Been a while since I messed with integrators. When one is inside a feedback loop to say, act as a "bias servo" to ensure some particular node is at ground potential, is it usually necessary to put a large value resistor across the integration cap to limit the LF gain?
It seems to work in Spice ok without it, but I haven't prototyped the circuit in the real world.
I've never needed to do that, although I've seen it in some older circuits.
I suspect that it has to do with op-amps that can't come off the rails cleanly, or that suffer from phase inversion when their differential voltage gets too high, or other such less than ideal behavior.
Usually not; the overall feedback loop should keep the integrator in its linear range.
Sometimes you'll see a clamp of some sort across the integration cap, back-to-back zeners or equivalent. That can keep the integrator from "winding up" from a sustained loop error. Windup can make recovery nasty.
A clamp can also limit the influence range of the integrator, which makes sense in some systems.
Sometimes there are two stable solutions and you need to encourage the control not to seek (say) the negative solution, but that's more of a limiting thing than a degeneration thing.
Copy of a copy of a copy....
Thank you all for the advice. Sometimes when I ask for advice here I'm often not able to reply as quickly as I'd like as I'm mobile a lot and on a tablet, but I do see the replies and appreciate it.
Thanks for the replies. Sometimes I don't respond to replies to queries as quicky or as often as I'd like here, because I'm mobile a lot these days and often using a phone or tablet. I can't thumb type anywhere near as well as the kids these days. But I do check the group regularly and appreciate it.
So what I want to know, from someone with lots more gray hair than me, is why in heck was this a somewhat-common practice in designs in the 70's and 80's?
Well, yes -- but what was the reasoning of the guy who designed the prototype?
What's the input offset voltage/current of the Spice model? In the real world, without the resistor, any DC offset will get integrated and the opamp will rail.
Presumably the integrator is inside a closed servo loop, and its input, the loop error, is actively driven to zero.
Yeah if it's inside a loop then an R across the integrator C just leads to a (small) offset. (In my experience.)
George H.
It makes the loop gain finite, in other words it's not a PID loop any more.
There might be systems with stiction or backlash or something where infinite gain would be undesirable and in fact useless, so limiting the integrator gain might help. Especially as a kluge.
Not inside a loop it won't.
Just a wild guess: Taming the integrator at high (and varying) offset voltage.
Have something to add? Share your thoughts — no account required.
Ask the community — no account required