With some hacking, I can build blue-side images, which demonstrates that the spot data (SIM2D-93) works.
To do this properly, we need:

• Transmission curves
• Verify blue detector size
• Blue bias frames

After a bunch of mucking around, I managed to process both blue and red data through mergeArms.py. An example pfsMerged file is plotted above. There are a bunch of problems evident in the plot, but the main thing to notice is that the blue and red arms have been merged.

I'll push some changes to ticket branches once I've cleaned them up.

Currently the fiber assignments don't match between the blue and red images. I would therefore like to delay adding blue data to the integration test until we can get the simulator properly producing coordinated blue and red images (SIM2D-99 + SIM2D-94).

merge.png shows that merging works, but the continuum is not continuous over the merge region (~650nm). I believe this is due to an incorrect normalisation in the flat field and/or fiber trace.

How do we know what's right here?  Should we generate some fibres with bright objects with simple spectra (e.g. constant nu I_nu;  power laws; black bodies) to help assess our success here.

That's what I'm doing right now.

coadd.png: two simulated exposures each with red+blue arms, run through the pipeline, and this is the resultant coadd spectrum (black) compared with the input simulator spectrum (constant F_nu @ 22 mag; red). The sky line subtraction is awful because it’s a braindead placeholder algorithm, but the continuum level looks about right. Looks like we've fixed the discontinuities!

from lsst.daf.persistence import Butler
butler = Butler("INTEGRATION/rerun/integration/pipeline")
coadd  = butler.get("pfsCoadd", catId=7, tract=0, patch="0,0", objId=3, numExp=4, expHash=0x1eab2e92)
fig, axes = coadd.plot(plotSky=False, plotErrors=False, show=False)
axes.axhline(3631e9*10**(-0.4*22.0), color="r")
fig.show()

merged_fixed.png: demonstrates that the merged spectrum (pfsMerged: single exposure) is good too. The higher orange line is the flux standard (18 mag), while the lower lines are the science objects (22 mag).

from lsst.daf.persistence import Butler
butler = Butler("INTEGRATION/rerun/integration/pipeline")
merged  = butler.get("pfsMerged", expId=32, arm="r")
fig, axes = merged.plot()

I'm finally happy with this. hassan, would you please take a look? This ticket was the home for simulator work as well, so there are a bunch of changes in several packages.

Ok overall. Separate tickets can be raised later on points that could be missed.

Merged to master.