Hello,
I am testing a 12 bit dac. I do not have enough memory to exercise all the 4096 digital codes.  It is a segmented architecture. It has a 6 bit R-dac and a 6 bit capdac.
So what I do is to step thru the coarse settings only which are 2^6-->64 codes. I then calculate inl/dnl. For inl, my question/concern is should I normalize the error with respect to the coarse step or should I normalize it with respect to the 12 bit level lsb. Of course, the 12 bit level lsb is lot smaller and gives me pretty bad inl. Is this valid ?

In the literature, everyone goes thru all the codes so their normalization is a no brainer, i.e. they normalize it with 1lsb of the entire converter.

Any comments ?
thanks
analog2000

Hi analog2000,

Sure you should normalize the INL to 1LSB of the 12 bit DAC. If it is pretty bad, that probably caused by poor mismatch of the coarse DAC. Now that you have not enough memory to handle all the 4096 codes, you'd better measured a part of codes each time. In one word, all codes should be excercised eventually.


Yawei