Analog-to-Digital Converter (ADC)
An ADC represents an analog input with a digital code. An N-bit result provides 2^N possible codes, but the useful resolution and accuracy also depend on noise, distortion, reference quality and the input circuit. The permitted input range and data interface are device-specific.
A successive-approximation (SAR) converter compares its input with trial levels, resolving the code step by step. A sigma-delta converter combines oversampling, noise shaping, digital filtering and decimation. Neither architecture name imposes one fixed speed or bit-depth limit: ADI's SAR tutorial gives an 18-bit, 2 MSPS example, and sigma-delta devices trade bandwidth, filtering and noise performance according to their design.
In plain terms
Imagine a height-marked staircase next to a continuously rising ramp. The ADC tells you which step the ramp is currently touching — the more steps you carve in, the closer the number matches the real height.
Also called: A/D converter, Analog-to-digital converter.
Why designers use it
- Converting sensor voltages into data for processing.
- Choosing input channels, bandwidth and noise performance that meet a measurement requirement.
- Providing a documented analog-input interface where a digital system needs measurements.
Best for
- Sensor reads (temperature, pressure, light)
- Audio input sampling
- Battery / current monitoring
- Strain-gauge and load-cell measurement
- Data acquisition systems
Key specifications
- Nominal resolution: Output code width, not a noise-free accuracy guarantee
- Noise performance: Compare at the selected range, gain, bandwidth and data rate
- Rate and settling: Check throughput, latency and channel-switching settling
- Input and reference: Check range, common-mode limits, driver and reference requirements
- Architecture and interface: Select for the complete measurement task
When not to use it
- When an existing ADC already meets the complete measurement requirements.
- When only a threshold decision is needed and a suitable comparator would suffice.
- When the selected converter's input, bandwidth, settling or noise requirements cannot be met.
Common mistakes
- Treating a 24-bit output word as 24 noise-free bits of measurement.
- Confusing output data rate with the internal modulator sampling rate.
- Assuming channels selected through one multiplexer are sampled simultaneously.
- Ignoring reference noise, input settling, aliasing and the selected device's input limits.
- Comparing ENOB, effective DC resolution and noise-free resolution as if they were identical metrics.
Where you will find it
- ADI's MT-021 tutorial describes AD7641 as an 18-bit, 2 MSPS SAR example, so SAR is not universally limited to 16 bits.
- A sigma-delta measurement system can trade output data rate against filtering and noise. The usable result must be checked at the actual configuration, not inferred from its advertised word length.
A short history
ADC development includes several architectures rather than one universal design. ADI's SAR tutorial describes a 1947 experimental voice system with a 5-bit, 8 kSPS converter. Its sigma-delta tutorial explains the later combination of oversampling, noise shaping, digital filtering and decimation. Those architectural differences still shape measurement trade-offs.
Good to know
- An ideal 10-bit code has 1,024 possible values, while an ideal 16-bit code has 65,536.
- Noise-free resolution uses peak-to-peak noise, while effective DC resolution uses RMS noise.
- A multiplexer selects among inputs; by itself it does not make their sampling simultaneous.