Thanks,
I've been looking through the code and I'm beginning to think that it
relates to my choice of State Covariance. In Montenbruck the state and
example are specified in terms of cartesian coordinates (standard deviations
of 10km and 10m/s in each axis) so I had to guess at a covariance.
As you say, I can fix that if I have the jacobian relating the two
coordinate systems. So, I am working on the cartesian to equinoctial
jacobian and I had one follow up with respect to your Latex
dv = \frac{\xi^2}{\varepsilon^3} dM + \frac{(1+\xi)\nu}{e\varepsilon^2} de
===
where
\xi = 1+e\cos v
\varepsilon = \sqrt{1-e^2}
Since \lambda_v = \Omega + \omega + v and \lambda_M = \Omega + \omega
+ M one can easily use the first equation to compute d\lambda_v from
d\lambda_M.
I noticed you used both v and \nu. Is that just a quick typo and did you
really mean the true anomaly for both?