ideally, I want this to run off coin cell batteries. I found a circuit that uses transistors and capacitors, but the voltage requirements are higher than I want. Is there any way I can make this circuit circuit low power?

enter image description here

Any other ideas would be awesome!

  • \$\begingroup\$ coin cells are pretty thin. any reason you couldn't just stack two cr2032's for 6V? \$\endgroup\$
    – JustJeff
    Aug 9, 2011 at 21:29
  • 2
    \$\begingroup\$ IMHO your question is a bit unclear. The circuit you posted is very low power, but it is not low voltage. I'm assuming you're thinking about some low voltage joule thief circuit? You'll find flashing Joule thief circuit here and here \$\endgroup\$ Aug 9, 2011 at 21:52
  • 1
    \$\begingroup\$ I would get a low-pin count, low power micro like the PIC 10F and go to sleep mostly waking up on watchdog to toggle the LED on....but that's why I'm a software guy. \$\endgroup\$
    – kenny
    Aug 9, 2011 at 22:19
  • 1
    \$\begingroup\$ @sfhhd Comments? \$\endgroup\$
    – Russell McMahon
    Aug 12, 2011 at 0:23
  • 1
    \$\begingroup\$ Nice joule thief circuit here too: watsonseblog.blogspot.com/2008/04/… \$\endgroup\$
    – Linker3000
    Aug 15, 2011 at 22:59

6 Answers 6


The first circuit on the page you linked to is already pretty efficient in that just about all supply current is used to light the LEDs. So the issue isn't lower power as much as to be able to run from a lower voltage.

Typical coin cells, like the common CR2032, put out about 3V. This mean you have to use LEDs that light at somewhat lower voltage. Fortunately common green LEDs typcially light at around 2.1V. Switching to green LEDs allows for using a 3V supply.

To cut down power, you have to cut down the current thru the LEDs. This directly effects brightness. If this device is for indoor use, even 1mA may be good enough to noticeably light the LEDs. Getting decent LEDs with decent efficiency definitely helps here. Let's say you want to draw only 2 mA from the coin cell. The voltage drop accross the resistor in series with the LED will be roughly 800 mV. 800mV / 2mA = 400Ω. That's a starting point. Then you can trade off battery life with brightness from there. Lower resistors will give higher brightness and shorter battery life.

  • 7
    \$\begingroup\$ Note that when Olin says "green LEDs" he means the old-style green gallium-phosphide (GaP) LEDs which emit around 560 nm, giving a yellow-green light. True green LEDs (530-550 nm) use a different chemistry which has around a 3 V forward drop, which would not work for a 3 V source unless a more complex power supply were used. \$\endgroup\$ Aug 9, 2011 at 22:10


  • (1) Advice on running supplied circuit on lower voltage.

    Reduce R2x slightly more than ratio of reduction in supply voltages for same LED current.

    Max supportable LED current set by transistor beta (current gain) and R2x.

Increase C1x as R2x reduced to keep time constant constant.

  • (2) Low voltage & component count any-LED flasher with other uses.

........ Details below enter image description here

  • (3) Other options

(1) Advice on running supplied circuit on lower voltage.

Your circuit can be made to run from lower voltage. Play with components and see what happens. See notes below for guidance.

With care you should be able to get it down to under 1 Volt without the LED loads. With red LEDs more like 3 Volts would be the lower limit.

This is your circuit. I've added component labels to the original.

enter image description here

The circuit is symetrical with two identical halves - so I've named components with a and b suffixes. eg R1A and R1B serve the same function in the two halves, as do R2A/R2B, C1A/C1B, Q1a/Q1b. The *values of eg R1A and R1b may be different (see text) so the circuit may oscillate with different on times per half etc.

This is an astable oscillator. The oscillation periods are set by setting the OFF times of each half, rather than the on times. When eg Q1a is off Q1b is on so it may seem like you are setting on times but it is useful to know that in fact off times are being set.

  • OFF time of Q1B is set by the time constant of C1A x R2B. Note carefully the mix of left hand and right hand parts involved.

  • OFF time for Q1A is set by time constant of C1B x R2a

LED current is set by ~ (Vsupply - VLED - Vsat_ Qx)/ R1x

Supply voltage needs to be the higher of about 1 Volt or Vf_LED + 0.5 to 1 V. eg with a RED LED with Vf of 1.8V then Vsupply >= ~ 1.8 + 0.5 = 2.3V. So, operation from 3V is practical.

Use term "Beta" = transistor_current_gain ((=hfE))

Ic_max = I_Base x Beta. As shown base current Ib ~~= (Vsupply-Vbe)/R2x = (9-0.6)/100k = 84 uA.
For Beta = 100 then Icmax = 84 uA x 100 = 8.4 mA.

ie for transistors with Beta of 100 (= realistic value for many but not all "jellybean" transistors) max LED current =~ 8 mA.

