In short: The four mainstream ADC architectures — SAR, flash, pipeline, and sigma-delta — each trade speed against resolution in a different way. Flash is the fastest but burns area and power; sigma-delta is the slowest but reaches the highest resolution; SAR and pipeline sit in between and dominate most real designs. Choosing the right one starts with your required sample rate and effective number of bits.
An analog-to-digital converter turns a continuous voltage into a stream of binary numbers. The adc architectures we compare here are simply the circuit strategies used to perform that quantization. They all sample an input and compare it against known references, but they differ in how many comparisons happen, whether those comparisons run in parallel or in sequence, and how the result is encoded. Those structural choices decide the headline specs an engineer actually cares about: sampling rate, resolution (bits / ENOB), power, latency, and silicon area.
Two numbers frame every decision. Resolution is how finely the input range is divided — an N-bit converter resolves 2N levels. Conversion speed is how many samples per second (SPS) the converter produces. The central law of data conversion is that you rarely get both cheaply at once, and each architecture represents a different point on that speed-versus-resolution curve.
A SAR ADC uses a single comparator and a binary search. It guesses the most-significant bit, compares, keeps or discards it, then moves to the next bit — one comparison per bit. An N-bit conversion therefore takes N clock cycles. The internal DAC and comparator are reused each cycle, so a SAR is compact and extremely power-efficient. This makes SAR the workhorse for 8–18 bit converters from kSPS up to tens of MSPS: battery sensors, data acquisition, and medical front-ends.
A flash ADC is pure parallelism. For N bits it uses 2N−1 comparators, each wired to a different tap on a resistor ladder, all firing simultaneously. The result appears in essentially one clock edge, so flash is the fastest architecture in existence — the basis of GSPS oscilloscope and RF front-ends. The cost is brutal: comparator count doubles with every added bit, so power, area, and input capacitance explode. That is why flash converters are almost always 4–8 bits.
A pipeline ADC is a chain of low-resolution stages. Each stage resolves a few bits with a small flash sub-ADC, converts them back with a DAC, subtracts, amplifies the residue, and hands it to the next stage. Because stages work on different samples at the same time (like a factory assembly line), throughput stays high while resolution builds up across the chain. Pipelines deliver 10–16 bits at tens-to-hundreds of MSPS — the standard choice for imaging sensors and wireless receivers. The trade-off is latency: a sample needs several clocks to travel the full pipe.
A sigma-delta ADC trades time for precision. A 1-bit (or few-bit) modulator samples the input far faster than Nyquist requires — this is oversampling — and a feedback loop pushes quantization noise up to high frequencies (noise shaping). A digital decimation filter then averages that coarse, fast stream into a slow, very high-resolution output. The result is 16–24 effective bits for audio, precision instrumentation, and weigh scales, at the price of low bandwidth and high latency.
The clean way to compare adc architectures is to plot resolution against sample rate and see where each one lives:
Real projects and industry tool flows, with certification.
Explore CourseTron Courses →Flash is the fastest because all comparisons happen in parallel in a single clock cycle. It is used for gigasample-per-second applications, but its comparator count grows as 2N, which limits it to roughly 4–8 bits.
Sigma-delta, reaching 24 effective bits. It achieves this through oversampling and noise shaping followed by digital decimation, at the cost of limited bandwidth and higher latency — ideal for audio and precision instrumentation.
SAR converters reuse one comparator and DAC across a binary search, making them small and very power-efficient while still delivering 8–18 bits at useful speeds. That balance fits most sensor and data-acquisition systems, so they are the default choice.
A SAR resolves all bits of one sample sequentially with a single comparator, so it is compact but slower. A pipeline splits the work across many stages that process different samples simultaneously, trading extra area and latency for much higher throughput.
Related: PLL design basics · Glossary · all courses.