Separate tables for just one field can be logical?

If I have one table which has a City fie开发者_如何学编程ld and as expected this field will repeat itself so only separate it to another table can be ok for normalizing concepts? I mean separated table who has only one field can be meaningful and can help performance gaining?

It is unlikely to help much. You would need to invent a 'City ID' which would, presumably, be an integer type for cross-referencing to the table of city names. You would complicate the data; when presenting the address information, you'd have to do a join - so simple data dumps would be less readily meaningful (you need two dumps - the main table and the city table - to make sense of a simple data dump).