GainMatcher

[AdamGarnsworthy]

-The spectra take a long time to download. Sometimes it is fast, others it is either really slow or maybe it is hung up.

-Auto calibrate does not succeed. All numbers come out as NaN. I think the issue is that the auto-attempted fit has a preconceived expectation for the rough gain. This does not work well as we have 4 different gain ranges that we can use. If the gain is quite different then the fit simply fails. Can we make it more tolerant? As we define that the spectrum is 60Co then we expect there to be two strong peaks in the spectrum, maybe we can look for the highest-intensity channel (force it not to be noise in chan0 or overflow at the top) and say it is the 1173, then calculate a rough gain assuming 0keV is chan0. Then try this fitting routine already implemented.

[katieannemills]

auto-attempted fit has a preconceived expectation for the rough gain

This is no longer true as of the #31. Now, the user is asked to draw a line intersecting all the peaks they want to calibrate on, and identify what position along the line those peaks fall, and their energies (ie "the second and fifth peaks intersected, which have energies of x and y respectively"). This allows us to correctly identify the same peaks with a wide tolerance of uncalibrated gains.

we define that the spectrum is 60Co then we expect there to be two strong peaks in the spectrum

60Co is really easy, but simple solutions fall apart for more complicated spectra - hence the construction above.

I'm in San Francisco for the rest of the week, but will see if I can dig into this on the weekend. Is there any chance we could have a calibration source in the array over the weekend? 152Eu would be good.

[katieannemills]

See https://www.youtube.com/watch?v=UFEZ-f5vfZY for a demo of how this is all supposed to work - let me know if this doesn't work for you, or if we can get a more complicated calibration spectrum going sometime this week.

[AdamGarnsworthy]

The video demo is really helpful