I don't know what the issue is here but you could use ngrams8k to perform an character by character analysis like this: 

-- sample data
DECLARE @table table(someId int identity, someString varchar(100));
INSERT @table VALUES ('xxx123'),(CONCAT('abc',CHAR(0xC2),CHAR(0x92),'!'))

-- character-by-character analysis
SELECT
t.someId,
t.someString,
ng.position,
ng.token,
ascii_value = ASCII(ng.token),
unicode_value = UNICODE(ng.token)
FROM @table t
CROSS APPLY dbo.ngrams8k(t.someString,1) ng;

Returns: