Current source

Howland is not especially suited to that wide range.

How accurate does it have to be?

How much voltage compliance?

What will the load be?

Can you ground one end of the load?

John

Even with a CMOS OpAmp that low end is going to be difficult. Board leakage alone will be biting you.

Further describe the application. ...Jim Thompson

[On the Road, in New York]

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
I love to cook with wine.     Sometimes I even put it in the food.

Maybe tell us more about what are you trying to do? Does it have to "cover" that range all in one go, or can there be range switching? How is it controlled? Etc. etc.

Best regards, Spehro Pefhany

--
"it's the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

Le 23/07/2011 16:59, John Larkin a écrit :

Hello John,

why not ?

Ohh a few percent could match.

10V

A pure resistor

Yes the this resistor is referenced to 0Volt.

Habib

Le 23/07/2011 17:12, Jim Thompson a écrit :

Jim,

I'm not expecting using a low end opamp for that kind of application. The argument of board leakage would be an issue at the time of PCB routing.

Thanks.

Le 23/07/2011 17:14, Spehro Pefhany a écrit :

Yes of course it should be 3 or 4 decades switching mA, 1/10 mA, 1/100,

1/1000. controlled by indenpendent switchs like adg1419 or so.

Habib.

Conceptually, you could make an instrumentation amplifier with a couple LMC6001s, and measure the voltage across a series (to the grounded load) resistor to control the applied voltage (maybe one or two additional op-amps). The series resistor could be switched to change the range. Say 100M/10M/1M/100K/10K/1K for 1nA-10nA,

10nA-100nA,.., 100uA-1mA

ADG1419 (and basically all other similar devices) have orders of magnitude too high leakage for this purpose, but you could use mechanical relays.

Remember that 1% of 1nA is only 10pA.

Best regards, Spehro Pefhany

--
"it's the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

This is a pretty standard way to do this:

ftp://jjlarkin.lmi.net/Isrc2.JPG

Both the opamp and the analog switch need to have low leakages. Use a CMOS opamp with low input bias current. As Spehro points out, low resistance analog switches are leaky; an old CD4051B would be better.

Since the mux resistance is in series with the sense resistors, that would cause some error, especially on the 1 mA range. You could trim the sense resistor on the high range or two to compensate, or use a dual mux (4052 type, or a couple of 4051s for more ranges) to sense the resistor voltage without putting the mux resistance in the loop.

ftp://jjlarkin.lmi.net/Isrc2B.JPG

PCB leakage shouldn't be a problem at 1 nA, as long as the board is clean. If you're paranoid, use the low end of the reference (+12.5 volts) as a guard. Picoamps and femtoamps require more care, but nanoamps aren't bad.

Lots of variations are possible, like varying the reference voltage to generate some of the ranges. Watch the common-mode range of the opamp.

Instead of the bandgap, you could put a resistor there and drive that from a grounded current sink circuit. That would allow DAC control, or practical low-side calibration pots if you need more precision. Sort of like this:

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

John

Le 23/07/2011 18:17, Spehro Pefhany a écrit :

Good idea (i had the same). I was afraid of the effects of the Loop Gain and Margin Phase of the instrumentation amp in the loop ... but switching the serie resistor to 100M..1K would maintain the loop gain to a resonable value.

I'll try this. Many thanks Mr Spehro

mmhh.. bloody leakage current !i'm pretty sure that mechanical relays will make my client screaming ! i will investigate on opto mos relays.

Once agaib many thanks.

Le 23/07/2011 18:42, John Larkin a écrit :

Hi John !

First of all many thanks giving me your advice on my request.

I take a look at the suggestion and i'm wondering if

1- Maintain +15V at an accuracy value is easy 2- +12.5V at the + of the op am isn't too near to the 15V rail in order to regulate I current on the whole swing (1mA..1nA)

So i will give it a try by simulation.

Thanks

[...]

You can get low leakage devices, not sure where to buy them in France though:

--
Regards, Joerg

http://www.analogconsultants.com/

The +15 doesn't have to be accurate at all. But if it is accurate, you can replace the bandgap with a resistor.

That should be OK to at least 12 volts at the output.

Things this simple don't need simulation!

John

CD4051 types, and even HC4041 types, have very low leakages. And there are some DG408 style mux's that are spec'd for very low leakage.

John

Leakage data is often (in contrast to some other data such as power for FETs ...) very pessimistic. Especially when there is only a max entry. I've had cases where I wanted to measure leakage and almost couldn't because it was miniscule, orders of magnitude lower than expected.

--
Regards, Joerg

http://www.analogconsultants.com/

Interesting part. Appears to be very expensive (>$20 each).

Despite the leakage of 30% of the signal on the lowest range, I think I can see a way of using this kind of part with a bunch of low resistance switches to do it, but it won't be especially simple.

BTW, it's not clear from the CAlogic or the Linear Systems datasheets, but the case lead is connected to the body.

Best regards, Spehro Pefhany

--
"it's the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

I have the schematic of the Agilent 34410A DVM. It's full of HC4066 type switches. It clearly wouldn't work if they had the max datasheet leakages.

John

Le 23/07/2011 19:17, John Larkin a écrit :

John,

Not sure to well understand but in this

ftp://jjlarkin.lmi.net/Isrc2.JPG

the current I = (15-12.5)/R

R is the resistor on the Pmos source.

Why did you say +15V does not to be accurate at all ?

Habib.

30% is the stated max value. It is usually much, much less in most such devices. But this gets into a range where you'd seriously have to consider humidity levels and such.

The SD5400 series also has a substrate connection and woe to those who treat that haphazardly or, worse, forget about it. No joke, I've seen that ... "Well, every time the tech takes it into his hands it works and keeps on working, and then it gradually quits".

--
Regards, Joerg

http://www.analogconsultants.com/

Yep, even the big guys do it, use parts outside the ranges the datasheet says they can do. A lot of progress in the world wouldn't even have happened otherwise.

--
Regards, Joerg

http://www.analogconsultants.com/