时间:2024-07-28
, , ,
(1.Shenzhen Graduate School,Peking University,Shenzhen 518055,China;2.Institute of Microelectronics,Peking University,Beijing 100871,China;3.Shenzhen Hai Li Tech.,Inc.,Shenzhen 518129,China)
Abstract:Gallium nitride (GaN) field-effect transistors have low ON resistance and switching losses in high-frequency (>MHz)resonant wireless power transfer systems.Nevertheless,their performance in the system is determined by their characteristics and operation mode.A particular operating mode in a 6.78-MHz magnetic resonant wireless transfer system that employs class-D GaN power amplifiers in the zero-voltage switching mode is studied.Two operation modes,the forward mode and the reverse mode,are investigated.The nonideal effect under the device-level dynamic resistance and thermal effect are also analyzed.The dynamic resistance under different operation modes is demonstrated to have different generation mechanisms.Finally,the device characteristics with system operating conditions are combined,and the effects of temperature and dynamic resistance under different operating conditions are evaluated.
Keywords:GaN,wireless power transfer,dynamic resistance,thermal effect
Wireless power transfer (WPT) is recognized as an efficient and resourceful technology for wireless charging in radio frequency identification,electric vehicles,buried sensors,portable electronic devices,and medical devices[1-2].Near-field WPT systems have a high power-conversion efficiency and are widely used in short-and mid-range applications.There are two categories of near-field WPT systems:the inductive and capacitive coupling technique for short-range applications(Wireless Power Consortium standard and Power Matters Alliance standard)[3]and the magnetic resonance coupling technique for mid-range applications (AirFuel standard)[4].
Wireless power can be transferred with an operating frequency in the megahertz region via the AirFuel standard (A4WP),for example,6.78 MHz.Owing to its high operating frequency,this standard has greater requirements in terms of device characteristics.However,the fundamental limitations and device structure of silicon with its relatively narrow bandgap make it challenging to use this material to handle high-power and high-frequency applications.
Gallium-nitride based high-electron-mobility transistors (HEMTs) have tremendous potential for high-power,high-frequency,and high-efficiency power switching applications owing to the outstanding properties of the gallium nitride (GaN) material (e.g.,wide bandgap,high-electron saturation velocity,and large breakdown electric field)[5-8].Many scholars have studied and reported on the properties of GaN field-effect transistors (FETs) at the device level.The doping,structure,and material quality can all be adjusted to improve the device characteristics.However,the device characteristic presentation in a real system is also related to circuit design and operation mode.
Pulsed current-voltage (I-V) measurements are widely used to research device characteristics.They can be used to describe the trap response and extract the activation energy of the deep level.However,using such measurements does not capture device behavior in realistic switching conditions[9].Several questions arise:Do the special operation modes affect the device’s characteristics in a real circuit? Does a particular circuit topology enhance or suppress the nonideal effects of a GaN FET?
At present,the application of GaN devices in practical circuits is little known.Most applications of GaN FETs operate in the forward mode (see Fig.1a).Some devices operate in the reverse mode (see Fig.1b),such as for synchronous rectification[10].One characteristic that has drawn the most attention is the change in the ON resistance (Ron) of a GaN FET during circuit operation.Rondoes not remain constant throughout the circuit operation.It changes with the biasing condition and operation time scale.Generally,an increase inRonincreases the power loss or reduces the reliability of the GaN FET[11].Therefore,the so-called dynamicRonhas been extensively studied from the transistor perspective.
Fig.1 Two different operation modes for a GaN device
In this work,we used an accurate measurement method to detect the value ofRonof the device in an WPT system operating at 6.78 MHz,as detailed in Section 2.In Section 3,a particular operation mode is investigated in the WPT system.Two nonideal effects,dynamic resistance (Section 4) and the thermal effect(Section 5) at the device level are investigated.In Section 6,we evaluate the influence of nonideal effects on the device in the WPT system,and Section 7 concludes the paper.
Determining the exactRonrequires obtaining an accurate current measurement.A new measurement setup is introduced to measureRon.It involves sensing the voltage (VDS) and current (IDS) of a GaN FET in a circuit in real-time operation.Given the fast switching speed of a GaN device (e.g.,an EPC2107 transistor),a high-bandwidth accurate current testing measurement is preferred.Four kinds of test methods are compared in Fig.2.
