Function get_hpc_info should return interpolated values when time range fully between two measurements

[FredSchuller]

As first noticed by @jajmitchell the function get_hpc_info sometimes returns wrong results that show strong deviation compared to results obtained one minute earlier or one minute later.
The reason is: when that function doesn't find any valid pointing measurement within the input time range, it falls back to the spacecraft pointing. But it should perform an interpolation between the previous and the next measurements (if available and marked as good, i.e. sas_ok=1) for better results.

[FredSchuller]

Here is a small script to illustrate that issue:

from astropy.time import Time
from stixpy.coordinates.transforms import get_hpc_info

t1 = Time("2022-03-28T18:43:00")
t2 = Time("2022-03-28T18:44:00")
t3 = Time("2022-03-28T18:45:00")
t4 = Time("2022-03-28T18:46:00")

roll, solo_xyz, ptg_1 = get_hpc_info(t1, t2)
roll, solo_xyz, ptg_2 = get_hpc_info(t2, t3)
roll, solo_xyz, ptg_3 = get_hpc_info(t3, t4)
print(ptg_1, ptg_2, ptg_3)

All three results should be very similar...

[samaloney]

So the code does do the interpolation but ignores it because there is no SAS solution in the time range and the code mistakenly interprets this as a bad SAS solution. I can fix this relatively easily but it does raise another question what you end up interpolating between two bad SAS solutions?

In the second case in the example L123-124 indices is empty so good_sas is empty which cases the spacecraft solution to be used via L137

stixpy/stixpy/coordinates/transforms.py

Lines 123 to 147 in 95c342b

	sas_x = np.interp(x, xp, aux["y_srf"])
	sas_y = np.interp(x, xp, aux["z_srf"])
	good_solution = np.where(aux[indices]["sas_ok"] == 1)
	good_sas = aux[good_solution]
	# Convert the spacecraft pointing to STIX frame
	rotated_yaw = -yaw * np.cos(roll) + pitch * np.sin(roll)
	rotated_pitch = yaw * np.sin(roll) + pitch * np.cos(roll)
	spacecraft_pointing = np.vstack([STIX_X_SHIFT + rotated_yaw, STIX_Y_SHIFT + rotated_pitch]).T
	stix_pointing = spacecraft_pointing
	sas_pointing = np.vstack([sas_x + STIX_X_OFFSET, -1 * sas_y + STIX_Y_OFFSET]).T
	pointing_diff = np.sqrt(np.sum((spacecraft_pointing - sas_pointing) ** 2, axis=0))
	if np.all(np.isfinite(sas_pointing)) and len(good_sas) > 0:
	if np.max(pointing_diff) < 200 * u.arcsec:
	logger.info(f"Using SAS pointing: {sas_pointing}")
	stix_pointing = np.where(pointing_diff < 200 * u.arcsec, sas_pointing, spacecraft_pointing)
	else:
	warnings.warn(
	f"Using spacecraft pointing: {spacecraft_pointing}" f" large difference between SAS and spacecraft."
	)
	else:
	warnings.warn(f"SAS solution not available using spacecraft pointing: {stix_pointing}.")

[FredSchuller]

The interpolated value should be used only if the aspect solutions before and after are both "ok".
In the IDL version (see stx_create_auxiliary_data), we first compute the time of the nearest neighbours t_before and t_after and then evaluate sas_ok as the logical operation: sas_ok[before] AND sas_ok[after].

[samaloney]

I might just filter out all the bad SAS solutions earlier, is the a case where they might be needed?

[FredSchuller]

No, I think it's a safe approach.