Conceptual Question: Is the sampled value under 1 LSB still accurate in ADC?

[elberlee]

It is always defined that the LSB is the resolution of ADC.

Here I have one example, which confused me a lot.
Example, two different signal are sampled and the difference is made afterwards. When the difference value is less than 1 LSB, is it the result still counted as an accurate one?

Additionally, the two sampled value are averaged 1000 times separately before subtraction, can I think the "sampling + averaging" process improve the sampling resolution?

Many thanks,

Elber

 

[andre_luis]

Once accumulated on a single big value resulted from sum of multiple acquired samples, theoretically we are able to determine an extended precision value, due statistically occurrences must be unequally distributed.



+++

 

[zorro]

Signals below 1 LSB can in fact be detected by means of dithering.
Look for "dithering".
Regards

Z