Since

2^93 = 9,903,520,314,283,042,199,192,993,792 and, of course, 10^28 = 10,000,000,000,000,000,000,000,000,000

twenty-eight digits can represent 93 bits with fairly good efficiency.

The method devised by IBM for representing three decimal digits in 10 binary bits which we saw previously will serve as part of the basis for doing this.

Let ooo (or ppp, qqq) represent a digit from 0 to 7 represented by the following code:

000 0 001 1 010 2 011 3 100 4 101 5 110 6 111 7

and let d (or e, f) represent a digit from 8 to 9 in the following code:

0 8 1 9

then, any three decimal digits can be represented within 10 bits as follows:

0ooopppqqq (3 digits with 3 bits - 1 way) 100dpppqqq (2 digits with 3 bits, 1 digit with 1 bit - 3 ways) 101oooeqqq 110ooopppf 11100deqqq (1 digit with 3 bits, 2 digits with 1 bit - 3 ways) 11101dpppf 11110oooef 1111100def (3 digits with 1 bit - 1 way)

with 24 unused combinations remaining.

As, again, previously noted, the table illustrating the decimal to binary code is based on ooo, ppp, and qqq representing the first, second, and third digits respectively (if they are from 0 to 7) and on d, e, and f representing the first, second, and third digits respectively if they are instead either 8 or 9. This same notational convention will be used in outlining the method to be explained on this page.

Since 2^10 is 1,024, slightly greater than a power of 10, this method cannot serve directly for storing binary bits in decimal digits; it is instead useful for performing the reverse task of storing decimal digits in binary bits.

The coding scheme I propose works as follows:

oAAABBBCCCDDDEEEFFFGGGHHHIII

Here, o represents a digit from 0 to 7, representing the first three bits of the 93 bits to be represented. AAA through III each represent three digits, from 000 to 999, representing 10 bits of the 93 bits to be represented, provided that each group of 10 bits belongs to one of the 1,000 codes out of 1,024 that can be so represented.

nnn0BBBCCCDDDEEEFFFGGGHHHIII nnn1AAACCCDDDEEEFFFGGGHHHIII nnn2AAABBBDDDEEEFFFGGGHHHIII nnn3AAABBBCCCEEEFFFGGGHHHIII nnn4AAABBBCCCDDDFFFGGGHHHIII nnn5AAABBBCCCDDDEEEGGGHHHIII nnn6AAABBBCCCDDDEEEFFFHHHIII nnn7AAABBBCCCDDDEEEFFFGGGIII nnn8AAABBBCCCDDDEEEFFFGGGHHH

Here, nnn is a series of three digits from 800 to 999. This involves 200 possible values, 192 of which are used, to represent the 8 possible values of the first three bits of the 93 to be encoded, times 24 possible values for one of the groups of 10 bits in the remaining 90 bits. The value will be generated by the formula 800 + 24 * a + b, where a is the value of the first three bits from 0 to 7, and b is a number from 0 to 23 representing each of the 10 bit codes, which did not correspond to three digits in the coding of three digits to ten bits given above, in the numerical order of these codes, as shown in the table below:

0 1111101000 7 1111101111 8 1111110000 15 1111110111 16 1111111000 23 1111111111

Thus, these codes cover the case where exactly one of those groups has one of the 24 values which is not covered by the first encoding. Once we cover all possible cases where any combination of those groups, even all of them, can have such a value, we will have our complete encoding of 93-bit values into decimal digits. Note that the eight nnn values of 992 through 999 are not used so far.

nnn90ppCCCDDDEEEFFFGGGHHHIII (00) nnn95ppAAABBBCCCDDDEEEHHHIII BBBDDDEEEFFFGGGHHHIII (25) BBBCCCDDDEEEFFFGGGIII AAADDDEEEFFFGGGHHHIII (50) AAACCCDDDEEEFFFGGGIII BBBCCCEEEFFFGGGHHHIII (75) AAABBBDDDEEEFFFGGGIII nnn91ppAAACCCEEEFFFGGGHHHIII nnn96ppAAABBBCCCEEEFFFGGGIII AAABBBEEEFFFGGGHHHIII AAABBBCCCDDDFFFGGGIII BBBCCCDDDFFFGGGHHHIII AAABBBCCCDDDEEEGGGIII AAACCCDDDFFFGGGHHHIII AAABBBCCCDDDEEEFFFIII nnn92ppAAABBBDDDFFFGGGHHHIII nnn97ppBBBCCCDDDEEEFFFGGGHHH AAABBBCCCFFFGGGHHHIII AAACCCDDDEEEFFFGGGHHH BBBCCCDDDEEEGGGHHHIII AAABBBDDDEEEFFFGGGHHH AAACCCDDDEEEGGGHHHIII AAABBBCCCEEEFFFGGGHHH nnn93ppAAABBBDDDEEEGGGHHHIII nnn98ppAAABBBCCCDDDFFFGGGHHH AAABBBCCCEEEGGGHHHIII AAABBBCCCDDDEEEGGGHHH AAABBBCCCDDDGGGHHHIII AAABBBCCCDDDEEEFFFHHH BBBCCCDDDEEEFFFHHHIII AAABBBCCCDDDEEEFFFGGG nnn94ppAAACCCDDDEEEFFFHHHIII AAABBBDDDEEEFFFHHHIII AAABBBCCCEEEFFFHHHIII AAABBBCCCDDDFFFHHHIII

