Some problems about LNA measurement

Hi, everyone. I designed a differential wideband(50MHz-1GHz) LNA with 75 ohm input impedance, and there are several problems about how to measure it:
1. The input impedance match to 50 hom test machine: Which is the best way? Should I adopt a resistor matching network? Or are there some better choices?

2. The output buffer choice. The LNA output impedance is designed to be several hundreds ohms, so I have to use a output buffer for testing. I searched the previous topics and decided to use an open-drain topology when measure NF and Gain. But I don't konw how to measure the IIP3 because the IIP3 of the LNA is >10dBm while both the open-drain and source-follow buffers can not reach this value, so the simulated IIP3 is totally restricted by the buffer!

3. Is there any paper or book about how to measure such a LNA? Thank you very much!

I find the resistive min loss pad is very boradband and the loss is very predictable. However, that won't get you from single ended to diff. For that you may need a wideband balun transformed (Toko, MACOM, etc). I use the balun at 50 ohms and then do a 50:75 ohm differential min loss pad. This has worked the best for me up to 2 GHz.

Thanks. Could you pls tell me more about the "50:75 ohm differential min loss pad"? Is this a pad on the chip?