Atomic case:

'  26.000            '    1       2342
'Fe I    NLTE'
  4200.087  3.884 -1.130   -7.420    7.0  5.01E+07  0.000  'p' 'd'   0.0    1.0 'Fe I LS:3p6.3d6.(5D).4s.4p.(3P*) z3D* LS:3p6.3d7.(4F).4d f3F'  82 0  'z3D3*' 'none'  'c' 'x'
  4200.463  4.154 -3.374   -7.290    9.0  2.45E+08  0.000  'p' 'd'   0.0    1.0 'Fe I LS:3p6.3d7.(4F).4p y5D* LS:3p6.3d6.(5D).4s.\ (6D).5d 7F'  90 484  'y5D3*' '6s5F'  'c' 'a'
  
Header: 
26.000 is element.isotope (000 means no specific isotope). For Iron we could have, e.g., 26.056 for 56Fe.
1 is for neutral (2 for first ion, etc)
2342 is number of lines for that element in the file.
Fe I is the element, NLTE is a flag that is not necessary.

For each line that follows:

col 1: lambda(A)  
col 2: Elow(eV) 
col 3: loggf 
col 4: fdamp (see below)
col 5: gup
col 6: gamma_rad (if =0, gf-value is used to compute gamma_rad)
col 7: gamma_Stark (may be omitted)
col 8: s,p,d,f etc for upper level (or X), see fdamp
col 9: same for lower level
col 10: equivalent width, when needed (abundance determination in eqwidt run)
col 11: error in eqw
col 12: (in quotes) some text describing levels or whatever you like to include
The following columns are only used in NLTE run:
col 13: lower level number in NLTE model atom
col 14: same for upper level
col 15: lower level id
col 16: same for upper level
col 17: not used (may be omitted)
col 18: not used (may be omitted)


and for fdamp:
================================

... in the fourth column of the line list (called fdamp):
* 1) use ABO theory (Anstee, Barklem, O'Mara) for collisional damping with H,
*    with data taken from line list: fdamp contains sigma.alpha.
*    This number is available starting with VALD3 version of the VALD database.
*    See : http://www.astro.uu.se/~barklem/howto.html
* 2) if (1) not available check if something can be computed in the anstee.f
*    routine using generic broadening recipes from the ABO formalism
* 3) if (2) not available, check in linelist for a gamma6 at 10000K
* 4) if nothing else worked, compute Unsoeld approximation, using a fudge factor
*    read from column 4.

Example for 
case 1:
  6103.649  1.848  0.361  836.274    6.0  1.05E+08 'p' 'd'   0.0    1.0 'Li I LS:1s2.2p 2P* LS:1s2.3d 2D'
case 2:
  8858.071  3.879 -4.036    2.500    4.0  6.76E+07 'd' 'p'   0.0    1.0 'Li I LS:1s2.3d 2D LS:1s2.11p 2P*'
case 3:
  4305.083  7.685 -4.221   -7.160    5.0  4.79E+08 'x' 'x'   0.0    1.0 'C I LS:2s2.2p.3s 1P* LS:2s2.2p.5p 3P'
case 4:
  8858.071  3.879 -4.036    2.500    4.0  6.76E+07 'x' 'x'   0.0    1.0 'Li I LS:1s2.3d 2D LS:1s2.11p 2P*'

One example:
  5349.465  2.709 -0.428   -7.652    7.0  2.69E+06 's' 'p'   0.0    1.0 'Ca I LS:3p6.3d.4s 1D LS:3p6.3d.4p 1F*'    8   26  '13D1D'       '14P1FP'
fdamp = -7.652, and the calculation will be made according to the generic  ABO recipe for s-p transition (case (2)). 
If I replace 's' 'p' by 'x' 'x' in the line list, broadening will be computed using -7.652 and the gam_6 recipe from Kurucz (case 3), 
if in addition I replace -7.652 by 1.0, The Unsoeld recipe with a fudge factor = 1 will be used.

================================

