Limiting Op-amp Output

Use the CMOS version, or think like an engineer ;-) ...Jim Thompson

If the integrator swings negative, it can zener the b-e junction, then all sorts of weird things happen. (Like the infamous photoelectric inversion thing.)

Actually, I don't know what happens to the collector current if you zener the b-e junction, but a lot of carriers get poked into the junction, so it can't be nice.

A clamp diode in the other direction fixes it, if clamping a bit below ground is OK in the system context.

Back-to-back zeners aren't a bad way to do this if you need to swing/clamp symmetrically..

John

Thou shalt jot down stuff like that :-)

Try it again and if it acts up you might have to add a "gain spoiler", like the emitter resistor Jim suggested. And do the reverse Vbe fix, with opamps you never know what can happen at turn-on of the unit. They can behave like radial engines, spewing some flames when rumbling to live.

CMOS version?

That limits the negative swing to 2 junction drops, which may be OK.

John

Yes. Inquiring minds would know how to use Google ;-) ...Jim Thompson

adjustable

Didn't see a TL431 in CMOS. Can you point to a datasheet of a reasonably priced and widely available one?

It doesn't limit the negative swing. Like the emitter the other side of the diode sits at VOUT which can be seen as a hard rail in this context. Since there is a resistor in the base it won't present any significant load.

When the output swings negative, the diode gets forward biased, as does the c-b junction of the transistor. That's 2 diodes in series.

You can fix that with another diode in series with the collector. Or maybe just let it clamp 2 JDs below ground.

John

Using the transistor the other way round gives you a nice temperature-compensated breakpoint--as soon as the transistor saturates, the base impedance goes down by as much as 1000 times, depending on the transistor. Even a Darlington will work, but it's uglier to temperature compensate. The other nice thing about this is that it doesn't put a gain element into the loop, which some of the other suggestions do.

The -5V bias should actually be a bit below -5 + V_CEsat of the transistor, maybe -4.7 or -4.8V.

This approach does put some fairly significant conductance in parallel with the capacitor, so it's a design question as to whether it's good enough. When you start adding stuff to make it better, it starts to be more sensible to use a window comparator and an analogue switch to disconnect the integrator's input when it goes out of range.

Cheers

Phil Hobbs

adjustable

TLV431. Widely shown with the old original bipolar schematic, but it's actually BiCMOS. ...Jim Thompson

You can fix them up with series diodes, i.e.

GND----RRRRR---* | | / SJ--------|----0 /

The leakage doesn't bother you until it gets to be about 0.4V/R. I still like the Widlar approach, even though this bandaged Zener probably performs better.

If you return the other end of R to the negative supply instead of ground, the Zener will be a lot stiffer.

Cheers

Phil Hobbs

The general problem of fast, sharp clamping with fast recovery isn't trivial. I've wrestled with it for decades. It's worse in an integrator than in an amplifier.

We need an IC that does this for us. HFA1130 type clamp-amps work in amplifier situations but not so well in integrators. HFA1130 itself has quirks. The ADI part clamps better, but is slower.

John

Zetex sells dual b-b zeners in SOT23, and dual diodes, so one of each plus one resistor makes a nice integrator clamp. That reduces integrator errors from the zener capacitance.

John

Maybe the transistor saturated and the op amp's current limit made it latch like that.

Cheers

Phil Hobbs

One approach that works okay is to use an RRO amp to limit the output swing, and use an active diode clamp to force the SJ to stay near zero.

Cheers

Phil Hobbs

Well, an IC is a nice option, but a comparator and SPDT analog switch can pretty much do the task. Depends on what 'fast' and 'sharp' mean, of course, and there's always some control feedthrough.

adjustable

Of course I know the TLV431. According to the ON Semi datasheet it is not a CMOS device. But even if it was, it does not solve the leakage issue I mentioned above, take a look at figure 7:

They are in parallel, in opposite direction. The forward biasing of the diode does not matter because the opamp output is only loaded with the base resistor against the reference. Can be kept well under a milliamp, no big deal. I don't see why it can't be used as is.