Also known as: signal-to-noise ratio, SNR, S/N
Signal-to-noise ratio (SNR) is the gap, in decibels, between a signal’s power and the noise floor.1 It is the single best predictor of whether a signal will decode: enough SNR and the bits come out clean; too little and errors overwhelm the demodulator.
How it works
SNR = signal level − noise-floor level, with both measured in dBm so the subtraction gives a value in dB. A signal at −85 dBm over a −105 dBm floor has 20 dB of SNR. Because it is a difference of two dBm readings, any constant calibration offset in the receiver cancels — which is why SNR is meaningful even on uncalibrated SDR hardware that cannot report true absolute dBm.
The measurement depends on bandwidth. Noise power scales with the width over which you integrate it (the thermal noise floor rises ~3 dB each time bandwidth doubles), so an SNR figure is only well defined relative to a stated bandwidth. Measuring noise in a wide span and signal in a narrow one inflates the number; comparing SNR across systems means normalising to a common bandwidth. This is also why narrowing the receive filter to the signal’s occupied bandwidth improves SNR — it admits less noise while keeping all the signal.
Each digital mode has a threshold SNR below which the demodulator’s error rate climbs steeply. The relationship is a “waterfall” curve: above threshold the bit error rate is negligible, and within a few dB below it the link collapses. Forward error correction lowers the threshold by trading redundancy for robustness, letting a mode work at lower SNR than its raw modulation would allow.
SNR has several close relatives worth distinguishing. C/N (carrier-to-noise) measures the modulated carrier against noise before demodulation. Eb/N0 normalises to energy per bit and noise density, making a fair comparison across data rates and modulations. SINAD folds distortion in with noise for analog voice quality. All are dB ratios of wanted to unwanted power, differing in exactly what they count.
In practice
SNR is the number to watch on a site survey. Typical thresholds: analog FM voice becomes readable around 10–12 dB SNR; C4FM/CQPSK digital modes like P25 need roughly 15–20 dB at the demodulator for a clean lock, with FEC providing some margin below that. A signal 3 dB over threshold is usable but fragile; 10 dB over is comfortable.
The Shannon–Hartley theorem sets the ultimate limit: channel capacity grows with log₂(1 + SNR), so every doubling of SNR buys diminishing returns in achievable bit rate. Practical systems pick a modulation and code that decode reliably a few dB above their threshold at the expected SNR.
Relevance to SDR
Improving SNR — a better antenna, a higher or clearer site, correct gain, a low-noise amplifier, and matched filtering — is usually what moves a marginal signal from un-decodable to clean. GopherTrunk reports per-channel demodulator SNR and EVM so an operator can see, in decibels, exactly how much margin a link has and whether a failed decode is an SNR problem or something else (front-end overload, wrong tuning) entirely.
Sources
-
Signal-to-noise ratio — Wikipedia, definition, bandwidth dependence, and significance of SNR. ↩