Mercury Switch Inverter

Driving a fluorescent tube with the inverter

An inverter or DC-DC converter is a fairly straightforward circuit these days. However, this type of circuit relies heavily on the availability of suitable power semiconductors for converting the incoming DC into AC. I thought it would be interesting to investigate other ways of performing this function. Before the development of transistors, power conversion was generally performed using either automated mechanical switching of the incoming DC, or through an intermediate conversion to mechanical power in a motor-generator set.

Both of these solutions have the disadvantage of having moving parts, and suffer from wear due to sparking at the contacts. I wondered if it would be possible to avoid this wear by using mercury as a contact material. Special "mercury wetted" relays are available, although these are generally used with the objective of eliminating contact bounce. I didn't have any of these relays immediately available, and so I started looking at the possibility of using a standard mercury switch.

These switches contain a small bead of the liquid metal, which opens or closes a circuit according to their orientation with respect to gravity. If the orientation of the switch could be continuously changed, using a motor or vibrating solenoid, the switch could be used to chop an incoming DC source into an AC signal that could be stepped up or down using a transformer. However, this still has the disadvantage of having moving parts, and the connecting wires to the switch would be subject to continuous flexing, which could lead to a short life.

But then I had another idea. A conductor that is carrying a current in the presence of a magnetic field will be subject to a mechanical force. If a magnet was placed adjacent to a closed mercury switch, the bead would have a force applied to it. With suitable orientation of the magnet, and sufficient strength of the magnet and the current flowing in the switch, the bead would be forced away from the fixed conductors, breaking the circuit. The current would then stop flowing, the force would cease to act, and the bead would roll back under gravity and reclose the circuit. The cycle would then repeat.

Mercury switch oscillator

I knew I had a mercury switch somewhere, and after some rummaging in my parts collection, managed to find it. The switch was connected in series with a 12V lamp and power supply, and placed on top of a stack of old hard drive magnets. After carefully adjusting the angle of the switch, I managed to get the lamp to flash at around 5Hz (video, 1814k). Meanwhile, the mercury was sloshing around in a manner reminiscent of the mercury beating heart experiment (although the accepted explanation for this effect does not involve magnetism).

This proved the basic concept, so I then replaced the lamp with a 2155 (240:15V) mains transformer operating in reverse, with a neon lamp connected to the original primary. This worked well, with the lamp flashing brightly. Then, inspired by Cool386's article about a vibrator inverter fluorescent lamp, I tried a 15W fluorescent tube, which also gave a bright flash. The inductive kickback also caused a nice blue mercury vapour discharge in the switch... (video, 1815k)

The one limitation of this setup was the slow oscillation rate, resulting in a limit to the overall output power. I tried altering the current flowing through the mercury switch, the strength of the magnetic field, and also the orientation of the switch, but these only had a relatively small influence on the frequency of oscillation. It appeared that this was mainly set by the geometry of the switch and the amount of mercury present - I was using a large switch, with a fair sized pool of the metal, and the oscillations were actually setting up a wave in the pool.

I thought it would be interesting to see if it was possible to increase the oscillation frequency using a different switch, and perhaps light a discharge lamp continuously. I didn't have any more mercury switches, but I managed to find a few different sizes on ebay, so I ordered a few of each.

As soon as these arrived, I connected one up and tried it out. One thing became immediately obvious - the considerably smaller size of these switches (4mm and 6mm outside diameter) combined with the energy being dumped during the flyback period resulted in considerable heating of the switch. I actually cracked a couple of the switches from the heat (oops!), though at least they didn't explode and throw globules of mercury everywhere.

To alleviate this problem, a capacitor was connected across the switch contacts to absorb the energy and limit the voltage. It was necessary to experiment a bit with the value of the capacitor, as a large value will not only limit the voltage on the transformer primary, but will also suppress the high voltage spikes on the secondary, which are necessary to 'strike' the tube. I ended up with a value of 4uF. A 6V power supply was also connected to one of the fluorescent tube filaments in order to assist with starting.

The oscillation frequency of the new switches was considerably better, at around 18Hz. This provided quite a good level of light, with the tube ignited fully, though with an intolerable amount of flicker! To achieve a continuous light, it would be necessary to somehow store the energy, and release it in between each pulse from the transformer. A rectifier and electrolytic storage capacitor would be one possibility, but the negative resistance characteristic of the tube would make it hard to control the discharge of the capacitor.

