iraf.net

---------- Forwarded message ----------
Date: Mon, 05 Sep 2005 16:48:38 -0700
From: Ian Thompson <ian@ociw.edu>
To: Mike Fitzpatrick <fitz@tucana.tuc.noao.edu>
Subject: Re: FXCOR heliocentri corrections
Hi Mike,Thanks for your quick reply to my question about the FXCOR heliocentric
corrections. The table below is a summary of some measurements from a recent
MIKE run on Magellan. All measurements use the spectrum of HD193901
in the first row as the template, the first row represents the
template measured against itself. (1) (2) (3) (4) (5) (6) (7) (8) (9)
(10)
HD193901 30Jul05 0.000 -169.721 -172.000 -2.279 -2.47 -2.467 -172.00
-172.00
HD193901 31Jul05 -0.196 -169.917 -172.886 -2.969 -2.62 -2.624 -172.34
-172.35
HD193901 02Aug05 0.629 -169.092 -172.943 -3.851 -3.83 -3.832 -172.77
-172.74
HD193901 03Aug05 1.387 -168.334 -172.622 -4.288 -4.41 -4.408 -172.59
-172.55
HD200654 02Aug05 128.800 -40.994 -44.164 -3.170 -3.10 -3.098 -43.78
-43.83
HD200654 03Aug05 127.578 -42.216 -45.775 -3.559 -3.58 -3.577 -45.55
-45.53
SMC-001 02Aug05 358.396 188.472 188.111 -0.361 -0.18 -0.186 188.677
-188.533) velocity shift = Vobs(object) - Vobs(template) from FXCOR
4) observed velocity = Vtrue(object) - H(object) from FXCOR
5) heliocentric velocity from FXCOR
6) implied object heliocentric correction = column (5) - column (4)
7) heliocentric correction from RVCORRECT
8) heliocentric correction from BCVCORR
9) heliocentric velocity from XCSAO
10) Vtrue(object) = Vtrue(template) + Vshift + Hcorr(object) - Hcorr(template)Columns 3, 4, and 5 are from the FXCOR .log file (the last three rows of that
file). Column 6 is the implied FXCOR heliocentric correction. Columns 7 and 8
are the heliocentric corrections given by RVCORRECT and the BCVCORR routine
in the RVSAO package, these agree except for the final row, here I think
the difference comes from the fact that this was a long exposure, and the
effective time for the calculation is diffect, as you point out in your note.The FXCOR values are mostly different from the other two values, to a
surprinsingly large degree. If I use the BCVCORR heliocentric corrections
for the template and the program observations together with the FXCOR
velocity shifts (column 3) then I get the velocities in column 10. These
are very close to velocities measured with XCSAO (column 9) but not to
the final FXCOR velocities (column 5).Any idea of what is going on here, or where I might be making an error?
Since the RVCORRECT routine presumeably uses the KEYWPARS file and since
the heliocentric corrections for RVCORRECT and BCVCORR are identical,
the differnec with the FXCOR values is really puzzling.thanks for your help!IanMike Fitzpatrick wrote:
>
> Hi Ian,
>
> On Thu, 1 Sep 2005, Ian Thompson wrote:
>
> > In the FXCOR program RV package), should the heliocentric corrections
> > applied to get the object velocity agree with those calculated
> > with the RVCORRECT program for the same object star?
>
> To first order, yes. Small differences come about because
> FXCOR uses the mid-point time of observation (e.g. DATE-OBS + EXPTIME/2)
> whereas RVCORRECT simply uses the DATE-OBS keyword. If a time isn't
> supplied by date-obs the UT keyword is read, but the difference comes
> from taking into account the exposure time.
>
> Cheers,
> Mike Fitzpatrick

---------- Forwarded message ----------
Date: Mon, 26 Sep 2005 11:51:06 -0700 (MST)
From: Mike Fitzpatrick <fitz@tucana.tuc.noao.edu>
To: Ian Thompson <ian@ociw.edu>
Cc: Mike Fitzpatrick <fitz@tucana.tuc.noao.edu>
Subject: Re: FXCOR heliocentri correctionsHi Ian,
I apologize for the long delay in replying, I was travelling and
overlooked your message until just now.
Without actually having the data to play with all I can do is
confirm that, yes, the numbers in your columns are different. I'd be
careful about comparing the output of FXCOR and XCSAO directly. Even
though both tasks are based on the same general algorithm, the
implementation differences normally mean the results won't be exactly
the same for common input.
Within FXCOR itself, slight differences in the peak fitting can
produce the numbers you're seeing so a more relevant number might be the
computed shift. Adding a single point to the fit, or using a Lorentzian
instead of a Gaussian or a different weighting can dramatically change the
results in come cases. The CENTER1D fit option is usually the most robust
for simple peak-finding but you won't get the peak width and errors.
Still, it can be used to help eliminate the centering as the major cause
of the difference. If you do that, do the shifts settle down at all?
Are the pixel shifts between FXCOR and XCSAO the same, or enought to
account for the difference?Cheers,
-Mike

---------- Forwarded message ----------
Date: Thu, 29 Sep 2005 13:24:02 -0700
From: Ian Thompson <ian@ociw.edu>
To: Mike Fitzpatrick <fitz@tucana.tuc.noao.edu>
Subject: Re: FXCOR heliocentri corrections
Hi Mike,Thanks for your reply, I'm just back myself from Chile, and am off on a
week long vacation tomorrow. So I'll read your message in detail when I get
back. Thanks for your continued interest in this question...--Ian
Mike Fitzpatrick wrote:
>
> Hi Ian,
> I apologize for the long delay in replying, I was travelling and
> overlooked your message until just now.
> Without actually having the data to play with all I can do is
> confirm that, yes, the numbers in your columns are different. I'd be
> careful about comparing the output of FXCOR and XCSAO directly. Even
> though both tasks are based on the same general algorithm, the
> implementation differences normally mean the results won't be exactly
> the same for common input.
> Within FXCOR itself, slight differences in the peak fitting can
> produce the numbers you're seeing so a more relevant number might be the
> computed shift. Adding a single point to the fit, or using a Lorentzian
> instead of a Gaussian or a different weighting can dramatically change the
> results in come cases. The CENTER1D fit option is usually the most robust
> for simple peak-finding but you won't get the peak width and errors.
> Still, it can be used to help eliminate the centering as the major cause
> of the difference. If you do that, do the shifts settle down at all?
> Are the pixel shifts between FXCOR and XCSAO the same, or enought to
> account for the difference?
>
> Cheers,
> -Mike