Tungar Power Supply

Introduction

This project was less a circuit design exercise and more a fun construction project to display some tubes I've collected and make something out of parts that have been sitting too long in the 'parts bin'. One of my hobbies is collecting vintage vacuum tubes, particularly uncommon and interesting gas-filled tubes. Some of the earliest of the gas rectifiers were 'Tungar bulbs', a trade name which refers to a tube with a tungsten filament and an argon gas filling. Tungsten-argon rectifier tubes were first commercialised in the 1910's and became popular through the 1920's to charge lead-acid batteries and power arc lamps for cinema along with other other applications that required relatively high current DC (several amps) at low voltages (below 100V).

They use a thoriated tungsten filament as the cathode which in operation glows almost as brightly as a conventional incandescent light bulb. Physically they resemble an incandescent light bulb, with the exception that they have an additional electrode (the anode) made of graphite positioned just above the filament. The anode is hung from a lead-in wire that exits the top of the bulb.

The filament operates at low voltage and high current, nominally 2.5V and 25A for a tungar bulb that is rated to rectify 15A. This ensures even distribution of electrons emitted from the filament as the voltage across the filament is relatively small compared to the anode-to-cathode voltage.

The high filament current ensures that rectification currents up to the specified peak and average currents do not significantly influence the heating of the filament nor cause 'hotspots' on the filament which would result in excessive sputtering of the filament. Avoiding operation that causes excessive sputtering is critical to maximise the life of the tungar bulb, therefore, it is also important that the filament is powered at the designed voltage, the rectified/anode current remains within the rated peak and average values, and, the anode current is applied only after the filament has reached operating temperature and is removed before the filament supply is powered down.

Circuit Design

I had a 500VA 2x24V toroidal transformer that needed a home and figured I could make a low voltage high current supply using two Hangzhou EQ1-15/0.225 15A tungar rectifiers in a full-wave configuration. The powersupply is a simple unregulated design, made adjustable with an inexpensive 500VA variable autotransformer ('variac').

The tungar filaments are powered by a custom wound transformer (see following section) with the center-tap providing the DC output. Gas rectifier tubes require an inductive load to account for the negative resistance characteristic - the anode-cathode voltage required to ionise the gas is higher than the voltage drop maintained once ionised. Typically a choke is placed after rectifier cathodes before the first filter capacitor to achieve this, however, I had a Hammond 193U choke (200mH @ 2A) at hand and decided to place this on the primary side. Equivalently a 2mH @ 20A choke could be used on the secondary side which is approximately the same size and cost. The only operational difference is that by placing the choke on the primary side, current is forced to return to zero over each half cycle whereas a choke on the secondary side - if high enough inductance - may maintain continuous DC current, causing a current waveform approaching a square wave to be drawn from the power transformer. I sought to avoid the latter as this generally results in a harsh 100Hz buzzing sound emitted from the transformer/choke, due to the harmonic content of the 100Hz square wave.

Functionally the complete output filter is L-C-L-C. The first L-C (L1, C1, C2) can be considered to deal with the negative resistance characteristic of the gas rectifiers while also tuning the transformer load to be near-resisitive for maximum output power capability. The second L-C (L2, C3) filters 100Hz ripple and provides bulk output capacitance. The values of L2 and C3 were chosen to be as large as practically possible within size/cost constraints. The value of C2 was determined in SPICE simultation to optimise power output i.e. make the most of the power transformer current rating by reducing the crest factor of the secondary winding currents. C1, C2 is made from two banks of eleven 220uF 50V Panasonic FM series electrolytic capacitors connected back to back to function as a bipolar 1200uF capacitor. This is done for two reasons - firstly, the design capacitance value cannot be achieved with a single polarised bank of capacitors while also meeting the required ripple current rating of ~13Arms at 100Hz. Secondly, under low output voltages and maximum output current (12A) the voltage across C2 can dip briefly into negative voltages during each half-cycle, thus a bipolar capacitor is required. C3 is constructed from a bank of eight inexpensive 4700uF capacitors with comparatively low ripple ratings since the ripple current is significantlly smaller.

AC output terminals are also provided directly off the power transformer secondary windings, allowing the supply to alternatively be used as an adjustabled AC power supply

Overcurrent protection is provided by the 3A mains fuse, as well as a 2A circuit breaker between the autotransformer and power transformer, since if the autotransformer is set significantly below 240V it is possible for an overcurrent condition to occur on the output side of the autotransformer (such as if the powersupply output is shorted) without causing input current exceeding the 3A fuse.

Rewinding a toroidal transformer

A transformer with dual 1.25V @ 25A or 50A secondaries isn't readily available off-the-shelf so we will need to custom wind one. Conveniently I already had a 120VA, 2x18V toroidal power transformer on hand that I could use so I set about rewinding the secondary to 2x1.25V. As this represents 14x reduction in voltage and 14x increase in current, the secondary windings were unwound from the core and each of the wires (one from each secondary winding) were cut into 14 equal lengths. Groups of 7 parallel wires were wound back onto the core. This was done as winding 7 parallel wires at a time was easier to manage than 14. As the original 18V windings completed two full rotations of the core, the first two secondary windings of 7 parallel wires complete a single rotation and each become 1.25V @ 25A. Another two secondaries are layered over the first two, also each becoming 1.25V @ 25A. The first and second sets of secondaries are then wired in a series-parallel configuration to obtain 2.5V centre-tapped @ 50A. Two pairs of red 10GA wires provide the 2.5V for the tube filaments while a single black 10GA wire from the centre tap provides the DC output. Wiring both tubes to a shared 50A winding as opposed to dedicated 25A windings for each tube is advantageous as the resistance of the 50A winding and therefore DC resistive losses are lower.

Chassis Construction

The chassis is a generic aluminium plate and extrusion type I found on AliExpress. I tried to reduce the form factor as much as possible. The autotransformer was stripped of its stamped steel casing and mounted on its side, with the adjustment knob positioned front and centre. SO-45 analogue panel meters provide front panel V/I monitoring for the DC output. The toroidal power transformers and capacitors are also located inside the enclosure with the tungar bulb and E-I chokes mounted externally on top. The chokes are orientated 90degrees to each other to minimise magnetic coupling. I sunk the giant edison sockets as far in as possible to minimise how far the tubes protrude. I was skeptical if the sockets would handle 25Amps as these are sold ubiquitously as E40 or E39, 1500Watt rating which equates to 6.25 or 12.5Amps. Luckily these don't seem to have any issue at 25A, probably helped by the fact that the tungar bulb is at most dissipating only around 150Watts.

Cooling is convective with a mesh grille at the rear and holes drilled around the tungar tubes and in the bottom panel below the C3 capacitor bank. The anode caps are made from radial aluminium heatsinks with a set screw to grip the anode lead in wire.

One final tweak

Something I overlooked in the physical design was how bright the tungar filaments are - too bright to read the front panel comfortably. To remedy this I added a heavily tinted acrylic window infront of the tubes:

Bonus mercury-vapour rectifier content

I also have these NL-623 tubes which are a compatible replacement for a 15A tungar. Here's them in action: