stitching

[crawfordsm]

Hi Steve …

On 29 Jun 2017, at 17:06, Steve Crawford [email protected] wrote:

1. Okay, it might be a little hard to remove. I'll take a look at it and see if I can. Otherwise, what I can do, is adjust it so the data is processed twice -- once with the blaze removed and once without it removed (so returning the data to how they were originally for you).

The latter should work for me.

2. The final stitched product includes the heliocentric correction and converting from vacuum to air. So the second correction is likely the source of the 2 A shift.

That’s good to know about … I hope it will be flagged loud and clear in the FITS headers and the online docs !!

3. Yes, there is a some binning as the resolution is very oversampled in the observed data. Wasn't quite sure from your code, but do you have a good value at what it should be sampled as compared to the resolution and original pixel size?

I did this in ucles_resample … and normally do it in equal steps in log lambda.

So I convert all wavelengths in the 2d frame to log (lambda).
I find the mean log lambda interval in each order (wend-wstart)/(nw-1)
I select the largest interval
(not quite sure why I do that .. it might have been meant to be the smallest, but the range should be quite small).
and use that over all orders.
Then convert back to natural wavelength.
So this just uses the pixel sizes (in wavelength) and assumes that lambda / delta lambda is roughly constant over
the entire spectrum.

I don’t necessarily assume the instrumental resolution ‘a priori’ and so don’t degrade. The astronomer can make that adjustment later …

This is the code snippet that gives me ‘w0’ and ‘dw’ for the resampling; all orders are resampled onto that common grid.

echelle_size=size(w)
nw = echelle_size(1)
no = echelle_size(2)

wb = w
wi = intarr(no)
if (keyword_set(logdw_set)) THEN wb = alog10(wb)

dwx = (wb(nw-1,)-wb(0,)) / (nw-1)
dw = max ( dwx )
w0 = wb(0,0)
PRINT,'uniform grid, dw =‘,dw

Simon