Skip to content

AetherDSP Settings

The AetherDSP Settings dialog tunes the advanced parameters of AetherSDR's client-side noise-reduction engines (NR2, NR4, MNR, DFNR, RN2, BNR), letting operators dial in the tradeoff between noise suppression and speech fidelity. The six DSP modules are selectable via a toggle row at the top; clicking a toggle also activates or bypasses that engine.

Before you start

  • AetherSDR does not need to be connected to a radio to adjust DSP settings.
  • The selected NR engine must already be active on your receive slice for these changes to have an audible effect.
  • Some DSP engines (MNR, BNR, RN2, NR4) are platform-dependent or informational only. For details see Choosing the right noise reduction: NR2, NR4, DFNR, MNR.

Open AetherDSP Settings

  1. Open Settings > AetherDSP Settings....

The dialog appears with a frameless 18 px gradient title bar showing a grip glyph (โ‹ฎโ‹ฎ) on the left and the dialog title. Click and drag the title bar to move the dialog. Double-click the title bar to toggle maximize/restore. Click and drag any edge or corner to resize the dialog. The 6 px resize hit zone surrounds the inner content widget; the cursor changes to indicate the resize direction.

The title bar contains three window controls: - โ€” (Minimize): Minimizes the dialog. - โ–ก (Maximize): Maximizes or restores the dialog. - ร— (Close): Closes the dialog.

The dialog geometry is persisted automatically to the AetherDspDialogGeometry setting and restored on next launch.

The dialog uses the color theme defined by the current AetherSDR theme (setting key theme/aetherDsp). Sliders render with the applyPrimarySliderStyle style applied by the theme manager.

Select a DSP engine

At the top of the dialog, six toggle tabs act as both selector and enable/disable controls:

  • NR2 โ€“ Musical-noise-reduction engine
  • NR4 โ€“ libspecbleach noise suppression (dimmed on Windows builds without LLVM/clang-cl)
  • MNR โ€“ macOS MMSE-Wiener (dimmed on Windows/Linux builds)
  • DFNR โ€“ DeepFilterNet3 neural noise suppression
  • RN2 โ€“ RNNoise
  • BNR โ€“ NVIDIA Broadcast (dimmed on builds without NVIDIA Broadcast SDK)

Click a toggle to select its parameter page and activate or bypass that engine. When NR2 is activated, AudioEngine cascades exclusion, disabling DFNR and other mutually exclusive modules.

NR2 parameters

Control Default Valid values Setting key
Gain Method Gamma Linear | Log | Gamma | Trained NR2GainMethod
NPE Method OSMS OSMS | MMSE | NSTAT NR2NpeMethod
AE Filter (artifact elimination) Enabled (checked) Checked / unchecked NR2AeFilter
Reduction: 1.50 0.50โ€“2.00 NR2GainMax
Smoothing: 0.85 0.50โ€“0.98 NR2GainSmooth
Threshold: 0.20 0.05โ€“0.50 NR2Qspp
  • Gain Method selects the gain-curve mapping used by NR2. Stored as integer 0-3 matching the order Linear, Log, Gamma, Trained.
  • NPE Method selects the noise power estimator. Stored as integer 0-2.
  • AE Filter (artifact elimination) toggles the anti-artefact post-filter.
  • Reduction: sets the maximum NR2 reduction depth. Slider stores value*100 internally.
  • Smoothing: controls how smoothly the noise estimate tracks changes.
  • Threshold: sets the speech-presence-probability threshold.

Click Reset Defaults (โ†บ icon) to restore NR2 defaults (Gamma/OSMS/AE on, 1.50/0.85/0.20).

NR4 parameters

Control Default Valid values Setting key
Noise Estimation: MMSE MMSE | Brandt | Martin NR4NoiseEstimationMethod
Adaptive Noise Estimation Enabled (checked) Checked / unchecked NR4AdaptiveNoise
Reduction (dB): 10.0 0.0โ€“40.0 dB NR4ReductionAmount
Smoothing (%): 0 0โ€“100 NR4SmoothingFactor
Whitening (%): 0 0โ€“100 NR4WhiteningFactor
Masking Depth: 0.50 0.00โ€“1.00 NR4MaskingDepth
Suppression: 0.50 0.00โ€“1.00 NR4SuppressionStrength
  • Noise Estimation: selects the noise-floor estimator used by NR4. Stored as integer 0-2.
  • Adaptive Noise Estimation enables continuous re-estimation of the noise floor.
  • Reduction (dB): sets maximum NR4 noise reduction in dB. Slider stores value*10.
  • Smoothing (%): applies time-domain smoothing of the NR4 noise estimate.
  • Whitening (%): flattens residual noise spectral shape.
  • Masking Depth: controls spectral-masking depth.
  • Suppression: sets overall NR4 suppression strength.

