[SalesForce] How to get Base64 String from Binary String or UTF-8 String

String.fromCharArray is going to create a UTF-8 encoded string, which will alter the bytes in the stream. For example, 242 will be encoded as two bytes instead of one. Any value outside 0-127 will be encoded as multiple bytes when using this function, including all negative integers. Since we can't really work with binary data directly in Apex Code, just store the base-64 encoded value directly in a string or some persistent storage (e.g. a custom setting).

public class Base64 {
    static String[] codes = 'ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+/='.split('');

    // Encodes integers 0-255 into base 64 string
    public static String encode(Integer[] source) {
        // Preallocate memory for speed
        String[] result = new String[(source.size()+2)/3];
        // Every three bytes input becomes four bytes output
        for(Integer index = 0, size = source.size()/3; index < size; index++) {
            // Combine three bytes in to one single integer
            Integer temp = (source[index]<<16|source[index+1]<<8)|source[index+2];
            // Extract four values from 0-63, and use the code from the base 64 index
            result[index]=codes[temp>>18]+(codes[(temp>>12)&63])+(codes[(temp>>6)&63])+codes[temp&63];
        }
        if(Math.mod(source.size(),3)==1) {
            // One byte left over, need two bytes padding
            Integer temp = (source[source.size()-1]<<16);
            result[result.size()-1] = codes[temp>>18]+(codes[(temp>>12)&63])+codes[64]+codes[64];
        } else if(Math.mod(source.size(),3)==2) {
            // Two bytes left over, need one byte padding
            Integer temp = (source[source.size()-2]<<16)|(source[source.size()-1]<<8);
            result[result.size()-1] = codes[temp>>18]+(codes[(temp>>12)&63])+(codes[(temp>>6)&63])+codes[64];
        }
        // Join into a single string
        return String.join(result, '');
    }
}

For the bytes you shared, the text came out as:

STAMP_DATE;CUSNO;FIRSTNAME;LASTNAME;COMPANYNAME;ROWTEXT1;ROWTEXT2;ROWTEXT3;CUSTOMER_TYPE;CUSTOMER_TITLE;CUS

Which means that the problem isn't at the beginning. What's going on is that you've got a stray byte somewhere that isn't UTF-8 compatible, and the upper range of ASCII characters (0x80-0xFF) have a different meaning in Unicode (mostly invalid combinations, like 0x80 0x00 would be invalid).

Without seeing the entire file, we really have no idea what's broken, but if I had to guess, it's most likely a character that has some sort of decoration (e.g. Ñ), which in ASCII are one byte, but in Unicode are two (or more) bytes. When invalid characters appear in a Unicode string, Apex code fails with that exception, as it cannot understand ASCII/ISO-8859-1 encoding except in some special circumstances.

If your file is in ISO-8859-1 format, you'll first need to convert it to be Unicode compliant. Most likely, that means opening it up in a text editor (not Notepad, but a good one, like Programmer's Notepad or Notepad++), change the encoding to Unicode, upload the new file to Salesforce, and try again. Or, you can try to decode it using something like in this answer. Your results may vary.