Gate Drive Design

One important parasitic in power electronics is the parasitic capacitance formed by the gate metal layer, the gate oxide, and the body of the transistor. The gate charge of a power device can ranged from a few nanocoulombs to several microcoulombs, corresponding to capacitances up to the microfarad range. It is important to understand the intricacies of gate drivers in order to design an efficient and reliable DRSSTC.

The Necessity of Gate Drivers

The existence of the gate capacitance means that the gate of a power transistor can never be driven by the output of a logic IC. Because of the low current capabilities of these logic outputs, charging the gate capacitance would require an inordinate amount of time, most likely longer than the duration of a switching cycle. Instead, a gate driver, which is an IC or circuit which can source significant amounts of current for a brief pulse, must be used.

The amount of current necessary depends on the transistor and switching frequency in question. For example, consider the following table, taken from the datasheet of a large IGBT module:

Gate Charge - physics demonstrations

This IGBT is quoted as having 7.2 μC of gate charge. In order to charge this in 150nS, we must provide 7.2x10-6/150x10-9=50A of drive current, far beyond the capabilities of any logic-level IC. 

Gate Drive ICs

The simplest way to drive a gate is to use a gate drive IC. This IC contains the necessary circuitry to source and sink high-current pulses, and usually simply requires a power supply and a logic-level input. Driver ICs typically come in an 8-pin DIP, SOIC, or power TSSOP package; the larger ones may come in a 5-pin TO-220 package.

A gate drive IC has two limitations - the peak current, which limits how short the rise and fall times can be, and the average package power dissipation, which limits the drive frequency. For this reason, the TO-220 packages are preferred for driving large IGBTs at high frequencies.

A common practice in the community is to use a discrete push-pull MOSFET pair to drive an IGBT gate. While this does provide cost savings, it adds significant complexity to the circuit and may suffer from shoot-through. For this reason, such a driver is typically less optimal than a driver using a high-current IC. With ICs capable of sourcing and sinking 30A available for a nominal price, the only reason to use a discrete driver is if more than 30A of current is required (and this is extremely rare, for reasons discussed later).

High-Side Isolation

In most power converters, there is at least one transistor whose emitter is not ground-referenced; typically, this transistor has its emitter connected to the load network, and swings at the same voltage as the load. For this reason, it is necessary to provide a floating output to drive this transistor. Furthermore, isolating the control circuitry and driver from the power components has an added benefit of minimizing ground noise in the low-voltage circuitry, as well as providing an additional level of user safety. This floating drive signal can be provided in several ways.

1. Bootstrapped Drivers

Bootstrapping - mini tesla coil

These drivers use a small capacitor to provide the necessary floating supply. The capacitor is charged through the bootstrap diode during the portion of the cycle when the low-side switch is on. Boostrapping has the advantage of requiring a minimum of external components, but suffers from poor performance as well as a lack of isolation. This technique is typically not used in Tesla coils, but is common in low and medium voltage motor control applications.

2. Discrete, Isolated power supplies

By far the highest performing and most flexible of the techniques mentioned here. A power supply with sufficient isolation is used to supply a standard driver circuit, whose input is optically or magnetically isolated from the rest of the control circuit, This technique has the advantage of providing the best rise times, and can operate at arbitrary duty cycles. It is sometimes used in large Tesla coils, and was once the technique of choice for driving large inverters built from IGBT modules, but this has largely been supplanted by gate drive transformers.

3. Gate Drive Transfomers

The gates involved are driven by the output of an isolation transformer. This technique is what the oneTesla product line uses, and is what is typically used on Tesla coils due to its simplicity and reliability. Gate drive transformers typically provide worse rise and fall times than other drivers, but due to the resonant nature of the Tesla coil this is less important than in other applications.

In order to design a good gate drive transformer, several things must be necessary:

1. The leakage inductance must be low. Any leakage inductance manifests itself as additional inductance in series with the gate. This reduces performance and causes excessive ringing.

2. The primary inductance must be high enough to keep magnetizing current (given by V/2πfL) low. For a given GDT core, primary inductance can be computed as A_L*N^2, where N is the turn count and AL is given in the datasheet. However, leakage inductance generally increases with increasing magnetizing inductance, so do not make the primary inductance too high.

3. The core must not saturate during operation. For ferrite cores, this imposes the constraint V/2fNA < 0.3, where A is the cross-sectional area of the core.

Designing a good GDT is not hard; typically, any ferrite core designed for transformer usage will work, but toroids are preferred. An alternative is to purchase a pre-fabricated GDT; in this case, leakage inductance, magnetizing inductance, and maximum "volt-seconds" (which governs saturation for a given geometry) will be given in the datasheet.

A Word on Switching Times

Standard power converter design paradigms say to make switching times as short as possible in order to reduce switching loss (bounded, of course, by dI/dt and dV/dt constraints of the switches used, and EMI considerataions). This is untrue for a resonant converter. Faster switching times correspond to increased switching transients, and in a DRSSTC, the phase lag caused by the switching delays contribute more to switching loss than the rise and fall times. Furthermore, switching loss is often minimal compared to conduction losses, and it is often more in the designer's interest to reduce transients and noise in the circuit.

For this reason, do not push for the shortest switching times possible; using the datasheet values for the gate resistor often results in more than sufficient performance. In fact, using a larger value of gate resistance can sometimes improve reliability. In a similar vein, using greater-than-datasheet values of gate voltage is discouraged. 

However, there is one exception to this: in cases of extremely high currents, the IGBT may fall out of saturation for a given voltage, resulting in a rapid increase in the forward voltage of the device and consequent device destruction. In this case it may be necessary to use a higher gate voltage to keep the device in saturation. Datasheets typically have graphs indicating forward voltage versus current for a given gate voltage.