INSTRUCTIONS FOR CROSS-CORRELATING SPECTRA - Pat Hall, November 1998
www.astro.utoronto.ca/~hall/cnoc/pipeline/howto_xcor
						Last modified 20000124

Details on & locations of the scripts/programs used are found below
(all are in this WWW directory), after the outlined reduction steps.
Everything is bundled together in the "CNOCXCORPKG" IRAF package.
Enable this package in IRAF by typing (after altering the path if needed):
task $cnocxcorpkg = /d/http/hall/public_html/cnoc/pipeline/cnocxcorpkg.cl

For re-extractions / 3rd-pass: skip steps 3-5 since they've been done
before, and instead do "getarch 0223 c5 A".  This will copy the contents
of "/d/cnoc3/cnoc/0223/c5/specred/xcor/A/" to the directory, and do
the necessary last 2 lines of reduce.2.cl as well.


1. Using xgterms (not xterms), open the following IRAF windows.
-- one IRAF window for reduce.2.cl and mmultimake.  Load CNOCXCORPKG.
-- one IRAF window for xcem2.  
	Load the CNOCXCORPKG and RVSAO packages, then type "stty xterm".
-- two IRAF windows (for splot) with the NOAO and ONEDSPEC packages loaded.
-- one IRAF window for xcdisp5; no packages need be loaded.
-- Finally, start saotngxc or ximtool (w/ "-display <machine>:0.0" if needed).
	Saotngxc should be defined as follows in your .mycshrc or wherever:
	setenv SAORD_ROOT "/d/cnoc2/lin/temp/saotemp/saord"
	alias saotngxc '/d/cnoc2/lin/temp/saotemp/saord/bin.sun4/saotng 
			-fifo /dev/imt1 -winwidth 1250 -winheight 700 
			-framewidth 1024 -frameheight 1024 -maxsize 2500000
			-accel /d/baobab/lin/saotng/accel.cmds -flip y'


2. Make a directory (e.g. Y2148c7A) and from it run "getxcor.scr":
   /d/http/hall/public_html/cnoc/pipeline/getxcor.scr 2148 c7 A
   This script copies the following items to this directory:
	- the .axf.extr file created by Simon (copied to "mask")
	  ...this is just the .axf file without its two-line header.
	- the *_final.ms.fits spectra file (renamed to "netall.fits")
	- the direct images of the field
	- the 2-D spectra of the field (2 raw images and 1 coadded)
	- the comments file called *.notes 
   The following are copied from /d/lungsampa/hall/cnoc/xcor/:
	- the reduce.2.cl IRAF script
	- the xcdisp5.in file (and/or xcdisp.in for xcdisp instead)
	- the template file (called template.imh and template.pix)
	- the template velocity file (called vcorr)
	- the optional plotcors.cl IRAF task
   If available, also copy over:
	- the *_nointerp.ms.fits spectra file (copied to "nointerp.fits")

   Convert any *.fits files to *.imh files using "rfits x.fits * x".


3. Execute reduce.2.cl in IRAF (type "cl < reduce.2.cl"):

hedit netall dispaxis 1 add+ ver- show+ update+
hedit netall fields=te* del+ ver-
hedit netall fields=wind* del+ ver-
hedit netall fields=relhumid* del+ ver-
hedit netall fields=*probe del+ ver-
hedit netall fields=ccd* del+ ver-
hedit netall fields=ppp* del+ ver-
hedit netall WAT0_001 "system=equispec " show+ update+ ver-
hedit netall WAT1_001 "wtype=linear label=Wavelength units=Angstroms " show+ update+ ver-
hedit netall WAT2_001 "wtype=linear " show+ update+ ver-
hedit netall WAT3_001 "wtype=linear " show+ update+ ver-

imdel rr* v-
imdel ff,ww v-
imdel cleancal,unweighted,sky,weights v-
imslice netall rr 3
imrename rr001 cleancal
imrename rr002 unweighted
imrename rr003 sky
imrename rr004 weights
hedit netall wat* "($//' ')" show+ update+ ver-
hedit cleancal wat* "($//' ')" show+ update+ ver-
hedit unweighted wat* "($//' ')" show+ update+ ver-
hedit sky wat* "($//' ')" show+ update+ ver-
hedit weights wat* "($//' ')" s