Fig.2a shows a schematic of the method using a current clamp.However,its large size and narrow bandwidth (200 MHz) severely limit its application in high-frequency testing.Moreover,based on the principle of electromagnetic coupling,the measurement results are easily affected by the position of the testing circuit.
Fig.2 Schematics of detecting current circuit
Fig.2b shows a schematic of the method using a surface mount resistor[12].The current is obtained by detecting the voltage on the surface mount resistor.The main issue here is the parasitic inductance,which is mainly coming from the packaging Fig.3a shows the test result.The surface mount resistor is 0.2 Ω.The current curve exhibits considerable oscillation.Increasing the value of the surface mount resistor can reduce the amount of oscillation,but it could affect the operation state of the device.
Fig.2c shows a schematic of the method using a GaN transistor (in which the gate connects to the drain)to detect the current.It has a very small size and parasitic inductance.Fig.3b shows the measurement result.When the device turns on,Voutdoes not oscillate.However,when the device turns off,Voutovershoots negatively.We consider that the parasitic capacitive(Cgs) of the device (see the inset figure) is the reason for this phenomenon.The voltage-on capacitance cannot be altered.When the device suddenly turns off,Cgscannot discharge through the underlying transistor rapidly,which causes the negative overshoot.Reducing the test frequency eliminates the negative overshoot.To get an accurateRon,an accurateI-Vcharacteristic is also needed at high-frequency operation.This poses another challenge.
Fig.3 Measurement results using a surface mount resistor and a transistor
The most suitable method is to use a coaxial current shunt,as shown in Fig.2d.The coaxial current shunt can be used to detect current flowing through itself and hence the GaN FET in real time (Fig.4).The part we used was an SSDN-10 manufactured by T&M Research Products,Inc.It has accurate resistance(within 0.1 Ω),small parasitic inductance,and high bandwidth (2 GHz)[13].We can achieve accurateRon=VDS/IDSmeasurements,similar to what can be achieved by using expensive commercial equipment (e.g.,a Keysight B1505A).
Fig.4 Test results using a coaxial current shunt when Vamp=70 V
The system level measurement was performed with a WPT module using a class-D amplifier[14]at 6.78 MHz under zero-voltage switching operation.Two EPC2107 GaN FETs were used to form a half-bridge.Fig.5 shows a simplified schematic and an image of the WPT system.The value ofRonof the low-side GaN FET (Q2) was measured.To simplify the experiment,we used a resistance (Rloadin Fig.5a)instead of the receiver.
Fig.5 Our WPT system
Fig.6a shows the output curve of a GaN FET in an operating system.The value ofVGSwas set to 5 V(based on the vendor recommendation).The waveform distortion has been reported in Ref.[15].To ensure an accurateRon,we selected the average value of the middle part (in the box) as the result in Fig.6b.
In Fig.6a,it is worth noting the two overshoots that appear when the device switches instantaneously.The negative overshoot makes Q2operate in reserve mode and the positive overshoot makes Q1operate in reverse mode.Meanwhile,the voltage spike of the positive overshoot is higher than that of the negative overshoot.
Fig.6 Test results for a GaN FET operating in a WPT system
The same phenomenon was discovered by simulation using LTspice software (Fig.7a).Fig.7b shows a schematic of the overshoot generation.
As Q2turns on and Q1turns off,the LC resonance circuit will discharge through Q2.The presence of dead time in the circuit and current through the inductance (coil) must be continuous.When Q2turns off and Q1has not turned on (dead time),the voltage makes Q1work in reverse mode (Fig.7b,condition I).When Q1turns off and Q2does not yet turn on,the negative overshoot makes Q2work in reverse mode(Fig.7b,condition Ⅱ).
In Fig.7b,the two conditions explain the reason for the overshoot.For condition I,current flows throughRload,Q1,and the power source (Vamp).For condition Ⅱ,current flows through Q2andRload.Normally,the power source has internal resistance.The value of the resistance for condition I is higher than that for condition Ⅱ.We conclude that this difference is the reason for the different voltage spikes.
Fig.7 Simulation of the device operating in a WPT system(6.78 MHz) when Vamp is 5 V
Two operation modes are presented in Section 3:forward mode and reverse mode.To more easily understand these modes,we study them in the hard-switching mode.By changing the different working conditions (such as frequency,VGS,andVDS),we can observe the variation in dynamicRon.
