As described in the introduction to this series, the pre-regulator used in the Elektor 400W power supply project is very similar to a light dimmer, but designed to handle the inductive load of the transformer. I will attempt to describe the theory of operation below, based on my understanding of the circuit and the description provided in the Elektor articles. Additionally, this application note from SGS-Thomson describes a similar topology.
Theory of Operation
The original schematic of pre-regulator is shown below.
To start, assume the AC input terminals are at the zero crossing – in other words, the same potential. Now, let the the top terminal (marked COM above) start rising above the bottom terminal (B). C1 slowly charges toward B through R30 + P1 in parallel with R31 and the diode bridge / opto (more on that in a moment). When C1 charges enough that the breakover voltage of the diac is reached, the diac turns on, triggering Tri1 and allowing current to flow through the transformer.
For resistive loads, the circuit will function correctly because the load current will keep Tri1 enabled until just before the zero crossing when the operation will repeat, only in reverse (the other AC phase). By varying the charge rate of C1, the trigger instance can be varied, thereby changing the amount of power delivered to the load. A longer delay results in less of the AC waveform being applied to the load. In the circuit above, the trigger timing can be varied by changing the drive to the optoisolator. The diode bridge essentially turns the optoisolator into an AC variable resistor.
Everything works well until an inductive load is placed on the output. In this case, the current is no longer in phase with the voltage. When the current reaches the zero crossing, Tri1 will turn off prematurely before the voltage zero crossing. To keep Tri1 enabled until the voltage zero crossing, additional trigger pulses are used. These are created by R33 and Tri2. Initially, Tri2 is disabled. Once the first trigger pulse fires and turns on Tri1, the current through R32 causes a voltage drop that triggers Tri2. This then allows the lower valued R33 to control the charge rate of C1, resulting in a much faster charge rate and creating a sequence of pulses to keep Tri1 enabled.
The rest of the components serve secondary purposes. R27, R28, and the associated diode bridge make sure C1 discharges completely at the zero crossings. D1-D4 set up a regulated voltage rail (biased by R29) to make the triggering more consistent. C2 and R34 form a snubber network.
To understand how a circuit works, it always helps to build a prototype. For years, I’ve had the pre-regulator built but never had the nerve to fire it up until recently.
WARNING: HAZARDOUS VOLTAGES EXIST IN THE FOLLOWING CIRCUIT. DO NOT ATTEMPT TO BUILD IT UNLESS YOU KNOW WHAT YOU ARE DOING. PROCEED AT YOUR OWN RISK!
All the following scope shots were taken with the COM terminal connected to the neutral line. This way, the GND of the scope could be connected there without the risk of dangerous voltages or nasty line-level short circuits. First, a look at the trigger pulses:
Trace A (20V / div) shows the voltage at the diac. Trace B (50V / div) is the line voltage. The trigger delay (from the zero crossing) can be seen, as can the repeated trigger pulses. For some reason, though, in this prototype Tri2 does not stay latched in the positive phase. The C1 charging time remains the same. In the negative phase, though, Tri2 behaves correctly and latches, allowing R33 to greatly speed up the charging cycles creating a fast train of trigger pulses.
Below is a scope shot with less current through the opto, resulting in a greater delay for the initial trigger pulse. However, the behavior afterwards is similar, continuing until the line voltage reaches the zero crossing.
Below is more detail on Tri2.
In the negative phase, it can be seen how Tri2 is latched on (trace C drops and remains low). In the positive phase, Tri2 fires when Tri1 is triggered (the spikes on C), but does not stay enabled. I am not sure why this is the case, but the triac used was an NTE replacement, so it might not be a true 4-quadrant triac. In any case, the basic behavior has been demonstrated.
Next is a series of scope shots showing the trigger pulses (A, top trace) and the voltage across Tri1 (D, bottom trace) with various control voltages on the optoisolator.
Since the load was a transformer with a small resistive load on the secondary side, no problems were seen on the positive side even though the trigger pulses were not as fast as expected.
Overall, the prototype worked better than I ever expected and led to a much better understanding of the circuit.
The next step is to build a better prototype using modern thyristors and surface mount technology. The original circuit has been drafted in gschem. The updated design will be released shortly – stay tuned.