Aliasing

Aliasing occurs when sampling maps high-frequency content into a lower apparent frequency. If a continuous signal contains components above the Nyquist frequency,

where f_s is the sampling frequency, those components cannot be represented uniquely by the sampled sequence. They fold into the measured band and may look like genuine low-frequency behaviour.

Engineering role

Aliasing matters whenever real-world signals are converted into digital data. It appears in vibration monitoring, motor-current analysis, sensor acquisition, audio, radio receivers, medical instruments, images, encoders, and control systems. In a test rig, aliasing can make a harmless high-frequency vibration look like a low-frequency instability. In an image sensor, it can create moire patterns. In a control loop, aliased sensor noise can trigger poor decisions or excite actuators.

Prevention

The standard mitigation is to limit the analog bandwidth before sampling. An anti-alias filter attenuates frequencies above the useful measurement band so that out-of-band energy cannot fold into the digital signal. Oversampling can make filtering easier, but it does not remove the need to control out-of-band content. Digital filtering after the analog-to-digital converter cannot remove aliasing that has already occurred, because the false low-frequency component is already mixed into the sampled data.

Representation and diagnosis

Engineers describe aliasing with sampling rate, Nyquist frequency, analog bandwidth, filter roll-off, record length, and spectral evidence. Frequency-domain tools such as the fast Fourier transform are useful, but only if the acquisition chain is understood. A suspicious peak that shifts when the sample rate changes is often a sign of aliasing. Time-domain plots alone may not reveal the problem.

Design considerations

A robust acquisition design starts from the highest physical frequency that may reach the sensor, not only from the frequency band of interest. Mechanical resonance, switching converters, electromagnetic interference, acoustic noise, and sensor internal dynamics can inject energy above the intended bandwidth. The anti-alias filter, sample rate, ADC resolution, clock stability, and signal conditioning should be specified together.

Common mistakes

Common mistakes include using the nominal sample rate without checking clock jitter, assuming a sensor’s stated bandwidth is an anti-alias filter, and relying on software smoothing after sampling. Another error is sampling exactly at twice the desired maximum frequency; practical filters need transition bandwidth, so real systems usually sample faster than the theoretical minimum.