ADC Resolution Calculator (LSB Size, Quantization Noise, ENOB, Oversampling & Sensor Resolution)
Work out an analog-to-digital converter's step size in millivolts from bits and reference voltage, the ideal SNR and quantization noise, the effective number of bits once your circuit's noise is included, the oversampling needed for extra bits, and what one count means in your sensor's units.
How it works
LSB = V_ref ÷ 2ᴺ · quantization noise = LSB ÷ √12; ideal SNR = 6.02 N + 1.76 dB · total noise = √(q² + circuit²); ENOB = (SNR_real − 1.76) ÷ 6.02 · noise-free bits = log₂(V_ref ÷ (6.6 × noise_rms)) · oversampling: 4ⁿ samples per extra n bits, output rate = f_s ÷ 4ⁿ
An ADC divides its reference voltage into 2ᴺ steps, and the step size is the finest change it can register — in volts, and after scaling in your sensor's units. Whether you can use that resolution depends on noise: quantization itself contributes LSB/√12, but the reference, the layout and the sensor usually contribute much more, so a 12-bit converter on a noisy board behaves like a 9- or 10-bit one. Averaging helps in two ways — it reduces random noise by √n and, if the noise exceeds a step, it recovers resolution below the step size (oversampling: four samples per extra bit). The calculator shows the ideal figures, the realistic ones, and the trade in sample rate.
Worked example
A 12-bit ADC with a 3.3 V reference has 4,096 counts of 0.806 mV — 0.024 °C per count for a 0–100 °C sensor. Quantization noise is 233 µV RMS (ideal SNR 74 dB), but with 1 mV of circuit noise the total is 1.03 mV: real SNR about 61 dB, ENOB about 9.9 bits, and the reading flickers over about 0.2 °C. Averaging 16 samples adds 2 bits (0.006 °C per count) and drops the 10 kHz sample rate to 625 results per second.
Frequently asked questions
Why does my 12-bit ADC reading jump around by several counts?
Because the noise on the input, the reference and the supply is larger than one step. Common fixes: a clean reference instead of the 3.3 V rail, decoupling and a ground plane, an RC filter at the input, sampling when the CPU and radio are idle, and averaging many samples.
How does oversampling add bits?
If the input has at least a step's worth of random noise, the ADC lands on adjacent codes in proportion to where the true value sits between them; averaging 4ⁿ samples and dividing by 2ⁿ recovers n extra bits. Without noise the average is just the same code — add dither. The extra bits are still limited by the circuit noise.
Is a 16-bit or 24-bit ADC worth it?
Only with an analog front end to match: 24-bit delta-sigma converters deliver 18–21 noise-free bits with careful design, and a 16-bit part on a noisy microcontroller board often gives no more usable bits than a 12-bit one. Spend on the reference, layout, filtering and gain before spending on bits.
Related calculators
Electronics estimates use simplified models and typical component values. Check the datasheet limits (voltage, current, temperature, gate charge, ESR), prototype, and measure — especially for anything connected to mains, batteries or motors.