The requirement for Ms was always a nuisance since it could potentially crash the code if Ms was larger than allowed by Kepler's 3rd law. The only reason it was added was so that starry would be able to accept both a/Rs and Porb, but Everett Schlawin pointed out to me that we can just set Mp=0 and then solve for Ms which is way easier and removes that nuisance parameter.
Some folks allow negative omega values, and sometimes its useful to use a prior like Intial Guess +/- 180 so that your fits don't end up bi-modal if they're near 0 or 360.
Finally, sometimes curve_fit will fail to converge with the default maxfev value, so I just increased it by 10x which solved my problems. Might be worth changing this to scipy.optimize.minimize and using Powell instead since that's been working so well for us elsewhere.
The requirement for Ms was always a nuisance since it could potentially crash the code if Ms was larger than allowed by Kepler's 3rd law. The only reason it was added was so that starry would be able to accept both a/Rs and Porb, but Everett Schlawin pointed out to me that we can just set Mp=0 and then solve for Ms which is way easier and removes that nuisance parameter.
Some folks allow negative omega values, and sometimes its useful to use a prior like Intial Guess +/- 180 so that your fits don't end up bi-modal if they're near 0 or 360.
Finally, sometimes curve_fit will fail to converge with the default maxfev value, so I just increased it by 10x which solved my problems. Might be worth changing this to scipy.optimize.minimize and using Powell instead since that's been working so well for us elsewhere.