#The above does bookkeeping first and then splits up the netall file
#(containing the extracted spectra et al.) into 4 files (cleancal, etc).
#Cleancal contains the (optimally) extracted spectra; Unweighted contains
#the (unoptimally) extracted spectrum summed across the slit.

dispcor cleancal ff w1=4395.4162 w2=6290.7149 dw=INDEF nw=520 log+ conf- flux-
dispcor weights ww w1=4395.4162 w2=6290.7149 dw=INDEF nw=520 log+ conf- flux-
hedit ff add+ field=w0 value=3.643 verify- 
hedit ff add+ field=wpc value=0.0003 verify-

background ff ff axis=1 sample=505 order=7 function=cheby nit=3 high=3 low=3 grow=1 interact-

imstack ff,ww finr

imdel cleancalm,unweightedm v-
imarith cleancal * 1.e19 cleancalm
imarith unweighted * 1.e19 unweightedm

#These commands rebin the spectra in log lambda, subtract the continuum
#and put the bits that the cross-correlation program uses into finr.
#(ff are the spectra and ww are the weights).


   If "nointerp.ms.imh" exists, also execute the reduce.3.cl file:

hedit nointerp.ms.imh dispaxis 1 add+ ver- show+ update+
hedit nointerp.ms.imh fields=te* del+ ver-
hedit nointerp.ms.imh fields=wind* del+ ver-
hedit nointerp.ms.imh fields=relhumid* del+ ver-
hedit nointerp.ms.imh fields=*probe del+ ver-
hedit nointerp.ms.imh fields=ccd* del+ ver-
hedit nointerp.ms.imh fields=ppp* del+ ver-

hedit nointerp.ms.imh WAT0_001 "system=equispec " show+ update+ ver-
hedit nointerp.ms.imh WAT1_001 "wtype=linear label=Wavelength units=Angstroms " 
show+ update+ ver-
hedit nointerp.ms.imh WAT2_001 "wtype=linear " show+ update+ ver-
hedit nointerp.ms.imh WAT3_001 "wtype=linear " show+ update+ ver-

imslice nointerp.ms.imh rr 3
hedit nointerp.ms.imh wat* "($//' ')" show+ update+ ver-
imrename rr001 nointerp
hedit nointerp wat* "($//' ')" show+ update+ ver-
imdel rr* ver-

imarith nointerp * 1.e19 nointerpm


4. If you had spectra short by more than 10 pixels at either end,
   then go into finr and edit:
	- look at the spectra and get rid of sudden drop-offs at the ends
	- look at the weights (using listpix finr[*,spectrum-#,2] and
	  get rid of gradual drop-offs by identifying the (usually
	  3-4) pixels of the drop-off and setting them to 0 
	  (imreplace finr[pix1:pix2,spec-#,2] value=0.0)


5. Run o3cora from IRAF.  This will take a few minutes so go and 
have some tea.  o3cora produces a cor (correlation) 1d image file 
(with 6 planes - one for each template), fort.2 (which contains the 
best redshift results for each spectrum for each template) 
and fort.3 (which contains the alternate oxygen-III redshifts).


6. After o3cora, run mmultimakec, xcem2, and xcdisp5/xcdisp6, which are,
respectively, the z-chooser, 1-D spectra examiner, and 2-D spectra examiner.
At the same time "splot cor" (the correlation functions) and "splot cleancal"
and edit the *.notes file.  (Yes, it's a lot of windows to have open.)
Instead of "splot cor" you can do "plotcors #" where # is the aperture number.
This will plot the xcor results for scl=2-5 for that aperture in one plot.
To get it to work first type "task plotcors = plotcors.cl".

--xcdisp5 (or xcdisp6): 
  You'll need to edit the xcdisp5.in file slightly to give it the 
  mask name; the spectral length and shift are 550 and 180 for CNOC2.

