A more robust way to get this is to use a DateOffset (which you can pass as freq to date_range
).
Since microseconds is the smallest unit for a timedelta:
In [11]: td = datetime.timedelta(minutes=5, microseconds=100)
In [12]: pd.offsets.Micro(td.total_seconds() * 10 ** 6)
Out[12]: <300000100 * Micros>
Note: I think it's a good enhancement request for this function to accept a timedelta object...