Fig.8 shows the test schematic in the forward mode,and Fig.9 shows the relationship betweenRonandIDSunder differentVGSvalues.The currentIDSwas controlled byRload.
Fig.8 Hard-switching mode test circuit
In Fig.9,one can observe thatRonis strongly dependent onIDSandVGS.To avoid the influence of these two factors,we setIDSto 0.25 A by adjustingRloadand sending a few pulse signals to the gate during the test to reduce the effect of heating.
Fig.9 Ron at different VGS and IDS values under dynamic conditions (6.78 MHz) (Vamp is 5 V and the driver signal is a square wave with a 50% duty cycle)
Fig.10 shows the effect ofVDSonRonwith different values ofVGS.VDSis the device’s OFF state voltage.The dotted line is theRonin an ideal state.AsVDSexceeds 6 V,Ronincreases rapidly whenVGSis 4 V.However,the resistance is relatively stable whenVGSis 5 V.This illustrates that the influence of dynamicRonis more serious whenVGSis low.
Fig.10 Relationship between Ron and VDS under different gate voltages (The dotted line is Ron in the ideal state)
Two factors affect the dynamic characteristics of the device:VDSandVGS.They affect devices in different states.VDSinfluences the device in the OFF state.It changes the electric field in the channel,leading to electrons being captured by traps and increasing the value ofRon.VGSmainly influences the device in the ON state.High positiveVGScan cause the electrons to be easily released from the traps and reducesRon.Therefore,whenVGS=5 V,Ronis slightly affected byVDS.In contrast,whenVGSis 4 V,increasingVDSleads to a more serious dynamicRonresponse.
Fig.11 shows the variation ofRonfor frequencies between 100 kHz and 1 MHz.ForVGS=4 V,the variation ofRonwith frequency can be understood as the traps’capture and release.VDSmakes the traps capture electrons in the OFF state andVGSmakes the traps release electrons in the ON state.With the increase in frequency,a lowerVGScannot immediately make the traps release electrons.When the frequency exceeds 1 MHz,Ronis not changed,indicating that the processes of capture and release are balanced.ForVGS=5 V,Ronis relatively insensitive to frequency.
Fig.11 Relationship between Ron and frequency under different gate voltages (The current is 0.25 A and VDS is 10 V)
Because the GaN device does not have a substrate electrode,Wang et al.[16]reported a test method to determine the effect of the substrate traps.In Fig.12,Ron,Dis the resistanceRonof the device working at 6.78 MHz andRon,Sis the DC value of the device.With the increase inVGS,the ratioRon,D/Ron,Sdecreases and gradually stabilizes,indicating that the traps are in the substrate.ForVDS=20 V,the stable ratio is not 1.We consider that this could be another mechanism for generating dynamicRon.
Fig.12 Change of the dynamic Ron under different VGS values(VDS is 20 V and 5 V)
Traps can be created by carbon doping in GaN to increase the insulation of the substrate[17-20].Many traps were introduced by carbon under the channel.These traps produce the acceptor level in the energy band.When the device turns off,VDScaptures the electrons captured by the traps.When the device turns on,VGSreleases electrons from the traps.
For the forward mode,the dynamic resistanceRonis mainly related to the substrate carbon doping.In addition,the influence of dynamicRonis strongly correlated withVGSandVDS.
Two types of reverse modes are illustrated in Fig.13.Fig.13a shows the reverse operation mode under ideal conditions of the WPT system (reverse mode I).If the gate drive produces an oscillation and causes the potential between the source and the gate to differ,reverse mode I could change to reverse mode Ⅱ(Fig.13b).We evaluated the dynamic effects of both reverse modes in an operational condition.
In Fig.13a,the gate connects to the source and current flows from the source to the drain.The device turns on whenVGD>VTH.Fig.14 shows the output characteristics for reverse mode I under the DC test.WhenVSDexceeds 2.7 V,the device is damaged.We attribute the device damage to the effect of heating.
