The circuit breaker test sets I have been involved with use an AC switch consisting of two SCRs wired in reverse parallel connection. For some test sets, a fixed voltage of 240 or 480 VAC is switched, while in others it is a variable voltage of 0 to 560 VAC. The currents involved may vary from only several amperes to about 200 amperes continuous, and pulses of 100 mSec or so up to at least 1000 amperes. The load is a step down transformer with an output of 6 to 15 VAC or so, and it is connected to circuit breakers up to 6000 amps. Breakers are typically tested at 3x for time delay (18,000 amps for 30-90 seconds), and instantaneous at 10x (60,000 amps for 0.02 to 0.05 seconds).
The load is primarily inductive, so the SCR firing circuit is adjusted for an initial firing angle of about 70 degrees. It is adjusted so that all half-cycles are approximately the same peak value. Also, the controller is set to produce an integral number of cycles (typically 5), when pulsing the output current below instantaneous trip value. If an extra half-cycle occurs, there is a net DC component, and the transformer is essentially magnetized (remanent magnetism). When this happens, the inrush current of the next pulse is extremely high, sometimes enough to trip a 200 ampere mains breaker, and causing a very audible noise in the conduit as the cables slap against the pipe.
I have used several designs for firing circuits. Originally we used commercially available controllers which used high frequency pulses to fire the gates, but quite often we would see waveform distortion because the current was not enough to keep the SCR in conduction. I designed several circuits that used DC for the gates, and made sure that the current was applied only when the anode had a positive voltage. However, because of the inductive load, there was current still flowing when the gate current was turned off, and distortion was seen.
Finally, about ten years ago, I designed a simpler board which kept gate current on continuously, and these boards have been used in hundreds of test sets with no apparent problems related to the gate drive. I use a simple constant current source with a PNP transistor and a 2.0 ohm resistor, and diodes, which limits the current to one junction drop (0.6V) on 2 ohms, or about 300 mA. It is sourced from about 12 VDC.
I am now designing a new board which will use a PIC to control the SCR firing. I plan to use DC/DC converters (12V to 5V at 200 mA), rather than transformers for the gate voltage supplies, to save size and also allow the circuit to run on 12 VDC. It will also have sensors to determine if the SCR is actually turned on when gate current has been supplied, and it will have other bells and whistles such as programmable phase angle, overcurrent detection, etc.
In researching gate drive requirements, I found a specification for the SCRs we now normally use, giving a guaranteed turn on current of 200 mA, at a voltage of 1.0 to 4.0 volts, but there was also a specification indicating that the gate should not have current applied when the anode is negative with respect to cathode. The previous design applies the 300 mA current during the full conduction cycle, which does not meet this specification. However, the other SCR is conducting at that time, so the anode to cathode voltage is only a couple of volts. I am assuming this condition does not do any harm except waste power. With my new board, I may be able to detect actual current flow and turn off the gate drive when current is flowing in the opposite direction, but it is easier to leave things as they are. Any comments on this?
I also found that the recommended gate drive consists of an initial high current pulse, followed by a "back porch" of lower current. I can do this fairly easily by adding a capacitor across the current limit sense resistor, but it will produce this waveform only for the first phase delayed firing. After that, I prefer to leave it on continuously. If I turn gate current off during the time of reverse conduction, I would need to retrigger at a very critical point just after the zero crossing, and any delay will cause distortion, and early firing will waste the peak current pulse.
Another problem I have seen is that, at very high current levels, there is often an additional half-cycle of current, which magnetizes the transformer and causes subsequent inrush problems on the next pulse. I found that this effect could be minimized by reversing the phase of the incoming power, or by reversing the gate connections to the SCRs. The 480 VAC supply is produced by an autotransformer on a 208 VAC source, so the voltage to ground on the two inputs is 360 VAC and 60 VAC. The case that works best is where the line side of the SCR is at the 360 VAC potential, and the load side switches from -60(off) to +360(on). The extra half cycle occurs when the line side stays at -60 and the load switches from +360(off) to -60(on). I think there is some turn-off transient that is being conducted into the firing circuit, and probably it is the one that is changing voltage to ground. If this circuit controls the gate of the SCR that sees a positive line voltage excursion, it fires and causes the extra half cycle. The SCR board is normally mounted on the SCR heat sink, which may be at the line or load side of the switch, but I think I have also seen this problem when the board has been mounted remotely. This may take more research, but, again, any comments or suggestions will be appreciated.
Thanks for taking the time to read this long post. Perhaps you may find it interesting, and responses may be helpful to anyone dealing with similar applications. More information on the general technology of high current primary injection testing is on my website.
Thanks!
Paul E. Schoen