Description:
------------
Writing a class to handle conversion of UTF-8 input into SJIS using usual 
$body = mb_convert_encoding($body, "SJIS", "UTF-8");
function, usually works, but so far by testing, I identified three kanji's which makes this function fail to convert these correctly and thus causing mojibake

The three UTF-8 characters is identified as follows 
chr(227).chr(136).chr(177)
chr(227).chr(136).chr(178)
chr(227).chr(136).chr(185)

which in SJIS corresponds to
chr(135).chr(138)
chr(135).chr(139)
chr(135).chr(140)

Reproduce code:
---------------
I created a "quick'n'dirty" solution as follows, which surely isnt optimal

$body = str_replace(chr(227).chr(136).chr(177), '#mojihack1#', $body);
$body = str_replace(chr(227).chr(136).chr(178), '#mojihack2#', $body);
$body = str_replace(chr(227).chr(136).chr(185), '#mojihack3#', $body);

$body = mb_convert_encoding($body, "SJIS", "UTF-8");

$body = str_replace('#mojihack1#', chr(135).chr(138), $body);
$body = str_replace('#mojihack2#', chr(135).chr(139), $body);
$body = str_replace('#mojihack3#', chr(135).chr(140), $body);