Open mperrin opened 6 years ago
mperrin
commented
6 years ago 
@kvangorkom This PR may be of interest to you. (but definitely feel free to completely ignore this if you're getting busy with other things and don't feel like looking at this now!)
mperrin
commented
6 years ago
Oh, and this PR also adds functions plot_noll_zernikes and plot_wyant_zernikes to make pretty plots of the polynomials. Mostly for debugging purposes but I figured I would add them to poppy.zernike since they seem like useful demos to have present.
mperrin
commented
6 years ago 
I put together some code a few months ago to do something very similar! I managed to cobble together an algorithm to map Fringe j -> (n, m) (and the inverse), but there's definitely room for improvement. You're welcome to adapt what I have (or else I can when I get the chance): https://github.com/kvangorkom/zernike/blob/master/zernike.py
One thought from looking at your implementation: Fringe Zernikes, in my experience, tend not to be Noll normalized. There's probably room to argue this, but I wonder if it might be better to disable the normalization implicit in zernike_basis_wyant (but add it as an optional argument). [Edit: woops, I see you raised this point above.]
mperrin
commented
6 years ago
oh nice! I figured there had to be an algorithmic way to do that mapping rather than my hack of a lookup table. I will look at merging in your code for that. Thanks much!
Fringe Zernikes, in my experience, tend not to be Noll normalized. [...] Edit: woops, I see you raised this point above.
That's not a "woops", that's specifically some of the feedback I was hoping for from you. 👍
Also I see your code has piston starting as index 1 as opposed to index 0. I wasn't sure what to do about that - I saw different starting indices in different references. Sigh.
mperrin
commented
6 years ago
I've seen both index 1 and index 0, and I'm not sure there's a preferred convention. (I believe both Zemax and CodeV adopt 1, but 4D's 4Sight adopts 0). I suspect being explicit about it is the only way to handle it.
Another complication (which you're likely already aware of) is that Wyant adopts a completely different (n, m) convention (see here). These map onto your (r, l) terms, but my impression is that it's not uncommon to generate Fringe zernikes based on Wyant (n, m, sin/cos) inputs. Robert Gray's MATLAB library handles it this way. I don't know if it's worth attempting to tackle this complication.
Adds functions for getting Zernike polynomial indices and Zernike basis sets using the so-called fringe or Wyant ordering of the Zernike polynomials. I think I've got this right, but gaaah the literature on this is an inconsistent mess. Same functionality as the existing Noll-ordering-based code (which this is calling behind the scenes); this just has the polynomials in a different ordering.
For now the polynomials are using the same scaling coefficients as the Noll indices, but I'm not sure if that's the correct thing to do, since Wyant himself has these on his web page without any scaling coefficients (see http://wyant.optics.arizona.edu/zernikes/zernikes.htm)
Could use some review, and still needs unit tests before merging.
mperrin included the following code: https://github.com/mperrin/poppy/pull/236/commits