Change page style: 

Example of the use of "mdefringe.pro"

Example of the use of <b>mdefringe</b>

Example of the use of mdefringe

The following is a short introduction of how to use the mdefringe.pro IDL widget. It covers only the most basic case. The user will need to experiment with the procedure to see how it works.

The assumption here is that the procedure will be run in the IDL Development Environment. The procedure file needs to be loaded in, compiled, and then started. That process is not descrivbed here.

Once the procedure is started it opens a separate GUI window with two display areas and function buttons. A screen shot of this is shown below.

A stacked Michelle spectrum image can be read into the tool via the File -> Load Data menu item from the upper left of the widget. One presently needs to read from a standard Gemini Multi-extension FITS file with one image extension.

Once the image is read in the raw image is displayed in the Image Display Area at the top of the widget. When an image is displayed then if the user moves the cursor onto the image area the line or column cut at the cursor position is plotted in the Spectrum Display area at the bottom of the GUI. An example is shown below. Wavelengths increase from left to right in these plots. There is no wavelength calibration; the x axis is just in pixels. The y axis is in raw ADUs. The plot is a line cut by default but it can be changed to a column cut via the Options menu. The plotting can also be tolggled on/off in the menu.

The slanted stripes across the image are the residual fringing in the detector that we wish to remove. Since the strength of the fringes is proportional to the signal strength, the fringing on the part of the detector where the source signal falls is larger than the fringing on the sky position in the other nod position, and there is a residual fringing as a result. There is also residual fringing off the source, positive and negative. The usual situation, as seen above, is that the negative fringes off the source correspond to positive fringes on the source spectrum.

The next step is to calculate the two-dimensional FFT power spectrum of the image. This is done with the Show Pwrspec button in the middle of the GUI. One can return to viewing the data with the Show Data button at any time. The FFT display is shown just below.

Note that the FFT image is shifted so that the zero frequency point is in the middle of the image, rather than down in the lower left corner.

The fringing, at least in this spectral mode, shows up as the two strong FFT peaks seen at an angle to the left and to the right of the central peak. The main spectral power is in the central few columns. As the spectrum is along the dispersion direction in the original image running from left to right the power spectrum extends up and down in the FFT image. There also apear to be several overtone frequency components in the FFT image. See the annotated image below.

When in the power spectrum image mode, one can select regions to mask out of the FFT using the mouse. Left clicking on the image starts a rubber-band box, and when the button is released a red box appears along with the reflected box on the other side of the origin point. What we have found is that the fringing needs to have the entire columns removed from the FFT image to produce ggod results, at least for the primary frequency component.

Each subsequent box is added to the image by the cursor. The second row of buttons allows one to undo or redo boxes as one goes along. Once at least one pair of boxes is defined, the FFT Filter button is active. When pressed it masks out the boxed areas and does the inverse FFT to get the corrected spectrum frame to display. At this point the Show Diff button with show the difference between the current image and the original one. Once can move back and forth between the various displays a number of times, adding or modifying the boxes, until one (it is hoped) achieves a good removal of the fringes without adversey affecting the real signal.

The next image shows the boxes defined for the example spectrum. Then following that is shown the resulting image after the inverse FFT transformation is applied.

The defringing does seem to introduce additional noise into the spectrum. Possibly by trial and error in the masking that situation can be improved.

The final image, and the associated mask image, can be stored via the File menu. The mask can then be read in later to apply to another spectral image, if desired.

Under the Options menu are several functions. Since these higher spectral modes tend to have a showdowed region at the top of the image, and since this affects the FFT somewhat, there is an option to mask off the dark regions. This option is fragile at the moment. If you do this and then define a box beyond the "active" region it will crash the procedure.

Finally there is an experimental "Select Level" Function option. This allows one to define a masking "box" by a level contour around a given feature in the power spectrum. It may produce a better masking than a simple box, but this has not been determined yet. If this is selected, then the level in the power spectrum to be used for the masking is selected by the cursor and then a closed contour at that value is defined by IDL for masking off.