In 1925, AT&T Bell Laboratories started investigating the sources of noise in its transatlantic radio telephone service.
By 1930, a radio antenna for a wavelength of 14.6 meters was constructed in Holmdel, NJ, to measure the noise in all directions.
[1][7] CCIR 322 provided seasonal world maps showing the expected values of the atmospheric noise figure Fa at 1 MHz during four hour blocks of the day.
[9] Unlike PRNGs, which rely on mathematical formulas, the numbers generated through atmospheric noise are not predetermined.
However, the complexity of weather systems and the difficulty in measuring variables like the position and velocity of every molecule make it practically impossible to predict these numbers.
[9] For this reason, even if atmospheric noise is theoretically predictable, it is practically unpredictable, making it useful for generating random numbers.