Stable Low Power Current Source/Sink without an ic

[Fcolor04]

Hi, I'd like to design a temperature consistent current source or sink that uses minimal current to drive output transistors,

I thought of using an op amp with pass transistor or current mirror with Darlington pair on one end but I need directions how to make it temperature stable.   

For constraints let's say the load current would be 400mA (I presume it's component dependent) and additional max source/sink current draw to be 100mA (Lower is better), please use many sources in your response if you could! Pretty please, thanks for help in advance.

[Fcolor04]

I've came up with these but I really am not sure if those are the solution, 1.3 is really "my crazy invention" and I believe 1.4 and 1.5 are true solutions found on the internet.

In 1.3 I tried to minimize current flowing through the feedback resistor.
In 1.4 and 1.5, Vref can use high value resistor so power lost wouldnt be an issue there however power would be lost in heat in the last resistor, is there a good solution for that?

[magic]

1.1 Sorta works, load current is similar to Vcc resistor current, but sensitive to Vcc variation and to thermal mismatch between the NPNs - the output gets hotter and runs more current for the same Vbe, then gets even hotter, and so on. Much improvement can be achieved by adding resistors between emitters and ground.
1.2 Nope, one Vbe voltage developed on the left is not enough to turn on a Darlington pair.
1.3 Opamp uses the upper NPN to set lower NPN's base voltage equal to Vref. Extremely sensitive to Vcc variation and transistor temperature, not gonna work at all.
1.4 Sense voltage is equalized to Vref, this works very well but sensitive to Vcc variation.
1.5 This also works, but Vref must be close to Vcc and the upper resistor is sense.

What accuracy is needed?

[Fcolor04]

Thank you for the review, the requirements are to have as good efficiency as it's possible, let's say VCC comes from battery so current won't lower the voltage but voltage will lower over the time.

Goal is to have as little current wasted in resistors and steering transistors, I know it might be impossible in the end but I'd rather ask than not.

For pointers let's say current can vary by at most 20%, but main goal is temperature stability and low operating power

I will run some ltspice simulations to really understand how these work

[mawyatt]

Check out little known Peaking Current Source, also known as "Wyatt Current Source".

https://books.google.com/books?id=03JmxpE39N4C&pg=PA87&lpg=PA87&dq=Wyatt+Current+Source&source=bl&ots=5zevJ8Ypzb&sig=ACfU3U3DvUbu9AotugZ_GWm2zi6zyEj6pA&hl=en&sa=X&ved=2ahUKEwicmoqB9Oj2AhWOSjABHRumDjYQ6AF6BAgYEAM#v=onepage&q=Wyatt%20Current%20Source&f=false

https://www.edn.com/peaking-current-source-has-high-rejection/

Best,

[Fcolor04]

I've came up with this circuit, It looks like it works, though I'm not sure I've fulfilled my requirements, It looks like it uses only 1.7mA to drive both opamp and Q1 base, however I'll try to get rid of the op amp and see how can it be done differently! I'll update here if I find any new solutions possibly using only discrete components.

[Benta]

20%?! Is that all? Please confirm.
A transistor, two diodes and two resistors can do that with no problem. Or simply a JFET current regulator diode.

[Fcolor04]

I don't have strict rules for this circuit, I want to learn how to make temp. independent current source or sink, I managed to build that 1W one, now I feel more ready to design more precisie current sources, 20% was first goal, I want to see how far can I get without using IC, in Dave's psu design series he uses LM317 in first design to create current source, however I'd like to do that without so much heat loss or with good temperature stability and reasonable heat loss

[Benta]

Dinner time here.
I'll be back in a couple of hours with a suggestion.

[rstofer]

w2aew has a video on current sources.  It may be too elementary for this discussion but it might be a reference for others.



There's a follow-on video where Alan shows how the Simpson 260 input impedance messed up the measurements.



There's a lot to remember in that second video.  Like DC 20,000 Ohms/Volt...

[Benta]

As promised:
These are the simplest I can think of.

The first can be partially temperature compensated by coupling the two transistors thermally (that's what the dashed rectangle implicates). The TC of the diode is still an issue, though.
Advantage is, that it can be configured both as source and sink (NPN/PNP)

The second is the best, but will only work as sink. You may also need to frequency compensate it due to the high gain from the TLV431. I'm too lazy to do that exercise, sorry.

