MAKEE uses calibrated template spectra to do the wavelength calibration-- one with CuAr and one with HgNe+Xe. The program looks at the Cu "LAMPCU1" header card and the Xe "LAMPNE1" header cards to figure out which is which.

MAKEE is limited to two arc lamps, if you don't have a combined HgNe+Xe arc lamp, you can add the raw exposures for Xe and HgNe together using opim . However, make sure the header shows the Xe turned on, since that is what the program looks for. Also, a Xenon exposure alone should work fine.

MAKEE does a correlation against the templates and then identifies and centroids individual arclines. It alters the 0th and 1st coefficients of the template calibration polynomials to match the new line centroids. The higher order coefficients are taken from the template calibration*. This usually gives about a +/- 0.2 pixel error, and should be fairly robust. Look at the Arc-###-fr.ps file produced by MAKEE to see the residuals.

Note from September 2017: the above paragraph is a little confusing, what the program does is fit a first order polynomial to the arcline shifts between the template arclamp spectrum and the new arclamp spectrum as a function of column number. It then applies the shift values from this fit to the pixel vs. wavelength of each arcline in the template spectrum and refits the 7th order polynomial wavelength solution for each echelle order. Please note that the values used in the first order 'pixel shift fit' are taken across all echelle orders.

In other words, if you are missing arclines in a given echelle order in your new spectrum, that might be ok, as long as there are enough lines from the other echelle orders. See the lower plot of the Arc-###-fr.ps file.

It is done this way, because generally the pixel shift across columns is larger than the shifts across echelle orders, and sometimes a given echelle may only have a few arclines.

To some extent the accuracy of the template effects all the reductions, so it is possible a better template calibration may be made in the future, however, the existing template (as of November 2000) should be adequate. The residuals are in general within 0.2 pixels, except at the edges where it may be as high as 0.5 to 1.0 pixel.

One problem that might occur is if the detector shifts for any reason in the future in wavelength space by more than about 20 pixels or so, then MAKEE might require another wavelength calibration template to give adequate results (and the program would have to figure out which one to use). The old template would still work, but the 2nd coefficient might not be optimal and you would probably see this in a curvature (of more than 0.2 pixels) in the Arc-###-fr.ps file.