Field Guide · term

Also known as: reciprocal mixing noise

Reciprocal mixing is the receiver degrading itself: every signal entering the mixer is convolved with the local oscillator’s phase-noise skirts, so each received carrier comes out wearing a copy of the LO’s noise pedestal.1 A strong signal near the tuned channel then splashes noise into the channel — the classic blocking mechanism — but the wanted signal also smears its own modulation with the LO’s close-in jitter. Either way the damage is done at the mixer: the noise is multiplied into the samples, and no amount of filtering, decimation, or clever demodulation downstream can take it back out.

wanted (weak) strong neighbour LO skirts land here the neighbour's spread is a mirror of the LO's own phase-noise shape
The strong signal is clean on the air; the skirts that bury the weak channel are the receiver's own oscillator noise, transferred onto it in the mixer.

How it works

An ideal LO is a spectral line; mixing with it translates every input frequency crisply. A real LO carries phase-noise sidebands falling away from the carrier (measured in dBc/Hz at each offset), and mixing is multiplication in time — convolution in frequency — so every received signal is smeared by that sideband shape. Two consequences follow:

  • Nearby strong signals raise the floor. The in-channel noise contributed by a blocker at offset Δf is the blocker’s power times the LO’s phase noise at Δf (integrated over the channel). This, not front-end overload, is what limits many receivers’ close-in blocking dynamic range: past a certain blocker level, more RF filtering does not help and more gain makes it worse.
  • The wanted signal degrades itself. Close-in LO jitter rotates the constellation symbol-to-symbol. Heavily averaged measurements — a spectrum-analyzer carrier peak, a long FFT — can still look clean while EVM and demodulated SNR are poor. Carrier-clean but modulation-degraded is the reciprocal-mixing signature, and it is how the effect is told apart from intermodulation or clipping, which show up in the amplitude domain.

The oscillator’s quality — crystal reference, PLL loop bandwidth, synthesizer architecture, and in an SDR the sample-clock/PLL configuration at a given rate — sets the effect’s size. Nothing after the mixer can reduce it.

Relevance to SDR

Reciprocal mixing is the standing explanation for a class of SDR mysteries where a capture is bad in a way no DSP setting changes. GopherTrunk’s canonical case (issue #764): the same TETRA site captured on the same Airspy decoded at ~19.7 dB demod SNR from a 2.5 MS/s capture but ~9.5 dB from a 10 MS/s capture — neither clipping, and the wideband FFT carrier SNR actually higher at 10 MS/s. Decimating the 10 MS/s file 4:1 with an independent resampler and replaying it through the proven 2.5 MS/s path reproduced the same ~9.5 dB, proving the ~10 dB deficit was baked into the captured samples: front-end phase noise at the device’s native 10 MS/s clock, not the decoder (pinned by TestDownconverterSNRInvariantAcrossRate in internal/scanner/ccdecoder/ddc_highrate_test.go). The practical rules that follow: prefer the sample rate at which your hardware’s clocking is clean, verify a “weak signal” problem against a capture at a different rate before blaming the decoder, and treat carrier-clean-but-modulation-degraded as an oscillator finding — a gain change will not fix it.

Sources

  1. Phase noise — Wikipedia, on oscillator phase-noise sidebands and their transfer onto received signals in mixing (reciprocal mixing). ↩

See also