My wife has restored an old kerosene lantern that the neighbours wanted to throw away. She really put a lot of effort to bring it back to its original glory. It is a Feuerhand Atom and probably was used somewhere in WW2. The red glass screen suggests that it was used as a signal light of some kind.

Beautiful old lantern

We have already tried it with petroleum and it works fine. For indoor use it would be nice to also be able to operate it with some kind of electric lighting. Since modifying the mechanical construction is out of question my idea was to put a board with a battery and some kind of LED inside the screen. But any kind of point light source would look totally wrong, so I thought about this nice filament LED bulbs. Such a filament is just a series of very small LEDs connected in series. Fortunately my favorite supermarket had a special offer with a pack of three filament bulbs. I immediately got one. To my surprise the screen was made of glass. I read that it is filled with some kind of gas for cooling purposes, so it is sealed. I used a metal saw to carefully cut the socket open, it worked quite well.

Filament LED bulb cut open

I removed the filaments and had to learn that they are quite fragile, I broke two in the process. Next I wanted to know, what voltage I needed to light them up. So I connected one to my Lab supply, and at 60V it finally started glowing. Since I want to supply the board from a 1.5V battery (and it should still work when the voltage is falling), I considered the circuit options. A standard boost converter would probably not work very well here. The input voltage is too for most ICs I saw. But I found another circuit I had not heard of before. It is called a “Joule thief” because it can suck most of the energy from a battery, even when the voltage gets low. It is a simple transistor circuit, that uses an inductor for boosting the voltage. It is often used in torches to get a reasonable battery life and boost the voltage to LED levels. I even found a project documentation where it was used to power a filament.

Joule thief circuit

To fit this into the Feuerhand glass screen I decided to do a two sided board. The battery holder goes to the front, the joule thief circuit is SMT on the back side. Since JLCPCB has some really nice discounts for assembly I decided to use their stock parts for the board.

SMT parts on the back side

The critical parts in this design are the transistors (which must properly operate at low voltage) and the inductor. To make sure that I find parts that work properly I decided to do a simulation in LTSpice, which is also something I have not done before. To do the simulation it was crucial to have a model for the filament. Luckily there is a research paper on that topic. The model presented in this paper was a little overkill and took endlessly too simulate. But at least there is a V-I curve and with the help of another article I was able to find matching diode parameters. The blue curve is from the research paper, the green curve is from my diode model.

Find matching diode model parameters

With this model I could rig a simulation and after some fails I finally found a combination that works reasonably well:

Simulation at 1.5V supply

The boards are in production now. I am really looking forward to do a reality check once they arrive. I will report.

This is all open hardware and can be found on github: