Consider a new floating point format that follows the

IEEE spec, except with $k=3$ exponent bits, $n=2$

mantissa bits, $s=1$ sign bit.

represent $\frac{1}{16}$ as a binary representation.