Is there any way to confirm this, like comparing a particular character on unix box and then on the dataset that we got through FTP.
You can manually inspect the records of the file to see if they have recognizable text in them. If there is recognizable text, the conversion occurred.
File transfers can be either binary or text. A binary transfer sends the data just as it is to the other machine -- in which case ASCII on the Unix box would remain ASCII on the mainframe. A text transfer automatically converts ASCII to EBCDIC (or, possibly, another code page).
One key point to remember is that ASCII and EBCDIC have the same collating sequence (for the first 128 characters, at least, and for all 256 if dealing with extended ASCII). There is therefore no automated way to determine if a given file is EBCDIC or ASCII.
For example, take the hexadecimal string '97994D'. This may be an EBCDIC packed decimal (COMP-3 in COBOL) field of -97994. However, it also represent 'pr(' if treated as characters. Which is it? Only the application documentation can tell you for sure. If you see a batch of hex '20' characters the file is probably ASCII whereas a batch of hex '40' characters would probably indicate the file is EBCDIC. But that is not a guarantee.