Here, nnn represents the first 3 bits, and the additional combination
from the set of 24 from the first of the two groups of 10 bits that are of
that form. pp, a series of two digits from 00 to 99, has 100 possible
values, 96 of which are used to represent the combination from the set of
**24** from the second of the two groups of 10 bits, and
which of the **4**
arrangements of which groups of 10 bits have these additional codes allocated
to the current numeric code is present. The numbers in parentheses after the
first set of four arrangements show the displacements to be added in the pp
field to the number from 0 to 23 containing the information about the second
group with an auxilliary value.

nnn990ppqqDDDEEEFFFGGGHHHIII (00 00) CCCEEEFFFGGGHHHIII (00 25) BBBEEEFFFGGGHHHIII (00 50) AAAEEEFFFGGGHHHIII (00 75) CCCDDDFFFGGGHHHIII (25 00) BBBDDDFFFGGGHHHIII (25 25) AAADDDFFFGGGHHHIII (25 50) BBBCCCFFFGGGHHHIII (25 75) AAACCCFFFGGGHHHIII (50 00) AAABBBFFFGGGHHHIII (50 25) CCCDDDEEEGGGHHHIII (50 50) BBBDDDEEEGGGHHHIII (50 75) AAADDDEEEGGGHHHIII (75 00) BBBCCCEEEGGGHHHIII (75 25) AAACCCEEEGGGHHHIII (75 50) AAABBBEEEGGGHHHIII (75 75) nnn991ppqqBBBCCCDDDGGGHHHIII AAACCCDDDGGGHHHIII AAABBBDDDGGGHHHIII AAABBBCCCGGGHHHIII CCCDDDEEEFFFHHHIII ... nnn995ppqqAAABBBCCCEEEFFFGGG AAABBBCCCDDDFFFGGG AAABBBCCCDDDEEEGGG AAABBBCCCDDDEEEFFF

Now we are dealing with the case where three of the nine groups of ten bits have a representation from the additional 24 combinations. Together, pp and qq mean that sixteen different arrangements of these three groups can be indicated. The displacements added to pp and qq are noted in the parentheses after the first group of 16 combinations. One additional digit lets us cope with 160 possibilities, but in fact we only need to worry about 84 possibilities.

nnn9990ppqqrrEEEFFFGGGHHHIII (00 00 00) DDDFFFGGGHHHIII (00 00 25) CCCFFFGGGHHHIII (00 00 50) BBBFFFGGGHHHIII (00 00 75) AAAFFFGGGHHHIII (00 25 00) DDDEEEGGGHHHIII (00 25 25) CCCEEEGGGHHHIII (00 25 50) CCCDDDGGGHHHIII (00 25 75) ... BBBCCCEEEGGGIII (75 75 75) nnn9991ppqqrrBBBCCCDDDGGGIII CCCDDDEEEFFFIII BBBDDDEEEFFFIII BBBCCCEEEFFFIII BBBCCCDDDFFFIII BBBCCCDDDEEEIII DDDEEEFFFGGGHHH ... AAABBBCCCDDDEEE

Here, four of the nine groups of ten bits use one of the additional codes. This involves 126 possibilities. pp, qq, and rr together allow sixty-four combinations to be indicated, so the additional digit need only be either 0 or 1.

nnn99990ppqqrrssFFFGGGHHHIII nnn999990ppqqrrssttGGGHHHIII nnn9999990ppqqrrssttuuHHHIII nnn99999990ppqqrrssttuuvvIII nnn999999990ppqqrrssttuuvvww

And there is no problem continuing the encoding, without needing to go back and use the eight unused nnn values, since we are now dealing with 126, 84, 36, 9, and 1 combinations again, which we have already seen comfortably fits in the space available, and we now have even more space with more two-digit fields, each of which contributes a factor of four.

[Next] [Up] [Previous] [Index]