> what the logic of the use of the 3819 was? if it was just because it has a sim model
Right, my sim has about one JFET in it.
> reasons not to use mosfets
I don't know. The same basic equation covers both. But commodity MOSFETs tend to be big. If you want a big limiter, OK. Dynamic range scales somewhat like square-root of power dissipation. And the tend to be Enhancement mode, which means a limiter would idle at a positive gate voltage; it is super simple and convenient to idle near zero bias and throw the gates down for GR.
> would this work if there was a mu-follower
Obviously not. What is the gain of a resistor loaded amp? Rl*Gm. As you change bias, Gm varies, Rl does not vary, gain does change. What is the gain of a Mu-follower? Mu/2. Period. Mu hardly changes with bias, or not until you are so near cut-off that your maximum signal is about zero. Anyway the Mu of a MOSFET is very-very high. For good S/N at the input, we want fairly large input signals; but if gain is high then output is HUGE. Don't think "good amplifier". A good gain cell may be a really bad amplifier. It is a different problem.
ANYWAY: what is all this complication? Start simple, get it working, then elaborate.
> the 3918 typical IDSS would not work with the 500 ohm resistors
There is nothing magic about 500 ohms. It happened to give an operating point at 1/2 of the supply voltage for the particular fake-FET that my sim models.
> I don't think I measured higher than 15ma
You got a lot of 5mA or 10mA parts? So use 1K or 2K drain resistors. That's what I was expecting for real JFETs, but the sim model wanted 500 ohms. As you say, real FETs vary by 10:1. In simple amps we can use self-bias and un-fussy circuits so that all FETs in a bag will amplify. In this case we do need matched pairs, matched for Idss and Vth.
The drain resistor is moderately critical because it is the source impedance for the next stage. If it goes to a transformer, high value resistors complicate transformer design, low values can drive most anything.
How fussy is the match? Build it and see. It will work, but for fast audio limiting it will thump. You can still do a lot of testing to see if the range of gain control and level is usable. If it is, then you can obsess the match and thump problems. If it sucks bad, not just for thump, then you can give up and not deal with matching.
> what your thoughts were for a starting point in picking the fets.
Pick two medium-size JFETs with similar Idss and Vth. Pick supply and load values to set the drains about halfway up the rail at zero bias. That's the easy part. Finding suitable input levels and sidechain will be the heavy work.
