parity of forward & inverse MFT transforms

[mperrin]

Issue by mperrin
Thursday Jan 15, 2015 at 21:42 GMT
Originally opened as mperrin/webbpsf#35

---

Begin forwarded message:

From: Anand Sivaramakrishnan anand@stsci.edu
Subject: troubles with parity of forward & inverse mDFT transforms...
Date: January 15, 2015 at 2:42:48 PM EST

Hi Joseph & Marshall

I use matrixDFT from WebbPSF as a standalone (to try to unify my code with WebbPSF utilities), and I'm having some trouble with coordinate flips when transforming forwards, then back.

I start with a psf in this example, transform it forwards to visibilities, then I try to transform it backwards as well as forwards - but get a flipped PSF both times. Running "matrixDFT_test.py" in the attached tar file will do these things, and give you two convenient ds9 commands to cut'n'paste to see the results.

This is a bit worrying, as matrixDFT is at the core of many things I do.

I also include matrixDFT.py copied from WebbPSF, with a print statement to show that I'm actually using this file from the *_test.py driver.

Any comments or suggestions would be great! If I should be using a different 'entry point' into the dft (such as SFT) please do let me know.

(I don't see any straightforward way to attach the file attachment to a github issue, but we have it in our inboxes)

[mperrin]

Comment by mperrin
Wednesday Jan 21, 2015 at 01:57 GMT

---

Reproduced by @josePhoenix from Anand's example code. Definitely looks like I missed this one somehow before.

[mperrin]

Comment by josePhoenix
Friday Jan 23, 2015 at 19:15 GMT

---

Really a POPPY issue, see mperrin/poppy#38 for remaining subtasks (basically writing test cases).