POWER
As we soon found out, the problem of driving multilevel switching inverters is still very much in the academic research phase of development. As such, we ended up looking through many research papers searching for an effective solution. The one we liked most was in a research paper, first shared with us by TA extraordinaire Ned Danyliw, titled Low-Cost Gate Drive Circuit for Three-Level Neutral-Point-Clamped Voltage-Source Inverter. A link to the paper can be found here. We really liked the second of the two drive circuits mentioned in the circuit because it allowed us to provide continuous power to the gate drivers regardless of the current state of the switcher or how fast we were switching. Many other solutions required the use of clever but unconventional switching patterns and switching frequencies in order to get any power up to the higher gate drivers. Below is a high level overview of the circuit we ended up using, pulled directly from the research paper mentioned above.
MGD stands for MOSFET gate driver. In the lowest high side driver, power is provided by running current from the high-side rail through a current source (we used resistors), through a reverse-biased zener diode to create a voltage drop and either to ground or through the other diode to the neutral point. The voltage difference through this path never goes below 50% of high side voltage being supplied, so power is supplied continuously. The symmetry of the switching cycle ensures that the lowest high-side gate drive puts the same amount of current that the highest high-side gate drive takes out. There are more complicated circuits called ‘Charge Control’ to power the higher two gate drivers because the corresponding FETs need to be driven at voltages above the high side rail while they are on. A diagram of the charge control circuit, also copied from the research paper, is included below.
At the center of the circuit is an oscillator (we used a 555 timer) what switches between the voltages of its power supply, which is supplied by a zener diode just like in the lowest high side FET. When the oscillator is low, charge flows into the capacitor CP1 through diode D5, charging it up. Then, when the oscillator goes high, it pushes the capacitor voltages up with it, and the charge can no longer go back through D5 so it overflows through D4, powering the gate drive at a voltage above the source. The one modification we made to this circuit was to add an extra diode (Schottky or something of the like) and resistor to the positive power rail of the middle high side gate driver. Without this, that power rail could only be charged once the MOSFET was already on, introducing a chicken and egg problem that we didn’t want to even think about. This modification made it so that truly every single gate driver was being powered at every single switching stage.