  If an error occurs, probably one of the direct images is a different
  size than the others.  Use "imheader" to figure out which one(s), edit &
  run "transimage.cl" for each, and then use "xcdisp6" instead of "xcdisp5".
  Note that the "xcdisp6.in" file has more lines than "xcdisp5.in".
  You can also use "xcdisp6" if only one 2-D spectrum exists.

  If the spectra are not displayed with the scaling, you can adjust it
  in "xcdisp5.in", but you might need to be sure that the combined 2-D
  spectral image is the sum of the sky-subtracted individual images.

  If the spectra are not centered, try using "xcdisp6", adjusting the
  lines in "xcdisp6.in" containing the phrase "2-D spectral" as needed.

  If you need to recompile xcdisp?, you'll need to install the IRAF CDL
  (client display library) and then from IRAF do:
	fc -c xcdisp6.f
	fc xcdisp6.o libcdl.a -lnsl -lsocket  (latter 2 options for Solaris)
  assuming "xcdisp6.f" is in the same directory as "libcdl.a".

--xcem2 
  Start this after mmultimakec.  See below for options; the basic ones
  are "x" (highest-R template) and "xt" followed by a template # (1-6).
  (An old version entitled "xcem" can be run from within IRAF V2.10.)

--mmultimakec:
	Multimake will ask for a few things.  The answers are "mask",
"fort.2", "fort.3", "Y2148c7A_1.coo" (e.g.), and (optionally) "old.coo".
	Go through the spectra one by one looking for the best R
values and checking them agains the cleancal spectra and the cor
correlation functions.  Anything that has an R value above 4.0 is
good.  If R < 4.0 the correlation might still be good.  Look at
cleancal by eye and decide if its OK to accept the result.  It is also
good if the redshifts from the various templates are off by no more
than 0.00010 or so.  
	Be careful with emission lines as the cross-correlator sometimes
thinks OIII is OII.  In that case you type "55" and multimake will take
the value from the fort.3 file instead of that from the fort.2 file.
	If there is a spectrum which the cross-correlator didn't get a
redshift for but you think you can get by had then do the following.
Type "99".  You will be asked a whole bunch of questions.  You get the
redshift from the cor function - expand near the peak that corresponds
to the redshift (using a a) and fit it with a Gaussian (k k).  Take
the antilog (base 10) of the centre of the peak thus fitted.  This is
the raw (1+z).  You need to multiply this by (1+z_template) where
z_teplate = v/c (v is the velocity of that template - look it up in the
file vcorr).  Now you have the true (1+z).  For uncertainty enter 0.00100.
	Try to limit the spectral class picked to only
scl=2, 4, or 5 as far as possible.
	In the *.notes file, note cases of scl=88, potential re-extractions
(e.g. due to emission lines removed during sky interpolation), and other
information which may be relevant later for 3rd pass.


8. Wavelength calibration check.  In IRAF load the TWODSPEC and 
   APEXTRACT packages.  epar aptrace and set nsum=1 and step=1.  
   Look at the sky file but display it in the other direction:
	hedit sky dispaxis 2 add+ ver- show+ update+
	apedit sky nsum=1 (answer "yes" to "Edit apertures?")
mark a sky line position by puting the cursor and pressing m:
	m - to mark a line
	t - to trace it
the positions of the lines in all the spectra are displayed.  Delete
the outliers from the fit (c for getting to the closest point, d
deletes).  Fit again (f).  Record the rms and which aperture was deleted
in "skydelt.list" and dump the coordinate of the sky line in each spectrum
to "trace.out" (using ":xyshow trace.out").
	Now need to bump up the z error estimates in the .coo file for
those spectra which are significantly off (see above).  This can be done
using the "skydelt" program (see below).  Find how far off the point is
from the best fit line (c to get to the closest point will also display
its position and the value of the line of best fit.  Take the difference
(=off).  Then 

	delta_z_extra = ( off(px) * 3.5(A/px) ) / 5577(A)

(5577 A or whatever the wavelength of the line you use is).  Add this,
in quadrature, to the error already present in the .coo file.


9. Making cute little plots.  Use the macro doplots (in IRAF):
	cl> doplots
send .mc to Howard, OR,
	cl> gkimosaic spec000* >G page1
			  001*
			  010*
			  011*

to print: gkimosaic spec000* dev=stdplot


10. Use "/d/baobab/lin/progs/cpxcor.scr" to copy everything to the
	appropriate archive directory, and update the summary file.
	It's also useful to run the coostats program to see if the
	% of identifications for the mask is reasonable (50-70%).


--------------------------------------------------------------------------
--------------------------------------------------------------------------

reduce.2.cl
-----------
/d/lungsampa/hall/cnoc/xcor/

Old example in "/d/cnoc1/lin/cnoc2/may98/xcor/Y2147c7A"
No longer need to change the *.ms file name in the script.


o3cora
------
task  $o3cora = "$!/d/larkya/lin/progs/o3cor_a.e"

Same as old "o3cor" except for minor modifications listed
at top of "o3cor_a.f". Should use this instead of "o3cor".
Default output is fort.2, fort.3, and cor.imh.
Troubleshooting: get rid of useless lines in the finr image header.


mmultimakec
-----------
task  $mmultimakec = "$!/d/larkya/lin/progs/mmultimake_c.e"

Similar to "mmultimake" but much more flexible.  First copy *.axf.extr
to "mask".  Enter an aperture number < 1 to end the program.  Output name
convention: "Y2148c7A_1.coo".  Use "old" to pick up where previously left off.


xcem2.cl  
--------
(Old version xcem.cl is in /d/larkya/lin/progs/xcem.cl)

IRAF script to simplify running xcsao and emsao in sync with mmultimakec.
Load rvsao in IRAF first, and stty gterm.
Cursor commands:
e (emsao) quick look
x (xcsao) calculate xcor peak
xt (xcsao) calculate xcor peak, choosing template


xcdisp/xcdisp5 
--------------
/d/larkya/lin/progs/xcdisp.e
/d/larkya/lin/progs/xcdisp5.e

Copy xcdisp[5].in and edit as needed.
Copy the 2 original spectral images and the 5 direct images.
Convert them to IRAF *.imh and imren 2148c7B 2148c7BB (B-band direct image).
Use saotngxc -display chooyu:0.0 or some such, turning off magnifier.
Hit ? for help when the cursor is blinking.

Saotngxc should be defined as follows in your .mycshrc or wherever:
	setenv SAORD_ROOT "/d/cnoc2/lin/temp/saotemp/saord"
	alias saotngxc '/d/cnoc2/lin/temp/saotemp/saord/bin.sun4/saotng
	-fifo /dev/imt1 -winwidth 1250 -winheight 700
	-framewidth 1024 -frameheight 1024 -maxsize 2500000
	-accel /d/baobab/lin/saotng/accel.cmds -flip y'


*.notes file
--------------

See, e.g., "/d/cnoc2/lin/cnoc2/feb97/xcor/Y0223c3B/0223c3B.notes".
I've been noting down cases of "scl=88", potential re-extractions,
and other information (I guess things which may be relevant later
for 3rd pass). 


/d/baobab/lin/progs/skydelt
---------------------------

Fortran program to add in additional redshift errors for apertures
with outlier sky line positions. Program will prompt for input
and output .coo files. Program also needs files "skydelt.list",
which gives the list of outlier apertures, and the file "trace.out",
the IRAF output file from "apedit", which gives the sky line and
fit positions (use ":xyshow trace.out" in "apedit" to produce
this file). Examples were in "/d/cnoc4/lin/cnoc2/oct96/xcor/Y0223a2B".

Format of "skydelt.list" is as follows:

		  # blank line at beginning
0.05579 rms       # the rms of the fit in pixels
5577.             # the sky line wavelength
5.1 A/pix         # the dispersion for the STIS chip
106 points        # the number of apertures fit
2 deleted         # the number of deleted outlier apertures
97                # the list of outlier apertures, one per line, **IN ORDER**
106               #   ...

