Based upon the data in the airs_freq_oct98, JPL want five AIRS
channels added to those in the Mar98 fast model.  The five new
channels are:

  ID    A   I      freq         resol   module 
 9911  7  192   1337.1950    1315.666    3
 9510  4  150   2169.4250    1211.239    2b
 9361  3  130   2309.5125    1266.269    1b
 9232  2  116   2432.5700    1200.279    2a
 9117  1  118   2552.5375    1259.962    1a

ID = fake channel ID number.  Subtract 9000 and add 0.5 to
find the channel position within 1-2373 using the numbering
scheme specified by George in May96 (file airsfreq.pdr) and
also used in the Mar98 AIRS fast model.  Note: there is a
new numbering scheme in the airs_freq_oct98 file from JPL,
where the channels are numbered 0 - 2377.

A = array number.  The array number is just another way of
refering to the modules.  There are 17 arrays/modules.

I = index of channel in module/array.  All 5 new channels
are on the low freq side of their respective array/module.

freq = the channel frequencies are exactly as given (JPL has
already rounded them to the nearest 0.0025 cm^-1).

resol = the resolution comes from a linear interpolation of the
two nearest channels in the module (JPL did not supply the
resolution for the new channels).

Scott, 20 Oct 1998.
---
 Chan=9911, f=1337.1950 (1306.7050 to 1367.6850 by 0.0025), R=1315.6660, #FWHM=30
 Chan=9510, f=2169.4250 (2115.6925 to 2223.1575 by 0.0025), R=1211.2390, #FWHM=30
 Chan=9361, f=2309.5125 (2254.7975 to 2364.2275 by 0.0025), R=1266.2690, #FWHM=30
 Chan=9232, f=2432.5700 (2371.7700 to 2493.3700 by 0.0025), R=1200.2790, #FWHM=30
 Chan=9117, f=2552.5375 (2491.7600 to 2613.3150 by 0.0025), R=1259.9620, #FWHM=30
---

The closest matching nearby channels to the 5 new channels are:
   new chan    fake with chan:
   freq cm-1   freq cm-1, chan #
   ---------   ----------------
   1337.1950   1337.5775 = #913
   2169.4250   2177.5150 = #502
   2309.5125   2309.4950 = #364
   2432.5700   2432.5825 = #234
   2552.5375   2552.6700 = #120
These 5 old channels can be used to fake the 5 new channels: just use
the coefs for the indicated old channel as the coefs for the new channel.
I've not tested to see how well this works...I can only guesstimate, but
it might be good to within 1K or so.

Scott, 29 Oct 1998.