Besides gain changes, another potential effect of temperature drifts in the focal plane will be a
linear drift 
in the reference load temperature, appearing in the signal as a
baseline change with typical time scale 
hours. We have simulated the presence of
a baseline drift 
per hour, a
conservative estimate based on the current thermal study of the
mission, to demonstrate the ability to accurately remove their
potential effects.
We simulate the observations in one year in nominal configuration.
As the spin axis moves every two hours
by 5', in the thermal time-scale (
hours) the beam observes 
different circles, with some fraction of overlap among them. The signal measured in each
elements of the circle is
where 
is the difference between the true sky
signal and the reference at an arbitrary time t=0, and 
is
the statistical noise of the receiver. We assume that the elements of
3 adjacent circles observed the same signal on the sky, a reasonable
approximation because of the slow gradients of the dipole and of the
Galactic emission. So we average Eq. (23) over the 3 circles obtaining
Now we
fit this average with a linear law
Thus, 
provides the
estimate of 
, while
is the estimate of the thermal drift. The accuracy of the
estimate improves by a factor 4 by using the 
estimates of
to reach the 100 hours baseline (corresponding to 
circles).
Note that while absolute and relative calibration use the observed signal modulation, here
the result is independent of the observed signals, and it depends on the LFI frequency channels
only because of their different intrinsic noise and angular resolution. The accuracy in
reconstructing the thermal drift is determined by the signal to noise ratio in each pixel. From our
simulations we conclude that we can recover the correct value of the thermal drift with an accuracy of
(at the 1 sigma level). This introduces errors which are
on the estimate of the CMB anisotropy
amplitude.
The most potentially dangerous thermal effects are those with a component synchronous
with the spin rotation. In fact, in this case any spurious signal variation will be integrated
over long time periods and will easily exceed the final integrated statistical noise. An
accurate thermal study shows that for the COBRAS/SAMBA payload in L2 Lissajous orbit
and in total-shadow condition the only spin-synchronous effect can come from
small asymmetries of the solar panel plane with respect to the spin axis. Even in
the most critical configuration (which happens to be when pointing 
away from
anti-sun direction) the synchronous component of the physical
temperature modulation is expected to be
less than 
, and any effects on the signals observed by
the COBRAS/SAMBA instruments can be considered negligible. This is a
major advantage of the choice of the Sun-Earth L2 orbit, which
provides the best possible thermal environment for the mission.