Fig.13 Different operation modes and GaN FET simplified parastic paraneter model
Fig.14 Device output characteristic under reverse mode I
Fig.15 compares the relationship betweenRonand current under the forward mode and reverse mode Ⅱ.In reverse mode Ⅱ,Ronincreases rapidly when the current exceeds 0.3 A.This can be explained by Fig.13c.Because of the asymmetric design of the device,in general,the gate-source distance is less than the gate-drain distance.This leads to parasitic resistanceRs<Rd.Under the same conditions,the high parasitic resistance will reduce the efficiency gate voltage and increaseRon.To avoid this influence,we chose the current to be 0.25 A.
Fig.15 Ron under different currents in a dynamic system(Vamp=5 V and the working frequency is 6.78 MHz)
Fig.16 comparesRonfor different operation modes.The gate voltage was set to 5 V.The dynamicRonis more obvious when the device operates in reverse mode Ⅱ.We also performed a frequency test as was illustrated in Fig.11.As can be seen,Ronis essentially independent of the frequency.A mechanism can explain this phenomenon.Traps exist between the passivation layer and the barrier layer[21-22].When the device turns off,the gate injects electrons to the source-side barrier layer and the gate dielectric layer in reverse mode Ⅱ[23].Electrons are captured at the interface between the source-side barrier and the passivation layer.As the device turns on,gate voltage will extract the electrons from the traps.
Fig.16 Dynamic Ron under reverse mode Ⅱ and the forward mode (In the forward mode,the X axis is VDS;In reverse modeⅡ,the X axis is VSD)
In the experiment,the gate driver signal was a 50% duty cycle square wave.The device had the same turn-on and turn-off times.We consider that these two processes,electron injection (OFF state) and extraction (ON state),cancel each other so thatRondoes not vary with frequency under reverse mode Ⅱ.
Fig.17 shows the two mechanisms of dynamicRongeneration.We studied a commercial GaN transistor(EPC2107),but accurate determination of the mechanism of dynamicRonmust rely upon a combination of the device structure and fabrication process.
Fig.17 Schematic of dynamic Ron generation mechanism(Two different types of traps produce nonideal effects)
The effect of temperature onRonis shown in Fig.18.To limit the dynamicRoneffect,we set the drain voltage to 10 V in the forward mode.The temperature was controlled by the duty cycle of the square wave,and it was taken on the backside surface of the GaN FET by using an infrared camera.
Fig.18 Effect of temperature on Ron (The test circuit is Fig.8 and VGS=5 V;The point marks system operation in an ideal condition)
An increase in temperature reduces the mobility of the two-dimensional electron gas,which increasesRon[24].When the temperature rises from 50 ℃ to 70 ℃,Ronincreases from 0.52 Ω to 0.92 Ω.The device may fail when the temperature rises above 70 ℃(whenRonincreases by >100%).
Two nonideal effects,dynamicRonand thermal effects,were studied.In a WPT system,different operation conditions will lead to different nonideal effects becoming the dominant factor in restraining system efficiency.
Clearly,the open-load or high-Vampcondition can damage the GaN HEMT in a WPT system.In many cases,the device can fail instantaneously,and we cannot capture the value ofRon(such as whenVDS=50 V with no load).We consider that the damage is caused by heat.
In a WPT system,the level of stress on the GaN FET is closely related to the working conditions.For example,when the transmission distance is long,the system will generate a high current flowing through the GaN device,which will produce heat and could even cause device damage.This effect needs to be addressed at the system level.
The impact of the GaN FETRonduring operation in a WPT system was studied.Four test methods were compared.Using a coaxial current shunt can most accurately measure dynamicRonin real time in a circuit.Because of the particular operation mode for the WPT system,the dynamicRondevice under forward and reverse modes was studied.We found that the traps caused by carbon doping play a major role in the forward mode.The traps located at the interface between the barrier layer and the passivation layer have more influence on the reserve mode.Both trap-induced and heat-induced could causeRonchanges,with the latter being more significant at lowVDS.
In our research,we found thatRoncan saturate because of the trap,while heat dissipation tends to increaseRonin a positive feedback manner.Therefore,for commercially ready GaN FETs dynamicRoncould not be a bottle neck in WPT applications at present.
我们致力于保护作者版权,注重分享,被刊用文章因无法核实真实出处,未能及时与作者取得联系,或有版权异议的,请联系管理员,我们会立即处理! 部分文章是来自各大过期杂志,内容仅供学习参考,不准确地方联系删除处理!