The first bug will only affect some people and concerns the “smile” distortion correction.
After moving up to an 1800 lpm grating I noticed that my calibration lines now appeared slightly curved
and I decided I should be correcting for this “smile” instead of just a slant.
I have discovered that ISIS only measures a positive radius for the smile regardless of the way the smile actually faces.
If the smile curves to the right (as mine does) ISIS applies the wrong correction making the smile even worse.
I discovered this quite by chance when I viewed the temporary @1.fit file that is produced. If the smile curves to the left everything is fine.
Fortunately you can change the sign of the radius number manually on the calibration screen but unless you have looked at the @1.fit file
it would not be obvious that the wrong correction had been applied.
My second observation may not be a bug but when a tilt is specified the object spectrum is corrected for slant/smile and tilt
but the calibration spectrum is only corrected for slant/smile. I am judging this from looking at the @1.fit and the @calib.fit files.
If the tilt is large it seems to me that this could result in a calibration error.
My last finding is more complicated and concerns the polynomial fitting for the dispersion.
I normally use the “file mode” for wavelength calibration with nine or more lines and a forth order polynomial specified in the .lst file, like this:
4
0.33
6143.063
6266.495
6334.428
6402.246
6506.528
6598.953
6717.043
6871.289
6929.467
I discovered what may be a bug when processing a relco emission spectrum and using a copy of the same spectrum for the calibration.
Using file mode with the .lst file above ISIS reported an excellent fit on the "Go" screen:
Wavelength fitting error
Line #1 x = 446.033 lambda = 6143.063 dlambda = -0.003
Line #2 x = 809.096 lambda = 6266.495 dlambda = 0.010
Line #3 x = 1009.610 lambda = 6334.428 dlambda = 0.001
Line #4 x = 1210.386 lambda = 6402.246 dlambda = -0.015
Line #5 x = 1520.366 lambda = 6506.528 dlambda = -0.001
Line #6 x = 1796.759 lambda = 6598.953 dlambda = 0.011
Line #7 x = 2152.761 lambda = 6717.043 dlambda = -0.002
Line #8 x = 2623.494 lambda = 6871.289 dlambda = -0.005
Line #9 x = 2803.043 lambda = 6929.467 dlambda = 0.003
--------------------------------------------------------------------------
Wavelength file : c:\users\tim\documents\atm stuff\astro spectroscopy_1\ch cyg\2014-12-19\calib\file mode_1x2_1800_585.lst
Coefficient a4 : -2.152690E-14
Coefficient a3 : -5.627464E-10
Coefficient a2 : -6.029153E-07
Coefficient a1 : 0.34140
Coefficient a0 : 5990.960
--------------------------------------------------------------------------
RMS : 0.011224
The residuals are all small, however, when I used the FWHM tool to measure the line at 6143.063 on the final spectrum it reported
a position of 6143.007 which is off by -0.056 not -0.003. This is small but it led me to investigate further and I discovered
something strange on the “compute spectral dispersion" screen under the profile tab. This is what it shows.
The same polynomial coefficients are listed but the positions of all the lines on the left are now 1 unit less.
For example line#1 at 6143.063 is listed as measure at 445.033. ISIS had previously reported this line at 446.033.
If I then hit the “compute polynom” button the polynomial coefficients change to:
Coefficient a4 : -2.148211E-14
Coefficient a3 : -5.630147E-10
Coefficient a2 : -6.024832E-007
Coefficient a1 : 0.3414014
Coefficient a0 : 5990.960
The residuals, however, are unchanged.
Not knowing which positions and coefficients are correct I decided to process without calibration and measure
the line positions myself and then use those positions on the “compute spectral dispersion screen”.
The positions I measure with the FWHM tool are:
Line #1 x = 446.867 lambda = 6143.063
Line #2 x = 808.064 lambda = 6266.495
Line #3 x = 1008.586 lambda = 6334.428
Line #4 x = 1209.367 lambda = 6402.246
Line #5 x = 1519.350 lambda = 6506.528
Line #6 x = 1795.732 lambda = 6598.953
Line #7 x = 2151.741 lambda = 6717.043
Line #8 x = 2622.463 lambda = 6871.289
Line #9 x = 2802.021 lambda = 6929.467
These are different from both sets above and after clicking the “compute polynom” button the listed residuals are:
Lambda - O - C
6143.065 -0.002
6266.488 0.007
6334.425 0.003
6402.258 -0.012
6506.530 -0.002
6598.941 0.012
6717.048 -0.005
6871.291 -0.002
6929.465 0.002
RMS = 0.010
Finally after clicking the “calibration” button the 6143.063 line on the calibrated spectrum measures at 6143.0635.
This time the measured position of all the lines agree with or are less than the residuals so I am inclined to believe this is the correct calibration.
The differences are all very small considering the dispersion is 0.33 ang/pixel but I still think this is something that needs looking into.
Tim