Current source [Result]

Give or take offset voltage etc. of the op-amp.

Sure, unlikely that the 1G will be anywhere near that accurate anyhow.

But a more general solution is to alter the ratio so that you get the desired curent.

If the top resistor is Rt, the bottom is Rb and the series resistor Rs, then the desired top resistor is just

Rt = (Rb||Rs)*(V/(i*Rs) - 1)

For 0 < i < V/Rs

Eg. Rs = 10M, i = 100nA, Rb = 1M, then Rt = (200/1.1)K

If you set i = V/Rs, then the divider disappears (Rt = 0) (eg. use Rs=1.2G in JL's example to get 1nA). The divider may be worth keeping though- a nice stable voltage divider incorporating a shunted 10K trimpot could be used to trim a 1nA current.

Best regards, Spehro Pefhany

Why am I not surprised? Works for me ;-)

What in the world have you done to your system to make it so... so... so... function-proof ?:-) ...Jim Thompson

That can be very hard to take. When I had the heart attack in 1998, Aaron came to the hospital looking like he was the one who had the heart attack :-( ...Jim Thompson

Some of the work I do requires certain precautions. Think of it this way: Riding a bicycle with a helmet is uncomfortable because it can make you sweat profusely. Riding a bicycle without a helmet can ...

Sheeeesh! ...Jim Thompson

Digikey carries ceramic ones that have inside threads:

Also comes in other lengths such as 1/2"

In Europe you can buy relatively cheap "chandelier terminal block" that are ceramic:

If you live in or near a larger city a big electrician supply store should have those. In France it's probably called "bornier ceramique" or somthing like that. I like the shiny white ones, they pick up less dirt.

It's not an approximation to anything. It's a bootstrap current source.

to a more traditional IA bootstrap

Traditional? Bootstrap current sources pre-date the opamp. WWII radars were full of them, and I'm sure there were earlier examples.

What you've drawn is equivalent to a Howland current source, and has its problems.

John

You'd be surprised how many companies do not allow the use of certains scripts and stuff within their premises. There's a whole slew of vendors that misses out on sales because of that. And they will never know that others are eating their lunch.

Right. It's supposed to be a follower. I sort of drew it upside down.

How? Some people can't see it. Never figured out why.

John

One nice thing here is that the error in the reference and its divider and the offset voltage of both opamps all contribute to a single current accuracy error, which a single trim in the divider can fix. 1 volt is sort of nice, too, because you can use standard values, like

There are no resistor matching requirements, and only one low-bias-current opamp.

It's the 1 mA range that has the accuracy problems!

John

Thanks. I fixed it.

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

Some fraction of the errors we catch in checking designs is either the inputs or the power pins swapped on opamps. That has propagated to a few actual boards, too. We try to triple-check this nowadays.

Inverting amps like the - input on top, and followers look best if the

• input is on top. So if you Y-mirror the symbol to get the inputs nice, you flip the power pins, too.

John

Foothill Flea Market, obviously removed from old Tek scopes.

But at 1 nA, a PCB will be fine. Just don't use water-wash flux.

John

We have the same problems. I almost always put '-' on top. There are a few times where '+' on top looks better. I use the alternate representation in the schematic symbol library for that, so the power stays in the right place.

Normally you should at least provide series resistors at IN+ that are equal (or at least close) to the impedance in front of IN-. It's a cheap way to muffle the effects of the input offset current. Which with this opamp is almost as large as the input bias current. And there's no guarantee that this offset current will be the same for U5 and U6 because the LT1012 comes in singles.

Of course if the load R9 can be hi-Z that will be a wee problem with this circuit.

What accuracy at 1nA would you think it's fine for? 1%? 0.1%?

Best regards, Spehro Pefhany

I'd expect a clean board with a decent layout to have under 1 pA leakage. Guarding would be a good idea if high accuracy were needed. The output of the CMOS follower is the ideal guard drive.

If the range resistors are switched with CMOS multiplexers, they will probably dominate the error on both the 1 nA and 1 mA ranges.

John

In a lot of modern opamps, there's no particular reason why the input bias currents would match, even in sign. Lots of precision bipolar amps have active bias cancellation whicg tries to drive each input current to zero. CMOS amp bias is just leakage.

In this case, for the lower opamp, there's no "correct" value anyhow.

Which with this

I rarely bother to try to match the impedances of opamp inputs.

John

If they have active cancelling built in it might indeed not work.

For hi-rel stuff I usually do, especially when it has to work over a large temperature range.

I use Firefox and NoScript, yet never have the weird issues you seem to have all the time. ...Jim Thompson