Adjustable constant current source 0.5 uA microAmps to 2 Amps

What's wrong with a TIP31? Put an op-amp around it if you need. I'm willing to bet leakage will be plenty low (as long as you keep it cool). I've used a 2N4401 current mirror from nanoamps to 10mA, made a function generator that ran as slow as 1Hz and as fast as 4.5MHz. :- D

Tim

Hmm 2A at 10V is getting up there. Is it single sided? Then an opamp driving a transitor might work. The LM675 always seems to have some wiggies near zero crossing. There's an OPA power opamp I like, not sure 548 is the right number.

George H.

Hmm 2A at 10V is getting up there. Is it single sided? Then an opamp driving a transitor might work. The LM675 always seems to have some wiggies near zero crossing. There's an OPA power opamp I like, not sure 548 is the right number.

George H.

ic

I like the LM12. It's input bias current is to high - at 1uA max - for it to be a one-stop solution, but a monolithic dual FET front end might get it there, if you can find one - Farnell have still got a few Vishay 2N5912 parts, but only until the stock runs out.

-- Bill Sloman, Nijmegen

Yes, it is single ended. And it is DC.

Be>>> I need an adjustable constant current source than can be set to

Something like that circuit, with a MOSFET follower (and probably some compensation), ought to work, assuming you're talking about 0.5uA resolution, not generating an accurate 0.5uA current. If it's the latter, I think you'll have a great deal of trouble covering that range. If your sense resistor is 1 ohm (4W at 2A) then you'll only have 500nV at 0.5uA, and the offset of the LT1167 is 40uV (A version) which represents 40uA. 500nV is possible, but will be noisy and you might have to wait something like half an hour after running at 2A for the bulk of the thermal gradients to go bye-bye. Best avoided if possible.

If you can put switches in there, two 2000:1 circuits (or three 200:1 circuits) are a lot easier than one attempting to cover 4E6:1.

For a unipolar current source, it's silly to float the sense resistor on top of the load, then work hard to take out the resulting errors. It's more reasonable to nail one end of the sense resistor to a supply rail.

But floating the sense resistor does sell a lot of high-performance instrumentation amps.

Something like this, maybe:

ftp://jjlarkin.lmi.net/Isrc.JPG

To get your huge dynamic range, you'll need chopper opamps and maybe a trim or two, or switch Rs to give two or three ranges.

John

I should have mentioned that more than one range is fine. I do need to generate 0.5 microamps accurately.

EDN has a Design Idea from 5/14/09 showing transistors for higher current. with some errors in connections of IC2

Yes, John it is written by a mfr of >But floating the sense resistor does sell a lot of high-performance

Fig 2 is insane. It must be drawn wrong. Most EDN "design idea" schematics are.

John

Nowadays EDN is so short of contributors that most "designs" come from third world countries ;-) ...Jim Thompson

The inputs of IC2 are obviously supposed to connect across "RL".

It's perhaps of some interest because it is bipolar.

I do wonder how stable it is..

These days you don't have to flip through a lot of ads to get to the Design Ideas..

I wonder how many employees they can support per page of advertising.

In addition to the overall loop getting schitzoid, I'd be concerned about the complementary transistors having delusions of being RF oscillators.

John

Is it for testing where manual adjustment is allowed? If yes, you can use a transistor controlled by a potmeter.

How much money can you spend? I like the lm395? 'npn' power transistor. It's got thermal shutdowns if you over drive it so it's hard to kill. 2A at 10V leaves 5V to drop on the current sense resistor. So 2 ohms might work... at the low end you've got to set the voltage to a microvolt to get your 0.5ua spec. That might be doable?

George H.

Yeah, I if you can float the load, then tie the sense resistor to ground. If not you have to tie it to the positive rail. And then use the positve rail as the voltage reference...(can be a pain.) If you want it to be voltage programable then the easist is to use ground as the reference.

Switchable resistors as John L. suggests are a good solution to the resolution issue. George H.

..

OK, I was going to use one opamp and one transistor. That does look crazy. Every once in a while I find something that works in the EDN archives. but hardly anything from this decade.

George h.

Now _that_ is a riot. Two instrumentation amps, and all it's measuring is the offset voltage of the buffer. It does null that out reasonably well, but not as well as an op amp and a feedback resistor....

Cheers

Phil Hobbs