If running at 3V and if LED current target was say 10 mA and Beta = 100 then

  • Ib = Ic/Beta

  • Ic/Beta ~~= (Vsupply-Vbe)/R2x

or R2x = (Vsupply-Vbe)/Ib x Beta = (3 - 0.6)/0.010A x 100 = 24 kohm.

R2A of 22K is closest value and perhaps 15k or even 10k would be wise.

Flash rate may now be set by calculating required time constant.

Say R2A = 15k. Say half flash time = 0.5 seconds.

RC = t or C = t/r = 0.5/15k = 33 uF.

This is only a starting point for reasons which can be explained if people are interested but gives some idea of what values to use. Note that at lower voltages R2x will get smaller to supply enough base drives so C1x will increase in size for the same time constant.

(2) Low voltage low component count LED flasher with other uses.

'Russell's' one-cell "any or many LED" flasher circuit.

  • This can be not only a LED flasher or LED driver but a -ve or +ve low power voltage generator.

So also potentially a programmer supply, LCD bias supply, -ve opamp supply etc.

This circuit will flash an LED of any colour and forward voltage (or potentially even several LEDs in series) or will pulse a load using one cell - probably about 1 volt will be enough to operate it. I "designed" this circuit but it is based on a design that has not only long been used in transistor form but existed in pre-transistor thermionic valve days and, while I have never seen it used elsewhere, I would be surprised if it has not been independently "developed" by many other people.

As shown Q1 collector is driven negative below ground when Q1 turns off until energy in L1 is dissipated. Swap ground and supply and transistor types for +ve supply. Add diode from output to use as a DC supply.
L1 - small potted "resistor like" inductor or many others - experiment.
Q1 Q2 - almost any "jellybean" small pnp & npn transistors.
C1 polarised only to get high capacitance per size.
Can be eg ceramic if capacitance high enough for needs. Use only LED2 (best) or LED1 at one time.

enter image description here

  • ...... Use either LED2 (most efficient) or LED1

Time constant ~= R2 x C1.

Long time constant leads to discrete flashes. Short time constant produce apparently permanently on LED.
Use resistor between Q1b-Q2c for higher supply voltages.
Resistor in series with C1 will extend pulse length.

This circuit is usually presented with a load of some sort in place of L1 - it may be an LED (depending on voltage or a transistor base (part of a following stage) or a light bulb etc. My 'innovation' was the very obvious one of using an inductor (L1) as the load. This provides a pulse of current into L1 when Q1 is on and when Q1 turns off L1 "flys back" and delivers whatever voltage is required to dump the energy from L1 into the load - here the load is one or other of the two LEDs shown. LED2 is the most efficient as it is powered by Vupply + V_L1 so part of the energy is stored in L1 and then released and part provided . LED1 if equipped is driven solely by V_L1.

(3) Other options

An LM393 dual comparator or its quad version can run on as little a 2 Volt and also do what you want. There will be flasher circuits using it on the internet.

LM339 quad version


Once you "allow" the use of an inductor you can operate any LED from down to under 1 volt. Here's one way. I'll post more later.

EDN White LED flasher operating from 1 cell enter image description here

Here are 4 flashers stated to run on 1.5V.
The LM3909 IC is designed to flash an LED from a single cell. May be hard to find.

The 74HC04 and 74HC14 are probably at the bottom of their spec sheet power supply range at 1.5V so a single cell falls below that voltage rapidly. Quite likely to work at lower voltage but out of spec.

Circuit at bottom right will work at 1.5V and below.
Note that it is a variant of "my" 2 transistor + inductor flasher but they have added a transistor output buffer and do not have an inductor. Replacing the 330 ohm resistor with an inductor and removing the right hand 220 uF cap and 330 ohm resistor would produce a pulse at the right hand transistor's collector that would drive any colour LED.