This is just to give you some ideas.

EDIT: on thinking about it, replacing the diode in the first circuit with a TLV431 (anode/ref shorted) would remove the diode temperature problem and make both sink and source possible.

[Terry Bites]

What about LED contstant current ICs. eg ADP8140 with 4x programmable chanels; 124uA to 500mA or a LM3410. Zillions to choose from. Effciencies in the 90% range.
Still, including resistor and opamp drift you can get better than 0.01% stability over 0-70'C.
The equivalent source require more components and is less thermally stable.

You might want to see this digital controller that uses PWM to achive high efficiences[attach=

[Benta]

Somehow, "without an IC" or "simple" looks different. But that's just me...

@Fcolor04: nice of you to "casually mention" the current levels we're talking about. That was not in your original post. :-(

[magic]

Check out little known Peaking Current Source, also known as "Wyatt Current Source".

https://www.edn.com/peaking-current-source-has-high-rejection/

That's an interesting circuit, thank you for posting the analysis.
Unfortunately, its output is (ideally e times, apparently) lower than the input, so not great for driving heavy loads. But the independence of input current is pleasantly surprising.

I have seen similar things in some TL431 implementations, but off the top of my head I don't think they made use of this output current stability. On the contrary, when your circuit is biased with a PTAT current, it looks like the output will simply be a PTAT current too at some fixed ratio to the input.


I noticed that OP mostly tries to exploit the dependence of Ic on Vbe. This is not very reliable in discrete implementation because of thermal drift: 1°C of heating on the current generating transistor has the same effect as increasing its Vbe by 2mV, which increases Ic by about 6%. Current effectively doubles at 10°C temperature increase, for the same applied Vbe.

You may actually get more stability by feeding a constant base current into the transistor. β also increases with temperature, but it takes some 100°C difference before it is doubled

The circuits posted by Benta could be better because they sense load current in resistors. Perhaps a Wilson mirror would be decent (the output "cascode" transistor takes all the load-dependent heat). ICs have an advantage of being able to match temperature of multiple transistors to a fraction of a degree.

[magic]

Here's a more useful kind of SPICE simulation. I step temperature of the circuit (de-facto: Q2) from 0 to 70 in 1°C increments. At the same time, Q1 has its temperature fixed at 20°C. I also step V1 from 1 to 10 in 3V increments.

It's an attempt at a simple current mirror with 10x current gain, from 10mA to 100mA. Despite the use of emitter resistors in appropriate ratio (dropping ~100mV on each side), despite the addition of R0 which provides additional Vbe "boost" to Q2, the circuit still doesn't deal with temperature mismatch very well.

[Terry Bites]

OMG a TL431 is a chip!

[Terry Bites]

Just a dumb old current limiter will be good to 1% over 0-70.
The RED LED has an approximately opposite tempco to the BC547.
Costs Sweet FA and it lights up!

[Benta]

OMG a TL431 is a chip!

Yes. With three pins... you know, like a transistor.

[Fcolor04]

Do you think T stability will be okay in this circuit? It looks great and has low parts count, though I'm not sure if no IC is the right choice if I want efficient circuit... that BD911 is going to get hot

[Terry Bites]

So, with a bit of unravelling, we have a current boosted Howland current pump. This will work well if R4=R5 and R3=R6, if not both temperature drift and current stability will be poor.
Usually R3=R4=R5=R6 to get the best performance. Precision reistors are needed. Bog standard 1% wont give a satisfactory result. Adding a curent control pot to R6 violates the R3=R6 condition. To make this work you need a buffer amplifier. . The thermal stability of the the voltage referrence must be taken into account also.The purpose of C1 is a mystery.
I've used 10V, its the maximum Vcc for this opamp. With Ic=1A and a voltage drop of about 5V across Q1, the transistor has to dissipate 5W. The BD 911 is a little over spec'd at 90W. You have a base current of about 18mA which excedes the Io max for the opamp (15mA). Consisder using a Darlington transistor to avoid cooking the AD8031. eg a TIP122 see https://www.ti.com/lit/pdf/sboa441?keyMatch=SBOA441   Right up your chipless street: www.edn.com/three-transistor-current-source-covers-wide-range/ I got 3.9% over 0-70. 1.2% 15-40 @1A

Still, for an LED this all seems a bit OTT.  TPS92200?