It can be done with 2 Jfets, 2 bipolar if the lower input impedance is useful, or 1 of each (HyCas).
IF one wants to be retro a 12AT7!
They all work. The two J-fets is really old and appear in ths Allied Radioshack AX-190 from
the early 70s. Two transistors was over 10 years before. Hycas actually was in the same
time frame. HyCas was published as a complete IF with RF derived AGC and it works
I've used all the chips ua703, MC1550, CA3028 and their variants and follow ons
like the 1350/1590. They work but require care or you get of what you asked for.
All of them where differential mode is used suffer an increase in noise as agc is
increased. The exception is maybe the CA3028 as it can be used in cascade
mode as well and will avoid that. That noise increase degrades signal quality.
Designs like dual Jfet, Dual gate mosfets like the BFR998 tend to be quieter.
They are more difficult in incorporate in some topologies. I think they are worth
it. A high dynamic range receiver puts a very high requirement on succeeding
Like K9HZ I've built enough high performance radios for hobbe and some commercial
and I have one general comment. AGC can be an aid to a smooth performing radio.
You can calculate and model it to death, then you build it, most importantly test it
and listen to it. If it sounds good you done all to often it will not. Many times its
easier to do it empirically cut, try and test. Right wrong and what I want doesn't
matter if it meets your criteria.
One thing, filter ripple is hard to hear for most people and unless its more than 6DB
most will not. Most commercial filters can easily have 3Db of ripple. I doubt one
made form less than hand picked parts is much better than that. I've never made
a satisfactory quality filter without hand picking the crystals for both frequency
and Q and same for the parallel caps. Data can be picky and not like that but
our ears are not the same (we have different preferences) and we are building
to suit the ears.