GaN SYSTEMS 网络研讨会: GaN器件系统性能最优化的简单layout步骤

GaN HEMT的超低开关能耗使更高的系统效率和功率密度成为可能

卢俊诚, 应用工程经理

观看此视频 YouTube >

在这个研讨会,您将

  • 回顾

    回顾功率半导体行业在过去十年中不断学习和完善的设计理论和实践

  • 学习

    了解设计高效可靠的基于氮化镓系统的关键要素

  • 检查

    检查初级设计人员在氮化镓电源布局中最常见的错误

  • 发现

    发现设计规则和基本步骤,以避免最常见的设计错误,例如:振荡,附加损耗,EMI差,寄生导通/关断,器件过应力或开关转换期间的硬故障

问题与解答

We do not prefer film capacitors. Ceramic capacitors are usually preferred for decoupling compared to film capacitors because of their smaller package, which introduces less stray inductance to the power commutation loops.

The test boards are designed for pulse test only. In double pulse tests, very little heat less than 1mJ  is generated, so we did not apply a heatsink. For practical use, the need for a heatsink should be evaluated. In many cases, customers use the GaN solution because the heat sink can be eliminated.

Yes, there will be some parasitic capacitance between the top layer and the first middle layer. In practical design, we minimize the area of the switching node to reduce the parasitic capacitance between the drain and the source, and to avoid capacitance between the drain and the gate.

Although there is no body diode in a GaN HEMT, a GaN HEMT does freewheel current during the deadtime. In reverse conduction mode with Vgs < Vth, the Drain of the GaN will behave as a source, while the source will behave as a drain. When Vgs>Vth, the GaN device will be fully turned on like a Si MOSFET. Please refer to page 5 of our application note GN001 for more information.

Usually, the smaller the package the smaller the inductance. So we choose 0603 in our designs.

Yes. As we have seen Si MOSFETs in this application. A GaN circuit should have higher performance since the GaN does not have reverse recovery and operates with a much faster switching transition.

Usually, a Cgs is not needed. However, a Cgs can help filter some noise on Vgs, if the layout is not optimum. So, it may be good to have a footprint for Cgs, and then, after testing, determine if it should be populated.

We use the bottom layer of the driver board to do flux cancellation with the single layer of the aluminum PCB. If you have IEEE access, refer to this link for more information:  https://ieeexplore.ieee.org/document/8096647

We recommend 2220 package, 650V, 0.2-1.0uF ceramic capacitors for decoupling.

Yes, GaN is quite easy to parallel. We have reference designs with 2 and 4 GaN HEMTs in parallel. The only difference versus a single GaN device applications is that, for paralleled GaN, we need to make sure the power commutation loop for each paralleled transistor is minimized and symmetric. Refer to the application note GN004 for more information:

Yes. For example, the common source inductance will slow down the switching transition or even cause false triggering so as to affect the switching loss. Also the power commutation loop inductance will increase the voltage overshoot on Vds, which will also affect the switching loss. Generally speaking, minimizing parasitic inductance minimizes system loss.

The maximum transient voltage  less than 1uS  for our GaN is 750V. So, as long as the voltage spike is less than 650V, there is no concern.  In practical designs, we’ve seen 550V bus-voltage systems using 650V GaN HEMTs without issue. For 100V, the design rule is the same, ensure the peak Vds (greater than) 100V.

For GS66516T, the junction to heatsink thermal resistance is about 2°C/W.

The fastest switching transition is achieved in the Lidar application: the device current increases from 0 to 250A within 1.7nS.

氮化镓系统 (GaN Systems) is continuously working on new product developments. Please keep watching this information on new products.

Although there are no body diodes in GaN HEMTs, GaN HEMTs  freewheel current during the deadtime.  In reverse conduction mode with Vgs < Vth, the drain of the GaN will behave as a source, while the source will behave as a drain. When Vgs>Vth, the GaN will be fully turned on like a Si MOSFET. Please refer to the page 5 of GN001 application note for more information.

Correct,  there are no body diodes in GaN HEMTs, GaN HEMTs freewheel current during deadtime.  In reverse conduction mode with Vgs < Vth, the drain of the GaN will behave as a source, while the source will behave as a drain. When Vgs>Vth, the GaN will be fully turned on like a Si MOSFET. Please refer to the page 5 of GN001 Application Note for more information.

No. All these devices mentioned in the webinar are HEMTs – unipolar devices.

Yes, the bandwidth of Rogowski coil is only 30MHz. If there is some high frequency oscillation, the measurement will miss some details. However, it is good enough to tell if there is abnormal beheviour on the switching waveforms. We compared with 2GHz current shunt SDN-414-10, and the measured waveforms were close enough.

Yes. i_DS is measured by Rogowski coil ( CWTUM/1/B  ). V_DS is measured by isolated probe ( THDP0200 ). The V_GS_L is measured by non-isolated probe TPP1000 with MMCX connectors. Please refer to page 14 of this Layout webinar for more information on the test setup.

We suggest using the high-side driver to turn off the high-side switches to emulate the conditions the power stage will see in the system.

During the switching-on transition, the di/dt of Ids will cause a voltage drop on Vds ( L*di/dt ). We adjust the deskew of probes by matching the start of di/dt of ids and the beginning of the Vds voltage plateau.

We only use Q3D for qualitative analysis.

从本研讨会下载演示文稿

    卢俊诚, 应用工程经理

    本科毕业于浙江大学,硕士毕业于凯特琳大学(Kettering University)。2016年起,加入氮化镓系统 (GaN Systems), 主要研究兴趣包括 宽禁带半导体器件的应用,电力电子封装,高功率密度电源,以及车载汽车充电器。发表IEEE/SAE论文20余篇,持有13个美国专利。

    友情赞助:

    论坛:电力电子技术与新能源https://21micro-grid.com/