I'm probably showing my age calling it a "bandgap reference" - they just seem to be called voltage references these days.
Originally they were only available in 1.2V, (the bandgap voltage of silicon) but now there are versions for any voltage you might fancy.
The basic premise is to take two semiconductor junctions and make sure that the current through one is ten times the current through the other. Because of the exponential relationship between I and V it follows that if I2/I1 = 10 then V1-V2 = 60mV.
There's a good explanation in Horowitz & Hill and also on Wikipedia https://en.wikipedia.org/wiki/Bandgap_voltage_reference

 

Sounds like something I've been looking for years. Thank you.

That would have worked on my original concept.

 

Look here: https://sound-au.com/ism.htm They don't explain the FET version very well. They brushed it off.
turn that into a low leakage fast 38 kHz LED driver at 50-100 mA that I can understand, I'm interested.

Analog devices has a 2T current source. The LM334 is very temperature dependent.

I made a +-50 mA variable current source.

 

Sounds like something I've been looking for years. Thank you.

That would have worked on my original concept.

I remember when this one
https://datasheets.maximintegrated.com/en/ds/ICL8069.pdf
was about the only one available.

Did you try the old JFET circuit?

 

Gotta have a JFET to do it. I don't see a value added buying one.

 

Two things ...

1. 555 output drives a Howland current pump opamp circuit (which is bi-directional by nature), and that output drives the timing capacitor at the 555 input.

2. As above, but the other half of the dual opamp package replaces the 555 as a hysteretic comparator.

https://www.ti.com/lit/an/snoa474a/snoa474a.pdf

http://www.philbrickarchive.org/1964-1_v12_no1_the_lightning_empiricist.htm - page 7

ak

 

As I said, I'm not after a circuit per se but a really low CC circuit. Parts are on order, so I will get back to this thread.

 

To drive a sensor requiring a very small current <500 uA, I used the LM 334 with one of the reccommened circuits in the datasheet to reduce the effects of temp / voltage variations.

That was maybe around 20 years ago. Should look for details if required. Circuit was extremely simple.

 

Hello,

Perhaps the following page will give you some ideas:
http://www.recursion.jp/prose/bicrrnt/index.html

Bertus

 

So, the parts arrived yesterday. I'm testing this design 1st, because the parts are something a home experimenter might have. I'll post the results when I have them. I give this circuit the smallest chance of working, but their will be more contenders.

​

The Data

Volts Ic Q1bv
2.0 0 50.6µa 0.442V
3.0 90.0µa 0.490V
4.0 140.7µa 0.524V
6.0 250.0µa 0.571V
8.0 359.0µa 0.604V
9.0 1033µa 0.617V

I would call this one a bust, NEXT she yelled.
Q1bv is Q1 Base Voltage.

 

OK. Referring to post 11 I will try this circuit...

​

Guess I'll be doing the math check tomorrow (this is not the final circuit.)

 

Disappointing
2V 12µa
4V 19µa
6V 25µa
8V 51µa
10V 63µa

It didn't help when I blew a 600ma fuse in my DVM. So I dusted off a spare, ordered more fuses from Amazon. Tomorrow I will try a 16KΩ resistor and will post the results again.

 

Here's a sim using the 2 BJT constant-current circuit in a bridge configuration.
The circuit needs a supply voltage of at least 7V for reasonable linearity.

 

See

 

Helps if you install the chip correctly, so after reorienting it correctly and updating my personal Paint CAD files it now regulates to 0.8V! I did have to adjust the formula empirically.

It seems to be (for this individual chip):

\(Icc=\frac{0.06513}{R1}\) or \(R1=\frac{.06513}{Icc}\)

​

So, for 620Ω I got a flat 106 µA across the board. Regulation stopped right under 0.8V, so a conventional diode bridge should work down to almost the full range of the 555 CMOS or not. For complete results I will test the bandgap reference next.

We have a winner in any case! I love the simplicity!

I attached a worksheet if anyone is interested.

 

I've always liked this circuit, each half puts a constant current through the opposite reference.
For really low current you can't beat a bandgap reference, but I think LEDs might workView attachment 233391
For the 555 oscillator you need a bidirectional source, but if you use two similar circuits in anti-parallel you risk their not being quite similar, but you can put a single circuit inside a schottky-diode bridge.

So I'm just about ready to try this. to limit current in only one direction what Resistor values do you recommend? Which one programs the current? So if I use x100 for the other would that work?

 

So I'm just about ready to try this. to limit current in only one direction what Resistor values do you recommend? Which one programs the current? So if I use x100 for the other would that work?

Both R values are the same. The current in each side is half the total, which is (Vref-Vbe)/R.
So for 1.2V reference and Vbe of about 0.6V, total current = 1.2/R

 

I was really tired last night, and not thinking straight. Starting over.

 

OK, Figuring a minimum current to bias the LM4041 is 60µa And minimum bias voltage is 2V. R = 13KΩ, call it 10KΩ. So I will try this schematic...

<drawing>

 

Next experiment:

​