ps_kspan moves power to eor window

[lmberkhout]

First noticed in data and then reconfirmed in simulations. Using the ps_kspan keyword changes power in the EoR window by a few OOM.

ps_kspan not set in wrapper, uses default

identical simulation with one extra keyword, ps_span=200. (wavelength units):

[lmberkhout]

The first test I did was to try going straight to power spectrum through uvf cubes. The keywords were identical to the first plot in the thread (no kspan restriction), with an additional keyword to save uvf cubes, and a keyword to point eppsilon to the uvf cubes to use. This is the result.

[lmberkhout]

Next step: Do a bit of code development, interpolate to an orthoslant grid from the zenith observation, take an obs at a different time, and interpolate to that same zenith orthoslant grid

[lmberkhout]

linking to Nichole's issue on this topic #108

[nicholebarry]

@rlbyrne Could this be similar to the effect that you talked about in Figure 4 of your paper? Essentially a wedge-dependent horizontal throw: the more wedge you include, the higher up in kpar it goes.

@lmberkhout I assume you aren't using the modified gridding kernel in these sims? I wonder if the effect is mitigated by the modified gridding kernel (my gut feeling is yes)

[lmberkhout]

@nicholebarry I believe I am using the modified gridding kernel (at least I am very much intending to). This was replicating a simulation with keywords sent by @mwilensky768

[rlbyrne]

@nicholebarry The effect I was talking about in my paper only comes in in calibration. My understanding is that ps_kspan doesn't affect calibration and only enters when the Healpix cubes are made.

[nicholebarry]

@rlbyrne gotcha

@lmberkhout Okay cool, just checking. Would it be possible for you to plot the UV-plane at a frequency slice for both the default kspan and the 200 wavelengths kspan? I bet we'll see some spatial messiness at longer baselines. I believe Bryna has an in-house uv plotter, but I forget what it's called. But you can just plot it yourself with quick_image, the data are located in ps/data/uvf_cubes under something_something_something_model_uvf.idlsave

[bhazelton]

The one in eppsilon is called uvf_slice_plot.pro, it's easiest to invoke by calling ps_wrapper with the following keywords: /plot_slices, slice_type='raw' or /plot_slices, slice_type='divided' depending whether you want them before or after the weights are divided out. You can also use the uvf_plot_type keyword to control what is plotted -- the options are ['abs', 'phase', 'real', 'imaginary'], the default if you don't set it is 'abs'.

[lmberkhout]

Slice type = 'abs' (default)
Default Kspan UV plane

Kspan = 200. UV plane

Some discussion with Nichole highlights that the background contamination is evident in these plots