Abstract
This study examined the output electrical characteristics—current-voltage (I-V) output, threshold voltage, and parasitic capacitance—of novel gallium nitride (GaN) power transistors. Experimental measurements revealed that both enhanced- and depletion-mode GaN field-effect transistors (FETs) containing different components of identical specifications yielded varied turn-off impedance; hence, the FET quality was inconsistent. Establishing standardized electrical measurements can provide necessary information for designers, and measuring transistor electrical characteristics establishes its equivalent-circuit model for circuit simulations. Moreover, high power output requires multiple parallel power transistors, and sorting the difference between similar electrical characteristics is critical in a power system. An isolated gate driver detection method is proposed for sorting the uniformity from the option of the turn-off characteristic. In addition, an equivalent-circuit model for GaN FETs is established on the basis of the measured electrical characteristics and verified experimentally.
1. Introduction
Metal-oxide-semiconductor field-effect transistors (MOSFETs) have been widely used over the past 30 years. As silicon approaches its performance limits, wide-bandgap semiconductors, such as gallium nitride (GaN) and silicon carbide (SiC), are emerging technologies that can supersede silicon MOSFETs as next-generation power transistors. Novel wide-band III-nitride semiconductor materials are being rapidly developed because of their unique properties, such as high electron mobility, saturation velocity, sheet carrier concentration at heterojunction interfaces, and breakdown voltages [1, 2]. These properties make III-nitrides feasible for high-power, high-temperature applications. Compared with SiC, GaN has low turn-on and switching losses and is less expensive. In addition, GaN wafers are produced by numerous manufacturers, thus negating any monopoly concerns. Furthermore, GaN has been widely used in light-emitting diodes and wireless applications. GaN power FETs are suitable for high-voltage, high-current, and motor-control applications as well as for industrial automation systems and automotive electronics [3–5].
Because both the commercially enhanced-mode (E-mode) and depletion-mode (D-mode) GaN FETs manufactured by National Chiao Tung University (NCTU) [6, 7] are relatively new types of power transistors, few related studies are available in the literature. In addition, few manufacturers discuss them because commercial applications are not yet prevalent. The electrical characteristics of commercially manufactured power transistors differ because of the differences in cutting, wiring, wire bonding materials and diameters, and packaging. Before using such power transistors in circuit applications, their electrical characteristics must be extracted and sorted to match similar electrical properties in circuit designs. Unfortunately, extracting similar electrical properties is time-consuming and expensive. Thus, rapid and easy extraction of the electrical characteristics of GaN FETs to sort similar electrical properties is essential.
Moreover, GaN FETs and the design of their gate drivers are relatively new. When using GaN FET power transistors in circuit applications, their unique electrical properties must be considered: (1) no intrinsic body diode [8–10], (2) low gate-to-source voltage limits [9, 10], (3) full-conduction voltage of the gate and uncommon power supply voltage (e.g., 7 V, 8 V) [11–13], and (4) low threshold voltage [9, 10, 12, 13]. The following properties are applicable to bridge-leg architecture power transistors: (1) floating source of the high-side power switch [14, 15] and (2) faulty turn-on [9, 15]. Therefore, GaN-FET-based power transistors require appropriate gate drive circuits and methods to prevent overload. Although numerous studies have examined these properties, none have focused on the turn-off characteristic.
This paper first reports the output electrical characteristics of novel GaN power transistors and standardized electrical measurements to provide necessary information for designers. Second, an isolated gate driver detection method is proposed for sorting. Finally, this paper presents a simple and accurate equivalent-circuit model of GaN FETs for circuit simulations, established on the basis of the measured electrical characteristics and verified experimentally.
2. Materials and Methods
2.1. Measurement of GaN Electrical Characteristics
On the basis of MOSFET and GaN-FET datasheets, the following characteristics were used in this study: (1) breakdown voltage 200 V, rated current 9 A, on-resistance 0.4 Ω, and E-mode MOSFET [16]; (2) breakdown voltage 500 V, rated current 6 A, on-resistance 0.5 Ω, and D-mode MOSFET [17]; (3) breakdown voltage 200 V, rated current 12 A, on-resistance 25 mΩ, and E-mode GaN FET [18]; and (4) D-mode GaN FET manufactured in the laboratory as testing devices. The electrical characteristics measured were characteristics, threshold voltage, and parasitic capacitance.
