"JFET stickiness" in constant current source

[David Hess]

I also have the Linden T Harrison book and have mixed feelings about it. The relentless flagwaving irritates and he's clearly not seriously technical. But he clearly has technical people around him. Perhaps he's a mananger in a technical environment. Despite my reservations, the book makes some good points and I have acted on tips in there.

I think it is one of the better books with a good compromise between math and practical examples.

For instance, rather than temperature compensating an LM334 with the data sheet's recommended diode, I now use an MPSA42's base-emitter junction. Harrison claims this is a better match to the LM334's behaviour. Maybe it is, maybe it isn't. But what's definite is that you can epoxy the flat face of an MPSA42 to the flat face of an LM334, achieving thermal bonding that is way better than can be achieved using a glass-bodied diode.

Any small signal high conductance low leakage non-gold doped silicon diode should work well.  When the LM334 was first released, this diode was the 1N457 and it is still readily available.  The 1N458 or FDH300 should also work well.  A base-emitter junction of a non-gold doped small signal transistor makes an excellent high conductance low leakage diode, albeit at low voltage, and like you mention, it is easier to attach for better thermal tracking.  Little s-clips or aluminum blocks used to be available for thermally bonding TO-92 and similar small packages.

[free_electron]

i use the 1n457 with lm334 to stabilize reference zeners.