Talk:Cross-correlation analysis of FEROS spectra

New masks
How about other spectral types? jpritcha 06:01, 28 November 2006 (CLST)

I sent you by email the masks of almost all FEROS orders (G star template). The very blue orders were so crowded with lines, and remembering the low S/N generally found here, I did not bother to go bluer than 4000AA/9th order. For early type stars it may be quite different of course...

The masks have in general slightly fewer lines than the first batch.

Let me know how this works out. Happy testing! -- Tdall 17:33, 16 November 2006 (CLST)

First CCF between VSOP FEROS Scientific data and template
Hi hi Thomas,

Great, I eagerly await the further templates.

Indeed my script is derived directly from your test-xcor.pl script in still uses the simple max value estimate for the RV. I already fixed the conversion from pixels to km/s to use the mean wavelength of the order and an accurate value for the speed of light and now the agreement is much better:


 * 1) RV =  55.962 km/s at r = 0.535
 * 2) RV =  56.400 km/s at r = 0.583
 * 3) RV =  55.089 km/s at r = 0.535
 * 4) RV =  55.409 km/s at r = 0.512
 * 5) RV =  50.878 km/s at r = 0.472
 * 6) RV =  54.258 km/s at r = 0.577
 * 7) RV =  51.357 km/s at r = 0.585
 * 8) RV =  54.419 km/s at r = 0.435
 * 9) RV =  54.381 km/s at r = 0.396
 * 10) RV =  52.871 km/s at r = 0.544
 * 11) RV =  52.713 km/s at r = 0.528

Mean RV: 53.976 ± 1.812 km/s at 0.518 ± 0.061

But indeed there are still a number of improvements to be made including gaussain fits to the CCFs to better estimate the RV and indeed weighted combining of the results of the individual orders into a single final result. (The above means are unweighted).

jpritcha 10:11, 10 November 2006 (CLST)

Hi Thomas, John,

Thanks for keeping me in the loop. John, you seems very close to a final working results. This is great. I'll continue to follow the story, and remains open for help in the wiki side.

Cheers,

Cedric

Hi John,

That's excellent news, and the CCF looks good! I'll go on with the masks when I'm back in the office next week (still at CS14 in Pasadena).

If you used anything like the script I sent you, the RV is taken from the highest r-value and so will be pixel-ised. Ideally, you'd fit a gaussian to the profile to determine the RV. First, however, you'd make a weighted (by S/N and/or contrast) average of all orders to produce the final CCF, from which you'd then calculate the 'final' RV. Sorry if I'm stating the obvious, but I don't have access to our previous correspondence here.

I just realized that we ought to post our discussions on the wiki. I completely forgot that myself... so a reminder for both of us ;-)

Cheers, Thomas

Thomas,

You'd better get cracking on the rest of those masks, things will progress fairly quickly now I think... I have just managed my first CCF between real data and you G template... see attached CCF for order 20 of DI Psc.




 * 1) Order 20: RV = 43.8823708669816 at r = 0.534609128907715
 * 2) Order 21: RV = 45.2536949565747 at r = 0.583196626860828
 * 3) Order 22: RV = 45.2536949565747 at r = 0.534845473720434
 * 4) Order 23: RV = 46.6250190461679 at r = 0.512124984442109
 * 5) Order 24: RV = 43.8823708669816 at r = 0.471977384021546
 * 6) Order 25: RV = 47.9963431357611 at r = 0.576820979492326
 * 7) Order 26: RV = 46.6250190461679 at r = 0.58513470797665
 * 8) Order 27: RV = 50.7389913149474 at r = 0.435493041091199
 * 9) Order 28: RV = 52.1103154045406 at r = 0.396012783316446
 * 10) Order 29: RV = 52.1103154045406 at r = 0.54395848982183
 * 11) Order 30: RV = 53.4816394941338 at r = 0.527845775214038

Not entirely sure why some of these numbers are exactly equal, or why the are varying so much, will have to look at exactly how RV is being calculated... but we are on the right track I think...

jpritcha 10:11, 10 November 2006 (CLST)

The rules of engagement
Remember that the team leader should organize the work, and make sure dates are assigned to the deadlines. These should be published on this page before May 24 (for Paper1).

Team: John (lead), Cedric, Fernando, Thomas

Goal: Make (semi-)automatic cross-correlation with suitable masks to find the radial velocity and reveal SB's. This may be integrated into the FEROS DRS as a standard feature.

Methods: Suitable masks for a wide range of spectral types should be constructed from linelists and synthetic spectra (with MOOG). These masks are used to cross-correlate each individual order, yielding not only the radial velocity, but revealilng SB's too. The individual orders are then finally combined into one master-CCF (HARPS-style). This may then be included as a step in the FEROS DRS, offered to all users.

Deadlines:


 * 1) format specifications for masks
 * 2) first masks ready
 * 3) cross-corr working with first masks
 * 4) all masks ready
 * 5) cross-corr fully functional
 * 6) cross-corr done for all targets
 * 7) integrated in DRS

As team leader I have preliminarily assigned the tasks to the following team members with the given preliminary deadlines. I have (to the best of my estimation abilities) been fairly conservative in the amount of time I have estimated to be required for each task, e.g. In an email Thomas wrote "Once we have the format specifications for the individual order masks (order size in pixels, starting wl's, etc.) I can make the first mask in a few days. Making masks for all orders will take about one week." Therefore I have allowed one and two weeks respectively for these tasks. Of course it will do NO harm if we complete ahead of deadline. With the deadline for deadline definition fast approaching I would like feedback from team members asap:


 * format specifications for masks : John : 2006-05-31 (one week)
 * first masks ready : Thomas : 2006-06-09 (one week)
 * cross-corr working with first masks : John : 2006-06-21 (two weeks)
 * all masks ready : Thomas : 2006-06-23 (two weeks)
 * cross-corr fully functional : John : 2006-07-28 (two weeks)
 * cross-corr done for all targets : Cedric & Fernando : 2006-08-18 (three weeks)
 * integrated in DRS : John : 2006-12-31 (stretchy...)

Unavailabilities Please add your known unavailabilites and update as such become known...


 * John on vacation from 22/6/2006 --> 12/7/2006 (summer vacation) (fixed typo in month of return 6 --> 7, jpritcha 15:57, 31 May 2006 (CLT))
 * John on vacation from 1/8/2006 --> 18/8/2006 (parents visit from NZ)
 * Thomas in limbo from 22/6/2006 --> 7/8/2006 (leaving Chile, holidays, moving to Hilo)
 * Cedric from 4/6/2006 --> 16/6/1006 (not completely unavailable, since it is confirmed, the castle will have a wifi, but vstill very busy with courses on VLTI). -- C édric [[Image:kmail.png|16px]]T alk  06:01, 2 June 2006 (CLT)