Nikola Boskovic, MSc

Prof. ir. Korneel Wijnands

Dr. ir. Maurice Roes

Prof. dr. Elena Lomonova June 14th 2018

Paralleling of Gallium Nitride transistors for

48V high current automotive applications

Outline

• Introduction

• Motivation

• Approach

• Measurement results

• Conclusion

1

Application – Mild hybrid system2

• Specificationso 48 V

o 500 Apeak

• Key performanceso Cost [€/kW]

o Efficiency [%]

battery pack48 V

dc-ac converter

electrical motor

combustion engine

Source: Federalmogul.com

Power converter topology

• Standard 3-phase 2-level inverter

• No transistors available for high currents (500 A)

• Type of low voltage transistors? Si MOSFET or GaN.

3

– solution: PARALLELING

Gallium Nitride transistors

• Lower losses compared to silicon devices

• Extremely fast switching

• Paralleling GaN transistors?

4

Source: EPC white paper

10 V/div

2 ns/div

80 V/ns

GaN transistor(100V, 90A)6,05 mm x 2,3 mm

• Static current sharing (DC)

• Dynamic current sharing(Transient)

Current sharing5

Static current sharing

• Steady state

• Resistance RDS(ON)

• Negative feedback

• Self-balancing

6

• Transients

• Transistor parameters

• Threshold voltage VGS(th)

• Positive feedback

• Avalanche effect

Dynamic current sharing7

Dynamic current sharing

• Asymmetric layout – unavoidable when paralleling more then two transistors

o Circuit parameters – mainly parasitic inductances

8

Power loopGate loop

Approach of modelling

• Extraction of the parameters values

• Simulation

• Measurements

• Verifying

9

• Wide frequency range

Model of passive components10

Si MOSFET

GaN

Relevant for:

• Models for needed frequency (GHz) range don’t exist!

11Models

40 Hz 110 MHz55 MHz 110 MHz

• Five slightly different circuits

• Most important two are:

Boost Converter

Transistor is in BGA package, 0.9 mm x 0.9 mm

12

Boost Converter13

Prototype14

Measurements15

~1 ns (1 GHz)

16Conclusions

• Gallium Nitride transistors are beneficial in terms of lower losses.

• Measurement uncertainty – 1 GHz oscillation!

• Impossible to verify parameters.

• Prediction of current for very high number of paralleled transistors.

• Too many uncertainties for the generalized conclusions.

• Different route should be taken.

Alternative approach

• Paralleling of half bridges by inductances

17

18

Thank you for your attention!

Nikola Boskovic

n.boskovic@tue.nl

Electromechanics & Power Electronics (EPE) group