Worked on documentation, mainly making a diagram to show the integration geometry.

Kept on documenting, reached a sort of end point. Also ran test with just black body, one shell and no actual source function in the integral, this gave good enough agreement that its important to do all the details properly. That is put the integral into physical units and make sure the MC and integral methods are compared on the same frequency grid.

Made comparisons between integrator and MC for one shell model of pure silicon. Shape agreement reasonable but normalization way off, factor of 10^18. . After poking around a bit I got out a pure black body spectrum, off by magnitudes.

After fixing regresson from the 20:th, kept on trying to find the source of the missing magnitudes.

Found the missing magnitudes, it was a missing dp. Still some magnitude missing, but turned out to be due to too few binns in impact parameter space

Fixed bugs in the part of the integrator I had until it actually ran the example. It gave wildly unrealistic results.

Found the source of the wildly unrealistic results. Now it gives sketchy results, but it still does not have the bit that is directly driven by a black body.

Working on getting in code to integrate on the inner half of the impact parameter space.

Finished up the integrator, in now does the complete impact parameter range. The results are still bad.

Wrote documentation and made a diagram. Also fixed bug found by @yeganer, which significantly improved the end result.

Finished up the documentation. Though about how the integral could be derived from a general volume integral. Had a meeting about the integral and started on the exercise about the integral. The integral.

Concluded the integral exercise. Though about how it all fit together, understood and felt enlightened. Scribbled notes about how it all should work.

Though more about the integral and wrote a rought outline of how the integral would work

Reached limit of what I could plan in vacuuo, started on a skeleton on which to hack to get a complete implementation.

Got sick.

Not as sick. Wrote out an integrator that manages impact parameters greater than the inner radius of the shell. That is the case with no blackbody providing initial intensity.

Tried to fix problem with varaiables for the unit test not working. Read code to figure out how to do the level absorption sums.

Implemented fix to the unit test after @yeganer found the cause. Discussed with @wkerzendorf how to do the level absorption sum, and how atomic data is stored. The conclusion was that doing the sum in pandas can be done at the cost of a few hundred miliseconds.

Implemented code to calculate the attenuated source function using pandas.

Wrote a unit test for the attenuated source function. Considered the details of the final impact parameter integral. The simplest interpretation is that the way the intensities are derived "flattens" the problem so that the integration simply amounts to summing up the intensities according to whitin which shell a given impact parameter lies. This of course does not have any interpretation in the case of p less than R, so it's probably not correct.

Read the existing documentation and started working on documenting my additions to TARDIS.

Trying to run tardis example on a new computer using my changes. Eventually stopped by #593.

Created PR for my changes so far, and update it after comments. Tried to exclude some unintended changes but failed.

Finished up the line emission estimator and looked at what it produced.

Moved the line emission estimator into the cyton code. Wrote a unit test and tried to run it, encountered unrelated problem.

Fixed problem with unit tests not running at all. My addition did however not work, seems to be a problem with initializing the storage: the code never takes the branch that would actually increment the sum.

Rereading Lucy 345. We will need to add a second line list, but how to structure it?

Had a meeting with Uill, spent the rest of the day following code around in cmontecarlo.

More studying of the verious papers and the TARDIS code. In the end we arrived at a firm understanding of what needs to be added in principle regarding the line absorption estimation. There is only very limited changes to be done to the monte carlo loops.

Implemented the changes in a local branch, renamed things as apropriate and ran test of the C code. It passed.

Using jekyll to simplify managing the weblog continued to be a bad idea. Gave up and accepted writing html by hand. Looked for branching ratio treatment in cmontecarlo.

Mostly worked on presentation.

Looked a bit more at how cmontecarlo does its thing.

Started reading Mazzali, agreed on skype meeting with Ulli. Tried installing TARDIS on local machine and realized I had to repartition.

Reading Mazzali. Had a skype meeting with Ulli, Wolfie and Markus about the start of the project. Then attended TARDIS meeting. Read Lucy 345.

Worked on setting up a devlog for the project. Got a starting point from Karan, tried to use jekyll to simplify it a bit.