Also known as: aliasing
Aliasing is when energy outside the bandwidth a sample rate can represent gets folded back into the captured spectrum, appearing at a wrong frequency — a phantom that looks like a real signal but sits nowhere near where its source actually transmits.1
How it works
Sampling maps every input frequency to a value modulo the sample rate. Two tones separated by an exact multiple of Fs produce identical sample sequences — they are aliases of one another and become indistinguishable the instant they are digitised. Formally, the spectrum repeats every Fs, so any component above the Nyquist frequency (Fs/2) mirrors, or “folds,” back into the band below it. This is simply the Nyquist theorem violated: the theorem guarantees a perfect reconstruction only if no energy exists above the Nyquist frequency, and aliasing is exactly what goes wrong when that precondition fails.
The corruption is irreversible. Once a strong out-of-band station has folded on top of a weak in-band one, no filter applied after sampling can pull them apart, because in the digital domain they occupy the same bin. The defence therefore has to come before the sampler: an analog anti-alias filter removes the offending energy while it can still be separated. In the digital domain the same logic governs decimation, which is why the rule is always filter first, then drop samples — decimating without filtering re-introduces aliasing by lowering the effective Nyquist frequency under the signal.
Variants
- Anti-alias (pre-sampling) aliasing — the classic case, prevented by the analog filter ahead of the ADC.
- Decimation aliasing — created by lowering the rate in software without an adequate low-pass stage first.
- Deliberate aliasing — bandpass sampling uses folding on purpose to bring a high band down to baseband.
- Image responses — the analog cousin: a mixer folds a band on the far side of the local oscillator into the IF, which is why superheterodyne front-ends need image rejection.
In practice
Oversampling gives the anti-alias filter a wide transition band to work with, making a gentle, cheap filter sufficient. On a waterfall an alias often gives itself away by moving the wrong direction when you retune, or by mirroring across the band edge as you change the sample rate — a real signal stays put in absolute frequency while an alias shifts. This is a standard sanity check when a station appears where none should be.
Relevance to SDR
Recognising an alias prevents chasing signals that are not really where they appear. GopherTrunk’s front-end and decimation stages include the low-pass filtering that keeps real control and voice channels safely inside the usable window; an operator who decimates a raw capture with a third-party tool that skips the filter can inject aliases that break the decode.