In all three cases, the altitude of the Sun above the horizon at the moment of lift-off has been within a range of less than one degree of angle.

The two launch attempts for this mission in 2012 April and December provided the pointer. Both were launched at times that were preset to the nearest second.

Usually with space launches, a time to the nearest minute is sufficient and there are often 'windows' extending to a number of hours when all launch constraints can be satisfied. It is only when it is necessary to meet a very small target does the button get pressed on a precise second. Examples are launching to the ISS where the location of the orbit plane in space is important, or launching to join an existing constellation of satellites where precise relative positioning is important.

For the failed launch attempt 2012 April, the sun was 16°.8 above the horizon at Sohae as the rocket lifted off. For the next, this time successful, launch 2012 December, the altitude was 16°.4. The values were so similar that it effectively ruled out coincidence. As with all theories, the way to test for a genuine launch constraint was make a prediction for the next launch which turned out to be Kwangmyongsong 4.

North Korea announced that the aim was to launch it 2016 February 7 between 07:00 and noon Pyongyang Time (between Feb 6, 22:30 and Feb 7, 03:30 UTC). Using the rule derived from the two Kwangmyongsong 3 launches, it was possible to say that lift-off would most likely take place between 00:25 and 00:29 UTC (08:55 and 08:59 Pyongyang Time).

In the event launch occurred at 00:30 UTC (09:00 Pyongyang Time), fractionally later than the predicted window and precisely on the minute. The very small difference between the prediction and actuality was probably a decision to round-off the calculated time in order to launch exactly on the hour local time. The altitude of the Sun at launch was 17°.1, close enough to the numbers derived from the earlier launches to show that coincidence was not in charge.

There is no guarantee that future launches will obey the same rule of thumb but, if they do then the potential launch time for any particular day of the year can be determined by formula to within about five minutes. The diagram illustrates how the launch time varies through the year.

Pyongyang is 8.5 hours ahead of UTC and does not implement daylight saving time in the summer. However, when making comparisons, it needs to be remembered that 2015 August 15, North Korea changed from being nine hours ahead of UTC.

When interpreting the times then, broadly, from October to February, launches have the same date in both UTC and Pyongyang Time. For Spring and Summer UTC the time is late in the day but the actual date is one day earlier than in Korea.

There are two major possibilities. Firstly, it may be a constraint required for launching the rocket. It is unlikely to be a requirement of the rocket's navigation and guidance system but it may be to do with visibility from ground-based tracking locations in North Korea as it recedes downrange. Another possibility is to do with the way the Sun illuminates the satellite's solar panels. Launching at this time of the day means that a satellite that goes into sun-synchronous orbit in sunlight for seventy percent of the time.

There are probably several competing requirements that lead to the constraint. Time, and future launches, will tell if any of them can be relaxed. There has already been a minor relaxation by moving away from a precisely calculated launch time to the nearest significant minute as happened with Kwangmyongsong 4.