We have used the method described in Flash 90 to determine the rotation axis of the rocket body 95- 32 B. What follows below is an extract from a much larger text, which may find its way in upcoming issues of Flash (or not). It may not be a bad idea to re-read the original article before reading on. A short synopsis of my method.
Input data are n timings of primary flashes of one object, orbital elements for that object, and the observer's location. My method basically takes each subsequent couple of timings and calculates the rotation period from those two times, for *all* possible directions of the rotation axis. I do this for each adjacent couple of timings, and thus come up with n-1 different rotation period values for each possible direction of the rotation axis. Obviously, the rocket only has one rotation period. So one would expect the n-1 different rotation period values to become equal for that direction that coincides with the actual rotation axis. In reality, because of measurement limitations (human reaction time, etc...) there isn't one specific direction for which the n-1 values coincide. There can however be a set of directions for which the dispersion about an average rotation period becomes minimal. In fact this usually happens for any set of observations of any object. However, only for a few special passes during which a 'synodic anomaly' happens will the minimum be significant. That is what I found for Walter's 91- 29 B observations last year. And now also for some observations of 95- 32 B obtained in April.
Note that this method can also be used to determine (with lower accuracy) the rotation axis in case no synodic anomaly is observed. In that case, quasi-simultaneous observations are needed. This application has also been used in what follows.