A barely stable throttle

As if Apollo ll’s descent wasn’t exciting enough, later analysis demonstrated that Eagle’s software had harboured a problem that could have forced an abort. In fact, Apollo 12 flew with the same problem coded into its software. Alhough it was most prominent during the final few metres before touchdown, when the commander was flying in P66, it could have afflicted any phase of the descent.

The story of the problem, as told by Don Eyles, came to light when an engineer, Clini Tillman, ran a simulation of a descent. Tillman worked for the LM’s manufacturer, Grumman. This gave him the ability to run his simulation using real hardware rather than having to mathematically model the operation of the engine. As a result, faulty assumptions in the model were avoided.

Tillman noticed that the commanded thrust was varying in an apparently random, stepped fashion that came to be known as throttle castellation, after the similarity of its waveform to castle battlements. The variation was only slight but when Tillman dug into the stored telemetry readings from Apollo 11 and 12. he discovered that during P66 these variations were not only present, they were unduly large and hinted at some intrinsic instability in the system as a whole. One aspect of the problem was uncovered by Eyles’s colleague. Allan Klumpp. Know’n as IMU bob. it came about because the accelerometers at the centre of the IMU did not reside at the centre of mass of the LM. So when the LM underwent significant rotation, the 1MIJ sensed a component of that rotation as being vehicle velocity, which it was not. When this fed into the thrust calculations, it caused a small degree of instability.

A more profound cause of throttle castellation was related to the time taken for the descent engine to respond to commands to change thrust so called ‘throttle lag’. As described by Lyles, the paperwork for the engine stated that its lag time was 0.3 seconds and it was his task to compensate for this in software. To determine the best compensation value to use. Hyles carried out simulations to model the performance of the engine. lie saw how unstable the throttle command was with no compensation, then how compensating for 0.1 seconds helped a lot and how compensation for 0.2 seconds of lag essentially eliminated the instability. He therefore programmed the flight software with compensation for 0.2 seconds of throttle lag instead of 0.3 seconds.

Now it turned out that the compensation was trying to hit a sweet spot and that overcompensation could also induce throttle instability. It also transpired that the documentation for the engine was wrong. The engine was still evolving and by the time it was installed in the LM. it could react to throttle commands in a mere 0.075 seconds. It was later demonstrated that if Hyles had followed the paperwork and programmed for 0.3 seconds compensation, the throttle would have been wildly unstable. Any further source of instability; for example the I MU bob problem and the mission would have had to have been aborted. As it was. the first two flights to the Moon landed with throttles that were barely stable.