I was amazed the other day when I found out that my brother, Dean, actually at least half listens to me even when his eyes glaze over if I get too scientific.  He had been watching a program on an attempt to find the remains of Amelia Earhart and/or her plane.  It seems as though everyone had searched in the wrong place because the International Dateline had thrown them out. 

This triggered a memory from more than 20 years before in Dean’s synapses when I had explained how one of our SA developed Satcom antennas had failed on the other side of the world – San Diego to be exact.  I was working for Omnipless at the time, a highly successful SA company that had developed a range of products for use with the Inmarsat satellites which provided global telephony anywhere in the world via satellite.  The products covered the full range of portable ground-based systems, through vehicle mounted systems to both small and large maritime systems and small and large aircraft systems. His memory was vague (I forgive him) and he thought our failure also had to do with the International Dateline.  He was a bit confused as it had nothing to do with the dateline but it indeed had something to do with being on the other side of the world.

   

The subject of this blog is the mysterious failure of one of Omnipless’s latest maritime products at the time, the F77 large shipborne antenna.  The company was still smarting from the abject failure of their Maritime-B development where we had completely underestimated the vibration levels experienced by antennas even in huge ships.  The follow-up F77 was bullet proof, mechanically at least.  I had made sure of that.  I had located every single vibration test spectrum available from institutions like the US Department of Defence through to private company specs from the likes of Germanischer-Lloyds and had tested it against them.  If it passed, I would incrementally raise the levels while keeping the frequency content of the spectrum the same until some failure occurred.  That aspect would then be strengthened.  Rinse and repeat until I obviously reached ridiculous levels.

We were therefore non-plussed when we were informed that an antenna refused to work correctly in San Diego.  At least the problem was not mechanical and the mechanical design team could breathe a communal sigh of relief.  We were not the whipping boys this time.  Numerous telephone calls and emails later, no reason could be found.  The installation and commissioning were all perfect.  Johan Gericke, an excellent electronics engineer and good allrounder, was sent to investigate.  The ship was located near the naval docks.  This was Simonstown x100 with every manner of naval vessel spewing out gobs of radiation in the normal wasteful American way.  Johan first had to eliminate electromagnetic interference as a cause which must have been a mission itself.

Eventually the culprit was located.  It was in the software, the stuff that makes the modern world go round.  To be precise, it was a rookie mistake in one line of code, in fact, one function.  But why had we never experienced the fault before given that we had already supplied many of these antennas?

The reason was prosaic.  Up to that date, our main clients were European and their clients plied their trade between Europe and Asia.  This should give the game away to the scientifically astute.  Our antennas had only ever operated between the longitudes of 90°E (+90°) and 90°W (-90°).  In other words, trigonometrically, the F77 had only ever operated in the 1^st and 4^th quadrants, but the longitude of San Diego at 117°W (-117° or +243°) is in the 3^rd quadrant.

In calculating various angles, the programmer used the inverse tangent function, ATAN(X) which does not distinguish between the 1^st and 3^rd quadrants (or, in fact, the 2^nd and 4^th quadrants).  However, he should have used the function ATAN2(Y,X) which does.  A similar situation exists for the inverse sines and cosines, except in different quadrants.

Eg:  ATAN(1) = 45° which is the same as ATAN2(1,1) = 45°

While ATAN2(-1,-1) = 225°

Given the known position of the satellite and the GPS position of the antenna, the software has to perform a series of trigonometric calculations to instruct the antenna where to point which it obviously got hopelessly wrong on the other side of the world.

This was a cautionary tale.  The embarrassed software designer was a friend who I knew as an extremely competent and solid engineer yet he fell into the trap.  I too had fallen into the trap early in my career when I was involved in designing seeker heads for missiles and had to simulate their performance.  Luckily, the mistake was at computer simulation level and was soon uncovered by myself without having to make any embarrassing admissions. 

Apart from the Boks stealing the knockout matches from under the noses of their opposition, remarkable was the winning scoreline in three matches in a row by a margin of ONE.  SMAC pays tribute to this number and elevates it to godlike status.

Continue reading