Isolated AC/DC power supply or testing station for 230 V devices

Introduction

Once upon a time, I spotted a good deal on OLX and bought cheaply a nice piece of equipment: the TST 280/1 variable transformer. It was slightly broken; the ammeter was not working. After inspecting the device more closely, I realized that it was not just a variac, but an isolated variable transformer: it combines an isolation transformer and a variac in one unit. Quite rare. Then I got an idea: since the ammeter was broken anyway and I probably wouldn’t be able to find a drop-in replacement, why not rebuild the device completely and make it much more versatile?

So, in 2021, I completed the build of this AC/DC power supply/testing station, and I’m pleased with the results. In this article, I will describe the design and the issues I faced. The project was featured in Elektronika dla wszystkich 12/2021, a Polish hobby electronics journal.

Making the testing better

Before I describe the device itself, let’s discuss the issues you can run into when testing mains-powered devices. When you are testing such a device, many things can go wrong. You can destroy a circuit, cause a small fire, or even electrocute yourself.

Protecting yourself with an isolation transformer

The neutral wire (N) is grounded (physically connected to the soil) in almost all electric power systems, except the highly uncommon isolé terre (IT) system. Therefore, touching the phase/hot/live/L wire carries a high risk of electrocution. To mitigate this risk, one can remove the galvanic connection between the grid and the device with an isolation transformer. Isolation transformers are commonly used to provide an additional layer of safety in places where there is a risk of contact with the human body, such as medical or laboratory environments.

In the lab, when working with mains-powered devices, it is a good idea to use an isolation transformer. Electrocution happens only if you touch two conductive parts with a voltage difference between them, not just one conductive part.

↑ click to enlarge ↑

There seems to be some controversy regarding the protective earth (PE) pin on the isolation transformer’s output socket. Intuition suggests that, on the transformer’s secondary side, the circuit should be completely insulated, forming an isolé terre (IT) system. However, some commercial isolation transformers have the PE pin connected to the output connector’s PE. This is discussed in this Electrical Engineering Stack Exchange thread. I went with unearthed sockets as the output connectors. This feels safer to me since the output is fully floating and, additionally, it prevents ground loops.

Moreover, you can connect an oscilloscope to the device’s output and have a floating-ground oscilloscope. Be careful: this is a janky solution! Since you already have an isolation transformer, it is safer to power the device under test through the isolation transformer, not the oscilloscope. Also, keep in mind that this setup does not provide insulation; it only eliminates the ground loop.

Smooth voltage regulation with a variable transformer

To start a device gently or test its behavior under abnormal supply voltages, one can leverage a variable transformer, as shown in the schematic:

↑ click to enlarge ↑

Most workshop-grade variable transformers are not isolated—there is only one winding, so the device effectively works as an autotransformer. Isolated variacs also act as isolation transformers: the primary winding is separate from the secondary. They are rare, though; fortunately, I managed to find one and avoided the need for a separate isolation transformer.

Limiting current with a lightbulb

Let’s say you have an unknown device: an old vacuum tube radio, a power supply not connected for thirty years, you name it. You fear exploding RIFA capacitors or the release of magic smoke. There’s been a trick since immemorial to mitigate this risk: connect your device in series with an incandescent lightbulb of appropriate power.

↑ click to enlarge ↑

To a first approximation, a lightbulb is a resistor. In reality, it is not fully linear. Looking at the chart (source), the current rises very fast up to ~5% of the bulb’s nominal voltage, then rises more slowly, roughly linearly:

According to this research paper, incandescent lightbulbs have a power law relationship between voltage and current. That’s why they usually give up the ghost the instant they are turned on, and almost never when they are already lit. The culprit is the large inrush current caused by the low resistance of the cold filament.

The resistance grows significantly as the filament heats up. This kind of negative feedback stabilizes the current and keeps the lightbulb at a roughly constant brightness. By the way, this current-stabilizing effect was leveraged in iron–hydrogen resistors (or barretters), the “tubes” used to stabilize the current of vacuum-tube filaments in the tube era.

How do you choose the rated power of the lightbulb? The conservative approach is to match the lightbulb’s rated power to the maximum current the device under test may consume. In that case, even in a short-circuit situation, there won’t be more than the maximum current flowing anywhere in the circuit. This approach has a disadvantage: the device cannot reach its full rated power because of the voltage drop across the lightbulb.

The opposite approach is to choose a much more powerful lightbulb, so the device’s rated current sits in the steeply rising part of the lightbulb’s I–U curve. This ensures only a small voltage drop under normal conditions.

Usually, the lightbulb’s rated power ends up somewhere between these two approaches. In my device, I have a 230 V 60 W lightbulb installed, and I’ve never had the need to change it. Of course, the lightbulb trick also works for DC; I leveraged it in Octoglow’s Geiger board.

Momentary activation with a pushbutton

During testing or experimentation, you may want to power the device for only a few seconds. A monostable pushbutton is more convenient than a bistable switch; it guarantees that when you remove your hand, the device switches off. Moreover, it raises the safety level a bit: even in the unlikely event of electrocution, your body would probably twitch, release the button, and stop the current flow. A similar approach is leveraged in the dead man’s switch.

↑ click to enlarge ↑

In my device, I installed both a pushbutton and a switch for energizing the output. The pushbutton allows momentary activation for quick testing, while the switch enables continuous operation.

The device description

With the issues that arise when testing mains-powered devices covered, I can finally describe my testing station design. It is based on a TST 280/1E variac. I kept the original cabinet but replaced the front panel with a new custom one. The testing station has the following features:

• Provides variable AC voltage 0–330 V and variable DC voltage 0–450 V, up to 1.5 A.
• Measures voltage, current, power, and power factor of the output AC voltage.
• Measures voltage and current of the output DC voltage.
• Optionally limits the output current using the series lightbulb method.
• Output contactors can be switched on by either a pushbutton or a switch; the presence of output voltage is indicated by neon lights.
• DC voltage is available on banana terminals. AC voltage is available on banana terminals as well as a standard European/Polish socket and an American socket.

Front panel

The front panel is made from a 4 mm aluminum sheet. I cut holes for the switches and other components. The components were labeled with Dymo embossing tape. The making of the panel itself is described in the section below.

The device’s controls are purely analog. There isn’t much circuit logic here, just switches and contactors. The only advanced digital circuitry is inside the factory-made meters. The front panel houses the following controls:

The labels and switch colors follow a convention described in the table below:

Electronic description

The testing station is powered from 230 V 50 Hz mains through cable P1. The main power switch SW1 has an integrated yellow neon bulb. The bimetallic overcurrent protectors F1 and F2 protect the isolated variable transformer T2 and other components from overheating.

The variac’s secondary winding provides two voltages: 280 V AC and variable 0–280 V AC. The variac’s lower leg is connected to ground, but this ground is purely virtual; under no circumstances should it be connected to the chassis or to the PE connector of the power cable P1. The variac’s secondary side must be galvanically isolated from the primary.

The small transformer T1 works as an autotransformer and steps the 280 V AC down to ~160 V AC, which serves as an auxiliary voltage. This auxiliary voltage, despite being significantly lower than the nominal 230 V AC, is high enough to power the meters (MES1, MES2), the contactors (K1, K2), and the neon bulbs inside the SW3 and SW5 switches.

↑ click to enlarge ↑

A DPDT switch SW2 (split into SW2A and SW2B in the schematic) selects the operating mode: AC or DC. It is wired so that it is impossible for both contactors to be engaged at the same time. The current-limiting lightbulb LA1 is connected to the appropriate outputs by the contactors, depending on the operating mode. The lightbulb can be bypassed with the SW7 switch.

In AC mode, the auxiliary voltage powers the MES1 meter. The meter measures the variac’s output voltage and current using a current transformer T3. With the base parameters (voltage and current) measured, it calculates and displays other parameters, such as power and power factor. The output connectors J2–J4 are energized by the K1 contactor, which has three sets of contacts. Two of them connect the hot side of the output, routing the current through the limiting lightbulb LA1. The remaining set of contacts connects the cold side of the output to the output terminals.

Contactor K1 can be energized by the bistable switch SW3 or the monostable pushbutton SW4. The pushbutton allows temporary activation of the output, while the switch allows continuous operation. The neon bulb inside SW3 lights up when the contactor is engaged.

In DC mode, the auxiliary voltage is routed to the K2 contactor coil and the PS01 power supply, which in turn provides +5 V for the MES2 meter. The AC voltage from the variac is rectified by the full-bridge rectifier D1 and smoothed by capacitors C1 and C2. Because of the high maximum output voltage of +450 V, the capacitors are connected in series. Resistors R1–R4 form a voltage divider and balance the voltage across the capacitors. Additionally, they discharge the capacitors after the device is shut down. As in AC mode, the switch SW5 and the pushbutton SW6 energize the K2 contactor, which routes the output DC voltage to the banana terminals J1.

Components

The newish-looking components were sourced from Aliexpress; the quality was usually acceptable. Others came from my personal collection, often salvaged from older equipment or bought locally years ago. Below, I discuss the rationale behind the choices I made, along with a few caveats.

The isolated variable transformer with its cabinet

The isolated laboratory variac (T2 in the schematic) is TST 280/1E, manufactured by East Germany’s VEB Technisch-Physikalische Werkstätten Thalheim. When supplied with 230 V AC, it can deliver output voltages in the range 0–280 V AC. Its rated current capability is 2 A below 100 V and 1 A above 100 V. The cabinet dimensions are 25 x 25 x 22 cm. More details are available on the radiomuseum.org website. For reference, the markings on the nameplate:

VEB TECHNISCH - PHYSIKALISCHE WERKSTÄTTEN 
THALHEIM (ERZGEB.) - MADE IN GERMANY - DDR

TYP TST 280/1
NR. 93 693 E 75 g
PRIM. 220 V
Uis = 350 V
LEIST. 300 VA 50 Hz
SEK. 0...280 V 1 A
NUR FUR TROCKENE RÄUME!

I bought mine on OLX, one needle meter was broken. Fortunately, it pushed me into building the testing station. Here are photos of the device:

And here are the markings on the transformer itself:

VEB Techn.-Phys. Werkstätten
Thalheim/Erzgeb.  RFT
Made in Germany DDR

Typ: TST 280/1E
Prim.: 220 V 50 Hz - E75g
Sek.: 0-280 V 1A - Sich.sek. 1A

AC voltage/current/power meter

MES1 is a combined voltage/current/power meter of type Sinometer SPM003, available on Aliexpress. To work with a variac, it needs to be modified. One must split the phase (L) terminal into two separate lines: one for powering the meter and the other for measuring the voltage itself. I described the modification in another article. The photo below shows how I installed the current transformer on one of the output lines:

Small auxiliary voltage transformer

Powering the contactors (K1, K2) and the MES1 meter theoretically requires an AC supply voltage of ~230 V. For the contactors, I could get away with powering them directly from the mains, but the meter has to be galvanically isolated from the mains. I didn’t have a small isolation transformer at hand, but I figured out I could get the required isolated voltage from the upper leg of the T2 variac.

However, the nominal 280 V (not to mention over 330 V, when the grid voltage nears its maximum) is too much. Fortunately, I had a small 24 V transformer T1 lying around. It had been extracted from a broken three-phase welding machine. The primary winding was for line-to-line voltage (400 V AC, 380 in the past), but it had a tap for the more familiar line-to-neutral voltage (230 V AC, formerly 220). Using the primary winding as an autotransformer allowed me to convert 280 V AC into ~160 V AC. This falls nicely into the wide range of allowed supply voltages for the MES1 meter. Any similar low-power transformer with taps would do.

As for the contactors, 160 V AC on their coils seems a bit low to close the contacts reliably. According to the contactor’s datasheet, the lowest acceptable voltage for a 230 V coil is 184 V. Mine worked without problems, however. This is not a proper design, though.

Contactors

In this article, I call the K1 and K2 elements contactors rather than relays. One might ask: what is the difference between a contactor and a relay? Basically, a contactor is a type of relay designed for heavy duty and almost always equipped with arc suppression features.

Arcing during disconnect is why a relay’s contacts are rated for much lower current on DC than on AC. An AC arc tends to quench when the current crosses zero. With DC, the arc may sustain for a long time and damage the contacts.

In my testing station, I used so-called industrial relays from the R15 series, manufactured by Relpol. For different kinds of load, the contacts are rated as follows:

As shown in the table, the rated current depends strongly on the load type. For 250 V DC and inductive loads, the rated current is as low as 100 mA. Assuming the power supply can deliver 3 A (150% of the nominal overcurrent protection limit), these R15 relays are too small for use in a DC circuit. A proper design would use specialized DC contactors.

But this is not an utter tragedy, because:

• There are three, not one contacts in the current path, so it is much less likely for the arc to sustain after disconnection. Arcs in series are very unstable and tend to quench quickly. That is why proper contactors usually have two contacts in series.
• The relay power ratings are for 1200 cycles per hour, orders of magnitude more than the switching rate in my testing station.
• After all, this is just a hobby device, used irregularly.

Summing up, there’s no concern about the relay’s reliability. The only valid concern is arc sustain after disconnection, which is very unlikely given there are three contacts in the path.

Switches and buttons

The switches are the contemporary Aliexpress variety, sized 31 x 25.5 mm, and commonly found in power strips and similar devices. Nothing unusual here. SW3 and SW5 have integrated neon bulbs, colored as described in the section above. The appropriate neon bulb lights up when the contactor is engaged. This happens when the switch or the pushbutton is activated.

I strongly discourage soldering directly to the switches’ terminals. The plastic melts almost instantly. Even if you don’t destroy the switch outright, the contacts may no longer be positioned correctly and can become unreliable. I recommend using 6.3 mm blade terminal sleeves instead.

Output connectors

DC output is available on laboratory-style banana connectors. AC is available on banana connectors too, and I also added a standard Polish unearthed socket (CEE 7/1) and a standard earthed American socket with the earth sleeve unconnected (NEMA 5-15). The American socket may come in handy if I stumble upon a device that requires it.

I warn against Aliexpress-grade banana connectors—they are of abysmal quality. I opted for high-quality connectors from a military surplus store. Mine are spring-loaded and allow attaching two cables and a banana plug at the same time.

Overcurrent breakers

Originally, the TST 280/1 was protected by two traditional blow fuses: 2.5 A on the primary side and 1.6 A on the secondary. Not wanting to keep replacing them, I opted for bimetallic overcurrent breakers instead. Chinese MR1 thermal breakers seem good enough for this purpose. I chose the closest nominal values to the original ones: 3 A on the primary side and 2 A on the secondary. The breakers’ tolerances are not great anyway.

Before installing them permanently, I tested their behavior by shorting the variac’s output. With the variac’s output shorted, the bimetallic breakers take several seconds to trip. When the variac’s output voltage was set to 100 V or less, the secondary-side breaker trips first. Above 100 V, the primary-side breaker trips first. This is logical, because the current capacity decreases as the output voltage increases.

Lightbulb

An E27 lightbulb socket was placed just behind a rectangular window cut in the front panel, so one can observe the filament. The window’s bezel was taken from a panel meter of unknown origin. I put a 230 V 60 W lightbulb there. More about limiting current with a lightbulb is in the section above.

There is an issue, though: the lightbulb in my setup is installed inside the enclosed cabinet and quite close to the other components. Running the testing station with the outputs shorted causes the lightbulb to run at full power. Doing this for a prolonged time might be a bad idea, because things may overheat.

DC power supply

The circuit is so simple that it didn’t require a custom PCB, so I assembled it on a universal prototyping board. The full-bridge rectifier D1 should be rated for high current. Not because of normal operating conditions, but because of the possibility of output shorts. It has to survive a large current until one of the bimetallic breakers trips. I took mine from an ATX power supply and mounted it on a piece of aluminum, which acts as a heatsink.

Electrolytic capacitors C1 and C2 are in series because of the high maximum voltage, which can exceed 450 V. In fact, there are four capacitors; each parallel pair provides ~220 μF, for about ~110 μF total. The caps are taken from the ATX power supply as well. Resistors R1–R4 discharge (bleed) the capacitors and balance the voltage across them. When the user sets the output voltage high with no load and then sets it low, the capacitors must discharge through R1–R4 before the MES2 meter shows the correct voltage. To limit the discharge time, C1 and C2 are of relatively low value.

I couldn’t speed up the discharge by lowering the resistances because they would heat too much at high output voltages. Keep in mind that, at the top voltage of 450 V, the resistors dissipate ~5 W of heat overall, so they should be properly rated. Another reason for the low capacitance is that there is simply no space for larger capacitors.

DC voltage/current meter

MES2 is a voltage/current meter of type YB4835VA. It comes in several versions; mine is rated for up to 600 V 3 A. It requires an external power supply of 3.5–30 V DC, I supply it with +5 V from a phone charger. Nothing unusual here. Phone chargers usually accept a wide range of input voltages, so it works perfectly with ~160 V AC, as described in detail in the previous section.

Making the front panel

For the front panel, I ordered a piece of 4 mm aluminum sheet, cut to 240 mm x 235 mm. It had the same dimensions and thickness as the original panel. Cutting all the holes for the components turned out to be a lot of work. For round holes, I could get away with a drill, but for the rectangular ones, I had to use a Proxxon scroll saw I had at hand:

I’m not sure if it makes any difference, but I used Niqua ANTILOPE Yellow 130 mm blades. 4 mm aluminum is quite thick for a small scroll saw, but despite breaking several blades, I managed. Different stages of making the panel are documented in the photos below:

Assembling the internals

Inside the cabinet, I installed a piece of aluminum to hold the components. They all came from scrap metal; one of the pieces is a 1U rack panel mount cut in half. The cables are usually 1.5 mm², recycled from a broken washing machine. It’s nice that they come in so many colors because that makes assembly much easier. When it made sense, I used blade and ring terminals. Different stages of assembly are documented in the photos below:

What can you do with it?

Over four years of use, the device has come in handy many times. Some examples are listed below:

Simulating extreme power conditions

Checking how your device behaves under unusually low or high supply voltages.

Varying drill speed and limiting the torque

In one setup, I used an old Soviet drill as a drive motor. You can easily regulate the rotational speed with the variac’s dial. Moreover, you can limit the torque generated by the drill with the lightbulb. The lower the lightbulb’s wattage, the lower the maximum torque. With the lightbulb in series, you can even stall the motor for a while. Probably not indefinitely, though, because a stalled motor is not cooled by its integrated fan, yet current still flows through it.

Powering vacuum tube devices

Such things as renovating old radios come to mind. The testing station can act as an anode power supply. It can deliver almost +450 V, which should be enough for most circuits.

Powering American-made devices

The American socket might come in handy if I put my hands on a device that requires it. For example, I heard that importing KitchenAid mixers from the US is quite popular in Poland, because they are much more expensive here. Remember: the US uses 120 V AC at 60 Hz, while Poland uses 230 V AC at 50 Hz. It’s not very practical to power a kitchen mixer from a laboratory-style power supply, though. People use small 110 V to 230 V autotransformers instead.

High voltage capacitor reforming

When electrolytic capacitors sit unused for a long time, their internal structure degrades. Just slamming them with full voltage might cause them to fail or even explode. More about this here. Under some circumstances, you can save a capacitor through reforming.

Reforming should be done with a lightbulb limiter active and, additionally, a series resistor that limits the current to 1–3 mA. Two or three resistors in series, something like 100 kΩ overall, should be enough. Voltage should be increased gradually over at least a few hours, or better—days. If the current doesn’t drop to zero within, say, a day, then the capacitor is broken.

When I finally redo the coilgun I made as a kid, I’ll surely need this.