Repeat Ex. 30.10 if 1.5 bits/stage are used. Assume the converter is ideal and the comparators switch precisely at VCM/2 (= 250 mV here) and 3VCM/2 (= 750 mV here). Assume all latches initially contain zeroes.
noting that b8=a1.57⊕c7=1. We can reorder the bits so the MSB is on the left, the LSB is on the right, yielding 1 0001 1001 and subtract 0 0111 1111 yielding
1 0001 1001 (281)
– 0 0111 1111 (127)
0 1001 1010 (154)
This is the result given in Ex. 30.10 (1001 1001, or decimal 153) plus 1 LSB. The 1 LSB discrepancy can be traced to Eq. (30.66) where we used 0.5 LSBs. Because our resolution is at best 1 LSB, sometimes the result will experience a round-off error. To understand this in the subtraction above, the more correct decimal representation of VCM−0.5 LSBs is 127.5 and the more correct decimal output is 153.5.
vin=a1.5N−1⋅2VCM+(a1.5N−1b1.5N−1+a1.5N−2)⋅VCM+(a1.5N−2b1.5N−2+a1.5N−3)⋅2VCM+(a1.5N−3b1.5N−3+a1.5N−4)⋅4VCM+…+a1.50b1.50⋅2N−1VCM–(VCM–0.5LSB)(30.66)