Access Official Technical PDF

Share Summary:LinkedInEmail

For power-conversion designers, faster switching is not simply a matter of selecting a wide-bandgap material. Device capacitances, gate-driver capability and the way transistors are connected all influence switching time and energy loss. The all-GaN integrated cascode is a device architecture intended to combine enhancement-mode operation with the reduced Miller-effect constraints of a cascode, while avoiding the use of a separate silicon transistor.

The configuration pairs an enhancement-mode (E-mode) GaN device with a depletion-mode (D-mode) GaN device in an integrated structure. Its switching behavior depends on how those sections share current and voltage, as well as on capacitance at the internal node. Understanding these relationships helps engineers assess the design’s potential and the trade-offs that must be managed.

Why consider an all-GaN integrated cascode?

AlGaN/GaN heterostructure field-effect transistors (HFETs) use a high-density two-dimensional electron gas (2DEG) as their channel. This supports high current density, while GaN’s material properties make it attractive for high-voltage and high-frequency power switching. However, conventional HFETs are generally depletion-mode devices: they conduct at zero gate bias and require a negative gate voltage to turn off. Power applications commonly favor normally-off, E-mode operation for safer behavior when control power is lost.

A cascode places two transistors in series. In the conventional GaN-plus-silicon version, a high-voltage D-mode GaN HFET is paired with a low-voltage silicon MOSFET. The silicon device sets the gate-controlled, normally-off behavior, while the GaN device supports the high blocking voltage. But the silicon transistor brings its own input capacitances and body diode, and connections between separate devices can add parasitic inductance.

An integrated all-GaN approach replaces the silicon transistor with a GaN E-mode section. Integration can remove the discrete interconnect between the two devices and avoids relying on a silicon MOSFET’s body diode. The key design challenge is to preserve the cascode’s switching advantage while controlling the added capacitance and matching the current capability of its two sections.

How the cascode reduces the Miller effect

In a standalone power transistor, the drain-to-gate capacitance, commonly called the Miller capacitance, couples drain-voltage changes back into the gate circuit. During a switching transition, this feedback current can limit how quickly the gate voltage and drain voltage change. The effect becomes especially important when the gate driver cannot supply or sink current comparable to the load current.

In a cascode, the high-voltage device is separated from the gate by the lower-voltage transistor. This shifts the relevant Miller capacitance to an internal node that experiences a comparatively low voltage rather than the full drain swing. With less voltage change acting through the gate-drain coupling, the cascode can achieve a faster output-voltage transition than an equivalent standalone device under suitable conditions.

Analysis of the integrated configuration identifies an important qualification: the advantage is influenced by the ratio of gate-driving current to load current. A relatively small gate-driving current compared with the load current can make the cascode’s reduced Miller-effect limitation more beneficial. This is not a universal guarantee of faster switching; the device, driver and operating conditions must be considered together.

Balancing switching speed and internode energy loss

The internal connection between the E-mode and D-mode sections introduces an internode capacitance. During switching, energy can be stored in and transferred through capacitances at this node. If the associated capacitance is too large, the resulting energy loss can offset part of the benefit gained by reducing the Miller effect.

Designers therefore need to consider more than the gate-drain capacitance of either section in isolation. Relevant factors include:

  • Current matching: the E-mode and D-mode sections need compatible current drivability. In the reported device, matching these parts was used to optimize output current.
  • Internode capacitance: reducing the D-mode device’s drain-source capacitance can suppress additional capacitance-related energy loss at the internal node.
  • Gate-driver capability: the available gate current affects how strongly Miller-related feedback constrains the transition.
  • Turn-off control: the connection of the field plate can change how gate resistance influences turn-off behavior.

These dependencies make integrated-cascode design a system-level task. A device that appears favorable based on one capacitance or switching metric may still require careful evaluation of the complete gate loop and switching conditions.

Field plate design and dynamic on-resistance

Field plates shape the electric field in a GaN HFET and can help address both high-voltage operation and capacitance. Their geometry and electrical connection matter because changes that improve one characteristic can affect another. Field-plate design is also relevant to dynamic on-resistance, the increase in on-resistance that can occur after a device has been exposed to off-state high voltage.

The reported work found field-plate geometries effective in suppressing dynamic on-resistance and reducing Miller capacitance. In the integrated cascode, a source-connected field plate can reduce the D-mode device’s drain-source capacitance, helping limit additional energy loss at the internode. A field plate connected to the E-mode gate offers a different design benefit: improved control of turn-off switching through the gate resistance.

These options serve distinct purposes. The source-connected arrangement targets internode capacitance loss, while the E-mode-gate connection provides a means of adjusting turn-off behavior. Selecting between them requires weighing switching loss, dynamic on-resistance and the desired degree of turn-off control rather than treating field-plate geometry as a standalone optimization.

What the demonstrated device establishes

The fabricated all-GaN integrated cascode used fluorine treatment and a metal-insulator-semiconductor gate structure to achieve a threshold voltage of +2 V. Its output current was optimized by matching the current drivability of the E-mode and D-mode sections. Switching-speed behavior was demonstrated with a double-pulse tester at 200 V, including a comparison with equivalent standalone GaN devices.

The results provide evidence that the cascode’s switching advantage can be demonstrated in an all-GaN structure, and the accompanying analysis explains the role of reduced Miller effect. They also highlight design constraints: the internal-node capacitance must be minimized, and the device sections must be appropriately matched. These findings describe a device demonstration, not a guarantee that every integrated cascode will outperform every standalone GaN transistor across all operating conditions.

Practical implications for power-electronics teams

For engineers evaluating GaN switches, the all-GaN integrated cascode offers a way to combine normally-off control with cascode switching behavior without a separate silicon device. Assessment should include the gate driver, load current, output capacitances and internal-node behavior, rather than focusing only on nominal switching speed.

Field-plate connections add another design lever: source connection can help reduce internode capacitance loss, while an E-mode-gate connection can improve turn-off controllability through gate resistance. The practical takeaway is to treat these features as a coordinated set of choices. The architecture’s potential depends on balancing Miller-effect reduction against capacitance-related loss and matching the electrical behavior of the integrated device sections.

Access Official Technical PDF

About The Author