Neon Flasher

neon flasher prototype

OK, there is nothing special about flashing a neon, but this little inverter will run from as little as 1.2V (a NiCd or NiMH cell).

The circuit is based on the Joule Thief circuit which is credited to Z. Kaparnik on that site. The circuit is a simple oscillator, which in the case of the Joule Thief has sufficiently high voltage spikes at the collector when running from less than 1.5V to forward bias a LED. In the practical circuit I built to test the idea (very neat I might add!), the spikes are about 35V peak at 70kHz into an open circuit. Add the LED and the frequency drops to about 40kHz and the voltage is clamped to the drop of the LED, with a duty cycle of about 75%.

For the neon flasher I have used the same Hartley oscillator configuration, but instead wound it on a xenon flash lamp trigger transformer to facilitate a voltage step-up to the 70-90V needed to ionize the neon. The raw AC output of the transformer will light the neon quite well from a 3V supply, but adding a rectifier and storage capacitor gives bright flashes from 1.2V or less.

The trigger transformer is the model available from Jaycar Electronics, it has the secondary wound on the ferrite core first, the primary is wound over that with thicker wire. It is a simple matter to remove the primary winding and wind 400mm of bifilar 0.2mm winding wire onto the core to make the new primary and feedback windings (take about 800-1000mm of wire, halve it, twist it up with a drill, and wind that, then separate out each filament with a continuity tester). It isn't really practical to save and re-use the primary, or even leave it in its auto-transformer configuration, unless you redesign the oscillator to have the primary in the emitter (probably worth a try some day).

neon flasher circuit

While not really suited to the frequency the 1N4004 works almost as well as an 'ultra-fast' device of the same current rating, the difference isn't worth being concerned about, I measured it at 2% percent (which was surprising). Picking a neon that ionizes at a low voltage allows the circuit to operate usefully to very low input voltages. It will continue to oscillate to at least 600mV, but won't produce enough voltage to fire a neon below about 1V. I probably wound too many turns on the primary, using a few less could extend the operational voltage lower, but would make the efficiency even worse. The whole transformer is 'wrong', but very easy to work with.

My prototype unit pulls 33mA from 1.5V (about 50mW), charging the output capacitor to 82V before the neon fires. That means with each flash about 500uJ is delivered to the neon gas per shot, at 4Hz flashing rate, that places the efficiency at a woeful 4%. Less than 1mW is lost in the bias resistor, the bulk of the losses are in the transformer magnetizing current and the capacitive load of the diode and storage cap series combination across the high impedance secondary. A resistor to control the charging rate might help with those losses, as might an RF rated rectifier with lower reverse capacitance and a fast response time. I am not sure if decoupling capacitors would help, but the problem is fundamental, the transformer and rectifier are too lossy because they are operating at around 200kHz.

There is no discharge resistor across the cap. I tried a 10M unit, but it loaded the circuit too much to flash at 1.2V. Instead I just potted the entire device in epoxy. To display how poorly designed and feeble the transformer output is, you can place your finger on the epoxy over the hot side of the transformer near the diode and the extra capacitance will drop the flash rate. If your unit seems even more pathetically inefficient than mine, try reversing the diode, the drive is unipolar and far from square so you should pick the best half of the cycle for charging and leave the least for the losses.

Although I've made it sound really bad, it works well, even if it isn't very efficient. In glow mode it is quite a bit more efficient, the RF output ionizing the neon easily, but the primary still has too few turns and a very large magnetization current. Winding your own transformer but keeping the circuit the same you could achieve much more efficiency, but the wire would be very thin and take considerable effort to wind without a machine.

1 comment.


title type size
circuit postscript source application/postscript 12.795 kbytes