SQL Server - Enforcing uniqueness in one column depending on another column

Apologies if I get the terminology wrong. How do I define a constraint in MSSQL Server 2005 to enforce uniqueness in one column depending on another column?

E.g. considering the last two columns:

1    A    1
2    A    2
3    A    2 <- Disallow because '2' has already appeared alongside 'A'
4    B    1
5    B    2

From stackoverflow

• Try this:

  CREATE TABLE tTable
  (field1 CHAR(1) NOT NULL,
  field2 INT NOT NULL,
  UNIQUE (field1, field2)
  )

  Jeremy French : I don't think that solves the problem. (If I am reading this right) in that row three will have unique field 1 and 2 but still fail

  Brent Ozar : This does solve the problem. When you try to insert the second ('A', 2) record, SQL stops you with a violation of a unique constraint.

• Create a unique constraint on the 2 columns ?

  This is the most logical thing to do, since it seems that this one column is not unique, but the combination of the 2 columns must be unique.

• Doesn't need to be a primary key, all it needs to be is a unique composite index.