 24 kHz), latency-predictable.

Key parameters

Applet tile and floating editor run the same parameters; changes in either view are mirrored to the other at 30 Hz.

Example: subtle broadcast "space"

1. Size = 35 % — small-room character, intimate not cavernous.
2. Decay = 0.8 s — audible tail but clears quickly between syllables.
3. Damping = 70 % — warm tail, no metallic HF ring.
4. Pre-delay = 15 ms — separates direct from reverb enough to keep consonants crisp.
5. Mix = 8 % — barely audible, just enough to glue the voice.

Turn Mix to 0 and click ⏺/▶ in quick succession to A/B against dry; the effect should be noticeable only on a direct comparison.

Example: AM broadcaster hall

1. Size = 70 %, Decay = 1.8 s — big room.
2. Damping = 40 % — brighter tail.
3. Pre-delay = 30 ms — pronounced separation.
4. Mix = 20 % — audible wetness.

Pair with Mode = Even on PUDU and light COMP makeup for that classic AM broadcast character.

Pitfalls

• Latency: Pre-delay adds exactly its value to the wet path's group delay, and the comb tail adds more. With Pre-delay at 50 ms you'll hear a noticeable gap on keyed TX — not a problem for rag-chew but unwelcome for contest/DX timing.
• Reverb into compressor: if you drag VERB ahead of COMP in the chain, the compressor sees wet peaks and squashes them. Interesting for some effects, not usually what you want. Keep VERB last.
• High Mix + high Decay + PUDU Air stacks up HF energy and tail length simultaneously — the signal gets wash-y and muddled. Dial one of the three back.

The record (⏺) and play (▶) buttons next to the TX/RX toggle in the CHAIN header let you hear your processed audio without keying the radio.

1. Click ⏺ — record up to 30 seconds of post-PUDU audio (the signal that would go out the radio).
2. Speak into the mic.
3. Click ⏺ again to stop, OR wait for the 30-second auto-stop.
4. Playback starts automatically, through your PC speakers.
5. Click ▶ during playback to cancel.

During recording AND during playback the RX slice is muted — otherwise you'd hear the band under your voice during playback. The mute lifts automatically when playback finishes.

Tuning a voice chain by listening to a remote station is impossible — you can't hear what your own audio sounds like without a second receiver. The monitor gives you:

• Immediate feedback on EQ/Comp/Tube/PUDU changes without TX.
• A/B comparisons via the BYPASS button between takes.
• Safety — no carrier emission while tuning. Don't block the frequency just to tweak Attack from 20 ms to 30 ms.

1. Set the chain to your best starting point (or disable everything with BYPASS).
2. Record a test phrase: "The quick brown fox jumps over the lazy dog. Five or six big jet planes flew over the quiet countryside." (Covers most phonemes.)
3. Listen. Note what's off — too much sibilance? Muddy low-end? Harsh presence peak?
4. Click the relevant stage in the CHAIN widget, adjust, record again.
5. Iterate. Five minutes with the monitor is worth an hour of flagging strangers on 40m.

Every record also writes /tmp/pudu_monitor.wav — a standard 24 kHz / stereo / int16 PCM WAV. Drop it into Audacity, sox, or any audio editor to analyse spectrum, peak levels, LUFS. Overwritten on every new recording.

Aetherial settings persist across restarts via AppSettings. Each stage's parameters live under ClientGate*, ClientEq*, ClientDeEss*, ClientComp*, ClientTube*, ClientPudu*, ClientReverb* keys in ~/.config/AetherSDR/AetherSDR.settings. The TX and RX sides use distinct keys (e.g. ClientCompTxThreshold vs ClientCompRxThreshold) so the two paths never trample each other.

The chain order itself is persisted under ClientCompTxChainStages (TX) and ClientCompRxChainStages (RX).

To share a chain configuration with another operator, copy the relevant Client*Tx* block (or Client*Rx* block) from your settings file. Sending only the TX block is the usual case — the recipient probably has their own RX preferences.

Quick starting points — tune from here, don't just load and transmit.

All presets are starting points — the point of Aetherial Audio is to listen and tweak.

"I changed a knob, nothing happened." Check that stage's tile — is it dim (bypassed)? Single-click the stage in CHAIN to enable it. Also verify the [MIC] endpoint is green.

"My audio sounds worse, not better." Hit BYPASS and listen. If the dry signal is OK, you're over-processing. Start with just EQ + gentle COMP; layer slowly.

"The monitor plays silence." Record will only capture when [MIC] is green. If DAX is on or mic source isn't PC, there's nothing to tap.

"Reordering the chain did nothing audible." Position matters most between dynamics stages. GATE-before-EQ vs GATE-after-EQ changes where the noise floor sits relative to the EQ's gain stages. COMP-before-TUBE compresses the dry, COMP-after-TUBE compresses saturation harmonics. Experiment with the monitor running.

"BYPASS didn't restore my previous state." BYPASS only restores stages that were ON when you pressed it. If you toggled a stage manually while bypassed, that's a manual change and won't be clobbered on un-bypass.

• Audio Settings — PC audio device selection + sample rates.
• TX Controls — the surrounding TX applet chain.
• RX Controls — RX panel controls (filters, AGC, squelch).
• DSP Noise Mitigation — client-side noise reduction (NR2, NR4, BNR) — these feed the [DSP] indicator on the RX chain.