Passphrase Encoding and Entropy

[Oskar L.]

Hi,

If I'm not misinformed the passphrase can be encoded using different
character sets. Can I in gpg change witch one is used, or does it depend
on witch operating system I use? How does it affect the way you calculate
entropy if a character is encoded using 16 or 24 bits (as some characters
are in UTF-8) or as a 8-bit character, if at all?

Also, let's say it is known that the characters in a passphrase has been
selected from the 64 ASCII characters A-Z, a-z, 0-9, # and $. This will
give each character an entropy of 6 bits (log2(64)), witch if I understand
correctly means that 6 of the 8 bits used to represent the character are
unknown. But can you in real life tell witch six, for example for the
character A, witch in binary is 01000001? The first zero will of course be
known, but is there a second known digit?

Oskar