sci_to_tel and tel_to_sci transforms buggy for HST

[Witchblade101]

Slack message from Tri:

Hi there folks, I've been working with pySIAF to transform between coordinates, and run into something... when using an HST aperture (in this case, ACS/WFC) and transforming between sci and tel frame, the transformation from sci to tel works fine. But when transforming the tel coordinates back into sci, they're not the same as the original input sci coordinates.
With Roman apertures, it seems to work fine. This is pySIAF version 0.24.1 and using the latest HST PRD version. A notebook with minimum working example is attached. Is there a file somewhere that I need to update?

[Witchblade101]

Comment from Paul Goudfroij

Looks like there is an issue in aperture.py:

`import pysiaf
import numpy as np
hst_siaf = pysiaf.Siaf('HST')
hst_aperture = hst_siaf['JWFCENTER']
hst_siaf.plot()
Traceback (most recent call last):
File "", line 1, in
File "/Users/goudfroo/opt/miniconda3/envs/jwstpipe/lib/python3.11/site-packages/pysiaf/siaf.py", line 460, in plot
ap.plot(frame=frame, label=label, ax=ax, units=units, mark_ref=mark_ref,
File "/Users/goudfroo/opt/miniconda3/envs/jwstpipe/lib/python3.11/site-packages/pysiaf/aperture.py", line 666, in plot
x, y = self.corners(frame, rederive=False)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/goudfroo/opt/miniconda3/envs/jwstpipe/lib/python3.11/site-packages/pysiaf/aperture.py", line 1792, in corners
return self.convert(corners.x, corners.y, corners.frame, to_frame)
^^^^^^^
UnboundLocalError: cannot access local variable 'corners' where it is not associated with a value'

[mfixstsci]

import pysiaf
from pysiaf.aperture import SiafCoordinates
import numpy as np


hst_siaf = pysiaf.Siaf('HST')
hst_aperture = hst_siaf['JWFCENTER']

print(hst_aperture.a_shape)

x_Idl = np.array([hst_aperture.v1x, hst_aperture.v2x, hst_aperture.v3x, hst_aperture.v4x])
y_Idl = np.array([hst_aperture.v1y, hst_aperture.v2y, hst_aperture.v3y, hst_aperture.v4y])
corners = SiafCoordinates(x_Idl, y_Idl, 'idl')

print(corners.x, corners.y, corners.frame)

hst_aperture.convert(corners.x, corners.y, corners.frame, "sci")

returned results:

QUAD

[-101.9206  102.3036   99.4038  -99.6281] [-109.2273  -95.5362  111.6168   93.0262] idl

(array([  -7.68741804, 4133.18782711, 4112.02889197,  -13.0652571 ]),
 array([1.76084477e+00, 2.33841245e+01, 4.13203003e+03, 4.12716344e+03]))

Corners will be assigned in SiafCoordinates since it satisfies the if/else self.a_frame == 'QUAD'. Let me know what you turn up @Witchblade101

[tonysohn]

Good news and bad news.

Good news is that I found and fixed a big bug. In pysiaf/aperture.py line 1927, the following has to be changed

ORIGINAL:
return super(HstAperture, self).idl_to_tel(v2_arcsec, v3_arcsec,

FIXED:
return super(HstAperture, self).tel_to_idl(v2_arcsec, v3_arcsec,

Bad news is that even after doing this, there is still a small level (compared to before) of discrepancy. See below for example:

import pysiaf
siaf = pysiaf.Siaf('HST')
aper = siaf['JWFCENTER']
v2, v3 = aper.sci_to_tel(200, 200)
aper.tel_to_sci(v2, v3)
(168.71650162561696, 302.58418001540645) # Of course, this is supposed to be (200, 200)

I'll dig deeper to find out why this is happening.

