RF FrontEnd Codesign Sayf Alalusi, Professor Bob Brodersen

1
RF Front-End CodesignSayf Alalusi, Professor Bob
Brodersen
Incoming Signals -- Only in-band signals reach
this point unattenuated. -- Out-of-band signals
may reach this point, but will be attenuated. --
All signals, are reflected at this point.
Differential RF Filter -- At the full carrier
frequency, before the LNA, we should only need a
2- or 3-pole filter. -- We want a fully
differential filter for performance reasons
common-mode rejection as well as not needing a
balun, and therefore not having to incur its
insertion loss where we are most sensitive to
loss.
Differential, Resonant Antenna -- A differential
antenna does not require a balun to couple to the
(differential) RF filter. -- A resonant antenna
provides a first level of frequency selection,
and can tolerate a reflection.
Fully Differential LNA -- This is the input to a
differential LNA, again, to avoid a balun and
achieve better common-mode rejection. -- We need
to know the input parasitics because the RF
filter will see the LNA input as its termination
impedance. Filter operation must take this into
account. -- We also need to know the MOS noise
parameters for the particular process that the
LNA is to be designed in. LNA Design Set max.
power level in LNA (Id) size devices to achieve
desired noise figure check gain repeat if
necessary.
Antenna / Filter Interface -- The antenna
terminal impedance must match the characteristic
impedance of the first filter section only, not
the steady-state impedance seen looking into the
filter input. -- This will deliver the maximum
possible signal to the filter, dont lose power
to a reflection. -- This also ensures that power
reflected from the LNA input will be totally
re-radiated and not bounce around inside the
front-end.
Filter / LNA Interface -- The filter must be
designed to be terminated by the LNA input
impedance, including all parasitics. -- The
filter must also present the correct
driving-point impedance to the LNA to achieve a
noise match. -- This will cause a reflection
which travels back toward the antenna.
Shielding Zones -- Various parts of the front-end
must be shielded. -- The Antenna needs to be
able to see the air, but should be shielded from
the active circuits underneath -- The RF filter
should not be able to see the air, or else it can
act as an antenna as well. The RF filter should
also be shielded from the active circuits
underneath. -- The active circuits of the LNA,
for example, should be shielded from the
air. This can work because -- Worst case is the
we use 3 layers of metal (in the RF filter area
1 signal 2 plane layers), in a six-level-metal
CMOS process, this still leaves 3 layers for
routing. -- The shields could double as power and
ground plane layers.
Area Concerns -- At a 60 GHz carrier frequency
and a 0.1µm minimum channel length, the antenna
and RF filter could potentially cover an area
equal to that of hundreds of thousands of
transistors. -- If shields can be constructed,
this area could possibly be reclaimed, possibly
with some routing constraints.