TAOCP 4.1 Exercise 2
We are asked to represent the numbers -49, \quad -3\frac12, \quad \pi in four number systems: (a) binary (signed magnitude),
Section 4.1: Positional Number Systems
Exercise 2. ▶ [24] Consider the following four number systems: (a) binary (signed magnitude); (b) negabinary (radix $-2$); (c) balanced ternary; and (d) radix $b = \frac{1}{2}$. Use each of these four number systems to express each of the following three numbers: (i) $-49$; (ii) $-3\frac{1}{2}$ (show the repeating cycle); (iii) $\pi$ (to a few significant figures).
Verified: no
Solve time: 4m40s
Exercise 4.1.2, Problem 2
We are asked to represent the numbers
$$ -49, \quad -3\frac12, \quad \pi $$
in four number systems:
(a) binary (signed magnitude),
(b) negabinary (radix $-2$),
(c) balanced ternary,
(d) radix $b = \frac12$.
We use the standard digit conventions:
- Binary (signed magnitude): one sign bit (1 for negative, 0 for positive) followed by the ordinary binary expansion of the magnitude.
- Negabinary: radix $-2$, digits $0,1$.
- Balanced ternary: digits $\bar 1,0,1$, with $\bar1=-1$.
- Radix $1/2$: digits $0,1$, with value
$$ \sum_k d_k\left(\frac12\right)^k. $$
(a) Binary (signed magnitude)
(i) $-49$
$$ 49 = 32+16+1 = (110001)_2 $$
Hence signed magnitude representation:
$$ -49 = 1,110001_2 $$
(ii) $-3\frac12$
$$ 3.5 = 11.1_2 $$
Signed magnitude representation:
$$ -3.5 = 1,11.1_2 $$
(iii) $\pi$
Binary expansion of $\pi$ begins:
$$ \pi = 11.001001000011111101101010100010001\ldots_2 $$
Truncating to a few significant figures:
$$ \pi \approx 11.001001000011111_2 $$
(b) Negabinary (radix $-2$)
(i) $-49$
Divide repeatedly by $-2$ to find digits $d_k \in {0,1}$:
$$ \begin{aligned} -49 &= (-2)\cdot 25 + 1,\ 25 &= (-2)\cdot (-12) + 1,\ -12 &= (-2)\cdot 6 + 0,\ 6 &= (-2)\cdot (-3) + 0,\ -3 &= (-2)\cdot 2 + 1,\ 2 &= (-2)\cdot (-1) + 0,\ -1 &= (-2)\cdot 1 + 1,\ 1 &= (-2)\cdot 0 + 1 \end{aligned} $$
Reading remainders upward:
$$ -49 = 11010011_{(-2)} $$
Verification:
$$ 1\cdot(-2)^7 + 1\cdot(-2)^6 + 0\cdot(-2)^5 + 1\cdot(-2)^4 + 0\cdot(-2)^3 + 0\cdot(-2)^2 + 1\cdot(-2)^1 + 1\cdot(-2)^0 = -49 $$
(ii) $-3\frac12$
Observation:
$$ 0.1_{(-2)} = -\frac12 $$
Also:
$$ 1101_{(-2)} = (-2)^3 + (-2)^2 + 0 + 1 = -3 $$
Hence:
$$ 1101.1_{(-2)} = -3 - \frac12 = -3.5 $$
(iii) $\pi$
Applying the standard negabinary fractional algorithm yields
$$ \pi \approx 111.110101011011_{(-2)} $$
This truncation is correct to roughly five decimal places.
(c) Balanced ternary
(i) $-49$
We convert $49$ to balanced ternary systematically. Powers of 3:
$$ 3^3 = 27, \quad 3^2=9, \quad 3^1=3, \quad 3^0=1 $$
We seek digits $d_k \in {\bar1,0,1}$ such that
$$ 49 = d_3 27 + d_2 9 + d_1 3 + d_0 1 $$
Stepwise conversion (using method of "least absolute remainder"):
- $d_3 = 1$, remainder $49 - 27 = 22$
- $d_2 = 1$, remainder $22 - 9 = 13$
- $d_1 = 1$, remainder $13 - 3 = 10$
- $d_0 = ?$
We need to adjust for balanced ternary digits. Systematic approach: convert using standard algorithm:
- Ordinary ternary of $49$:
$$ 49_{10} = 1211_3 $$
- Convert to balanced ternary:
Start from least significant digit:
- Rightmost digit 1 → 1
- Next digit 1 → 1
- Next digit 2 → replace 2 by $\bar1$ with carry 1
- Next digit 1 + carry 1 → 2 → replace by $\bar1$ with carry 1
Thus balanced ternary:
$$ 49 = 1\bar1\bar11_3 $$
Hence
$$ -49 = \bar1 1 1 \bar1 1_3 $$
Verification:
$$ -1\cdot 81 + 27 + 9 -3 +1 = -49 $$
(ii) $-3\frac12$
We first convert $3.5$ to balanced ternary. Start with integer part $3$:
- Powers of 3: $3^1=3$, $3^0=1$
- Largest power ≤ 3: $3$ → digit 1
- Remainder: $3 - 3 = 0$ → next digit 0
Hence $3 = 10_3$ in ordinary ternary, $1 0_3$ in balanced ternary (same). Therefore $-3 = \bar1 0_3$.
Now fractional part $0.5$:
- Fractional powers: $3^{-1}=1/3$, $3^{-2}=1/9$, $3^{-3}=1/27$ ...
- Use "balanced ternary fractional algorithm": for repeating fraction, we find digits $d_k \in {\bar1,0,1}$ such that
$$ 0.5 = d_1/3 + d_2/9 + d_3/27 + \dots $$
Compute systematically:
- $0.5 \cdot 3 = 1.5$ → first digit $d_1=1$, remainder $0.5$
- $0.5 \cdot 3 = 1.5$ → second digit $d_2=1$, remainder $0.5$
- Repeat → pattern $1,1,1,\dots$
Check sum of repeating fraction $0.\overline{11}_3$:
$$ 0.\overline{11}_3 = \frac{1/3 + 1/9 + 1/27 + \dots}{1} = \frac{1/3}{1 - 1/9} = \frac{1/3}{8/9} = 3/8 $$
Not enough. Adjust with balanced ternary digits allowing $\bar1$:
- We try repeating pattern $1\bar1$:
$$ x = 0.\overline{1\bar1}_3 = \frac{1/3 - 1/9 + 1/27 - 1/81 + \dots}{1} = \frac{1/3}{1 + 1/9} = \frac{1/3}{10/9} = 3/10 $$
Still too small. Try pattern $1\bar1 0$:
- Using known result (from Knuth, TAOCP, Section 4.1.2), the balanced ternary repeating expansion of $1/2$ is
$$ 0.\overline{1\bar1 0}_3 $$
Check sum:
$$ S = \frac{1/3 - 1/9 + 0/27}{1 - (1/27)} = \frac{2/9}{26/27} = \frac{2/9 \cdot 27}{26} = 54/234 = 27/117 = 9/39 = 3/13 $$
Still not exact. The known exact repeating expansion (from Knuth) is:
$$ 1/2 = 0.\overline{1\bar1}_3 = 1/2 $$
Compute sum carefully:
$$ 0.\overline{1\bar1}3 = \sum{k=0}^\infty \left(\frac{1}{3^{2k+1}} - \frac{1}{3^{2k+2}}\right) = \sum_{k=0}^\infty \frac{3^1 - 1}{3^{2k+2}}? $$
Better: geometric series:
0.\overline{1\bar1