Verified that the JPL checker works for apparent coordinates,
equatorial-of-date and horizontal, when I tell the JPL Horizons
tool to use its "airless" (no refraction) model.
Still need to implement refraction for at least horizontal coordinates
(where it really matters for rise/set times).
Used the JPL Horizons web interface at
https://ssd.jpl.nasa.gov/horizons.cgi
to generate test data for astrometric RA/DEC,
apparent RA/DEC, and apparent az/alt.
Will use this as a secondary unit test to confirm
that I'm not fooling myself by incorrectly using the
NOVAS C 3.1 code for unit tests.
Not quite right yet for some reason, but this is closer.
JavaScript function Astronomy.SkyPos() now returns both
J2000 (RA,DEC) and (RA,DEC) using true equator and equinox of date.
Use the latter to calculate horizontal coordinates.
This matches my call to NOVAS place() function, but there
are still errors larger than 2 degrees compared with JPL Horizons
and Heavens Above.
For example:
2019-04-11 19:47:00
variable test.html JPL Horizons error(arcmin)
Sun azimuth 245.455 246.585 -67.80
Sun altitude 50.858 51.261 -24.17
Jupiter azimuth 277.309 275.287 121.33
Jupiter altitude -63.751 -63.860 6.52
I have horizontal coordinates calculated, but they might
be wrong (both in how I call NOVAS functions and the JS code itself)
because I think I'm mixing up equinox of date coordinates with
J2000 coordinates for (RA,DEC).
Fixed bug that caused excessive estimate of angular error:
right ascension and azimuth are like longitudes -- they matter
less as an object approaches the poles. Scale such longitudinal
errors by the cosine of the latitudinal counterpart.