2.1.1. Characteristic Curve
curve measurements detect the maximum output current of the power transistor when the gate voltage is applied as the full-conduction voltage, and the full-conduction on-state resistance is applied between the drain and source (). The characteristic curve reveals the linear and saturation regions of the circuit, where the device can operate properly. According to the test circuit in [19], the output characteristics are drain current versus drain-source voltage measured under different gate voltages ranging from the gate turn-off to full-on voltage at intervals of 1 V. The range of the E- and D-mode MOSFET is 0 to 12 V and −12 to 0 V, respectively, the range of the E-mode GaN FET is 0 to 5 V, and that of the D-mode GaN FET manufactured by NCTU is −5 to 0 V. Drain voltage generally uses a pulse mode input, which prevents excessive heat that affects the output characteristics. According to the datasheets, test pulse properties for E- and D-mode MOSFETs and E-mode GaN-FETs are a pulse width of 300 μs and a duty cycle (duty) ≦ 2% [16–18]. In this study, a test pulse width of 300 μs, pulse period of 300 ms, and duty cycle of 0.1% were used to measure the curve characteristics of the four aforementioned power transistors.
2.1.2. Threshold Voltage
Threshold voltage () is the minimum gate bias required to turn the device on and produce a drain current specified in the datasheet. In the threshold voltage measurement circuit for E-mode power transistors, the drain and gate terminals of the power transistors are shorted () [16]. The voltage to the gate terminal is gradually increased from the turn-off voltage (0 V) until the measured current equals the specified drain current. For D-mode power transistors, the drain terminal is connected to a fixed DC voltage source. Similar to the test procedure for E-mode transistors, the voltage to the gate terminal of the D-mode transistor is gradually increased from (−5 V) and changes in its drain current are observed. As specified in the datasheet [17], a DC voltage of 25 V is applied to the drain terminal of D-mode MOSFETs. D-mode GaN FETs, which applying to drain terminal’s DC voltage value refer to characteristic curve while the drain voltage attains the saturation region under the power transistor, are fully opened ( = 0 V). In this study, laboratory-manufactured D-mode GaN FET saturation voltage was set to the voltage measured from the characteristic curve. The conduction threshold current of E- and D-mode MOSFET is 250 μA [16, 17] and that of E-mode GaN-FET is 3 mA [18]. The laboratory-manufactured D-mode GaN FET has no datasheet for referencing its conduction threshold current. Therefore, to obtain the gate terminal input voltage at which the drain current increases instantaneously, the experimental measurements of the drain current are transformed logarithmically [20]. The thus obtained is considered the threshold voltage .
2.1.3. Parasitic Capacitance
A power device analyzer/curve tracer [21] was used to measure the capacitance because it supports the measurement of the three nonlinear capacitances: , , and . Figure 1 depicts the , , and measurement circuits, respectively. A multiple frequency capacitance measurement unit with four ports (Hp, Hc, Lp, and Lc) was used for capacitance measurements. Hp and Hc were shorted together (hereafter, CMH), and Lp and Lc were shorted together (hereafter, CML). The CMH outputs an AC test signal through the circuit under test, which is detected by the CML. The CMH operates in the 100 kHz–1 MHz AC frequency range. As specified in the MOSFET and E-mode GaN-FET datasheets, is measured at 100 kHz, and and are measured at 1 MHz; the oscillation level is 30 mV for all measurements. The CML receiving port potential is equivalent to the ground terminal.
(a)
(b)
(c)
The parasitic capacitance is measured using the circuit shown in Figure 1(a). The AC test signal is output from the CMH to the drain terminal, the source terminal is connected to a high-voltage source/monitor unit (HVSMU), and the source terminal grounds the AC signal to the AC guard. Therefore, the AC test signal from to the source terminal is grounded by the AC guard, preventing the signals from being received by the source terminal through the path. When measuring the parasitic capacitance , the HVSMU should provide a DC bias relative to the CML gate terminal, and the power transistor should be kept off. Both the enhanced MOSFET and E-mode GaN-FET power transistor have the same (0 V), whereas that of the D-mode MOSFET power transistor is −12 V. However, the curve revealed that the tested D-mode MOSFET turns off at −5 V. Therefore, in this study, −5 V was used as the for the D-mode MOSFET power transistor. The of D-mode GaN FET is −5 V. Because the source terminal has a bias voltage, −, the CMH should apply an AC bias voltage in the range to 100 − to drain the terminal. Therefore, the bias voltage of measurement can range from 0 to 100 V.
