Help! - Constant Current Sources

[ehill03]

Hi All,

I'm currently working on a signal conditioning/isolation design for our products. Part of the design utilizes a constant-current source so that I can deal with bi-polar voltages. I am using a modified Howland, which incorporates a JFET and a LM313 instead of all the resistors. This circuit outputs 100uA and I need stablilty to as close to perfect as I can get (preferably less than +/-5%). The problem is that the LM313 seems extremely sensitive to temperature and does not seem to hold my calibration settings.

Does anyone know how to build an ultra-precise current source? Do any pre-manufactured ones exist for the PCB integration level?

 

[melone]

Try

http://www-s.ti.com/sc/psheets/sbva001/sbva001.pdf

 

[Lewish]

Well, sorry, but a Howland isn't going to do it for you if you need high stability or accuracy. Way too many sensitivities to try to account for.

A Widlar topology has much better stability. And there are other topologies better than that, I just can't remember which ones at the moment. Senior moment! Argh!!!!

 

[ehill03]

Thanks!

I think the Burr-Brown part will work much better than what I have now. I guess I'll also have to go through my small "library" and look up the Widlar sources too. If you can think of anything else I should know, please feel free to beat me over the head with it.

Sorry for the dumb question, but 90% of my work now is Microprocessor/FPGA/Digital logic. This is the first chance I've had to mess around in the world of current sources...

 

[dumbo]

To start with
What kind of voltage reference do you have ?
Or which Vreference would you use to make a CC source ?

 

[jbartos]

Suggestion: Visist

http://www.sandtron.com/scan/modselgd.htm

http://www.yacoub.de/e_modul4.htm

http://www.chipcenter.com/analog/products_700-799/prod716.htm

etc. for more info

 

[ehill03]

Thanks again for the info! I've had a chance this week to implement the Burr-Brown REF102a part. Even with a cheap op-amp, this circuit was far better than what I had. Before I was seeing 50-100mV of drift, now it is almost not measurable with our Tek scope!!

<The original circuit was a modified Howland that had the anode of the LM313 tied to the op-amp (OPA177G) output and the cathode tied to the negative input. A 12k resistor was then connected between the output and the positive input. 100uA of current then flows out of the positive input connection.>

You all have managed to save my job, thanks! ;-)