enter image description here

  • \$\begingroup\$ The rest is interesting but under 3) "There will be flasher circuits using it on the internet" your answer seems to be just the word "comparator", not even how it's to be used for a solution. \$\endgroup\$ Aug 14, 2011 at 12:42
  • \$\begingroup\$ @Federico Russo - hand feeding only goes so far. Knowing that a cheap available IC like the LM339 is used for flashers and access to a search engine is all that is required. There are many other comparators but none (I think) cheaper than the 339/393. But, build the others that I describe in more detail instead :-). \$\endgroup\$
    – Russell McMahon
    Aug 14, 2011 at 16:16
  • 1
    \$\begingroup\$ FWIW - I must say the feedback is disappointing - not for the "points" but because it shows people haven't appreciated the utter brilliance of the simple 2 transistor plus inductor circuit. A large number of really useful applications beckon. Whatever. \$\endgroup\$
    – Russell McMahon
    Aug 14, 2011 at 16:18
  • \$\begingroup\$ About the feedback: your answer is (as often) rather long, and may suffer from the tl;dr syndrome. If you think part of an answer deserves special attention, why not highlight it by making it a separate answer? \$\endgroup\$ Aug 14, 2011 at 18:00
  • \$\begingroup\$ @Federico Russo- I do rather gain the impression that you are trolling. You make comments on various of my posts without it being evident that you have read them. You may, however, just be trying to be helpful :-). Your "separate answer" suggestion is wholly covered by the summary at the top of my response. My point 2 is 5 lines down the answer and has a diagram, inviting description and note that more follows below. You can lead a child to Euclid, but you can't make him think. if people can't read to line 5 and look at a picture then no amount of body shortening is liable to help. Alas. \$\endgroup\$
    – Russell McMahon
    Aug 15, 2011 at 0:35

All very nice ideas, but the real solution is of course the microcontroller :-). The TI MSP430 series is known for its low-power modes,


the MSP430F110(*) for instance uses less than 1.9\$\mu\$A in LPM3, where one clock remains active which can be used to clock a timer. A CR3032 cell has a capacity of 500mAh, so not counting the LED the MSP430 will run at least 30 years on it. Means the microcontroller's power usage is negligible.
Put the MSP430 in LPM3, and upon timer compare switch on the LED for a short time, after which you switch off the LED and go to LPM3 again. Average power usage and therefore battery longevity are determined by LED current and duty cycle.
Example: at 0.5% duty cycle and 70mA you can run the circuit for 1430 hours on a CR3032 cell, that's 2 months.

(*) Yes, I know, we don't need all those pins, but I chose the MSP430 for its low power, not its I/O.

  • \$\begingroup\$ Most of the current of the original circuit was already going into the LEDs, so even if the micro took no power it wouldn't be much of a improvement. \$\endgroup\$ Aug 14, 2011 at 14:10
  • \$\begingroup\$ @Olin - That's right, this is just another way to solve it. If the circuit minus LED doesn't use any power there's nothing you can do, apart from playing with duty cycle and series resistor. This is also a low component count solution, compared to the AMV. \$\endgroup\$
    – stevenvh
    Aug 14, 2011 at 14:16
  • \$\begingroup\$ Using a micro could be a reasonable solution. It has the advantage of allowing you to more deliberately set the timing and blink pattern, maybe with a little dead time between the LEDs to save power. If I were to do this with a micro though, I'd use a PIC 10F200. It's the same physical size as a single transistor. \$\endgroup\$ Aug 15, 2011 at 12:18
  • \$\begingroup\$ @Olin - I had thought of the F200, but it still has too many I/Os! :-). No, I don't have any experience with PIC and I guess I was too lazy to read through a complete datasheet to learn about low-power modes. The MSP430 isn't easy for first-time users either, but I've worked with that one before. BTW, at 2x3mm the F200 is pretty small, but NXP has a Cortex M0 only 5mm\$^2\$. With 16 pins cq. balls. Now that's damn small! \$\endgroup\$
    – stevenvh
    Aug 15, 2011 at 15:10

There's also these circuits based on the (out of production) LM3909 LED flasher/oscillator:


Side note: My first-ever home made PCB was for an LM3909 circuit (in 1976 when I was 11). Hand-drawn straight onto the copper with a resist pen!


This charge pump circuit operates from a 1.5v cell...


"This electronic circuit uses only a single inexpensive C-MOS IC and flashes the LED for a full year on a single 1.5 volt AA alkaline battery cell. The circuit uses a charge pump technique to provide the LED with the needed voltage. This electronic schematic will only work on red, green and yellow LEDs. It is not be able to flash white LEDs."

  • \$\begingroup\$ This sounds impressive, but it only has a normal LED current of 20mA at 0.4% duty cycle. I wonder how visible this is in daylight. \$\endgroup\$
    – stevenvh
    Aug 14, 2011 at 12:04
  • 1
    \$\begingroup\$ 20 mA is "plenty bright" with a suitable modern LED. Available white LEDs are available in the 30 to 40 Candela range and some monochromatic LEDs will be higher again. Note that even though the circuit is from the great Dave Johnson it still "breaks the rules". The IC is specified at Vcc = 1.5V minimum, so will be below spec for essentially all of the battery life. Apparently it works. Note that the TI version is specified at 1.5V whereas several other manufacturers parts are specd at Vcc = 2.0 V min. A transistor output driver could be added to provide more current. \$\endgroup\$
    – Russell McMahon
    Aug 15, 2011 at 7:04

Here's a low power multivibrator that may be a useful design start:

NMOS Multivibrator

It runs at 79 Hz with about 1 mA from a 2.5V source, and about 2.4 mW.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.