The lines need to be in the above order, though the program doesn't
really use the rms or the number of apertures fit for anything.

Input:  Y..._1
Output: Y..._1.coo


/d/larkya/lin/cfrs/sk3a.com
---------------------------

smongo file that plots B-r vs. z for CWW SED's (E/S0, Sbc, Im, and
Lilly's bluer than Im). In uncertain cases, I'm using this to check 
if the color of a particular galaxy is appropriate for its presumed 
redshift and spectral class (Howard does something like this for 3rd 
pass). To run, type

smongo
macro read sk3a.com
sk3a BRc 1.2 0.4  # the B-Rc color is "1.2" at z = 0.4, e.g.
 ...              # use "gr" instead of "BRc" or "Br" if you want g-r colors
era               # to erase plot
 ...
end               # to quit smongo


/d/baobab/lin/progs/coostats
----------------------------

Program prompts for the .coo file and outputs statistics for it
to the screen.


/d/baobab/lin/progs/cpxcor.scr
------------------------------

Do newgrp FIRST, then run this FROM the archive directory, then exit newgrp.

Script to copy files from a Y* directory to the cnoc archive.
Script needs two parameters: the first is the Y* directory name
and the second is the .coo file. To run, execute the 
following, say, while you are in the archive directory  
"/d/cnoc3/cnoc/0223/a2/specred/xcor/B":

  cpxcor.scr /d/cnoc4/lin/cnoc2/oct96/xcor/Y0223a2B Y0223a2B_1.coo

Also, I've been archiving the .coo file under the "cat" sub-directory
as well.

By the way, remember to get the cnoc group thing straight. I prefer
to use Teresa's suggestion of 

  newgrp cnoc
  umask 002
    ...
    ...
  exit

before I write anything into the archive (e.g., before running the
"cpxcor.scr" script). One thing to remember, though, is to "exit" 
before you leave the archive. 


"summary" file
--------------

When done, remember to update the appropriate summary file under
the patch directory (i.e. 0223/summary) with your initials.


--------------------------------------------------------------------------
--------------------------------------------------------------------------

(scl = 1 is a star, pretty much the same as #4)

scl = 2
-------
- E/S0 absorption features
   Ca H+K, CN (3833 A), G, Mg, CaFe
- color ~ E/S0 sed or redder
- nominal min. R = 4

(scl = 3 is an E+A)

scl = 4
-------
- 3727 + absorption features, or
- stronger presence of Balmer features relative to scl = 2
  - H zeta (3889.05 A), H delta (4101.74 A), H gamma (4340.47 A), 
    Hbeta (4861.33 A); occasionally also H theta (3797.9 A)
  - note H epsilon (3970.07 A) is blended with Ca H 
     and H eta     (3835.38 A) is blended with CN
- color ~ between E/S0 and Sbc
- nominal min. R = 4

scl = 5
-------
- strong emission or emission + ratty/no absorption
- color bluer than ~ Sbc
- nominal min. R = 5

scl = 77
--------
- strongest features may be Mg and CaFe, plus funny continuum in red
- check axf file for interloping stars

scl = 88
--------
- potential/possible z but not quite convincing enough
- e.g., weak 2D eline but noisy 1D spectrum
        weak absorption features but relatively high R value
         or xcor peak

AGN 
---
- can have broad MgII emission (2798 A rest);
   set z error to be half-width as measured using, e.g. splot k,k
- set scl = 6

