Lex wrote on Jul 23rd, 2010, 12:18am:May I ask why analog techniques are not good enough to get to 1mV offset?
What you got going on here is 5 different people
saying you can use 5 different methods to do the same thing.
That said - there are multiple ways to get this done.
Geometry scaling (big devices and better matching) will only work up to a point. You need the specific matching data for the foundry process to tell you if this is viable.
The 16X increase to get better matching is an academic perception that is not valid, because it is based upon a number of older papers that are no longer valid at smaller geometry.
Static offset compensation thru a digital adjustment works.
Dynamic compensation with capaciive offset cancellation works too.
The quick decision between the above two is whether or not its a clocked system. Sampled time systems lend themselves well to doing dynamic offset cancellation.
OP-amps make poor comparators. You are gain-BW response limiting the device so it can work inside a feedback loop. Not needed in a comparator.
As for the signal you need to respond to? Sounds like you need more gain in the system.
There are lots of papers and books out there on this subject.
Time to do some research.