Data Converters: ADC & DAC

resolution (bits)

Resolution is how finely a converter can split its voltage range — the number of distinct levels it can tell apart. It is quoted in bits, and each bit doubles the number of levels: n bits give 2^n levels. Think of it as how many marks are on the ruler. A coarse ruler with a few big marks gives rough readings; a fine ruler with thousands of tiny marks lets you read small differences. More bits means a finer ruler.

The single most useful number that resolution gives you is the size of one step, the LSB (least significant bit): LSB = full-scale voltage / 2^n. Work an example: a 12-bit converter over a 0 to 3.3 V reference has 2^12 = 4,096 levels, so one LSB is 3.3 V / 4096 = 0.806 mV. That is the smallest voltage change it can resolve. Go to 16 bits over the same range and you get 65,536 levels, an LSB of just 50 microvolts — sixteen times finer than 12 bits, because you added 4 bits and 2^4 = 16.

Why this matters: resolution sets the best-case precision of every reading and, together with the reference, the smallest detail you can measure. But here is the honest catch that beginners miss: the bits printed on the datasheet are only the POTENTIAL resolution. If the converter's own noise is larger than an LSB, the lowest bits are just flickering randomness and carry no real information. A 24-bit converter buried in noise might deliver only 20 honest bits — which is exactly what ENOB, the effective number of bits, measures.

Over a 5 V reference: an 8-bit converter resolves 5/256 = 19.5 mV steps; a 12-bit resolves 5/4096 = 1.22 mV; a 16-bit resolves 5/65536 = 76 microvolts. Each extra 4 bits makes the step 16 times finer.

More bits = a finer ruler = smaller LSB step.

Resolution is not the same as accuracy. A converter can have 16 bits of resolution yet be wrong by many LSBs because of offset, gain error, and nonlinearity. Fine steps mean little if the steps are in the wrong places.

Also called
bit depthLSB size解析度位元數