Last current pump question, sorry!

There are many programmable current-source configurations, and most of us believe the Howland is usually not the most suitable one to use. Go explore some of the others.

Or if you'd like advice about which one might be better, tell us more about exactly what you're trying to do.

If R1, R2, R3, R4 are >> R5, you'd get pretty good correlation (not absolutely perfect, but reasonable), by mirroring the collector current of the NPN with a PNP mirror that uses a "helper" to supply the base current, and emitter ballast resistors.

If you have a large dynamic range there are other ways that take another OpAmp, but don't require a full Howland architecture. ...Jim Thompson

Try this...

But, as Win advises, WHAT are you trying to do? There are many easy (and stable) implementations.

Howland Pumps were useful in the days of no PNP's, see, for instance, my patent...

in the Patent Listings near the bottom of my Home Page... circuit actually integrated ~1968, Motorola scurried to patent all my stuff after I up and laid myself off >:-}

There is a "dot" junction missing from Figure 5 ;-) ...Jim Thompson

It's a little difficult to explain, but essentially what I'd really like is a voltage controllable current source with two separate CV inputs. The Howland seems to have that advantage - I'm not sure how to do it with the more standard resistor transistor current source without using a second op amp as a summer.

If the Howland is so objectively bad, why does it seem like there has been so much written about it? I'm not necessarily disagreeing, just curious.

Thank you for the reply, that looks interesting. Please see my message to Win for further information.

?! Differential, additive, multiplicative? You can only have one output!

Which I'm also not sure about because you've been talking about two outputs, as well...

That's fine, and probably has better accuracy since you're less dependent on differences in resistors. Or instead of adding voltages, add currents, or subtract them.

Seems like the kind of thing academics would like?

Very easy to say "resistors are equal" and poof let it be... doesn't work so nicely in the real world.

Also, sourcing current through a resistor means you lose I*R of voltage compliance range.

And the slowness of the amp (finite BW plus slew rate) manifests as capacitance in the output, which will be considerable compared to a discrete version.

Tim

Howlands are commonly used as audio drivers for duplex circuits. A Howland into a hybrid, to the line is a very common configuration.

They do have the nice feature of being four-quadrant devices, quite unlike the usual transistor plus op amp approach.

Cheers

Phil Hobbs

Differential.

I'm most interested in temperature stability than the absolute output impedance of the current source. The load impedance is essentially fixed. I think I got into a discussion here about how the Howland compared in that regard but I am blanking on the results.

Could you elaborate on the method that requires another opamp? I think I'd actually prefer to not have to use discrete transistors.

Actually uses an OpAmp and one PNP transistor, but is more accurate/more headroom than my transistors-only approach.

Again, the question is: WHAT are you trying to accomplish? ...Jim Thompson

I've been experimenting with using diode bridges as variable resistances for large AC signals. So, the impedance across two nodes of the bridge is proportional to the DC current sourced/sunk through the other two nodes.

So Id like a set of current sources/sinks that can bias up the bridge at some DC current level and then move the current up and down around the bias point in response to an AC control voltage. The input and output nodes of the bridge for the signal need to be kept at 0 potential (with split supplies) and not source DC current into the bridge so that the signal can be applied without disturbing the bias. And I'd like multiple bridges whose dynamic impedances track a single CV. And I'd like the dynamic impedance of everything to be very temperature stable.

It's not an easy task, but I seem to have a solution that may work using howland pumps. But for the sources for the multiple bridges it currently uses an individual Howland to look at the same CV and source current into each bridge, and that's wasteful because for the others I really just need a mirror that follows the first.

Sorry if you thought I was beating around the bush or tryING to obfuscate what I'm trying to do, it's just that I'm on a tablet much of the time now, and my thumbs get tired!

There is an opamp circuit I played with once that is basically a current controlled current source. I tried googleing that and didn't find what I was recalling. Resistor ratio's scaled the output current to the input. (Perhaps someone here will know what I'm talking about. :^)

George H.

Well it did no look like the "obvious" current mirror here.

Geo

Maybe this:

or its equivalent made with four matched transistors and a few op amps.