It’s been a while since I posted a wholly new article on this site and that’s because of some exciting shenanigans keeping me busy at this end. In fact, relating to the exciting shenanigans was a particular project which recently arose requiring the use of a flashing white LED. Of course, I’m sure nobody reading this is interested enough to want to know just *why* there would be a need for a flashing white LED so I won’t presume to bore you with the yawn-inducing details of how this project came into being when I can simply cut to the chase by skipping straight to how it is to be achieved!

The requirements are that it needs to be low power as it would be running on battery for hours at a time, and the flash rate is to be once every few seconds.

Now, there are other projects I’ve previously documented to flash LEDs using transistors, timers or chipsets capable of producing a square-wave output, but those circuits were designed for a constant DC supply transformer and ideally this one would only have a single battery.

Last year I had a bit of a play with the Joule Thief, a homemade (in my case) inductor capable of ramping up a source voltage through the induction between two coils to enable a 1.5v battery to drive a 3v LED. Since meddling with the Joule Thief 18 months ago, I haven’t had cause to do anything useful with it since, and my homemade inductor found itself slung into the bottom of my deep, dark drawer of yesteryear porn where it was subsequently forgotten about.... until now.

The Joule Thief is an ideal candidate for this job. Not only will it allow me to drive a 3v white LED from a single 1.5v AA battery, but it’s also super efficient so the LED will continue to light even when the battery eventually drains until its output is perhaps as low as 0.2v. This means the circuit will get the maximum life out of the supply battery and can even use batteries that would normally be thrown away as ‘spent’ because they no longer have the wallop to drive higher current applications.

By its very nature a Joule Thief flashes a LED, however it is at such a high rate (say, 40kHz), that the human eye cannot detect the transitions and the light appears continuous. What I needed in this instance was a way to modify the existing circuit so that a slow, visible flash occurs. In order to do this, I’m not going to change how the Joule Thief operates - i.e. it will still be flashing the LED at a high frequency making it appear to be continuously lit, however I will be turning the Joule Thief itself on and off at a much slower rate so that the LED illuminates only for a short time before extinguishing.

To accomplish this, I’m adding an R-C timing element to the circuit as below.


At switch-on, the 1000uF capacitor begins charging through the 10k resistor, however there isn’t enough voltage to forward bias the 1N4148 diode so the secondary coil of the inductor and base input to the transistor are off. Although the battery is connected to the LED via the primary coil of the inductor, there isn’t enough forward voltage to drive it so the LED also remains off.

Eventually the capacitor builds up enough charge to bias the 1N4148 diode and current now flows across the secondary coil and through the transistor. As explained on my Joule Thief page, the inductance resulting from current flowing through the coils generates a voltage high enough to drive the LED. As the circuit operates, the LED pulses at a high frequency, turning electricity into light in response to the magnetic fields building up and collapsing across the inductor’s coils.

Eventually, the capacitor discharges below what can pass over the diode which closes the path so that the inductor, transistor and LED cease to operate and the capacitor begins charging up via the 10k resistor again. And so the cycle continues.

Using these particular components, my LED illuminates for about one second and then extinguishes for about another three seconds before repeating. The timing of the circuit can be altered by adjusting the values of the R-C components, i.e. the 1000uF capacitor and 10k resistor. Reducing the resistance will allow the capacitor to charge quicker and increasing it will slow the charge. In my case I found a resistance of over 20k was too high for the circuit to operate so a 10k resistor worked out well.

Decreasing the capacitance will have a greater effect on the timing. A 10uF capacitor will produce a fast visible flash rate, a 4700uF capacitor a much slower rate and anything in-between will be... well... in between. Again, for my particular application, I found a 1000uF capacitor producing a duty cycle of about four seconds to be the best candidate.

The choice of LED also had a significant effect on the circuit. Although this works like a charm with my particular 3V white LED, experimenting with different LED flavours of varying colours and forward voltage requirements could mean altered timing, no flashing or no operation at all. It’s a good job I breadboarded this circuit and dicked around with different components before reaching for the soldering iron.

Y’know, all this reminds me of the LM3909 chip I used to play with in the early ‘90’s. It’s long been out of production now and internally it didn’t work in the same way as this circuit, but it did have similar characteristics. It could be driven from a single 1.5v battery, was power efficient and could drive LED’s with a forward bias requirement higher than the circuit power source. If I remember rightly, apart from a LED and a battery, all you needed to add was a capacitor, the value of which would determine the flash rate. Although this circuit isn’t a discrete version of the LM3909, it largely performs the same task for me.

I guess they stopped producing the LM3909 because it wasn’t capable of driving the newer high-brightness blue/white 3v LEDS and besides, if you wanted a flashing LED, you could pick one up with its own in-built flash driver that didn’t need any external components - but where the hell is the fun in that??

I might even have a twenty year old LM3909 lurking in the bottom of that same porn drawer... although I doubt it has the minerals to drive my white LED. Y’know, I’m actually tempted to fish it out along with a 1992 copy of Razzle for.. uh.. nostalgic purposes....