ADC Transition Noise and DNL / INL calculation

[nirc]

Hi All,

I have designed a 8 bit SAR ADC. (1 LSB ~ 10 mV)
In order to test my DNL and INL, I have an outomatic system with ramp input.
(The input resolution is ~ 2mV - in order to reduce estimation errors)

My typical DNL is ~+/- 0.1 LSB and INL ~ +/- 0.2 LSB.
Every ADC has a transition noise with an associated transition region.
Due to the above I can see every ~ 500 (random) conversions 1 LSB toggle.
(The output code word goes up to the next code and immediately going down)

The DNL calculation depends by definition by the actual transition voltage.
Should I ignore this noise in my DNL / INL calculation?
Should I take the random transion noise in my estimation? if yes how?

Thank you.

 

[nirc]

Hi All,

From analyzing few articles, I think that I got a solution.

The DNL / INL estimations depends on the actual transition point.
The actual transition points should be estimated. (statistical measurement)
( If transition noise exist , one should repeat the measutement at the exact input
in order to locate the exact transition point)

The transition noise and transition region should be reported in a different
specification. Typical values are ~ +/- 0.2 LSB.