Incorrect light travel time correction for circular orbits

[JeffValenti]

For an eccentric orbit, juliet calculates a light travel time (LTT) delay of +51 seconds for eclipses of WASP-17b and HAT-P-26b, as expected (2aRs/c). For a circular orbit, juliet calculates an LTT delay of minus 10 seconds for HAT-P-26b and minus 30 seconds for WASP-17b, which is unphysical and inconsistent with the eccentric case.

In utils.correct_light_travel_time, the code path for a circular orbit has an np.sin(params.inc), which assumes inclination is in radians:

    if params.ecc > 0:
        # Need to solve Kepler's equation, so use the KeplerOrbit class
        # for rapid computation. In the SPIDERMAN notation z is the radial
        # coordinate, while for Bell_EBM the radial coordinate is x
        orbit = KeplerOrbit(a=a, Porb=params.per, inc=params.inc,
                            t0=params.t0, e=params.ecc, argp=params.w)
        old_x, _, _ = orbit.xyz(times)
        transit_x, _, _ = orbit.xyz(params.t0)
    else:
        # No need to solve Kepler's equation for circular orbits, so save
        # some computation time
        transit_x = a*np.sin(params.inc)
        old_x = transit_x*np.cos(2*np.pi*(times-params.t0)/params.per)

But params.inc is in degrees. KeplerOrbit and batman assume inclination is in degrees and multiply by np.pi/180 before use in trigonometric functions. Mistakenly feeding degrees rather than radians to the cos function multiplies the LTT delay by a factor between -1 and 1. Changing the circular orbit code to np.sin(params.inc*np.pi/180) yields an LTT delay of +51 seconds, consistent with expectations and the eccentric orbit case.

Eureka! currently has the same bug: Issue #735

[JeffValenti]

Submitted pull request #124 .