Constant current source to drive three LEDs with 2xAAA batteries?

[magic]

Because I said so

Very crude but perhaps good enough for Australia
It relies on multiplication of R1 current by β of Q2.

I disagree about unpredictability. It's fairly predictable for any given pair of transistor and LED. Just measure the transistor with your DMM's hFE function, which typically tests common TO-92 parts at about 1~3mA - exactly as needed.

You have one of those, right?

[dazz]

Just breadboarded magic's circuit. Efficiency is up to 70% with fully charged batteries. Didn't need the 100K resistor and probably won't bother with the Tc diode. I'll do some tests in the fridge just in case. LOL. This one runs regulating the current for much longer if the simulation is right, which is good in a way, but on the other hand could cause me to overdischarge my Ni-MH batteries below 0.8V.

[langwadt]

The one part solution... https://www.mouser.com/datasheet/2/362/P22-23-CRD-1729293.pdf

not with 2*AAA, look at the voltage drop ..

[magic]

This one weird trick should reduce current draw at low voltages to less than 10µA

R3 bypasses a few µA of Q1 current around Q2 base before Q2 Vbe becomes high enough to seriously turn it on.
It also steals ~7µA from Q2 in normal operation, so R1 needs to be reduced a bit and efficiency decreases by 7µA.

I removed D1 as it doesn't seem to be necessary anymore. Reintroducing it could push the turn-on threshold even further up.

As for thermal compensation, I switched to a PNP model which includes the effect of increasing β with temperature and all is good again. But I wouldn't trust SPICE simulations too much in this regard because the models just aren't great. The exact real world TC of the LED is also unknown, my figure of -4mV/°C came from some ON semi application note which said that -3 to -5 is typical. I found other sources saying it could be -2mV/°C or that it varies with current and may even become positive at high currents.

[gbaddeley]

Sticking my neck out here. I haven’t built or sim’d it.

[dazz]

Sticking my neck out here. I haven’t built or sim’d it.

Thanks, I'll try that too 

[james_s]

But I would still need a way to set the current constant, right? BTW, does the Joule Thief do that? because I really need the LEDs to run mostly at a constant brightness.

The forward drop of an LED will remain fairly constant, so if you start with a constant voltage, all you need is a resistor in series with the LEDs to get a constant current. The *vast* majority of low power LEDs are powered from a constant voltage supply using only a resistor to limit the current. A boost converter will be regulated and will give you that constant voltage as long as the input voltage is within the acceptable range.

[Zero999]

What's the nominal forward voltage of the LED? If you don't know that, start with what colour is it? Assuming it's white, two AA cells don't really have a high enough voltage to run a white LED off. I have a set of cheep Chinese Christmas lights, which were just wired straight to a couple of AA cells. They were different colours, made with deep blue/violet LEDs with phosphors, so they all had the same voltage drop. I connected two sets in series, powered of a 5V power supply which output slightly more than 5V and gave them to my parents, for Christmas last year.

[dazz]

What's the nominal forward voltage of the LED? If you don't know that, start with what colour is it? Assuming it's white, two AA cells don't really have a high enough voltage to run a white LED off. I have a set of cheep Chinese Christmas lights, which were just wired straight to a couple of AA cells. They were different colours, made with deep blue/violet LEDs with phosphors, so they all had the same voltage drop. I connected two sets in series, powered of a 5V power supply which output slightly more than 5V and gave them to my parents, for Christmas last year.

I'm using red LEDs with a Vf of about 1.8V, three of them in parallel. I think I'm going with magic's circuit in the end because it's so simple and efficiency is pretty damn good at 70 to 80% depending on the
 batteries charge

[james_s]

If you run LEDs in parallel you need a separate resistor for each one. You may get away with running parallel LEDs if they are very well matched but often it will not work. In general it's better to run them in series if you have enough voltage available. That's why I suggested a boost converter.

[dazz]

If you run LEDs in parallel you need a separate resistor for each one. You may get away with running parallel LEDs if they are very well matched but often it will not work. In general it's better to run them in series if you have enough voltage available. That's why I suggested a boost converter.

Yeah, I know. Don't know why, but it's working fine without resistors... at least for now. And I didn't even bother matching them. I don't know what I was thinking.

[james_s]

You got lucky, and they're probably from the same batch. There are numerous examples of poor practice that you can get away with, sometimes, but then you might build another one later using a different batch of parts and wonder why it doesn't work right.

[dazz]

You got lucky, and they're probably from the same batch. There are numerous examples of poor practice that you can get away with, sometimes, but then you might build another one later using a different batch of parts and wonder why it doesn't work right.

I can still add them, if anything, for peace of mind. Will they work with the current source? I modded the red one in LTSpice to have mismatched LEDs and the resistors don't seem to do anything at all.

ETA: Oh, wait, I need to pick different resistor values for each led to match the current flowing through each LED, right? I think I'll do that because it would help if one of the LEDs is slightly brighter than the others and I can achieve that with those resistors

[james_s]

No you can use the same value resistor for each LED, assuming the LEDs are in at least the same ballpark. The idea is to provide some compliance to account for slightly different forward drops, say 1.75V for one LED, 1.77V for another, 1.68V for a third, etc. That handful of millivolts can make a significant difference in forward current if they are all just wired in parallel but if you have an individual resistor on each one you get a bit of wiggle room. If you're using a constant current source you don't much resistance at all, if you have a constant voltage source then pick a resistor that will result in the desired current and assuming the LEDs are all the same basic type the same resistor value will work for all.

True constant current sources are rarely used with small LEDs like this, it isn't until you get into power LEDs for illumination that it makes sense to worry about that.

[dazz]

No you can use the same value resistor for each LED, assuming the LEDs are in at least the same ballpark. The idea is to provide some compliance to account for slightly different forward drops, say 1.75V for one LED, 1.77V for another, 1.68V for a third, etc. That handful of millivolts can make a significant difference in forward current if they are all just wired in parallel but if you have an individual resistor on each one you get a bit of wiggle room. If you're using a constant current source you don't much resistance at all, if you have a constant voltage source then pick a resistor that will result in the desired current and assuming the LEDs are all the same basic type the same resistor value will work for all.

True constant current sources are rarely used with small LEDs like this, it isn't until you get into power LEDs for illumination that it makes sense to worry about that.

I added 100R resistors (two 100R, one 47R for the LED that needs to be a little bit brighter) and it seems to work well.