BFFT Techblog: High Voltage to C / Switched Current Source Precharge
Subject: Say good-bye to short circuits! – How to charge a high voltage capacitor without power loss.
Anyone who has ever tried to connect a capacitor to a running power supply unit knows the reaction and is familiar with the “sparkler effect.”
The reason this occurs is that as the capacitor is plugged in, it is suddenly charged with no resistance and thus experiences a short circuit. This is because capacitors and connecting cables have a very low series resistance by virtue of their design.
This generally also happens with every household appliance when it is plugged into an outlet and its input capacitors are charged.
It only becomes problematic when multiple devices are plugged in at the same time (e.g. with a multiple socket-outlet) and cause the fuse to blow.
This situation can definitely lead to problems when dealing with higher voltages of several hundred volts, larger capacities in the millifarad range and, accordingly, with higher currents.
The following explanations are supplemented with diagrams and wiring schematics from simulations (in LTSpice .) Such tools are very useful in hardware development for reviewing the basic functionality of circuits before they are implemented.
Example: Connecting a 1000V battery to the intermediate circuit capacity of an electric motor
C = 1mF, U=1000V, R= 100mOhm
Figure 1: Equivalent circuit for charging capacitor by simply connecting a 1000V battery with no pre-charge
Figure 2: Voltage profile, current profile, power loss through component resistances
During the short impulse, the energy released by the battery (in this case approx. 1kJ) is half transferred to the capacitor and half converted into heat energy. The losses occur in the line resistances, contact resistances and internal resistances of the components.
This situation always arises when the drive motor (along with its intermediate capacitors) of an electric vehicle is connected via protective devices to the high voltage battery.
The original reaction would be a huge flow of current (e.g. 10kA in this example). In most cases, this would cause the switch contacts to weld together, destroying the protective devices, since they are “only” designed to handle normal operating currents, depending on cost, weight and installation space.
So in order to prevent this process, the capacitor should ideally already be charged to operating voltage when the protective devices are to be closed.
This requires a pre-charging circuit.
In the simplest case, the capacitor can be charged with a resistor which is connected parallel to the opened protective device in a subsequent switching phase.
Due to the limited current flow, it takes a certain amount of time before the capacitor is charged. Depending on the permissible power loss (size and design of the resistor,) this can lead to a noticeable delay before the protective devices switch on.
This effect is also amplified by the fact that the maximum charging current is only present at the beginning of the charging curve (e-function,) after which it falls continuously.
.Old School 2.0
In attempting to optimize the process, we might start by allowing a much higher charging current and introducing it as short impulses. At first, we would only connect the resistor very briefly so as not to exceed the resistor’s permissible impulse energy. The impulse duration can be increased a little at a time, since the differential voltage, along with the current and power loss, will gradually decrease. Generally, the average power loss per time interval which occurs in the resistor can be kept constant, and a maximum charging speed is reached by the time the capacitor is fully charged.
Figure 3: Circuit for charging capacitor via pre-charing resistor with PWM control
Figure 4: Voltage profile, current profile, power loss via pre-charging resistor with PWM control
Disadvantage: With this method, too, we would still transform the same total amount of energy into heat through power loss in the resistor, which requires large components or long charging times accordingly.
Now here’s the trick
So, the energy needs to be transferred to the capacitor in a more controlled manner and not lost as heat in large amounts. To do this, we could use a theoretically lossless constant current power supply.
The energy can be buffered in a coil. This is the principle behind practically every simple switching controller for efficient voltage conversion.
For a suitable topology for using this characteristic for a power source, refer to the basic circuit in Figure 5.
Figure 5: Circuit for charging capacitor via constant current power supply
Figure 6: Voltage and current profile of the capacitor, power output of the battery
Figure 7: magnified view of a time segment from Figure 6
Now, all that is left is the “actual” power which the battery needs to provide, shown in the bottom diagram in Figures 6 and 7. In addition, there is also reactive power produced, which is needed to generate the field and thereby limit the current flow in the coil. But it then dissipates again. In this way, we transfer nearly all of the energy directly to the capacitor.
Requirements for implementation
Unfortunately, the reality is not as simple as it would seem at first glance. With this sort of circuit (and particularly with these high voltages,) we create new problems altogether.
Even the components selected for the switch and the coil are problematic if your goal is to pursue a compact design. For example, coils which have the right voltage resistance are normally designed for continuous operation with high voltage and are accordingly very large. We need to have appropriate insulation between the individual windings and large air and creepage distances. A wound coil with the available core materials and geometries would have to result in a relatively large design in order not to become saturated under the present operating conditions. For this reason, we decided on a planar air-core coil integrated in the circuit board.
A high voltage MOSFET is used as the switching element. The transistor is actuated with the help of a galvanically isolated gate driver.
A shunt is used to measure the current. The voltage drop which occurs there is converted into the control signal for the transistor by a galvanically isolated comparator circuit with hysteresis.
When the first prototypes were commissioned, several effects were observed which had not been present in the previous simulations with LT-Spice. These were caused, on the one hand, by parasitic characteristics of the circuit board and, on the other hand, by the actual properties of the transistor and other components. In the end, by optimizing the circuit board layout and the circuit itself (e.g. adding a snubber network to reduce interference when switching the transistor) we were able to achieve proper function.
There are different ways to achieve our goal. Which solution is right depends on the basic conditions. The last solution shown here is the best with regard to weight, size, speed and efficiency. The down-side is that it is more complex and more costly than the conventional version with simple resistance.
Author: Matthias S. (Concept development & Tooling)
Picture source: Matthias S.