English is currently available.
Arduino DISCUSSION
The UNO R4 supports higher ADC resolution, while the Arduino API retains a 10-bit default for compatibility. After selecting 14 bits, what determines whether the extra digits contain useful information?
For a nominal 5 V reference, how should quantization, reference error and input noise be compared?
The ideal step is Vref / 2^N. At 5 V and 14 bits that is about 0.305 mV, versus 4.88 mV at 10 bits. These are code widths, not accuracy specifications. Reference tolerance, ADC offset and gain error, nonlinearity and noise still affect the result.
Measure a stable input and inspect the spread of raw readings before converting to engineering units. Also measure or characterize the reference under the intended conditions. A 1% reference error can dominate many ADC counts. A sensor's own uncertainty may be larger still, so include the complete signal chain in the error budget.
Averaging can reduce some random noise but cannot remove a fixed reference or scaling error. Check source impedance and acquisition requirements, then compare known inputs across the useful range. State the selected resolution in the sketch and adjust conversion scaling accordingly; retaining a divisor intended for 10-bit data produces the wrong voltage.
Back to Arduino community