Click Reset Defaults (โ†บ icon) to restore NR4 defaults (MMSE/adaptive on, 10 dB, 0, 0, 0.50, 0.50).

Enable or disable NR4 adaptive noise estimation

With adaptive noise estimation enabled, NR4 tracks changes in the noise environment in real time; disabling it locks the noise-floor estimate to a static snapshot, which can suit highly stable noise conditions.

  1. Open Settings > AetherDSP Settings....
  2. Click the NR4 tab.
  3. Check or uncheck Adaptive Noise Estimation to enable or disable continuous noise-floor re-estimation.

The setting takes effect immediately and is saved automatically to NR4AdaptiveNoise.

Tips

  • If the noise floor on your band is stable and consistent, unchecking Adaptive Noise Estimation can prevent the estimator from following signal-level changes and misclassifying speech as noise.
  • If the noise floor varies rapidly โ€” such as during band openings or with impulse noise โ€” leave Adaptive Noise Estimation checked so NR4 can track the changing conditions.
  • The Noise Estimation Method selector (MMSE, Brandt, Martin) determines how NR4 builds its noise model regardless of whether adaptive mode is on or off. Changing the method can affect how well the static or adaptive estimate tracks your noise floor.
  • Click Reset Defaults on the NR4 tab to return all NR4 controls to their factory values (adaptive on, MMSE, 10 dB, 0, 0, 0.50, 0.50).

NR4 availability on Windows

NR4 uses the libspecbleach library, which requires the clang-cl compiler on Windows. If the NR4 tab is dimmed, install LLVM from llvm.org and rebuild AetherSDR to enable NR4 support.

MNR parameters (macOS only)

Control Default Valid values Setting key
Enable MNR (macOS only) Off Checked / unchecked MnrEnabled
Strength 100 0โ€“100 MnrStrength

The MNR toggle is dimmed on Windows/Linux builds. The Enable MNR checkbox enables MMSE-Wiener noise reduction with asymmetric gain smoothing. Strength adjusts MNR aggressiveness (0 mild, 100 max) and is persisted as a normalized 0.00โ€“1.00 value.

DFNR parameters

Control Default Valid values Setting key
Attenuation Limit 100 0โ€“100 dB DfnrAttenLimit
Post-Filter Beta 0.00 0.00โ€“0.30 DfnrPostFilterBeta

Attenuation Limit sets the maximum noise attenuation applied by DeepFilterNet3. 0 = passthrough; 100 = maximum.

Post-Filter Beta applies an additional post-filter for extra suppression. Slider stores value*100 internally.

RN2 tab

Selecting the RN2 tab displays RNNoise engine information and hosts the adjustable Noise Floor dry-mix slider.

Control Default Valid values Setting key
Noise Floor (RN2 dry mix) 0 0โ€“100 None (persisted by Rn2SettingsModel)

Noise Floor (RN2 dry mix) sets the percentage of the original signal RN2 leaves under the denoised audio. Zero yields full suppression (silent between phrases); 10โ€“20% keeps a steady quiet floor so the receiver still sounds alive.

The setting affects received audio only; the transmit denoiser is unchanged. It is exposed to the DSP chain via AetherDspWidget's rn2DryMixChanged signal. There is no setting key โ€” the value is persisted by Rn2SettingsModel.

Click Reset Defaults (โ†บ icon) on the RN2 tab to return the Noise Floor slider to 0.

BNR tab

Selecting the BNR tab displays NVIDIA Broadcast noise reduction controls. The BNR toggle is dimmed on builds without the NVIDIA Broadcast SDK. Intensity is controlled from the overlay menu.

โšก Athena AetherSDR Assistant
Hi! I'm Athena, the AetherSDR AI assistant. Ask me anything about installation, configuration, or troubleshooting.