Another possibility that I did explore was adding an (AC) capacitor directly across the transformer secondary, in an attempt to form a resonant circuit that would 'ring' during the intervals where power was not being supplied. However, this was unsuccessful, at least partly due the the Q of the resulting circuit being insufficient. It looked like a continuous source of light was impractical.

But after some very careful adjustment of the switch and magnet, including surrounding them with an open ended spanner in order to close the magnetic circuit, I found it was possible to excite a second oscillation mode in the switch, at around 75Hz. I am not sure of the reason for this - it could be a mechanical harmonic mode, and may be partly attributable to negative resistance behaviour in the mercury arc. When the oscillator ran at this frequency, the tube lit up continuously, with only a small amount of flicker due to jitter on the oscillator. Brightness was comparable to that of a similar 15W tube operating from a mains supply, and the input current was only around 1.5A at 12V, suggesting quite a good efficiency figure. The experiment was a success! (video, 3638k)

Small mercury switch oscillator with closed magnetic circuit

Unfortunately, while the circuit did work, it would be too touchy for use in a practical situation (especially in a portable lamp, subject to movement, where the switch could end up stuck closed, and the transformer would then quickly burn out.). The mercury switch required precise adjustment of its angle to achieve the desired oscillation mode, and sometimes it was not possible to make it run at the fast speed at all. And, although the oscillator would mostly run at the high speed for extended periods, it would occasionally spontaneously drop back to the 17Hz mode.

I thought about going back to the resonant circuit idea, hoping that I could force the oscillation frequency by tuning the electrical resonance to match the mechanical resonance. The 2155 primary measured around 50mH unloaded, which corresponds to a capacitance of 90uF for an undamped resonance at 75Hz (more capacitance would be required to account for the loading effects). I connected up an assortment of motor run capacitors in parallel across the primary, aiming for somewhat more than 100uF. However, despite trying several capacitance values, I did not notice any positive impact on the stability of the 75Hz mode.

Examination of the primary waveform with an oscilloscope showed a possible explanation for this: When the switch is closed, the resonant oscillation is very heavily damped by the low impedance across the supply rails, and cannot affect the motion of the mercury drop. When the switch opens, there is less damping, and a decaying oscillation is visible. But at this point, there is no current flowing in the switch, and hence no force acting on the mercury. I could not see any other easy ways of improving the reliability of the circuit at this point, and further experimentation was abandoned. However, I think that, given a mercury switch with suitably designed geometry, it would be possible to achieve stable oscillation at a reasonably high frequency.

It seems that this design is not a new idea - I had a look around and found references to one of Tesla's experiments using a transformer and mercury interrupter, although it is not clear if this was a self-oscillating configuration. A research paper from the 1960's appears to describe a similar configuration (though I have not had a chance to read the full article). In some ways, it is a bit strange this idea never really took off - if the problems were overcome, it would have been quite valuable in the pre-semiconductor era.

This should be a fairly easy circuit to get going if you want to play around with it - though you may need to try a few different types of mercury switch, and also experiment with the value of the capacitor to get optimum results. The main thing to watch is the power dissipation in the switch. When properly tuned, there is only a tiny flash of light inside the envelope, and the switch runs cold. A somewhat larger flash is still permissible, but sometimes the circuit can get into a mode with a very intense flash, producing rapid heating. This must not be allowed to occur for more than a second or so, or the switch may crack or burst.

Suggested circuit for experimentation

Another issue relates to the wetting of the contacts. As supplied, the mercury bead does not wet the contacts at all, and will break the circuit as soon as it has a chance to move downhill away from the contacts. However, sometimes after operation at high power, the mercury would start to wet the contacts, resulting in a greater or lesser amount of hysteresis between the opening and closing angles. This appears to be an irreversible change, and will obviously affect the oscillator characteristics significantly. I am not sure if this is an intrinsic problem, or merely a result of the contact material used in the cheap ebay switches. (The larger switch that I used originally did not suffer from this problem, although it obviously has a significantly greater capacity to absorb heat.)

One other note - it appears that mercury switches are still readily available at the moment (2016), though with all the RoHS stupidity, this may not always be the case. So if you do want to try this experiment, don't put it off too long!

This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.

loopgain.net