Watch how discrete sampling measures a continuous analog wave. Drag the Time Window Zoom slider to space out individual sample points and observe how phantom aliased waves emerge when the Nyquist limit is violated.
Nyquist criterion satisfied. Signal frequency (50 Hz) is below Nyquist (125 Hz). Sampling faithfully captures the original signal.
By zooming the time window to 0.08 seconds (80 ms), we are magnifying a tiny slice of time. At 250 Hz, one sample is recorded every 4 ms ($1 / 250\text{ s}$). The vertical yellow stems show the exact instant and voltage recorded by the analog-to-digital converter (ADC).
When the signal moves faster than half the sampling clock ($f > f_s / 2$), the yellow sample dots are too sparse to capture the rapid blue waves. The ADC only "sees" the yellow dots — and the simplest smooth curve passing through those exact yellow dots is the red dashed slow wave ($f_{\text{alias}} = |f - \text{round}(f/f_s) \cdot f_s|$).
Notice how the yellow sample dots lie simultaneously on both the fast blue wave and the slow red wave! Once sampled, a computer cannot distinguish whether those numbers came from a 180 Hz signal or a 20 Hz signal. This is why aliasing is permanent and uncorrectable after sampling.
Hardware EEG amplifiers use an analog low-pass filter before the ADC to block all frequencies above $f_s / 2$. Try toggling the Anti-alias Filter: ON button: high frequencies above Nyquist are attenuated to near-zero before sampling, ensuring no phantom red waves ever corrupt your recordings.
The Nyquist–Shannon Theorem proves that an analog signal $x(t)$ with maximum frequency $f_{\text{max}}$ can be perfectly reconstructed without loss if sampled at:
When $f > f_{\text{Nyquist}}$, the frequency component folds back into the passband $[0, f_s/2]$ according to: