# Variable transformer starter circuit

# Project description

After getting my big 3-phase variable transformer (“variac”), I was unpleasantly surprised by it tripping breakers at random at the moment I plugged it in. Curiously, it seemed to trip random phases, and rarely all three at once! If I attempted to reset the breakers by hand a couple of times, I could eventually get all phases to work, and then transformer would work fine after that! Obviously, some sort of inrush current problem was going on, but at first I had no clue what is going on. Only after a while of asking around, and a discussion with some folks on 4hv.org I finally reached the answer. It was something I should really have figured out myself!

The transformer can indeed draw a high “inrush” current upon turn on, which is caused by saturation of the core from excessive volt-seconds that can occur at startup, depending on the moment the switch has closed!

To better understand what’s going on, consider the transformer primary as a plain inductor for now, without any significant resistance (because there is no load on the transformer secondary). In stationary state, the primary current lags the voltage by 90 degrees, which means that it crosses zero when the voltage is at it’s peak. Let’s consider two situations of what could happen in the moment voltage is applied to the primary:

- When the transformer is off, the primary current is, of course, zero. Assuming the voltage has been applied to the primary at the peak of it’s waveform, the current will start from zero and continue as if everything was running for a long time in a stationary state.
- Now let’s assume the voltage has been applied to the transformer at it’s zero crossing. Now both the voltage and current start from zero, and this allows the current to integrate up to 2x higher than it would in normal operation! Another way to interpret this is to think of the first 1/4 of the mains voltage cycle as integrating into a DC component that slowly dissipates due to resistive losses. One would still not think what’s the big deal about this temporary double current – but that’s where transformer construction comes with a problem: The current, which is basically magnetizing current here, creates magnetic flux in the core, and the iron can only take a certain amount before it saturates. The transformer manufacturers don’t design their transformers to take this briefly doubled magnetic flux, because that would require 2x the iron – hence the transformer saturates upon turn-on! In saturation, the permeability of the core and the inductance drop dramatically, resulting in current peaks with value of tens or even hundreds of amps! These will persist for many cycles (until the DC component is dissipated by resistive losses) and is pretty certain to cause tripped breakers.

When a transformer is energized by a switch, voltage is applied to it at random phase anywhere between 1. or 2. above. But since 50Hz transformers generally go into saturation with little extra volt-seconds (guess 20-30%) over their designed rating, it doesn’t take much deviation from case 1. to trip breakers. Moreover, this explains why I often ended with only some of the 3 phases tripping breakers!

Once this problem was identified, it was easy to think up solutions. Some of them include:

- Using slow-blow fuses instead of the breakers – the simplest solution, but may not always be practical.
- Energizing the transformer over series resistors, and then shorting them over after a short period of time.
- Energizing the transformer over precisely timed switches (thyristors or triacs) that are controlled so that they apply the voltage to each phase at it’s peak.
- Having the transformer loaded during turn-on, which should quickly dissipate the DC-component and get the transformer out of saturation quickly.

Due to nature of loads such as Tesla coils I was going to power with this transformer, having load all the time on the transformer was out of the question – I wanted to be able to start the transformer with it’s output at 0 and then bring it up slowly. I could replace the breakers with fuses, but I also wanted to be able to run the transformer in other places where that may not be possible.

Hence I investigated the two of the proposed starter circuit solutions, and decided the phase controlled approach is too complex to bother with for now.

The idea was to have one relay that would first apply power to the transformer over a set of resistors, and then have another act in about a second to short the resistors out. The choice of the resistors depends on the expected magnetizing current required by the transformer; EI cored transformers will generally require lower values than toroidal transformers. This transformer seemed to require only a few hundred mA per phase, so resistors on scale of around 10o ohms were appropriate. I used 150ohms because I had those at hand. Resistors only need to stand high peak powers for a short time, so use at least a few watts of power rating in order to give the resistors enough thermal capacity. I used 10W rated resistors.

I only had 230V relays around, and I had to figure out a simple way to delay the run-on of the main relay. A simple solution was to use a fluorescent tube starter in series with the main relay winding, which delays it’s turn-on for a few seconds.

The circuit is shown on the following schematic:

The following picture shows how the circuit was implemented in the casing of the variac itself. The smaller relay is used to switch the resistors, and the larger one is the main switch.

# Results and conclusion

While the solution I impelmented was simple, robust, and worked very well, it has a significant drawback – it shouldn’t be ran with the load applied, because the resistors could burn out from high power dissipation during the few second delay created by the starter. In order to prevent this, one would either have to use resistors with much higher power rating (more like 100W or so), or, use a much shorter delay – few hundred milliseconds is actually enough here.

An improvement I’d recommend would be to use a NE555 monostable circuit along with 24V DC relays.