If even a small increase in efficiency could make a difference in product performance when using power converters, or if you have an application where power density is critical, you may want to consider switching from traditional silicon MOSFETs to gallium nitride (GaN) transistors. However, like silicon, GaN has limitations in the size of its production die, which limits the current capacity that a single device can handle.

To increase output power, parallel connection of switching devices is mainly used, but various technical considerations are required as there is a possibility of problems occurring during design.




Considerations when connecting GAN switching devices in parallel

The first thing to consider when operating switching devices in parallel is the RDS(on) of the devices you want to use. That is, device matching through equal static current flow sharing between parallel switches.

The second consideration is dynamic switching. If the gates of the switches are not symmetrical, not only will current sharing between devices be unequal, but circuit parasitics will be mixed in the current flow, which can cause undesirable high-frequency oscillations. This violates electromagnetic compatibility (EMC) requirements and, in severe cases, can even damage the switch itself.

While parallel operation of silicon MOSFETs is well known, the understanding of GaN HEMTs (High Electron Mobility Transistors) is lacking, even when driving a single device. Furthermore, GaN devices have very fast switching speeds, so a thorough understanding of these power transistors is required when trying to operate them in parallel.


Drive circuit design

Devices such as Infineon's CoolGaN™ 600 V HEMT use p-type impurity GaN at the bottom of the gate. This makes the threshold voltage of the device positive, which is typically very low, around 1.0 V to 1.5 V. As a result, the gate forms a pn-diode with a forward voltage of around 3.0 V and a resistance of a few Ω. Then the driving circuit is very different from that used in conventional MOSFETs.



The gate input of the CoolGaN™ transistor can be thought of as a diode with a forward voltage VF of about 3.5 V in parallel with the gate capacitance CG. This diode makes the device difficult to drive because the gate node is clamped to a value close to VF when the Miller plateau is reached. Therefore, a negative voltage is required to turn the transistor off in hard switching applications. There is also a difference in the drive required during steady-state and off/on transitions.

The circuit in Figure 1 helps to solve this problem easily. It provides a low-impedance, fast AC path for Ron to charge Con and CGS. To do this, the driving voltage VS must be at least twice VF (8 V to 10 V is typical). Then, a parallel DC path is formed through RSS. Therefore, when properly designed, the on transition current is determined by Ron, and RSS determines the steady-state diode current.

When the gate is turned off, the charges on CGS and Con quickly equalize. For this to happen, Con must be larger than CGS. If there is a charge difference, the gate voltage VG becomes negative, which turns the transistor off in hard switching applications.



When operating CoolGaN™ HEMTs in parallel, the same RC driving network can be used on the gate of each transistor along with standard gate drivers commonly used in silicon MOSFET applications. A single gate driver such as the isolated EiceDRIVER™ 1EDI20N12AF is required, which can implement transistor turn-on and turn-off using the source (OUT+) and sink (OUT-) outputs.

When using an isolated 12V power supply for the gate driver, the supply is split into positive and negative supplies, and the negative supply is regulated to -2.5V. This prevents the transistor gate threshold from being exceeded and minimizes reverse conduction losses. It also ensures good gate voltage control even at low duty cycles, and prevents the RC drive network from going to 0V.

The impact of alternative current paths

Even if each transistor uses its own RC driving circuit, a shared current path can still interfere with the gate driver loop (Figure 2). Ideally, all current should flow along the intended path from drain to source, but inevitably some current will flow along the Kelvin source path.

If these paths do not have equal impedance and wiring, the VGS voltage across the eT gate loop may vary. Even a few millivolts difference in the gates can result in current sharing unequal by several amperes, which can cause serious oscillations when switching between the two transistors.



This problem can be solved by using a high impedance common-mode (CM) path as the Kelvin source path. By using a CM inductor between the gate and the corresponding Kelvin source driver return path and adding a 1Ω resistor to it, only a small amount of leakage inductance is introduced into the gate drive loop. Meanwhile, the shared current path becomes the differential CM inductance of both inductors. Careful selection of the inductor is required to avoid interfering with the gate driver operation. Figure 3 shows the impact of the CM inductor through the SIMetrix simulation results.

- Check the GaN HEMT evaluation board
!applications" target="_blank">- SMPS application cases for servers and communications
- Case study of ultra-fast wireless charging application
- Class D Audio Amplifier Application Case
- Check the GaN HEMT Datasheet
ms/en/product/power/gan-hemt-gallium-nitride-transistor/?utm_source=ed4s&utm_medium=referral&utm_campaign=202104_ap_en_pss_+pss.pphv_gan.2021-kr&utm_content=paralleling+gan+article#!trainings" target="_blank">- Watch CoolGaN Transistor Video Lecture