Safety in a big, heavy package.
The Stancor transformer pictured earlier was not yet in my collection when this project was launched. Although an isolated supply can be effected by connecting two identically-rated transformers together on their secondaries, and then using the primary of each unit as the input and isolated output, I wanted a sustained current of at least 5A. I also wanted circuit breakers on the isolated outputs, with two different ratings. I wanted irrational, unnecessary complexity. In other words, I needed to build it myself.
First, I needed a big transformer, so I watched eBay for a couple months. Finally, a Canadian seller listed something that looked right. I got a bargain price for a unit this size and it arrived a couple weeks later:
Figure 3. Transformer for the isolated power supply.
Yes, that says 1380VA. Even better, it is rated at 50Hz, which means the core was derated by about 10% compared to 60Hz operation, making it theoretically capable of 1500VA in my application. The universal input taps allow the unit to be run from several common voltages at 1:1 or 2:1 by connecting the appropriate taps in series or parallel.
As expected, this unit is heavy. One can only guess at the postal carrier's thoughts between their truck and my front door. I knew my chassis was going to sag under the weight, so I set about reinforcing it with steel box tubing and a 3/16 inch (5mm) aluminum base platter. The chassis is a standard 19" rack width and three rack units (5.25") high, powdercoated steel with an aluminum front and rear. I started modifying it like so:
Figure 4. The chassis was reinforced to support the transformer mass.
In bare form it looks hackish, but it works. The assembled unit is stable and shows no signs of chassis torquing.
A large transformer is the electrical form of a loose cannon at power-up. The winding and core form a simple inductor. When the transformer is connected to power, the magnetic field must be re-established in the core before the current through the wire calms down, and a large inrush spike results. It then decreases exponentially to the excitation level, which is the current required to maintain the magnetic field. In small transformers the inrush current is a non-issue, but large transformers often have a primary winding resistance of less than 3Ω.
In toroidal transformers, the situation is made worse by the phenomenon of magnetic imprinting. Although all transformers can do this, the toroidal core is especially prone to "remember" the switch-off polarity of the AC waveform. If the AC polarity at switch-on coincides with the imprint's polarity, the core will saturate momentarily, causing a severe current spike. At minimum, substantial nuisance flicker will occur on the AC mains. At worst, protective devices may be tripped.
Consequently, larger transformers are often soft-started to reduce the impact on the mains and the startup stress on the transformer, which initially gets strangled by its own windings due to the magnetic forces generated. The steel core doesn't mind too much, but the windings and insulation are less resilient.
A soft-start places an impedance in series with the transformer primary winding when connecting power, then bypasses it after a brief delay. It may be resistive, reactive, or both, depending on the application. But it cannot remain in circuit permenently or it will impede the transformer's normal operation. The cheap-and-cheerful method is to place an NTC thermistor in series with the primary. On a 120V or 240V unit up to 1000VA or so, a cold value for the device might be 5Ω and the "negative" behavior means the thermistor's resistance is expected to decrease as it heats, becoming negligible under normal operation. (A typical resistor will show PTC behavior.)
The problem is knowing whether the device will heat up enough to "drop out" of the circuit. If the transformer magnetizing current plus any continous load on the secondary are insufficient, the NTC device may harm more than help. The superior approach is to place the ballast, whether NTC or conventional, in series with the primary and bypass it with a relay contact after a brief delay, as follows:
Figure 5. Simplified schematic of a soft-start application.
The timer can be a simple analog RC circuit driving a transistor switch, or it can be a digital control. Regardless, the timer must detect initial power-on, and when it expires it must switch on a device such as a relay that shunts around the ballast resistor. The limit of this approach is that a failure of the shunting device will leave the ballast in-circuit where it may become extremely hot.
In my application, I wanted to keep the high-voltage section fully separated from control section. Of course, that's pretty much why transformers are used in the first place, so what I was really trying to do was rationalize a bank of solid-state relays (SSRs) that I wanted to play with. These aren't ideal for toroidal applications (see this useful ESP article for an explanation). Fortunately, this transformer seemed to be comparatively immune to the DC offset, though I've had others that were not.
The incoming AC power would feed directly from the incoming safety breaker to the transformer via the SSRs, and the transformer output would feed through additional SSRs before actually leaving on the isolated outlets. The control section would then switch the SSRs in the appropriate sequence. The following schematic illustrates the basic design (click on the image for a larger view):
Figure 6. System control diagram. (Click image for a larger view).
Yes, you see that correctly: I was intending to use 120V relays to switch 15V to the SSRs to switch 120V to the transformer soft-start and outputs. Each SSR has an optocoupler at the control input, so the arrangement ensures a high degree of separation for the high-voltage section. (It's also completely unnecessary, but again, this is DIY. Go nuts.)
The four requisite SSRs are attached to a single heatsink, which should be adequate for full-current duty once aggressive fan cooling is added. A 65C thermal switch is included on the sink. It shuts down everything except the fan circuit if ever tripped, providing backup protection against an overheat.
Figure 7. The heatsink with four SSRs attached.
An SSR mimics the function of a single-pole, normally-open relay contact using electronic components. Physically, it is an opto-isolator coupled to a thyristor, and epoxy-potted into a heat-sinkable package with screw terminals that access the equivalent of "coil" and "contacts". The "coil" input drives the LED in the optoisolator, and the receiver side of the optoisolator is powered from the "contacts" side, where it drives the gate of the thyristor.
Unfortunately, thyristors do not have a perfect 'off' state and can produce a DC offset around the AC crossover point so this approach may not work for everyone. They also tend to fail short-circuit in a severe fault and then go slightly nuclear if the fault current persists, so the overall safety benefits are kind of a wash. Never omit the fuse when working with these things. Or just use a relay.
The soft-start ballast resistor is co-located on the opposite side of the heatsink:
Figure 8. The heatsink with the soft-start ballast attached.
Heatsinking is not required for the resistor's normal operation, since the ballast is in the circuit for less than a second. The intent of mounting it here is that it keeps it out of the way and provides a failsafe if the shunt circuit ever failed to operate, leaving the resistor in-circuit under load. I even added thermal sleeving to the ballast wiring at each heatsink penetration. Point of advice: if you ever have a coffee maker or toaster oven fail, strip it for parts. Or buy one you don't care about from the thrift store, and wreck it. There's a decent quantity of high-temperature wire and sleeving inside one of those.
The final component in the control system is the control board:
Figure 9. The main control board.
The master AC input feeds through a 15A double-pole circuit breaker before feeding anywhere else in the unit. After that, the isolation transformer circuit is fed directly, while the control circuit receives 120VAC for control power at the fuse and noise filter on the upper left and passes it out to a 50VA, 120:25V-CT transformer. The returning 12-0-12 feed is rectified and filtered. The negative rail drives the system cooling fans, while the positive rail circles out through the 65C thermal switch before returning to the control relays.
The relays are ordered and timed per the earlier control diagram. Each successive timer feeds from the contact of the previous, guaranteeing that the supply will not finish starting if an earlier relay fails to operate. The third relay, in addition to delivering power to the output SSRs, also illuminates an LED on the front panel to confirm the front panel outputs are active. Both timers are modeled after Elliott Sound Products Project 39, Figure 2. The 2W power resistor by each relay is a dropper to bring the coil voltage down to about 10V, from the 15V nominal that the supply delivers. The 5W resistor at lower right is a dropper for the fans, which are 12V rated.
If the 65C thermal switch trips, the fan circuit remains on but power is cut to the relay bank until the switch resets.
aaronv dot net -at- gmail dot com