Re: the relationship between harmonic distortion and INL or
There are is easy relationships between INL and HDn. But there si a worst case estimation which could be put into an expression.
Assume that the INL is a rect function with value transition in the middle of the ADC or DAC range. Then assume that the test signal is a full scale sinewave. So the time domain error function is a rectified sinewave. So you can do fourier analysis on this and also calculate the integrated time domain power. Because the fourier expression give a decay rate for the higher harmonics and the total error power in time domain and frequency domain are equal you can estimate the worst case for the harmonic spurs.
The above estimation case is a typical MSB transition error. Typical DNL are more smooth or a combination of some higher bit transition and a smooth curve. So the error bound is useable.
The detailed calculation is highschool math!