It became clear that we want to measure and characterize gain variations in the lab. This is important because it sets the differencing scheme, but it is also important because the frequency dependency of gain variations seems to be virtually unknown. We therefore want a setup that will measure gain variations simultaneously at two frequencies. At the end we want to express this in some sort of power spectrum of gain variations, both in time and frequency coordinate. This will allow us to very efficiently simulate everything in software.
At simplest, a temperature controlled experiment looking like this:
fix T white noise load (can be ice water) -> LNA -> splitter -> 2 bandpasses -> 2 diodes ->2 DC measurements
should do. A more complicated set-up has been seen in a paper by Weinreb using phase-locked setup, which should presumably be much more accurate. Interestingly, as temperature drops, LNA's gain goes up, but so do relative gain variations. Weinreb's setup did not measure two frequencies at the same time, but it should be easy to generalise his setup to this. If gain variations are very frequency independent, 1/f suppression within instrument by frequency differencing could be possible.
Temperature stability can be obtained by submerging everything in a dielectric liquid, further submerging everything into a bucket of iced water.
Ideally we want to have a clone of this setup at BNL, in addition to having it at wherever Hamdi wants to have it. :)
It became clear that we want to measure and characterize gain variations in the lab. This is important because it sets the differencing scheme, but it is also important because the frequency dependency of gain variations seems to be virtually unknown. We therefore want a setup that will measure gain variations simultaneously at two frequencies. At the end we want to express this in some sort of power spectrum of gain variations, both in time and frequency coordinate. This will allow us to very efficiently simulate everything in software.
At simplest, a temperature controlled experiment looking like this:
fix T white noise load (can be ice water) -> LNA -> splitter -> 2 bandpasses -> 2 diodes ->2 DC measurements
should do. A more complicated set-up has been seen in a paper by Weinreb using phase-locked setup, which should presumably be much more accurate. Interestingly, as temperature drops, LNA's gain goes up, but so do relative gain variations. Weinreb's setup did not measure two frequencies at the same time, but it should be easy to generalise his setup to this. If gain variations are very frequency independent, 1/f suppression within instrument by frequency differencing could be possible.
Temperature stability can be obtained by submerging everything in a dielectric liquid, further submerging everything into a bucket of iced water.
Ideally we want to have a clone of this setup at BNL, in addition to having it at wherever Hamdi wants to have it. :)