You need to transform analog input from the real world to something digestible to the digital world.

Some definition:

  • Signal to noise ratio: \(\hbox{SNR} = {P_\hbox{signal}\over P_\hbox{noise}}\)
  • Effective number of bits: indicated as ENOB

this holds (calculation)

Nyquist-Shannon sampling theorem

Signals that differs of \({1\over T}\) are sampled the same

Oversampling and decimation

In same cases you can improve the resolution of your ADC of \(n\) bits simply summing together \(f_\hbox{oversampling} = 4^n \cdot f_{\hbox{sampling}}\) and then scaling by a factor of \(s = 2^n\). This is possible if the following assumptions hold

  • the signal of interest should not vary significantly during a conversion
  • there should some noise in the signal with amplitude at least 1 LSB