Creating an ndarray with a given dtype.

dtype_ex1.py

import numpy as np

dt = np.dtype([
    ('time', [('min', int), ('sec', int)]),
    ('temp', float)
])

# create array with dtype
x = np.zeros((5,5), dtype=dt)

# Initialize values
x['time']['min'] = 10
x['temp'] = 98.25

# show array
x

# show temps
x['temp']

This example comes from http://docs.scipy.org/doc/numpy/reference/generated/numpy.fromfile.html except I create a 5 by 5 array.

Examples of more complex dtypes can be found here:

http://docs.scipy.org/doc/numpy/reference/generated/numpy.dtype.html