Sed not finding/ replacing all characters in a directory - sed

Question

I am trying to find and replace single characters in a number of text files in a directory. Apologies for the possible duplication but I have not been able to find an answer in other sed threads.
I used Homebrew to install gnu-sed, and I'm using the command:
find . -name "*.txt" -exec gsed -i -e 's/ñ/–/g' '{}' \;
I have a 'test' file containing the characters I need replacing in the directory, and these are all found and replaced correctly. But other characters in other text files are not. e.g. 'Weíre to Denmark ñ all' (ñ also isn't found/ replaced).
Why might this be? How can I fix it? Thank you!
Edit - Output of
$ od -c filethatworks.txt | head -2
0000000 – ** ** \n – ** ** \n “ ** ** \n “ ** ** \n
0000020 — ** ** \n — ** ** \n - \n “ ** ** \n “ **
$ od -c filethatdoesnot.txt | head -2
0000000 T h o s e b l e s s e d d a
0000020 y s o f s u m m e r a r e
For a file that works, the file command returns
test.txt: UTF-8 Unicode text
and for one that does not:
ca001_mci_17071971.txt: Non-ISO extended-ASCII text, with very long lines, with CRL
F line terminators

Answer 1

Characters are human concepts. When characters are to be represented in computer files, they need to be encoded. Encoding associates each character with an integer called a code point.
For example, take the character "ă" (that's lower case "a" with a breve on top, used in Romanian spelling for the vowel /ə/); in the old days of MS-DOS, we quite often used an encoding called "code page 852", where "ă" has the code point 199. Then Windows came, and on Windows we often used an encoding called "code page 1250", where "ă" has the code point 227. Then came Unicode, and in Unicode "ă" has the code point 259.
Since Unicode code points can have values larger than 255, there must be a way to represent them using bytes with values between 0 and 255. Those methods are called "Unicode Transformation Formats" (UTF), of which the most widely used are UTF-8 (very popular in Linux) and UTF-16 (of two kinds, little and big endian, and very popular on Windows). In UTF-8, "ă" is represented as two bytes, with the values 196 and 131 (by the rules of UTF-8, those two bytes together represent code point 259); in little endian UTF-16, "ă" is represented by two bytes, with the values 3 and 1 (by the rules of little endian UTF-16, those two bytes together represent code point 259).
The point is that in order to make sense of a text file you need to know (1) what encoding is used, and (2) in the case of Unicode, what transformation format is used. Now, on Linux and on the Web we are very close to a consensus that all text is represented in UTF-8; nevertheless, old files still exist, and occasionally new files come from Windows, so there is a very nice program called iconv (available both on Linux and on Windows) which is used to translate text files from one encoding into another.
For example, assuming that your problematic file is encoded in Windows-1252 (also called ANSI by the Windows documentation, although the American National Standards Institute had nothing to do with it), you could say
iconv -f windows-1250 -t utf-8 ca001_mci_17071971.txt | gsed -e 's/ñ/–/g' '{}'
Sadly, there is no way to use sed -i; you must write a temporary output file, then rename the temporary output file on top of the source file, of course after checking that everything went well.