Yoshi Uzawa writes:

There is now a Band 4 CRE to relax the polarization
efficiency specification after a careful study about
the influence of individual optical components on receiver
cross-polarization (XsP). Please see:
FEND-40.02.04.00-0233-A-REP

The requested change is

From 99.5 % (-23 dB) in any frequency

To 99.0 % (-20 dB) in the range 134 GHz <= Freq <= 158 GHz
   98.7 % (-19 dB) in the range 125 GHz < Freq < 134 GHz
                            and 158 GHz < Freq < 163 GHz

there are two major reasons why Band 4 XsP is high.

1) The warm optics are composed of only one ellipsoidal
mirror which generates XsP of about -28 dB and is not
compensated by any other XsP contribution. This high XsP
at the warm optics reduces the margin to the spec. (only
5 dB left). 

2) All optical components (IR filters, cryostat window and
ellipsoidal mirror in warm optics) are pretty close to
each other and the phase with which XsP is generated is
very similar. This results in almost-in-phase cross-polar
field combinations of their cross-polar components.
Even though the IR filters and window meet their specification
of -30 dB, their total XsP contribution becomes about -24 dB
(6 dB higher). Notice this is almost the level of the current
spec and the Band 4 cartridge XsP and warm optics will add to
this already high value.

>From our experiments and analyses, high XsP at the IR filters
and window could happen at the lower and higher ends of
the Band 4 range (especially at 131 GHz) due to the filters
and window dielectric structure. Additionally, XsP is increased
when these components are tilted with respect to the main
optical axis, because of an increase of the difference of
Fresnel transmission coefficients for parallel and orthogonal
incidence. We have experimentally confirmed that the XsP
at the IR filter and window become negligible (and without
any clear frequency dependence) at all measured frequencies
(including 131 GHz) when these components were attached with
no-tilt angle to the cryostat, as shown in the CRE Appendix II
and the report.

In conclusion, the Band 4 cartridge itself does not create
the measured XsP frequency dependence. The XsP degradation at
the lower and higher end of the band 4 range is mainly
introduced by the IR filters and window.

Finally, it should also be pointed out that if an RMS
combination of XsP contributions of individually compliant
components is considered, results of such an optimistic
analysis show that performance may be worse than -21 dB
when IR filters and windows are used.