Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com> Signed-off-by: Stan Grams <sjg@haxx.space>
2.6 KiB
trx-rs Manual
SDR Noise Blanker
The noise blanker suppresses impulse noise (clicks, pops, ignition interference) on raw IQ samples before any mixing or filtering takes place. It works by tracking a running RMS level of the signal and replacing any sample whose magnitude exceeds threshold x RMS with the last known clean sample.
Configuration (server-side)
Add a [sdr.noise_blanker] section to your trx-server.toml:
[sdr.noise_blanker]
enabled = true
threshold = 10.0 # 1 – 100; lower = more aggressive blanking
| Field | Type | Default | Range | Description |
|---|---|---|---|---|
enabled |
bool | false | — | Turn the noise blanker on or off. |
threshold |
float | 10.0 | 1 – 100 | Multiplier applied to the running RMS. A sample whose magnitude exceeds this multiple is replaced. Lower values blank more aggressively; higher values only catch strong impulses. |
The noise blanker is off by default. A threshold of 10 is a good starting point for typical HF impulse noise. Reduce toward 3–5 for heavy QRM; increase toward 20–30 if weak signals are being clipped.
Web UI
When the server reports noise-blanker support, two controls appear in the SDR Settings row of the web interface:
- Noise Blanker checkbox — enables or disables the blanker in real time.
- NB Threshold number input (1–100) with a Set button — adjusts the detection threshold. Press Enter or click Set to apply.
Both controls stay hidden until the server sends filter state containing NB fields, so they only appear when connected to an SDR backend.
HTTP API
POST /set_sdr_noise_blanker?enabled=true&threshold=10
| Parameter | Type | Required | Description |
|---|---|---|---|
enabled |
bool | yes | true or false |
threshold |
float | yes | Value between 1 and 100 |
How it works
The blanker runs on every IQ block (4096 samples) before the mixer stage in the DSP pipeline:
- For each sample, compute magnitude² (
re² + im²). - Compare against
threshold² × mean_sq(the exponentially-smoothed running mean of magnitude²). - If the sample exceeds the threshold, replace it with the previous clean sample.
- Otherwise, update the running mean with smoothing factor α = 1/128 and store the sample as the last clean value.
Because the blanker operates on raw IQ before frequency translation, it removes impulse noise across the entire captured bandwidth regardless of the tuned channel offset.