The offset error in the current stage is corrected by the succeeding stages due to the redundancy.I am looking at this in a simpler manner.
Suppose we wanted to input 0.375 to get 10 logic but because of an error 0.25 error we got it to input 0.125.
so at the output we have 0.125*2=0.25 whichis in the boundary between the two areas but suppose it will pick the 10 logic.
suppose we have an error of 0.25+0.125=0.375 so if my input is 0.375 then it will enter 0 and will stay in that logic level becase 2*0=0 .
But how can i see the exact value of the allowed error?