Zener Vs PN diode difference in forward mode

Probably the zeners have smaller junctions than whatever regular diodes you've selected. You could compare the forward drops.

Another way to reduce distortion is to add a base-emitter resistor. At low levels, the driver stage gets to the load through the resistor; at higher levels, the transistors help. This needs lots of loop gain to work well.

This is fun too:

Try some manufacturers' models for the zener instead of the hokey/idealized LTspice model. ...Jim Thompson

I'm surprised you use that. It's thermally unstable - ok if you keep the output trs cool, but that's inefficient use of parts for a linear amp.

NT

What's thermally unstable about it?

Easier to have tracking temperature compensation with diodes; zeners go the opposite way.

A bit of emitter resistance is usual, also.

A few important differences: the low-voltage Zener mechanism has a similar temperature coefficient to your Darlingtons' base bias, but 5V and up Zeners operate on avalanche (wrong sign of tempco). There's lots of diodes available, fewer choices for Zeners, fewer package options. Best performance requires your compensation diodes to be heatsinked right next to the Darlingtons.

Most important, the spec sheets of a typical diode don't cover forward voltage (it's not a controlled parameter) and you really WANT control down below a tenth of a volt (or an order of magnitude in bias current). You'll want to have a trim resistor and a procedure for setting the bias.

Another trick is to use big emitter resistors and some moderate quiescent current, and put big diodes across the emitter resistors. No need to tweak Iq and no chance of thermal runaway.

1. Assume class AB conditions (and all that implies).
2. Consider Vbe(T) dependence.

Indeed, because of the particular form of #2, such a circuit cannot be stable; it'll always have a runaway point. The best you can hope for is making that temperature out in the destruction region (150C+), but that doesn't prevent runaway, that only avoids it. Besides, you can't do that /and/ have class AB operation.

If there exist op-amps with supply bias IPTAT, you'd be set, but I haven't seen any that behave anywhere near that.

Another fundamental view: there's simply no feedback between output current and op-amp drive. If they were at least mirrors rather than CE BJTs, there would be a weak claim of that, if not a properly closed loop. As shown, output current is unbounded and hFE dependent, a huge no-no.

It's an old circuit, but that doesn't make it a good circuit. The 555 and

Tim

Models. Zeners are usually much /larger/, so have lower Vf. They also have thinner junctions (well, depending on which Vz you've got), reducing Vf even further. Test a few with a DMM and compare. :)

(Large by area, that is. Check the C vs. Vz plot in the On Semi 1N4733 etc. datasheet, for instance.)

Ideal is a couple diode-strapped (C to B) BJTs, running at similar current density as the output transistors. That will give nearly the same Vbe, and can be rearranged to show that it is, in fact, a current mirror configuration, which should give you ideas about biasing, sizing the VAS, controlling thermal drift and so on. :)

Tim

Vbe on the transistors should be small, maybe a couple of tenths of a volt, at no output. The transistors are quiescently OFF. The opamp drives small swings all by itself.

It has no possibility of runaway, if it's done right.

The best you can hope for is

I don't know what "class AB" means in this context, except that it sure sounds like I don't want it.

The transistors turn on exactly as hard as they need to, when they need to, to help the opamp drive the load. The load feeds back, through the opamp output back to the base drives.

I sold a couple thousand NMR gradient drivers with a very similar circuit, power fets driven from opamp supply pins. Worked great.

Tim totally misses the 'feedback' loop. I won't define it... let's see if Tim can get it >:-}

(I used a one-sided version of this, with uA741's no less, to make linear regulators for the TOW missile, 1970-73)

...Jim Thompson

Yeah. If it _were_ biased class AB it would be thermally unstable, because even with DC voltage feedback around the whole thing there's no way to bias the booster stage quiescent current reliably as drawn; the collector current could be anything. You'd have to add diodes in the supply lines and emitter resistors - which would compromise the output swing.

Whew. Good thing it isn't!

AND compromise the PSRR of the driver unless you used some kind of boostrapping back to both bases

AC stability needs to be considered, of course. Minor detail.

It works sort of like a very precise Sziklai pair, or two of them actually.

The helpers could be Darlingtons, too. Or, in the 21st century, mosfets. Or cascodes. Lots of variants.

Yep.

Exactly.

My TOW version used a PNP Darlington. ...Jim Thompson

Derp, there's two variants of this, one with the output _grounded_, that I had in my head for some reason!

In any case, I correctly answered "why it might be unstable", and you correctly noted that might be undesirable. ;-)

Tim

That's one. Or the output can go to a resistor to ground. Or to a voltage divider from the output, which provides higher voltage out but some feedback. But none of those allow the opamp to drive the load directly at low level, which brings back the class-AB biasing dilemma.

Here's the output stage from one of my old gradient drivers. The load is a coil, so we want to be a current source.

Vaguely the same idea, the signal path being from an opamp's supply rails.

shows you how to get the forward voltage you want - R1 and R2 should have a pot in between them. Obviously the Vbe drift with temperature gets multiplied by the same ratio as the Vbe drop, so you don't want too much multiplication.

Transistors do tend to have tighter Vbe specs than most diodes.

Yes, it does seem that the modelled zener forward drop is fixed, whereas a normal diode has a current dependency. That answers my query, thanks.

Yes, got that.

I did try something like that a while back, but couldn't get it stable over the temperature range (0 - 180'C with the output transistors at around 200'C) so went back to the more standard configuration.

Cheers