etc.
----
- "dssp" every spectrum in case of accidentally dcr'ed elines
   (this has happened a couple of times this run)
  spectral length and shift are 400 and 105 for STIS
  "set iorient 2" and then click zoom once for zoom 3
  "set dsup 5" (upper display limit) seems best
  need to add 2 blank lines at top of *.axf file from Simon in order
   for ppp to read it correctly (or get original *.axf file)
- "/d/larkya/lin/progs/xcdisp.e" now replaces "dssp" in ppp
- note cases for re-extraction; e.g., clean elines in 2D but
   surprisingly ratty 1D spectrum (perhaps forgot to trace distortion 
   for spectra in chip corners)
  worse case scenario hand measure elines off of 2D image, pending
   re-extraction
- note STIS dispersion is 5.1 A/pixel (this works pretty well but may be
   a bit off in the far blue)
  reference sky lines: 5577 and 5892 (Na)
- high-z H+K (or Balmer) sometimes indicated by noticeable "mottling"
   in the red
- check "unweighted" for pre-interpolation (and pre-flux-calibrated)
   spectrum in uncertain cases (e.g., eline near a sky line, and the
   interpolation may have chopped off part of the eline), and xcor
   using "unweighted" if necessary
- in such cases, may need to run emsao/xcsao directly, not xcem.

----------------------------------------------------------------------

- "ncl": can use up/down arrows directly, instead of "e" then up/down 
         arrows, to retrieve previous lines

- "imarith cleancal * 1.e19 cleancalm"         
  "imarith unweighted * 1.e19 unweightedm"         
  This is to get "emsao" to work correctly when reading the spectra
   (xcsao doesn't care)

- "xcsao cleancalm save_vel=yes template=templates.0005 czguess=120000.
    specnum = 1"
  cross correlate aperture 1 against template 5, with initial cz guess
   set at 120000 km/s, and write velocity into spectrum header
  keystrokes: ?      list keystrokes
              q      quit screen
              p      choose closest xcor peak or re-start xcor fit
              l      go to lines display window
              x      exit lines display window without quitting xcsao
              g      smooth spectrum (for display only); will prompt
                      for number of times (I typically use 3; 0 will
                      return to unsmoothed version)
              z, z   zoom between cursors
              u      unzoom
              d, d   interpolate data between cursors
              e      guess z from emission line location 
                      (choose from list or enter wavelength)
              a      guess z from absorption line location
                      (choose from list or enter wavelength)
              w      show observed and rest wavelengths at cursor
              0      if plot accidentallly goes off center
       
- "emsao cleancalm specnum=1"              
  emsao can be buggy
   it may not pick up obvious strong emission lines; e.g., you may need to 
    repeat e, f a couple of times or use xcsao instead 
   sometimes you have to immediately repeat the command if the cursor
    suddenly vanishes; typically happens for keystrokes v, g, e, or a
  keystrokes: ?      list keystrokes
              v      choose velocity to plot
                      x    from cross correlation, as saved from xcsao
                            (my default)
                      e    from emission line fit
              g      smooth spectrum (for display only); will prompt
                      for number of times (I typically use 3; 0 will
                      return to unsmoothed version)
              e      guess z from emission line location
                      (choose from list or enter wavelength)
              f      fit emission lines
              +      force add emission line to fit
              a      guess z from absorption line location
                      (choose from list or enter wavelength)
              w      show observed and rest wavelengths at cursor
              z, z   zoom between cursors
              u      unzoom
              d, d   interpolate data between cursors
              q      quit emsao
              0      if plot accidentallly goes off center

- splot  
  keystrokes: ?      list keystrokes
              )      forward 1 aperture
              (      backward 1 aperture
              #      select new aperture
              %      select new band 
              a, a   autoexpand between cursors
              c      clear (unzoom) and redraw
              r      replot current window
              k, k   fit gaussian between cursors
              space  type out current cursor values
              w      window commands
                      a     show full display
                      j     make current cursor position left display boundary
                      k     make current cursor position right display boundary
              x, x   interpolate between cursors
              j      fudge point to y-cursor value
              b      toggle base plot level to 0
              i      save current image

