Issue No. 05 - May (1986 vol. 35)
E.S., Jr. Swartzlander , TRW Electronic Systems Group
We thank the authors for taking the time to prepare a formal comment on our 1983 paper . Some of the confusion results from our use of b - 1 bits to represent the fractional part of the sign/logarithm instead of b bits as used by Hongyuan and Lee in their comment. The performance curves (Figs. 12-14 of ) are plots of output noise variance/output signal variance in units of 2-2 (number of fractional bits). Since there are b - 1 fractional bits, this is equivalent to 22-2b. With this correct interpretation, (48) agrees with Figs. 12-14 in .
H. J. Nagle, D. Satish Chandra and E. J. Swartzlander, "Authors' Reply," in IEEE Transactions on Computers, vol. 35, no. , pp. 484, 1986.