Electrical
Analog-to-Digital Conversion
ADC — sampling continuous signals into discrete digital values
An analog-to-digital converter (ADC) samples a continuous voltage at discrete time intervals and quantizes each sample into a finite digital value. Resolution measured in bits — an N-bit ADC produces 2^N levels. Sampling rate measured in samples per second (Hz). Nyquist theorem: sampling must be at least 2× the highest signal frequency. Common architectures: SAR, sigma-delta, flash, pipeline. Bridges the analog physical world to digital computation. Found in every microcontroller, audio device, sensor interface, and measurement instrument.
- Resolution8, 10, 12, 16, 24-bit common
- Levels2^N (12-bit = 4096)
- Sample ratekHz to GHz depending on architecture
- Nyquistf_sample > 2 × f_signal
- ArchitecturesSAR, sigma-delta, flash, pipeline
- Quantization error±½ LSB
Interactive visualization
Press play, or step through manually. The visualization is yours to drive — try it before reading on.
Watch the 60-second explainer
A condensed visual walkthrough — narrated, captioned, under a minute.
Why ADC matters
- Sensors. Pressure, temperature, strain — all analog at the source.
- Audio. Microphones to digital streams.
- Instrumentation. Oscilloscopes, multimeters, data acquisition.
- Communications. Software-defined radio, modems.
- Imaging. CMOS image sensors, medical scanners.
- Control systems. Closing the loop with digital controllers.
- Embedded. Every microcontroller integrates ADC peripherals.
Common misconceptions
- Sample at exactly 2× signal. Need margin and steep filters; aim 2.5×–10×.
- More bits = more accuracy. Noise and ENOB matter more than nominal bits.
- Skip the anti-alias filter. Out-of-band noise folds in undetectably.
- Reference doesn't matter. Absolute accuracy hinges on reference stability.
- Linear across full scale. INL and DNL specs reveal nonidealities.
- Faster ADC always better. Speed costs power and noise; pick to match signal.
Frequently asked questions
How does an ADC work?
Two stages: sample-and-hold captures a fixed voltage at a moment in time; quantizer maps that voltage to the nearest discrete code. SAR ADCs binary-search comparator decisions to settle on a code. Sigma-delta oversamples and filters digitally. Flash uses a ladder of comparators for instant conversion. Choice depends on speed, resolution, and power tradeoffs.
What's resolution?
Number of distinct output codes. A 12-bit ADC produces 4096 codes; 16-bit produces 65,536. With a 5V reference, 12-bit resolution corresponds to 1.22 mV per LSB. Higher resolution captures finer detail but requires lower noise, better references, and slower conversion. Effective number of bits (ENOB) accounts for noise and is usually 1–2 bits less than nominal.
What's the Nyquist theorem?
To reconstruct a signal of bandwidth B without aliasing, sample at greater than 2B. Audio CDs sample 44.1 kHz to capture frequencies up to 22 kHz. Below Nyquist, frequencies fold back as alias artifacts that cannot be removed digitally. Anti-alias filters before the ADC enforce bandwidth limits.
What's quantization error?
The difference between the true analog value and its discrete code. Bounded by ±½ LSB for an ideal ADC. Appears as noise with RMS value LSB/sqrt(12), giving theoretical SNR of 6.02N + 1.76 dB for an N-bit converter. Real ADCs add thermal, jitter, and integral nonlinearity errors that reduce ENOB.
When use SAR vs sigma-delta?
SAR offers moderate speed (up to MSps) at 8–18 bit resolution with low power and instant single-shot conversion. Sigma-delta achieves 16–24 bit resolution by oversampling and noise-shaping but is slower (kSPs range). Use SAR for multiplexed sensor reads, sigma-delta for audio, weight scales, and precision instrumentation.
What's a reference voltage?
The full-scale voltage that defines the ADC's input range. A 12-bit ADC with a 2.5V reference outputs 0 for 0V input and 4095 for 2.5V. Reference accuracy and stability set absolute measurement accuracy. Internal references are convenient but drift with temperature; external precision references (1–10 ppm/°C) are required for calibrated instruments.
What's oversampling?
Sampling much faster than Nyquist requires, then digitally filtering and decimating. Each 4× oversample gains roughly 1 bit of effective resolution (assuming white noise). Sigma-delta ADCs rely heavily on this — they sample at 1–10 MHz to deliver 24-bit results at 1 kHz. Tradeoff: more digital processing, longer latency.