----------------------------------------------------------
-------------- EDITING INDIVIDUAL SPECTRA ----------------
4. Clean cleancal.  Display (using splot)the cleancal spectra one at a
time (use #21 to go to the twenty first spectrum, ( and ) to scroll
through spectra in IRAF).  Do the following, while comparing with the
2d spectrum displayed with dssp in ppp (it is also useful to have the
*.notes file open to look at at the same time):
	- get rid of bad sky subtraction near sky lines (often, but
	  not always, characterised by a sine profile)
	- get rid of excessive noise in the blue part of the spectrum
	  (but be careful not to kill any emission lines)
	- look for emission lines in the 2d spectrum (put comments in
	  the *.notes file if you see them)
	- if the spectrum is a short one (more than 10 pixels at
	  either end) edit the continuum to make it more natural 
	  looking (important for continuum subtraction later on)
Editing of 1d spectra is done by typing x x to mark the endpoints of
the edit - EMACS will then interpolate between the 2 x's.  You zoom by
typing a a at the endpoints of the zoom.  Unzoom by typing c or ( ).
You can also smooth spectra by typing s.  Changes can be saved by
typing i.  Be carful what image you save to. Also be carful that you
don't save a smoothed spectrum!
----------------------------------------------------------------------


From hlin@qso.as.arizona.edu  Fri Nov  6 23:05:46 1998
Received: from lepus.astro.utoronto.ca (lepus.astro.utoronto.ca [128.100.89.2]) by manaslu.astro.utoronto.ca (8.6.12/8.6.2) with ESMTP id XAA16460 for <hall@manaslu.astro.utoronto.ca>; Fri, 6 Nov 1998 23:05:46 -0500
Received: from qso.as.arizona.edu (qso.as.arizona.edu [128.196.209.165])
	by lepus.astro.utoronto.ca (8.9.1/8.9.1) with ESMTP id XAA01529
	for <hall@astro.utoronto.ca>; Fri, 6 Nov 1998 23:09:38 -0500 (EST)
Received: (from hlin@localhost)
	by qso.as.arizona.edu (8.9.1/8.9.1) id VAA24759;
	Fri, 6 Nov 1998 21:05:36 -0700 (MST)
Date: Fri, 6 Nov 1998 21:05:36 -0700 (MST)
From: Huan Lin <hlin@as.arizona.edu>
Message-Id: <199811070405.VAA24759@qso.as.arizona.edu>
To: hall@astro.utoronto.ca, hlin@as.arizona.edu
Subject: Re: cnoc2: ir & xcor
Status: OR

Hi Pat,

> The main question is the difference between scl class 99 and 88
> objects.  I'm using 99 for believable redshifts that the xcor
> program missed for some reason, and taking the redshift from
> the peak in the correlation function, adjusted for the template cz.
> Per Howard's advice I'm using 88 for possible redshifts...
> e.g. we had a case with a possible eline which didn't quite
> make sense as OII or OIII, so I'm not giving it a redshift but
> I am noting the possible z's in the *.notes file.

> One last question, do you always choose the scl with the highest
> R value (among scl=2, 4, and 5 that is)?  e.g. in objects with OII
> and absorption I've been choosing scl=4 if R>4, even if R=10 for scl=5.

Your use of scl=99 and 88 is indeed the same as mine. One thing
that may make things slightly easier for scl=99 is to use
xcsao/emsao (via xcem or on their own; as opposed to reading off the 
correlation peak position, etc. manually) to get the correct redshift,
estimate errors and R values. And for scl=99, also specify the 
appropriate scl class when mmultimakec prompts you for it (which you're
probably already doing).

I don't always take the highest R value; e.g., sometimes I do choose scl=4 
if the spectrum obviously shows both emission and absorption, or if 
the clincher for redshift believability is the mutual corroborration
of both absorption and emission features. However, I don't override 
the highest-R-value rule unless the R value difference is small,
so for your R=10 for scl=5 example I would go with scl=5. (The scl=5 
template does show a bit of absorption I think, so it's not entirely
emission lines.) Another situation where I would override is for
absorption line galaxies: sometimes scl=4 gives the highest R value,
but if there are no emission lines I'd go with scl=2.

