Constant Current Source 1

[rmcdermo]

Please help!

I'm trying to build a constant current source that will drive between 5 uA and 200 uA through a variable impedance (50 kohms to 500 kohms) to ground. Once wired, the system will not need to be adjusted. (Multiple set-ups for different current levels will be made). I need an accuracy of +/- 5%. Tried working with a number of Howland type circuits using a TL074 op amp but have been unable to resolve oscillation/saturation problems.

Thanks

 

[nbucska]

give more detailed specs and FAX#. 200uA x 500KO = 100V
If V+ is less, it must saturate !!!
<nbucska@pcperipherals.com>

 

[IRstuff]

This subject has been discussed here:

thread240-38089

In your case, I don't see how you're going to get 200uA*500Kohm=100V out of the TL074. Can't find the specs, but it's supposed to be a low-noise JFET op-amp, and they rarely are designed for 100 V outputs.

Oddly enough, the following page suggests that the Howland circuit maximizes output impedance, which would seem to be the last thing you want for the output voltages you're trying generate.

Why not a conventional Wilson or Widlar source?

TTFN

 

[nbucska]

To use your Op.amp with 100 V, you must add a MOSFET
stage and use a source R for feedback.

You need stable Vref !!!

IR: Current source == infinite R ( I is independent from
load ). The VOLTAGE source is 0 Ohm.
<nbucska@pcperipherals.com>

 

[rmcdermo]

Allow make the following ammendment:

The testing I did was using an &quot;Improved Howland Current Pump&quot; configuration with a 15 V supply and impedances of less than 100K (max 150 uA). I wasn't testing at the upper end of my impedance limit.

However, even using resistances much lower (say 20K to 30K), I couldn't get constant current behavior (monitored by measuring the differential voltage across the load resistor).

Ultimately the circuit will be operated via battery power (around 12V). The upper-end impedances I gave in my original message are unreasonable (can you tell I'm not an EE!?). I would just like to get to the point were I can get the circuit to work for any impedance and any constant current.

What other information can I provide?

Regarding the Wilson or Widlar source: I am not familiar with these. I'll do some searching and get back.

Thanks for your continued assistance.

 

[IRstuff]

It's just that the Howland circuits I've seen have series output resistor to stabilize the output impedance of the op amp. The output current is I=Vin*Rf/(RiRs)

According to the following document, the overall accuracy is only 20% relative to Vin.

http://www.apexmicrotech.com/pdf/tech_seminar/currentoutputs.pdf

TTFN

 

[nbucska]

Engineering is the science of possible - the word &quot;any&quot;
is not in its dictionary. Any specification with the word
&quot;any&quot; is impossible.

If you have actual application, give spec's. This is an engineering forum, not for school projects.
<nbucska@pcperipherals.com>

 

[IRstuff]

If the circuit is oscillating, you need to clean up the layout and have proper power supply bypassing.

The circuit is a DC circuit, but the amplifier is a high performance amplifier, so a bad layout will convert it to an oscillator, pdq. It doesn't take much, you can even the venerable 741 to oscillate without trying too hard.

TTFN TTFN

 

[electricuwe]

There is a very easy way to build a two-terminal current source using a pair complementary bipolar transistors, two zener-diodes (or other voltage references) and a few resistors.

Bipolar transistors are easily available for the voltage range required and unless you operate the circuit at high ambient temperature the low currents shouldn't cause problems.

I will try to find some textbook describing that circuit in detail and post the reference within the next days.

 

[nbucska]

IRSTUFF:

Right on. Once an old tech has built for me a card.
Hi insisted on &quot;nice&quot; work, parallel wires, bent at
90 degrees.

Even 74LS74-s were oscillating... <nbucska@pcperipherals.com>

 

[CarlPugh]

A transistor, LED and two resistors make a pretty good constant current source.
Connect anode of red LED to +12V
Connect 1 k resistor to +12V
Connect other end of 1 k resistor to emitter of 2n3906
Connect 10 k resistor to cathode of LED and base of 2n3906
Connect other end of 10 k resistor to ground
Connect one end of load to collector of 2n3906
Connect other end of load to ground

LED works better than zener because tends to compensate for changes in 2n3906 with temperature

1 k resitor gives 1 mA. For 200 uA should use about 5 k resistor and for 5 uA should use about a 200 k (Instead of 1 k) resistor.

 

[Lewish]

Both the Analog Devices and the Linear Technology websites have app notes on building high accuracy current sources.

Why not take a look at those while you are trying to find out what a Wilson or a Widlar current source is all about!

 

[Lewish]

A Howland current source can be made to achieve 2% accuracy, but you need an Opamp with the divider and feedback resistors on the silicon with the Opamp.

Take a look at the datasheet for a INA105 and look at figures 19-23. You might find something useful there.

 

[IRstuff]

Got to go with Lewish here...

Another choice is the LM234 programmable current source:

http://www.national.com/pf/LM/LM234.html

It does pretty much everything you've asked for except the voltage range, it only goes to 40V.

AND it sells for $0.76 in 1K quantities. It so inexpensive that you could use it as the current reference in a discrete current mirror using a pair of matched NPN's.
TTFN

 

[electricuwe]

Unfortunately I can't find a suitable reference as stated in my post on Jan 8.

Maybe someone else knows this circuit and can post a reference. I used to work with that circuit but that was nearly twenty years ago.

I took a look on the Wilson and Wilard current sources and I would like to add the following:
These circuit are designed for beeing integrated on a single silicon chip and I do not think that they perform well when built with discrete transistors.

 

[IRstuff]

You'd have to start with a matched pair of transistors as a minimum.

But, you'd be able to accommodate the voltage range and the current range, particularly if you use a programmable current source as the reference current.

TTFN

 

[dumbo]

using ohms law the current is inversely proportional to voltage. The accuracy involves temperature coefficients.

If you start with tight voltage and compensated op-amp
the result is I=V/R +/- delta T
Some of the finest start very low and stable and amplify.

Example: you have 10uA that is nice and stable then the next stage will probably be nice also. But if you have 100mA and it is not so stable then later on you need a bandage approach which puts more temperature coefficients noise and more parts.

Maxim max4472 has an app note uses three common 1% resistors, an op-amp and a voltage reference it consumes 2.4uA runs very clean stable and very accurately on an ordinary battery for many years.

It is the next stage that requires drift and trimming.