Comparing ADC SNR, SINAD and ENOB Without Mixing Definitions

SNR, SINAD and effective number of bits describe related aspects of converter performance, but they are not interchangeable. A comparison can be misleading when harmonic distortion is excluded from one result and included in another. Establish the metric definitions before judging two ADCs or two boards.

Read what the calculation includes

Analog Devices tutorial MT-003 explains SNR, SINAD, ENOB and related spectral measurements. It also notes differences in conventions, making the manufacturer’s precise definition and analysis conditions important.

Record whether the calculation includes harmonics, excludes DC and integrates over the full Nyquist band or a narrower band. Identify which spectral bins are removed around the fundamental and harmonics. Keep these rules with the reported number rather than relying only on the metric label.

Match the signal conditions

Use the same input frequency, amplitude, sample rate and converter configuration for a meaningful comparison. Include reference voltage, supply and relevant digital filtering. Dynamic performance can change with frequency even when the nominal bit resolution is unchanged.

For illustration, the familiar full-scale sinewave relationship gives ENOB approximately equal to SINAD minus 1.76 dB, divided by 6.02. A 74 dB SINAD result corresponds to approximately 12 effective bits under that convention. It does not mean every DC reading has exactly twelve trustworthy bits.

Separate converter and test-source limits

Check the analog source’s noise and harmonic distortion and the sampling clock’s contribution. A measured harmonic may come from the generator or driver rather than the ADC. Document filtering and source verification before attributing the entire result to the converter.

Use an acquisition and FFT method appropriate to the signal, including coherent sampling or a suitable window with correct scaling. Retain record length and averaging settings. A visually lower displayed noise floor does not necessarily represent a lower integrated noise value.

Report the useful conclusion

Present SNR and SINAD separately when distortion is relevant, and state the bandwidth used. Include the dominant harmonics or spurs where they explain the difference. If the application uses a restricted signal band, report that result with its band rather than comparing it silently with a full-band catalog value.

Keep raw records and analysis settings so the calculation can be reproduced. Recheck definitions when changing software or measurement equipment. The result should show whether the converter meets the application’s noise and distortion requirements under stated conditions, rather than using one attractive ENOB figure as a universal replacement for resolution, linearity and every other aspect of measurement accuracy.