Packed Binary, An EBCDIC Conservationist
This is the most common and standard implementation of a Packed Decimal binary field. Some other verbiage used to reference this data type are "Computational-3", "Packed", and "Packed Decimal". It basically works by "packing" two digits into a single bytes effectively reducing storage for the number by half as compared to the default character (or "display") field data type.
The methodology for packing these digits down into a single byte is called Binary Coded Decimal, or BCD. BCD is where each digit of a decimal number is binary coded. Since each digit of any number must be between 0 and 9, it can be put into only 4 bits instead of the normal 8 bits, or 1 byte.
The last digit in a series of packed decimals is always the sign digit. It will be an "F" if it is unsigned, "C" if negative, and "D" if positive. Also, since there will always be an even number of nybbles (4 bits) that means that if a field had an even number of digits, not including the sign, then it would make for an odd number of total digits. If this is the case, then the upper most nybble should be filled with a zero, but to be safe, should just be ignored.