Yes, using "unweighted" is useful in the case of obvious elines which 
got zapped by the (sometimes too wide) sky interpolation. I basically do
these as I go along: I exit out of xcem and use xcsao/emsao on their 
own (then use scl=99), or in stubborn cases where the "unweighted" spectrum 
is also messed up, I just measure things manually off the 2D using xcdisp
(elines I can get to delta_z = +/- 0.001 using xcdisp). I also make
a note that these spectra may need to re-extracted, but I have not
yet sent Simon the list. However, I don't redo the redshift from
"unweighted" unless that's the only way, i.e., there's only 1 eline
and it got zapped by the sky. If you want to re-do xcor after editing
the spectrum, go ahead (and archive the edited "finr"); I haven't been
doing this because it becomes a subjective exercise trying to boost
up R values and it's an additional decision to make. I don't use R value
cuts anywhere in my analyses (nor does Howard for calculating weights,
I believe), so basically if I believe the redshift, I will accept it
regardless of the R value (which can be screwed up sometimes by a problematic
extraction; but if it's noise instead causing the low R, I will of course
be less likely to accept the redshift).

I've been very slow with the CNOC2 LF paper revisions, but I should
finish things up after the Steward IR proposal is done.

Huan



From hlin@as.arizona.edu  Thu Oct 15 21:45:28 1998
Received: from astro.as.arizona.edu (astro.as.arizona.edu [128.196.208.2]) by manaslu.astro.utoronto.ca (8.6.12/8.6.2) with ESMTP id VAA26442 for <hall@manaslu.astro.utoronto.ca>; Thu, 15 Oct 1998 21:45:27 -0400
Received: from qso.as.arizona.edu by astro.as.arizona.edu (SMI-8.6/SMI-SVR4)
	id SAA19186; Thu, 15 Oct 1998 18:44:31 -0700
Received: by qso.as.arizona.edu (SMI-8.6/SMI-SVR4)
	id SAA27525; Thu, 15 Oct 1998 18:44:28 -0700
From: hlin@as.arizona.edu (Huan Lin)
Message-Id: <199810160144.SAA27525@qso.as.arizona.edu>
Subject: extractions and xcor
To: hall@manaslu.astro.utoronto.ca (Pat Hall)
Date: Thu, 15 Oct 1998 18:44:27 -0700 (MST)
Cc: hlin@astro.as.arizona.edu
In-Reply-To: <199810052214.SAA14362@chooyu.utoronto.ca> from "Pat Hall" at Oct 5, 98 06:14:46 pm
X-Mailer: ELM [version 2.4 PL23]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Status: OR

Hi Pat,

I've ftp'ed the 2148c5B, c6B, and c7B extraction files
from DAO and have put them into the archives, so you
don't need to do any ftp'ing for those files.

I've double checked that the xcor stuff (o3cora specifically)
should run fine on the Ultras and in IRAF 2.11, so you 
can speed things up relative to manaslu. You may want
to recompile o3cora on the Ultras first
(use "fc o3cor_a.f" where "o3cor_a.f" is in /d/larkya/lin/progs/)
However, one thing to watch out for is to first multiply the
"finr" spectra by 1.e19 before running o3cora 
(you can do this in "reduce.2.cl"). What I've found is that
for some reason, when o3cora is compiled on the Ultras, the
program doesn't like the units of the flux-calibrated spectra
and wastes a lot of CPU time dealing with it I think
(no error messages are output, but you can see the wasted time if 
you run o3cor_a.e from unix with "time o3cor_a.e"). 
The compiler on manaslu doesn't seem to care however; I think
this is some peculiarity of the SunOS vs. Solaris Fortran compilers.
Also, on the Ultras things will also slow down if the
spectra are kept on manaslu's (rather than moonray's or 
makalu's own) disks.

One last minor thing to check is to have "flux=no" set for
dispcor (used in reduce.2.cl); the default seems to be "yes"
but I've always used "no". It doesn't make much difference
but I just want to keep everything consistent.

Huan

P.S. On the Ultras, we should also do a "set oifversion=1" 
     so that the .imh files produced by IRAF 2.11 (e.g. cleancal)
     will still be readable by older IRAF .