[tastraatmadja]

Thanks all for looking into this. Maybe also related, there’s another error when converting to det frame:

---

TypeError Traceback (most recent call last)
Cell In[19], line 2
1 ## Calculate the HST detector coordinates
----> 2 x_det_hst, y_det_hst = hst_aperture.sci_to_det(xx, yy)

File /opt/miniconda3/envs/hst/lib/python3.13/site-packages/pysiaf/aperture.py:1012, in Aperture.sci_to_det(self, x_sci, y_sci, *args)
1010 def sci_to_det(self, x_sci, y_sci, *args):
1011 """Science to detector frame transformation, following Section 4.1 of JWST-STScI-001550."""
-> 1012 x_model, y_model = self.detector_transform('sci', 'det', *args)
1013 return x_model(x_sci - self.XSciRef, y_sci - self.YSciRef), y_model(x_sci - self.XSciRef,
1014 y_sci - self.YSciRef)

File /opt/miniconda3/envs/hst/lib/python3.13/site-packages/pysiaf/aperture.py:858, in Aperture.detector_transform(self, from_system, to_system, angle_deg, parity)
855 if angle_deg is None:
856 angle_deg = getattr(self, 'DetSciYAngle')
--> 858 x_model_1, y_model_1 = linear_transform_model(from_system, to_system, parity, angle_deg)
860 if from_system == 'det':
861 # add offset model, see JWST-001550 Sect. 4.1
862 x_offset = models.Shift(self.XSciRef)

File /opt/miniconda3/envs/hst/lib/python3.13/site-packages/pysiaf/aperture.py:2077, in linear_transform_model(from_system, to_system, parity, angle_deg)
2050 """Create an astropy.modeling.Model object for linear transforms: det <-> sci and idl <-> tel.
2051
2052 See sics_to_v2v3 (HST):
(...) 2074
2075 """
2076 if type(angle_deg) not in [int, float, np.float64, np.int64]:
-> 2077 raise TypeError('Angle has to be a float. It is of type {} and has the value {}'.format(
2078 type(angle_deg), angle_deg))
2080 # check for allowed system pairs
2081 if (from_system == 'det' and to_system != 'sci') or
2082 (from_system == 'sci' and to_system != 'det') or
2083 (from_system == 'idl' and to_system != 'tel') or
2084 (from_system == 'tel' and to_system != 'idl'):

TypeError: Angle has to be a float. It is of type <class 'NoneType'> and has the value None

[mfixstsci]

@tonysohn nice catch! Let us know if you need anyone of to review your change. If you could tag this issue in the PR that would be great.

@tastraatmadja I did some digging around your issue too. Seems that this block is causing the failure: https://github.com/spacetelescope/pysiaf/blob/main/pysiaf/aperture.py#L855

When angle_deg is not defined, we try to define it by obtaining the attribute DetSciYAngle. Turns out this isn't a very clever solution in your case either because self.DetSciYAngle is also None which causes the failure in the linear_transform_model class method.

[tonysohn]

@mfixstsci Sure, I'll submit a PR once I figure out the discrepancy found between sci→tel and tel→sci .

[tonysohn]

Hi all, I filed a PR to fix the major bug in aperture.py file as mentioned above. The PR can be viewed here: #398

One important note, especially to @tastraatmadja : Even after this fix, there will still be discrepancies in the conversions between idl/sci→tel and tel→idl/sci. This is NOT a bug in the code itself but due to roundtrip errors inherent in the SIAF file (pysiaf/prd_data/HST/siaf.dat-Latest). The roundtrip errors are caused by the way inverse distortions are derived, i.e., generally if data points with bad sampling are used for the conversion, then higher roundtrip errors are introduced like in this case. Proper way to fix this is to use the forward distortion solution to create a grid of evenly+fine sampled output points and to derive a inverse transformation using the paired input/output points. However, this will require some time/effort, and also involve (complicated) HST SIAF changes. We may want to consider updating the SAIF in the future, but there are currently no firm plans.

[tastraatmadja]

Many thanks all for working on this! I’ll make note of pySIAF’s limitation in accuracy in my work.