Sometimes for my own edification I like to recreate the functional diagrams from IC datasheets in my iPad simulator and tinker with them to figure out how and why they work. Spot the current mirrors, the diodes, the diff pairs and you're usually off to a good start; figure out which parts make the device work and which parts are protection, compensation and correction for non-idealities and you've pretty much got it. But I'm stumped here and hoping someone with experience/industry knowledge can help.

This little bit of biasing circuitry is not uncommon; I've seen versions of it all over the place in op amps. But it doesn't work *at all* in my simulator. On the original datasheet for this particular instance (the LM318, https://www.ti.com/lit/ds/symlink/lm318-n.pdf ) the 500Ω resistor is marked as trimmable so I'm guessing it was tested and modded during manufacture. The trouble is, changing the value of that resistor--at least in my simulator--does absolutely nothing. It doesn't modify the output (the long wire passing 36.75μA) in the slightest. The only thing that *does* seem to change it is altering the specs of the JFET. There's a narrow range of threshold voltage and gain where I get the 37ish μA output and going above or below that I don't get anything at all. Given that the LM318 is famously one of the first fast devices I find it kind of hard to believe it kicks things off with a diff pair that under-powered...but hey, it's also possible I don't know what the heck I'm talking about. (There's also obviously no clues about the proper specs for the JFET which I know are grown a different way from the bipolars so maybe they act really different than discrete ones?)

Any hints, insight or links much appreciated!

 

Before anyone suggests it, yes I have tried very hard to Google this topic. If there's an answer out there it's swamped by the thousands of pages explaining how to create basic op amp circuits.

 

I think it is a bandgap reference. You can google that successfully, but I found an explanation here.
Bandgap reference across fixed resistance = constant current source that doesn't vary with temperature.
That's what you need as the tail current to the long-tailed pair.
It probably doesn't simulate because the JFET is peculiar. What voltage does your simulation give on the source pin?
Try replacing it by a constant current source and see if it works then. Or try JFET models with higher Vgs(cutoff)

 

I think it is a bandgap reference. You can google that successfully, but I found an explanation here.
Bandgap reference across fixed resistance = constant current source that doesn't vary with temperature.
That's what you need as the tail current to the long-tailed pair.
It probably doesn't simulate because the JFET is peculiar. What voltage does your simulation give on the source pin?
Try replacing it by a constant current source and see if it works then. Or try JFET models with higher Vgs(cutoff)

I agree that it's *some* kind of voltage reference, but I don't *think* it's a bandgap. Bandgap (at least as I understand it) is derived by multiplying the vbe of a transistor by a ratio that puts it in the 1.2V range and then thermal effects keep it there if it's fed the right current. That doesn't SEEM to be happening, here, at least not in any way I can discern, and it seems very imprecise to feed a bandgap through a plain old emitter follower that isn't compensated anywhere else.

I think you're probably right that the FET is weird. This simulator is pretty simple; it just lets you set the Vgs(off) of the FET and the gain. I have dragged Vgs(off) all the way up to 24 volts and all the way down to zero. The only place the circuit has output is with a Vgs(off) around 1V. I was able to get a very small amount of additional output by lowering the gain of the FET to almost zero. Are these things clues? I don't know. It still doesn't do anything much when I change the 500Ω, which was apparently important enough to be a trimmed value.

 

What I have learned from an actual applications engineer for one semiconductor company,, and have verified thru numerous conversations, is that the circuit in the data sheet IS NEVER the actual circuit inside the IC device. There is almost always a lot more. The temperature compensation is very seldom even shown, and greatly simplified iif it is shown. The presentation of a circuit is to help correctly interface to it.