The TIA is why it does work well. The input impedance does vary but not as much as you
might believe due to feedback and parallel loading by many resistors for both TX and RX
on the first device. The output impedance is stable due to the two emitter followers
(darlington pair) for isolation at the output replicating the driving impedance The
collector load resistor of the first device.
In practical use its not that much an issue even for the older bitx20 where the amp
was a single device. Do the math and see why its not a big deal. Hint a 3:1 change
is well within the performance range. That and it work despite the fact that the AC
Beta for 2n3904s undergo a 10:1 change from 3-30mhz. Plug that in too.
What happens to the filter if the impedance changes over a range?
What happens to the DBM with impedance changes in the 3:1 range, if the IF port is stable?
The 45mhz filter is 15khz wide if the impedance changes what is the impact?
What is the input impedance if there is no transistors at all?
Hint: its been tried. Doesn't catch fire.
What the TIA gets you in this design is a stable reproducible amplifier that can
tolerate component and device variation. It gets around the fact that filters have
unstable impedance outside their bandpass which leads to instability and unexpected
gain and impedance changes.
Farhan did a very good job in the design. It very tolerant of variation and is stable.
Use the spreadsheet to evaluate it. As to google and Wes, he knows me.
I've only built from the ground up maybe a dozen SSB transceivers that I use
for bands 80 through 432 but there are more than a few experiments in the
pile to answer what happens if....? I like those as They answer the questions
even if I didn't think afterwards it was such a good idea. Often it value is it
validates the math and topology that I may use for something else.