2010/2/26 Sylvain Munaut <246tnt@gmail.com>

Yes thanks a lot :) I'm working on a RF front end for openbts so
that's precious info :)

First, I'm a little surprised that moving the RX2 SMA before the
antenna is such a gain. After all, once C202 is removed and the pin
grounded, there is no more connection to the TX side through there.
What am I missing ?
My best guess is that the signal leaks through the common power supply. 
Second, when looking at the figures choosing a LNA, we noticed that we
couldn't put a very strong one if we didn't want to saturate the input
stage for close phones (think more picocell where the bts can be only
a couple meters from the antenna). My reasoning was such :

 - Phone TX power is regulated but the lowest then can transmit at is
0 dBm on a 0 dB gain antenna
 - Traveling trough 1 m of air: about 40 dB loss
 - Simple RX antenna 3 dB gain
 - Duplexer insertion loss 3 dB
 - Our external LNA : x dB gain
 - Our external RX filter insertion loss : 3.5 dB
 - The RFX LNA MGA82563: 13.5 dB
 - Maximum input level at the AD8347 (when set at minimum gain) : - 2 dBm

So :

-2 dBm = 0 dBm + 0 dB - 40 dB + 3 dB - 3 dB + x dB - 3.5 dB + 13.5 dB
=> x = 28 dB  absolute max for the LNA (after that -> saturation in the mixer)

And that's without much safety margin.
In the end, we chose a 15 dB LNA because there wasn't one ~ 25 (only
single stage 15-16 or double stage 31-33 ...).

Do you see anything wrong with my math here ?
No. 15 dB sounds like a good choise for close range operation. And, provided you have the required tx/rx isolation (two RFX cards), you can easily reconfigure for wide area coverage by turning up the AD8347 gain.
My reasons for using the 40 dB amp were:
1. It was the only amplifier I had available
2. I was aiming for long range and did not worry about phones close to the antenna
3. The system was limited by the 90 dB tx/rx isolation
BTW: What is the effective tx/rx isolation when using two transceiver boards?