delta_kuadrant |
04/06/2009 09:49AM (Read 4809 times)
|
|
|
Status: offline
Registered: 01/26/2008
Posts: 28
|
hello all,I've tried a fxcor package using template spectra (i thought it has single line data) on double line eclisping binary data. It's seemed working flawlessly, and the result (vhelio) came up. the thing is, i didn't know which component (secondary or primary) that has a those RV result. Can you give me some hints how to do cross correlation on double line data in order to obtain RV for each components. I usually using rvcorrect to do RV determination for each component and I have no experience before with fxcor package.thanks PJ
|
|
|
|
fitz |
04/06/2009 09:49AM
|
|
|
Status: offline
Registered: 09/30/2005
Posts: 4040
|
FXCOR has a d' keystroke to fit up to four gaussians in the correlation peak, essentially giving you a velocity determination for each component that can be resolved. This works the same as the deblending feature in SPLOT but see the help page for details.Once you have the velocities however, that's all you have. You'll need to use some other (probably non-iraf) software to determine the orbit and period information.-Mike
|
|
|
|
delta_kuadrant |
04/06/2009 09:49AM
|
|
|
Status: offline
Registered: 01/26/2008
Posts: 28
|
hello Mike,
thanks for your reply
I've tried the 'd' keystroke on correlation window
but now, I'm confuse how to mark the line because the spectra is very crowded.
is there a way to zoom in to the spectra so I can mark the selected lines?
also, I don't understand about the 't' keystroke. what value I must entered with this keystroke? where do I look for that value so I can entered the correct one?thanks and sorry for a bunch of newbie questions^^PJ
|
|
|
|
fitz |
04/06/2009 09:49AM
|
|
|
Status: offline
Registered: 09/30/2005
Posts: 4040
|
The 'd' keystroke is used to fit the peak(s) of the correlation function. You can use the ":window <N>" command to adjust the size of the main plot boundaries so you can see the CCF better. By default the window will find the max peak in the CCF and center on that, however the 'wincenter' parameter can be used to constrain it.The 't' keystroke lets you refit the template spectrum for continuum subtraction and is unrelated other than by normalizing the object and template you can produce a flatter CCF to fit the peak. The 'd' command is used to set the left/right side of the deblending region, from there you use 'm' to mark the peaks and other keys to determine what to fit. See the help page for details, post the images or a plot of the CCF if you still have problems.-Mike
|
|
|
|
delta_kuadrant |
04/06/2009 09:49AM
|
|
|
Status: offline
Registered: 01/26/2008
Posts: 28
|
Hello again,
thanks for your last reply, I've tried the :window command and it worked
I also tried to deblend my line, derived RV for each component, and compared it with rvcorrect command and it showed a consistent results (I hope I understood your last reply correctly)
but to make sure that I've done deblending correctly, here are what I did, correct me if I'm wrong :
1. I adjusted the main plot boundaries using :window command (and sometimes with 'wincenter' parameter too)
2. from the correlation window, I used 'd' keystroke to enter deblending mode
3. then, I marked 2 peaks from main window that has value above 'zero' level (green dashes (?)) <----- I'm not sure about this one, but, when I entered a large value in :window command, says 1000 or 2000, I saw 2 peaks above the zero level
4. I fitted the peaks with gaussian profile (I picked one of the options)
5. with 'v' keystroke, I derived RV for each peakfor fitting options, I usually pick 'f' one because I thought that the peaks is clearly resolved and have a gaussian shapes. Is it okay? probably you have a better suggestion for me.thanks a lotPJ
|
|
|
|
fitz |
04/06/2009 09:49AM
|
|
|
Status: offline
Registered: 09/30/2005
Posts: 4040
|
A single-lined spectrum should show an FXCOR plot something like[img:c6d7dcefff]http://iraf.noao.edu/projects/rv/ccfmode.gif[/img:c6d7dcefff]I'm not sure where you're seeing the green dashed lines unless the color of the lines used to do the fit is different in your case. A 'window' size of 1000 or 2000 is awfully large, are these two peaks even near each other? The upper plot is the entire correlation function, with a zero-shift in the middle. If the spectra were properly continuum normalized it should be mostly flat with some peaks indicating a correlation, a large hump in the CCF in normally due to unflattened spectra. Very low S/N spectra may not show any structure at all in which case it is up to you to pick the peak to be fit. Without seeing the spectra or at least a plot I'm not sure what else I can suggest.-Mike
|
|
|
|
delta_kuadrant |
04/06/2009 09:49AM
|
|
|
Status: offline
Registered: 01/26/2008
Posts: 28
|
hello Mike
here I attached a snapshots from my work
1. this is my correlation window when I started the cross correlation procedure
[img:1b0f29352b]http://s.itb.ac.id/~pj103/cor1.PNG[/img:1b0f29352b]2. then, after I defined a wavelength ranges and used a :window command, this is what I've got (I also adjust the plot with wincenter command too)
[img:1b0f29352b]http://s.itb.ac.id/~pj103/cor2.PNG[/img:1b0f29352b]3. I saw that two peaks, and then I employed a 'd' keystroke, marked the peaks with 'm' keystroke and applied fitting using 'f' keystroke. this is what I've got
[img:1b0f29352b]http://s.itb.ac.id/~pj103/cor3.PNG[/img:1b0f29352b]any suggestion?about 'green dashes', you can see that on my first snapshot above
thanks PJ
|
|
|
|
fitz |
04/06/2009 09:49AM
|
|
|
Status: offline
Registered: 09/30/2005
Posts: 4040
|
This all looks reasonable, the dashes simply means that there is no title string in the image. The two peaks are well-enough separated that you could fit them individually using some other function if you like, but the gaussian fit looks good so you might not gain much.-Mike
|
|
|
|
delta_kuadrant |
04/06/2009 09:49AM
|
|
|
Status: offline
Registered: 01/26/2008
Posts: 28
|
thanks a lot Mike
I'll post again if I have another problem ^^PJ
|
|
|
|