I am trying to determine DC errors for ADC128S102QML-SP. More specifically, I am referring to the gain error, offset error, INL error, and DNL error. It is a 12 bit ADC with a supply voltage of 5V.

From the datasheet, I concluded the following:

1. Offset error is 2.3 LSB which I believe translates to an error of
   (2.3212)⋅5=2.81mV.(2.3212)⋅5=2.81mV.
   
   
2. INL is 1.4 LSB which I believe translates to an error of
   (1.4212)⋅5=1.71mV.(1.4212)⋅5=1.71mV.
   
   
3. DNL is 1.5 LSB which I believe translates to an error of
   (1.5212)⋅5=1.83mV.(1.5212)⋅5=1.83mV.
   

Does the above calculations make sense? Also, how I can calculate the gain error?

 

You assume also that your operating on a supply voltage of exactly 5V. On a 12-Bit ADC we have at work there is a 1.1V reference upon which the ADC takes it's readings. Since the 1.1V reference is based on a band-gap voltage and can itself vary from one sensor to the next, we have to work that into the equation as well. For instance a 1.1V reference at 12-Bit should report back a value of 901 .... 1.1V/5V * 4095 = 900.9 .... Say for example we obtain a reading of 844 instead of 901. The compensation would be 0.9367 .... 844/901 = 0.9367 .... So a 5V reference would actually use 4.6835 V or 5.3165 V depending on which direction we needed to apply the compensation.

 

The ADC128S102QML-SP reference is the analog power pin, so all errors are referenced to that.
As noted, any error in that voltage will add to the listed data sheet errors, thus it should be supplied by an accurate voltage reference source.