An analog-to-digital converter, or ADC, is the bridge between the continuous world and the digital world. A sensor produces a voltage that varies smoothly, and the ADC reports that voltage as a whole number, called a code, which a microcontroller or computer can store and process. The fineness of that bridge is called resolution, and it sets the smallest voltage change the converter can distinguish. The ADC Resolution Calculator computes that step size from the number of bits and the reference voltage, but understanding the formula helps you choose a converter that is good enough without paying for precision that noise will hide.
Resolution starts with bits. An N-bit converter can produce 2 raised to the power N distinct codes. A 10-bit converter produces 2 to the power 10, which is 1,024 codes. A 12-bit converter produces 2 to the power 12, which is 4,096 codes. A 16-bit converter produces 2 to the power 16, which is 65,536 codes. Each additional bit doubles the number of codes and halves the step size, which is why the jump from 10 bits to 12 bits is a fourfold improvement in fineness, not a small one. The step size, often called the least significant bit or LSB size, equals the reference voltage divided by the number of codes. The reference voltage is the full-scale voltage the converter measures against, and it must be known and stable, because an error in the reference scales every reading.
Resolution is not the same as accuracy, and confusing them is the most expensive mistake in converter selection. Resolution says how finely the output is divided. Accuracy says how close a reading is to the true voltage, and it depends on reference accuracy, offset and gain errors, non-linearity, temperature drift, and noise. A 16-bit converter with a noisy input and an inaccurate reference can produce readings that are finely divided but wrong, while a well-designed 12-bit converter with a clean reference can be more trustworthy. Extra bits help only if the signal and the circuit are quiet enough for the smallest steps to mean something.
A worked example
Suppose a 12-bit converter uses a reference voltage of 5 volts. The number of codes is 2 to the power 12, which is 4,096. The step size is 5 divided by 4,096, which is 0.0012207 volts, or about 1.22 millivolts, because 0.0012207 multiplied by 1,000 equals 1.2207 millivolts. A sensor voltage of 2.5 volts would therefore produce a code of about 2.5 divided by 0.0012207, which is 2,048, exactly half of full scale, as expected for a mid-range voltage. A change of one code corresponds to a change of about 1.22 millivolts in the input, so a sensor signal that changes by only 0.5 millivolts cannot be resolved by this converter, however the result is displayed. For comparison, a 10-bit converter with the same 5-volt reference has 1,024 codes and a step size of 5 divided by 1,024, which is 0.0048828 volts, or about 4.88 millivolts, which is four times larger than the 12-bit step, because 4.88 divided by 1.22 equals 4. A 16-bit converter with the same reference has 65,536 codes and a step size of 5 divided by 65,536, which is about 0.0000763 volts, or 0.076 millivolts, but that fineness is useful only if the circuit noise is well below 0.076 millivolts, which requires careful design.
That example also shows how to check a datasheet claim. If a converter is described as 12-bit but its effective number of bits, a measure that accounts for noise, is only 10.5 bits, then its usable step is closer to the 10-bit figure than the 12-bit figure, and the extra codes are mostly noise. Effective bits, signal-to-noise ratio, and the reference specification are therefore worth reading alongside the headline bit count, especially for precision measurement.
Common mistakes
The first mistake is dividing by 2 to the power N minus 1 instead of 2 to the power N, or vice versa, without noting which convention a datasheet or calculator uses; the difference is one step in thousands, small but worth stating. The second mistake is using the supply voltage as the reference when the converter actually uses a separate, lower reference, which scales every step size wrongly. The third mistake is choosing bits for a display rather than for a measurement need: a readout that shows many decimal places does not make the measurement more accurate. The fourth mistake is ignoring input range: a sensor that only swings over 1 volt of a 5-volt range uses only a fifth of the codes, so its effective resolution is about two bits worse than the headline figure, because a fifth is close to a quarter, which is two bits. The fifth mistake is neglecting anti-alias filtering and sampling rate, which affect whether a changing signal is captured faithfully at all, regardless of step size.
Practical design follows from the arithmetic. Match the converter's range to the sensor's swing, using amplification where a small signal would otherwise waste codes. Keep the reference stable and the input quiet, with proper grounding and, where needed, filtering. Average several readings only when the signal is steady enough for averaging to reduce noise rather than blur a real change. And always verify a built circuit against a known voltage, such as a calibrated source or a trusted meter, at several points across the range, not just at zero.
A practical routine
A routine that works for most ADC choices has five steps. First, state the sensor's voltage swing and the smallest change you must detect. Second, choose a reference voltage and calculate the step size for candidate bit counts using the division shown above. Third, check the effective bits and noise specification, not just the headline bits. Fourth, match the input range to the sensor swing, amplifying a small signal if needed. Fifth, verify the built circuit against known voltages across the range. For related electrical calculations, the Voltage Drop Calculator estimates losses along a cable, and the Resistor Color Code Calculator decodes resistor values for divider and filter circuits. For another calculation walkthrough, the guide Data Mapping Transformer Guide explains how values are transformed between formats.
Resolution is a division: reference voltage divided by the number of codes. Know the reference, count the codes, and let noise, not marketing, decide how many bits you really need.