Fixes in optics code #925
Fixes in optics code #925
Conversation
sotodlib/coords/optics.py
Outdated
| @@ -330,7 +330,7 @@ def LAT_optics(zemax_path): | |||
| x = LAT["x"][gi].ravel() | |||
| y = LAT["y"][gi].ravel() | |||
| xi = LAT["elev"][gi].ravel() | |||
| eta = LAT["eta"][gi].ravel() | |||
| eta = LAT["xel"][gi].ravel() | |||
There was a problem hiding this comment.
Are we making an approximation here when converting "elev, xel" into "xi, eta"? If so ... does that matter? Elevation and cross-elevation are similar to xi, eta, if you're near elevation 0... but when you get a few degrees away from that origin, the difference starts to matter.
There was a problem hiding this comment.
The zemax model defines the origin to be at the boresite (since it doesn't know anything about sky coordinates and is modeled at a fixed configuration). So I think in this case el/xel is the same as xi/eta. So the only elevation dependent effect here is the LOS rotation.
I think even if it was actual el/xel then it would still be OK since we could just compute the template at el=0 and rotate to get the correct orientation (which is basically what happens right now anyways).
There was a problem hiding this comment.
The point is that "el, xel" is not the same as "xi, eta", even stipulating that el=0. The coordinate conversion to get (xi, eta) from (el, xel) is easy to do with so3g.proj.quat functions -- I guess I am proposing that we do that properly instead of ignoring the difference.
There was a problem hiding this comment.
Oh sure, I see what you are saying. I suppose el/xel is actually a lonlat pair? As usually my quaternion brain is under developed so a check on the transforms is appreciated. I want to say that I should have a sign change when doing xel but I'm only 51% sure...
There was a problem hiding this comment.
I suppose el/xel is actually a lonlat pair?
That is the concern. Is it? If so, you'll want to do something like (depending on the sign convention for xel):
xi, eta, _ = quat.decompose_xieta(quat.euler(1, -np.pi/2) * quat.rotation_lonlat(-xel, el))
That takes, for example,
>>> el, xel = .01, .02
>>> xi, eta, _ = quat.decompose_xieta(quat.euler(1, -np.pi/2) * quat.rotation_lonlat(-xel, el))
>>> xi, eta
(0.019997666768331163, 0.009999833334166508)
Though it's also confusing to me that you associate (xi, eta) with (el, xel) ... rather than (xel, el). That will need to be sorted. Draw me a picture and we'll work it out...
|
Ok I am reasonably confident in what is going on with these LAT updates (+ a unit test now exists). @mhasself can you take another look? |
No description provided.