The cause is really not clear. The formual using 2*Pi to convert
from radians to frequency is perfectly correct.
However, measurements with various test equipment (including Racal 6103)
have shown our frequency error estimate is always off by a power of two...