Non-Phool jellybean audio-frequency JFET

oise

hat it is

point

d to waste

t -

Very cool thanks Phil. I think I followed about 1/2 of that. (You make me feel like a cave man rubbing two resistors together at times.) I was thinking about using the jfets as a differential pair in front of a 'nice' opamp. (a B. Pease circuit fragment.) For a simple resistor only circuit, I've recently been turned on to this artifical resistor circuit by R.L. Forward. (US patent 4176331, or J. Appl. Phys. (53) 3365, 1982 Three resistors and an opamp. Just my speed! (I'm still working on the noise analysis.)

("Dragons Egg" (sci fi.) by R.L. Forward is a fun read.)

George H.

is

point

waste

Didn't mean to be mysterious about it. The main op amp runs with its noninverting input grounded, so its feedback will hold the BF862's source at ground. The remaining problem is to make sure that the gate of the BF862 is also at ground, for which you need a current sink in its source that's adjusted to exactly I_DSS.

The current sink should be a BJT with a couple of volts drop across its emitter resistor (which gets rid of its shot noise pretty well). The snooper op amp is connected as a slow integrator, with its inverting input connected to the gate of the BF862 through a sufficiently large resistor, and its noninverting input grounded. Its output controls the current source. (Make sure you can get any current from about 8 to 25 mA, and watch out that you don't crank up the current high enough to forward-bias the GS junction.)

The bad news is that you get the Johnson noise of the big resistor, but it goes away for frequencies more than ~10 times the loop BW. Another RC bypassing the base of the BJT to the negative supply helps with high frequency noise and PSRR.

That way the BF862 always runs at exactly I_DSS, and you avoid the offset, drift, and extra noise caused by using a BF862 diff pair.

The main problem with that, as with all composite amps, is frequency compensating it without getting all sorts of whoop-de-doos at late times in the step response. (Putting a pole-zero pair inside a feedback loop doesn't get rid of it entirely--it replaces it with two closely spaced pairs, and the error shows up as ~1%-ish ripples in the step response.)

Thanks. I'm an old Forward fan, ever since coming across some of his science writing when I was about 12. I've read Dragonfly, Rocheworld, and a few others. Good medicine. He also wrote a really great paper about using interferometers to detect gravity waves in about 1972.

Cheers

Phil Hobbs

BF862. About 0.8 nV noise, 1/f corner about 1 kHz. They parallel very nicely (we had a thread on that a couple of months back). When parallelled directly, you want to run them at I_DSS, so an op amp servo to force V_GS to be zero. If you're just using one, run it at about 12 mA.

Cheers

Phil Hobbs

Weird. This showed up on ES as a new post with no replies, and then suddenly all this discussion from 2012 appeared.

Ah well, even antique circuit discussion is better than zillions of posts about graduated cylinders and email clients, which in turn is better than some of the other stuff we get stuck in sometimes. New circuits discussion is better still of course.

Cheers

Phil Hobbs

You just don't appreciate the finer things in life >:-} ...Jim Thompson

What do you have against bread pudding recipes?

Hobbs probably doesn't like grits either ;-) ...Jim Thompson

Bread pudding is a crime against humanity.

Cheers

Phil Hobbs

Well, on my current ketogenic diet, it's in the same category as cotton candy. ;)

Cheers

Phil Hobbs

I suppose I could work up a version that has a lot of meat in it.

That would ruin it. Ignore those people who don't relish a good bread pudding. ...Jim Thompson

Or just save the bread, eggs, and sugar for something more like this:

Cheers

Phil Hobbs

This is my Custard Bread Pudding. Custard on the bottom, bread and fruit on top.

Phil doesn't like this sort of thing. Maybe I could make the top layer bacon. Or just take him to The House Of Prime Rib.

Last night I invented Shrimp and Bacon over fettucine, with a light garlic cream sauce. Sort of a Shrimp Carbonara. Got good reviews.

I have a really interesting oscillator phaselock problem with wild numerical things going on, vaguely like the old Tektronix "Random Sampling" thing, but I can't discuss it in public. Pity.

Looks good!

I regularly grill shrimp (and scallops) wrapped in bacon.

Most everything I do I can't talk about, until it's antiquated :-( ...Jim Thompson

. . .