Efficient current through 30 leds

danadak said:

Otherwise use an LDO....dropouts << 1V, plus you get short circuit and thermal protection.

Click to expand...

For a constant-current circuit it would be the dropout voltage plus the reference voltage which is often about 1.25V.

danadak said:

you do get the low dropout at the expense of circuit protection -

Click to expand...

The two-transistor circuit is protected against short duration shorts (since it limits at the constant-current), but its dissipation does need to be considered.
For a 12V short @ 50mA that would be 600mW.

crutschow said:

What do you mean "flip the transistor"?

Yes, the value of R3 determines the LED current, as the equation next to it shows.

No, each string needs its own current limiter circuit to avoid one string taking more current than the others.

If you understand how BJTs work, then understanding its operation should not be difficult.
R1 provides base current to Q2 to turn it on (current from emitter to base turns on a PNP transistor).
When the output current through R3 generates sufficient voltage across it to turn on Q3 (about 0.65V) then it starts to conduct and raises the voltage on R1 until Q2 starts to turn off, thus limiting the current to that value.

The circuit could also be built with NPN transistors, but the circuit would be put in series with the ground side of the LEDs instead of the high side.

Make sense?

If you simulate the circuit, then you can look at all the node voltages and currents to help understand how it works.
Might be a good time to try that with LTspice, since it's a simple circuit to start with.
I can post a copy of my .asc simulation file.

Click to expand...

I just mean switch the transistor on/off. Use the low voltage from a micro-controller to switch on a larger voltage source to the leds. Am I understanding a transistor correctly? You can use it as a switch, yes?

Okay, so each string will have to have it's own limiter. I'll have to start with one, build it, and see how it works.
I've pulled up a few videos and documents on BJT, so I can familiarize myself.

I believe I'm catching onto what you're saying. I would really love to get a copy of your simulation file. I've used NI Multisim before, but that was only due to class requirements. I've already downloaded and installed LTspice. I'll go through a couple of tutorial videos and jump right in. Thank you!

danadak said:

This uses a 3 Terminal adjustable regulator for a constant current source.
Has advantage it is both internal short circuit and thermal protected.

The current is Iled = 1.2V / Rout

Note I used regular diodes as I did not have a spice model, and modified their threshold to 1.6V.


Regards, Dana.

Click to expand...

Thermal protection is something that has been mentioned as a desired feature.
I've been reading over yours and Mike's conversation (thank you both) and I'm just going to have to get to work and see what suits me best.

Further, I am not necessarily limited by just one 18650. It would be ideal, but I can add more as needed. Other than extended run time from a higher Wh rating, are there any other benefits to having multiple batteries?

I'll definitely post some follow ups as I get more done.

Whatever circuit you decide on, it would be a good idea to test it with some visible light LEDs, maybe something like this:- https://uk.farnell.com/tt-electronics-optek-technology/ovfsrac8/led-super-flux-4-pin-red/dp/1497986

The reasons are that firstly you can see what is happening when the LEDs illuminate, and secondly the visible light LEDs less than 5% of the price of the IR ones, so you can afford mistakes.

They have a higher voltage, but you could compensate with fewer in each string. For instance, if you are intending to have 5 of the IR LEDs, that would total to 8 V. With those red LEDs, you could test with both 3 LEDs (6.9 V) and with 4 LEDs (9.2 V) and if your current control circuit works as expected in both cases, you can be confident it will work with the IR LEDs.

AmateurDesign said:

I just mean switch the transistor on/off. Use the low voltage from a micro-controller to switch on a larger voltage source to the leds. Am I understanding a transistor correctly? You can use it as a switch, yes?

Click to expand...

If you look at Crutshaw's constant current circuit:


Use an extra NPN transistor (with base resistor) or N channel logic FET controlled by the MCU output.

Connect the emitter or source to 0V and connect the 10K bias resistor (R1) in the current source to the collector or drain, rather than directly to 0V.
Add anther 10K from switch transistor collector to LED power positive, to ensure the current source switches completely off when the transistor is off.

That would allow the current source to be enabled or disabled via the MCU output, as the transistor was switched on or off.

You could link all the 10K resistors from the different current sources to a single switch, or each controlled by a separate switch.


If the LED current is really critical, you may need to fine-tune the current sources to make up for component variations?

The current source can be easily made adjustable by replacing R3 with two series resistors, eg. a 10 Ohm plus a 4.7 Ohm, or 12 + 2.2 ohm for a smaller range.

Use the lower value at Q1 base end and connect a 100 Ohm preset resistor across it, with the transistor base connected to the preset wiper instead of the original resistor.


Definitely stay with a step-up converter, whether you use one or two cells. The adjustable output voltage means you can set it so not too much over the minimum voltage is lost across each current regulator, so wasted power will be minimised, and the regulator power loss is contact, no changes with battery state..

Using "bare" cells to get enough voltage means a significant change from full charge to flat, which could affect regulation and will cause more power loss in the current regulators, possibly affecting their stability due to temperature changes.

If you are using a boost converter, some of those are designed to drive LEDs at a constant current, so you don't need to have a separate current limiter.

Here is an example:- https://uk.farnell.com/diodes-inc/pam2841gr/led-driver-boost-40-to-85deg-c/dp/3483114

Interesting part series -






Package a little challenging....

Certain parts preconfigured for 50 mA. So one part solution possibly.
System would still need one boost converter if doing strings to supply
constant current source(s). Then one of these per string.


Regards, Dana.

AmateurDesign said:

I would really love to get a copy of your simulation file.

Click to expand...

Attached is a copy of the LTspice .asc file showing both NPN and PNP circuits.
Clicking on the downloaded file should cause it to run directly in LTspice.

As rjenkinsgb noted, you can add a transistor in series with R1 to ground to turn the LEDs on and off.
You only would need one transistor to control all the strings by connecting all R1's to the transistor collector (NPN) or drain (N-MOSFET).
You can try that all in the simulation.