Serial correlation: When regression errors are correlated across observations.
Auto correlation: When a time series is correlated with its own past values.
Seems the definitions are a bit different, but are these things exactly the same?
And what kind of serial correlation can we have, except positive serial correlation?