plot "fringesp1.ps" shows:

top:  mean BT spectra of our 48 regression profiles
for nadir, night, sea surface

bottom: Errors due to shifting fringes.  The fringes are
shifted from T_filter=T_grating=147.5846 K to the same
T_grating but T_filter=T_grating + 1 K.

The cyan curve is the absolute value of the maximum error
over the 48 profiles.  The blue curve is the bias of the
difference over the 48, and the red is the std dev.

---
Update: 3:50 pm 18 July 2002

I took a look at the RMS difference for the other offsets and
they look fairly similar to the "fringesp1.ps".  The
0.5 K RMS errors are about half the 1K errors.  The
plus offset is similar to the minus offset, although in
the shortwave the errors shift phase.

The matlab file "bt_fringe_offsets.mat" includes the
BT spectra of the 48 regression profiles for the 5
offsets (-1, -0.5, 0, +0.5, +1), as well as the RMS
BT difference.

Scott.