• flatten

we definitely need ecidentify like module. No more wavelength val by manual.

• load Thar_ird2.dat
• allocate the line positions in pixcoord
• make a prediction by some interpolation or GP? #25
• (postpone) I would include uncertain estimate for wavelength position, probably GP is better than interpolation.

In my environment (astroquery v0.4.7.dev7761) Simbad.query_object('HR 8799') in get_radec returns a single object HD 218396 at 23 07 28.7157 +21 08 03.310 and test_getinfo.py fails. The coordinates written in test_get_radec() corresponds to 23 07 28.7156 +21 08 03.302, which are coordinates for one of the HR 8799 planets (presumably picked up by result_table['RA'][0] etc).

Now I am checking unit tests and found that test_aptrace.py does not work anymore.

shirochan:~/pyird/tests(stream2d_continue)>pytest
================================================= test session starts ==================================================
platform linux -- Python 3.8.8, pytest-6.2.3, py-1.10.0, pluggy-0.13.1
rootdir: /home/kawahara/pyird
plugins: anyio-2.2.0, asdf-2.8.1
collected 8 items                                                                                                      

image/test_aptrace.py F                                                                                          [ 12%]
image/test_bias.py .                                                                                             [ 25%]
io/test_iraf_trace.py ..                                                                                         [ 50%]
spec/test_wavcal.py .                                                                                            [ 62%]
utils/test_getinfo.py .                                                                                          [ 75%]
utils/test_stream2d.py ..                                                                                        [100%]

======================================================= FAILURES =======================================================
_____________________________________________________ test_aptrace _____________________________________________________

    def test_aptrace():
        import numpy as np
    
        npix = 2048
        step = 100
        nap = npix//step
    
        dat_test = np.zeros((npix,npix))
        for x in range(0,npix,step):
            dat_test[:,x-2:x+2] = 1000
    
        cutrow = 1000
>       y0, xmin, xmax, coeff = aptrace(dat_test,cutrow,nap)
E       TypeError: cannot unpack non-iterable NoneType object

image/test_aptrace.py:16: TypeError
------------------------------------------------- Captured stdout call -------------------------------------------------
Please set nap = 42 (for H band) or nap = 102 (for YJ band).
=============================================== short test summary info ================================================
FAILED image/test_aptrace.py::test_aptrace - TypeError: cannot unpack non-iterable NoneType object
============================================= 1 failed, 7 passed in 1.64s ==============================================

Substitution of apall in IRAF.
We plan to remove all the IRAF routine from pyird.

I reduced IRDA00041511.fits using IRD_stream.py in examples/python, and computed cross correlation with model telluric spectra to derive radial velocities (RVs) for telluric lines. The result is shown below: I expected ~0km/s but found ~-2km/s, which is close to the velocity shift corresponding to 1pixel.

I analyzed another higher S/N spectrum (IRDA00042313) and obtained a similar result: about -2km/s RV for telluric lines. In fact the same spectrum was reduced manually by Kawahara-san before, and this one does give ~0km/s RV for telluric lines. The difference is shown below for some deep telluric lines:

I haven't yet figured out what might be going wrong (or I may just be making some mistake) but thought I'd share it in case anyone has a better idea of what's going on.

goal: self-contained well-documented pipeline w/o IRAF

• bias
• readout pattern (for dark)
• readout pattern (for faint object)
• aperture extraction
• wavelength calibration
• connection to exojax analysis

Current implementation for dark frame

See examples/tutorials/image/Common Readout ... ipynb

Currently, we just have a function that processes from A to B.
Import these functionalities into instances in irdstream.Stream2D .

irdstream automatically does everything and the user cannot see what is happening in the pipeline. So, I think we need to make some figures during running irdstream. Here are some idea

• masking image used for image cleaning
• corrected 2D image
• aperture used for flatten
• wavelength cal info

I observed a non-common trend of the readout pattern in some order:
See #5 (comment).
This results in artifact in one-d raw spectrum for a faint object (can find it by performing REACH.py)

To remove this, some preparations are needed

• RSD (raw spectral detector) matrix (rsdmat.py in spec) to see the effect (by taking median axis=1)
• allocate channel position in a 1D raw spec

After that, consider how to remove it. probably GP?

I need it for #10.

Using sep, I constructed the hotpixel identification, though not well-tested.

I found that there were unintentional flux jumps in the order combined spectrum (ncw?????_m?.dat; normalized and combined spectrum, green lines on the figure). It seems to be caused by including the non-signal region at the edge of each order when adding the fluxes in overlap wavelengths. (I will fix it...)