#***********************************************************************************
#***********************************************************************************
#***********************************************************************************
#		CCD/TAURUS sequencing files required by charge shuffling
#***********************************************************************************
#***********************************************************************************
#***********************************************************************************
#
#	This program is available via WWW or FTP (aaoepp.aao.gov.au/pub/jbh/ttf),
#	a simple operation running under unix. There is a fortran file there if 
#	needs to be compiled on a different machine. It is essential to generate
#	the CSX file for a given exposure time and no. of shuffles.
#
#	Run: 	csx < csx.inp > csx.out
#
#       where csx.inp is exposure time (sec.) and steps on consec lines. The csx.out
#	file should be ftp's to disk$user:[observer.cs_files]. Note that this is the 
#       exposure time PER SHUFFLE STEP and not total exposure time. The remaining 
#	shuffle info is inserted into 
#
#	run-
#	  run_ccd-
#	    shuf_multi-
#
#	Assume Tek 1024x1024, and slit falling at 518 as I measured for the 4-pix
#	0.356mm slit. To understand the output CSX file, take the following 1 sec 
#	x 80 shuffle:-
#
#	PI
#	PR 0,65535,1000,2000,65535,474,6,0,0,1
#	PR 0,65535,1000,2000,65535,474,6,0,0,2
#	PR 0,65535,1000,2000,65535,474,6,0,0,3
#	PR 0,65535,1000,2000,65535,474,6,0,0,4
#	:
#	PE
#	
#	0,65535     - ignore
#	1000        - exp time in ms
#	2000        - end exp time with arbitrary offset
#	65535 or 1  - 65535 shift down away from read-out register
#	1 is shift up (so could reverse shift direction)
#	474         - shift charge down by 79*6 rows; then bring up slowly
#	6           - no. of rows in each shift
#	0,0         - ignore
#	1           - index needed by instr seq
#	
#	Therefore we have 80 x 6 rows = 480 rows.
#
#	So a jump of -474 puts first frequency centred at 518-474 = 44
#	centre of freq bands 44, 50......512, 518. But the first real row is 41.
#
#	How do we convert from row number to gap number. If this unsettles you, 
#	simply scan through a 5x5 window sausage cube using Heath's special window.
#	Let's say we stepped from 100 to 495 in steps of 5.  In the example above, 
#	we can write (where frequency is just an index from 1->80)
#
#		pixel = 6*(frequency-1) + 44
#		gap   = 5*(frequency-1) + 100
#	Thus
#		(gap-100)/5 = (pixel-44)/6
#
#	So what is the best way to calibrate wavelength vs. gap from the long-slit
#	mode? Take a column through your CuAr spectrum, say, and fit to the 3-4
#	known lines against gap. The column is indexed in pixels, but you can 
#	easily convert to gap with the above, i.e.
#
#		gap = (pixel-44)*5/6 + 100
#
#	So then the approach in Zodiac would be
#
#		spec=col(image,512)
#		pixel=seq(1024)
#		xfit,spec,gap
#		linfit,spec,gap,grad_lam,off_lam
#		wave = grad_lam*gap + off_lam
#
#	Exact calibration of order/gap requires that you fit to a few emission lines
#	to get all required constants:-
#
#		m * lam_1 = 2 * z_1 + off
#		m * lam_2 = 2 * z_2 + off
#		m * lam_3 = 2 * z_3 + off
#
#	and then you can solve for m, therefore true gap. More than 2 ensures single order.