A learning project by Huy · AI-assisted content · Updated
Digital12 min read
ADC Sampling & Aliasing
Understand how a continuous signal becomes numbers, why sampling rate matters and how aliasing creates false frequencies.
Mental model
A video of a spinning wheel can make it appear to rotate backward. The camera takes too few snapshots to distinguish the real motion—an ADC can make the same mistake with a waveform.
The ADC stores only discrete samples; frequencies above the usable Nyquist band require an analog anti-alias filter before sampling.UNDERSTAND THE BEHAVIOR
01
An ADC measures a continuous voltage only at discrete instants. The sample rate controls how closely spaced those instants are, while bit depth controls how many amplitude levels are available.
02
Nyquist's rule requires more than two samples per cycle of the highest frequency, but practical waveform quality usually needs more. Frequencies already aliased into the sampled data cannot be separated later.
03
An anti-alias filter removes frequencies above the usable band before sampling. It is analog because filtering after the ADC is too late to undo aliasing.
Why it matters
Sampling choices determine whether digital audio, sensors and control systems represent the physical signal faithfully.
Core ideas
The sample rate must exceed twice the highest signal frequency in the ideal theorem.
A real system uses an analog anti-alias filter before the ADC.
More bits improve amplitude resolution, while a higher sample rate improves time resolution.
4The stored samples look like a 1 kHz signal, so the original 9 kHz signal cannot be recovered.
Common mistake
Increasing ADC bit depth does not fix an inadequate sample rate, and increasing sample rate does not remove input noise by itself.
RememberThe sample rate must exceed twice the highest signal frequency in the ideal theorem.
DEEPER EXPLANATION
Why do high frequencies appear as lower ones?
An ADC sees the signal only at discrete instants. A 7 kHz sine sampled at 10,000 samples/s can generate the same sampled pattern as a 3 kHz sine with a suitable phase: samples cannot by themselves tell which continuous waveform was present. This folding of out-of-band content into the measured band is aliasing.
For an ideally band-limited signal whose highest component is fmax, the Nyquist boundary is fs = 2fmax. Sampling strictly above that boundary permits ideal reconstruction under the theorem's assumptions. A real anti-alias filter needs frequency room to attenuate components before fs/2, so designers choose a rate with margin rather than treating equality as a practical target.
An ADC's sample rate does not by itself define measurement accuracy. The analog input bandwidth, front-end filter, noise, resolution, aperture jitter and reference all matter. If you increase sample rate without reducing out-of-band energy, you have not guaranteed an alias-free result.
Walk through another example
Suppose useful sensor content extends to 3 kHz. The theoretical boundary is 2 × 3 kHz = 6 ksample/s.
At a chosen 10 ksample/s, the Nyquist frequency is 5 kHz. The anti-alias filter needs to pass the desired 0–3 kHz band and suppress unwanted energy as it approaches and exceeds 5 kHz.
A 7 kHz component entering that sampler can alias to |7 − 10| = 3 kHz, exactly inside the desired band.
Check your intuition
Before choosing an ADC, state both the wanted signal bandwidth and the highest unwanted frequencies reaching its input. The filter transition between them determines how much sample-rate margin is useful.
Where this model can mislead you
The calculator returns the mathematical boundary, not a safe sampling rate. Frequencies exactly at the boundary and real filters need more care.
TEST YOUR UNDERSTANDING
Three quick checks
Choose one answer per question, then check your work. Each answer includes an explanation.
Practice and check your understanding
Try each question before opening the worked answer.
1. A baseband signal contains frequencies up to 3 kHz. What sampling rate is required in the ideal band-limited case?
Show answer and reasoning
Choose fs greater than 2 × 3 kHz = 6 kHz. Exactly 6 kHz is a boundary, not a robust design. Real designs need transition-band margin and an analog anti-alias filter.
2. A 7 kHz sine is sampled at 10 kHz. What baseband alias frequency can appear?
Show answer and reasoning
It folds about the 5 kHz Nyquist frequency to |7 − 10| = 3 kHz. Once sampled, it cannot in general be distinguished from the corresponding aliased signal without extra information.
A related calculation tool, not a complete simulation of the circuit. Check its model and units.
Continue learning
Fourier Transform — Understand how a waveform can be described as a combination of frequencies and how to read an FFT without confusing bins with physical components.
Active Filters — Combine op-amps, resistors and capacitors to shape frequency response without using inductors.