Yeah, that wandering IDss (in a large sampling) is so inconvenient! I have designed circuits to auto-compensate for that, but they use so many parts that, "good enough" methods are obviously cheaper.

 

WBahn
It would be an active load for that stage.

Yes I expect an active load type of configuration but not for the output transistors, which are emitter followers.
The active load is for the preceding voltage amplifier stage, which is shown as a single transistor.
This needs quite high voltage gain and the current source/active load linearizes this and helps with the gain requirement.

 

My guess (and I don't know this for sure) would be that the availability of JFETs on a bipolar process would be even harder to come by. Although I know that National developed the Bi-FET process that allowed them to fab reasonable JFET devices on a bipolar process. I believe that nearly all bipolar processes that support JFETs are derivatives of their work (in the 70's). But I don't know how widespread such processes are. Even if the standard process physically permits JFETs to be realized, there would be additional masking steps, which are very expensive, that most people are not going to want to pay if they don't have to.

Here is an odd use of a J-FET in a bipolar transistor circuit:


Any idea what it does?

 

Hey, does anyone know the component I've marked out on this schematic? It's TI's LM386, my best guess is current source? I'm trying to analyze the internals of how this amp works, I'm unsure as to what's going on during the positive half of the waveform input...

http://www.ti.com/lit/ds/symlink/lm386.pdf
View attachment 82997

Its a current source - in that type of circuit, the *ACTUAL* current source is usually ground referenced and that particular symbol represents the current mirror that is driven by the real current source.

 

Here is an odd use of a J-FET in a bipolar transistor circuit:
View attachment 83276

Any idea what it does?

That P-channel JFET with its drain/source in parallel with the collector/emitter of the NPN with its base connected to the output, is a startup current source. Once the MC1403's output is up and running, the NPN takes over as the main current source for the current mirror up top (the two PNPs and the two 2.0K resistors) and the P-FET drops out of the picture. Bandgap voltage references usually need some kind of "kickstarter" circuit like this, at least according to the silicon wizard at National Semiconductor who was valiantly trying to explain bandgaps to me (with only partial success...).

 

That P-channel JFET with its drain/source in parallel with the collector/emitter of the NPN with its base connected to the output, is a startup current source. Once the MC1403's output is up and running, the NPN takes over as the main current source for the current mirror up top (the two PNPs and the two 2.0K resistors) and the P-FET drops out of the picture. Bandgap voltage references usually need some kind of "kickstarter" circuit like this

Good explanation.

The interesting thing about this commercial circuit is how simple it is.
Most of the learned papers develop far more complex circuits to do the same job.

 

The interesting thing about this commercial circuit is how simple it is.
Most of the learned papers develop far more complex circuits to do the same job.

My guess is that most of the additional circuitry is for curvature correction. Basic, no-frills bandgaps like the MC1403 (which at this point has got to be at least 30 years old) almost always have an output vs. temperature characteristic which looks like an inverted "U". The MC1403 data sheet, here:

http://www.onsemi.com/pub_link/Collateral/MC1403-D.PDF

shows just such a curve in Fig. 6. More modern (and complex) designs have extra innards arranged to compensate for this quadratic error characteristic. Some even have third-order correction, I've been told.