JFET gain stage with well-defined gain

[Fancy Lime]

Hi all,

I am currently noise-testing some different variants of preamp input stages. Up to this point mostly different opamps in non-inverting configuration with 40db of gain. Next will be the same thing but with a JFET common drain buffer up front (mostly to see how much noise a JFET buffer makes in the many situations where we use those). I am also pondering comparing the opamp stage to a discreet JFET stage or adding a JFET gain stage to the front of the opamp gain stage (with, say, 20db gain each). The whole endeavor is mostly academic. General consensus on the forum seems to be that a "good" JFET-input opamp will make the least noise in practical applications of the kind we usually discuss here. But I want to know how much practical difference (meaning as weighed by the subjective human ear, and therefore including frequency dependence, which tends to not be too well documented in data sheets) there is between opamps (is an OPA2134 noticeably better than a cheap ol' TL072) and how a buffer or discrete gain stage in front of a BJT opamp (which can be a lot quieter than most JFET opamps in their ideal input impedance range) stacks up here. For a fair comparison I obviously need identical gains for all tested stages and the whole procedure should be repeatable by anyone who cares to try. So I want as little device dependent parameters as possible.

My question is: how do I design a JFET gain stage with well-defined gain that does not (or "almost not") depend on the properties of the JFET (which are notoriously variable). A common source stage seems less then ideal because gain is set mostly by the Rd/Rs ratio and Rd needs to be adjusted for biasing. So I was thinking CCS/mu-amp/mini-booster (or whatever the kids call it these days), which simplifies the biasing and also provides more gain. But I have no idea how to calculate the gain on that one, let alone set it to a specific value. Is that even possible independently of the parameters of the JFET (for reproducibility's sake)?

I have so far ignored cascodes and other less common options. Is there some great alternative approach I'm missing?

Thanks,
Andy

[PRR]

WHY do you want 40dB gain in a high-Z input??

mu-amp gain depends on the mu of the devices, and will be highly variable.

In general, Rs does not matter because you must bypass it well for lowest hiss.

[Fancy Lime]

WHY do you want 40dB gain in a high-Z input??

...

Just for the tests, really. For practical application that *may* be a *tad much*. But I figured using very high gain for the test rig should demonstrate the noise contribution of the stage(s) in question most clearly compared to noise contributions by the audio interface (Focusrite Scarlett 2i4). When in use, these stages will of course be run at much lower gains most of the time.

Interesting schematic. What would that stage be called? I haven't seen this particular one.

Andy