Thanks Tom ... you are right it is not surprising -- but what I am surprised at is how good a job the builtin AGC does for most cases and maximizes the value of the limited dynamic range. Also, your explanation of the AGC matches my observations so its out.
This is what I'm thinking (out-loud so don't laugh) ... I have a channel that should be free of most signals (through FCC allocation and observation ... lets call it dark_channel). I can put that through an xlat_fir and get the resulting power of the noise floor in that channel ( should be relative noise floor of the sampled spectrum, call it dark_channel_power). I assume I can also measure the peak coming out of the source (call this peak_source_power). I'm envisioning a block that would set a gain based on the limiting peak_source_power to less than say 80% of signal peak and limiting dark_channel_power to some acceptable upper limit (probably derived from experimentation). The gain value would have to be filtered obvious to some reasonable time-constant.
Questions:
-- Does this sound reasonable? Can anyone poke a hole in it? Is there a more conventional approach?
-- What would be a good block would measure the power of the dark_channel and source signals?
-- For the near-far paradox ... dark_channel_power sets my max gain for far signals and source_power sets my min_gain for near signals -- is this a reasonable approach
-- Is there a good starting point for the block to convolve these two power measurements into a hardware gain setting? (remote, but had to ask)
Jim