4.7  Summary of Estimation Error

The simulations to determine the energy dissipation were carried out on the set of 10 random-generated test vectors shown in Table 4.21. In Table 4.22 we report the percentage errors for the energy estimation of the units presented above and the number of transistors for each implementation. The error is computed by expression (4.2) with confidence level of 99%. Table 4.22 also shows that the accuracy of the estimation is independent of the size of the circuit (number of transistors). This confirms the dimension independence property of this approach which is a common feature of the Monte Carlo methods.

n. vectors
1 x 10100101111110100101011011000111010010111111010010110
d 10100001100111000110110101001001010000110011100011100
2 x 10011001010001101111110010110111001100101000110111111
d 10001010110110011011011011111110000101011011001101110
3 x 10111110100010110000101111111101011111010001011000011
d 10011110011101001110011101011011001111001110100111010
4 x 11000101100101010100101101000000100010110010101010011
d 11000011101101000011110101110110100001110110100001111
5 x 11011001101010000101011001010110101100110101000010110
d 10110001100101111010011101100100011000110010111101010
6 x 11110110100110110111010101011001111011010011011011101
d 11000000001110110111011001000100100000000111011011110
7 x 11100011111001001000111010100011110001111100100100100
d 11111010101100011011001110011100111101010110001101101
8 x 10110111101100001101010101000010011011110110000110110
d 10100110101011000011111000000111010011010101100010000
9 x 10000110010001100001001111011101000011001000110000101
d 10001010101011001001001111000010000101010101100100101
10 x 10101111010011101111101010110110010111101001110111110
d 10110010100000011000011001000111011001010000001100010

Table 4.21: The 10 random vectors.

standard low-power
unit Eop [nJ] e % FETs Eop [nJ] e % FETs
radix-4 45.5 2.1 21000 26.0 1.7 16400
radix-8 47.5 4.1 31000 28.5 4.2 24900
radix-16 46.0 2.8 30400 30.0 3.3 25600
radix-512 66.5 2.1 83600 55.0 2.4 68300
comb. div/sqrt 46.0 3.6 25100 29.5 2.7 23800

Table 4.22: Percentage error in energy estimation.


File translated from TEX by TTH, version 1.1 and by ME. Last Modified : Fri Jul 9 11:14:37 PDT 1999