The parasitic capacitance is measured using the circuit shown in Figure 1(b). The circuit is similar to that used to measure ; the only difference is that the gate terminal was changed to the source terminal to connect the CML. The CMH outputs an AC test signal to the drain terminal and receives an AC test signal from the CML by connecting CML to the source of the device under test. The gate terminal grounds the AC guard to prevent the AC test signals from being received by the source terminal through the and paths. To ensure that the power transistor remains off during the measurement, the DC voltage is applied to the gate terminal. The CMH gradually increases the AC bias voltage from 0 to 100 V; therefore, the parasitic capacitance measurement is in the voltage range of 0–100 V.
The parasitic capacitance is measured using the circuit shown in Figure 1(c). The CMH outputs an AC test signal to the gate terminal, and the CML receives the signal by connecting to the source terminal. Because the CMH has a bias, the power transistor is kept off. To prevent the signals from being received by the source terminal through the and paths, the drain terminal is grounded to the AC guard. The HVSMU provides DC voltage in the range of 0–100 V; therefore, parasitic capacitance is measured at different voltages in the range of 0–100 V.
2.2. Mechanism of Isolated Gate Drive Detection
2.2.1. Conventional Gate Drive
Conventional power transistor gate drives use gate drive integrated circuit (IC) architecture. When used in half- and full-bridge-leg drive topologies, gate drives typically use an optical coupling IC to form an isolated floating-supply gate drive circuit architecture. Isolated gate drive circuit architecture consists of fast optical coupling IC, gate driver IC, and auxiliary supply voltage, as shown in Figure 2. The gate drive voltage for the E-mode GaN FET gate-to-source voltage is , and the for D-mode GaN FET gate drive voltage is to . The difference between the gate drive circuits for E- and D-mode GaN FETs is that the input supply voltages for the isolated gate driver amplifier are and , respectively. The external gate drive signals of the isolated gate drive circuit for the E-mode GaN FET is [0 to ] relative to the ground (Gnd). Through the fast optical coupling IC, the signal is isolated and converted to [0 to ] relative to . Finally, the isolated signal is amplified through the gate driver IC to drive the E-mode GaN FET. Similarly, the gate drive circuit for D-mode GaN FET is isolated to produce the gate signal [] relative to to drive the FET.
2.2.2. Isolated Gate Drive Detection
The isolated gate drive signal waveform can be measured by an oscilloscope to distinguish the waveforms of the turn-off impedance when GaN FETs are in the turn-off state. The proposed detection method is a relatively simple screening method to sort similar electrical characteristics of GaN FETs; by observing the switching waveforms, external stray capacitance in the circuit boards and internal parasitic capacitance in the transistors can be detected. The simple and accurate GaN FET model established on the basis of the measured electrical characteristics can be verified through experimental measurements of the isolated gate drive signal waveforms.
The proposed isolated gate drive detection circuit is illustrated in Figure 2, and the drive signal for the E-mode GaN FET is shown in Figure 3. Voltages , , and were measured relative to the Gnd; and were measured relative to the isolated supply ground . The drain terminal of the E-mode GaN FET test circuit connects to the power supply voltage relative to the Gnd. When the gate-to-source voltage of the E-mode GaN FET is , it turns on and shorts the drain and source terminals. Ideally, the source terminal voltage relative to the Gnd should be promoted to . The source terminal of the GaN FET and the isolated power source terminal are connected, indirectly causing and GaN FET Gnd to turn on and off; therefore, has a relative floating voltage of . When the gate terminal voltage of the GaN FET relative to the source terminal voltage is , the E-mode GaN FET turns on, the isolated power supply ground relative to the Gnd is , and the voltage between the GaN FET gate terminal and the Gnd is . When the gate terminal voltage relative to the source terminal is 0 V, the E-mode GaN FET turns off, the gate and source terminals are open, and and Gnd are 0. Because the GaN FET turns off, the drain terminal voltage no longer offers voltage to . The floating voltage discharges through circuit stray capacitance and load resistance . Therefore, floating voltage discharges from at the speed of the resistor-capacitor time constant until the next pulse width modulation (PWM) signal to the gate-to-source voltage is , which turns the GaN FET on again. When the drain terminal voltage is supplied to , the parasitic capacitance can be recharged to . The higher the PWM drive signal frequency entering the gate terminal, the more stable the in maintaining . When the PWM signal is no longer sent to the gate terminal and the GaN FET remains off for a sufficient period, discharges to 0 V through circuit stray capacitance [22] and load resistance , as shown in Figure 2 (). Concurrently, Gnd drops to 0 V. Compared with the turn-off impedance of Si MOSFETs, the turn-off impedance of the GaN FET is low. Therefore, the leakage current offers a load resistance to produce voltage . The turn-off impedance of the GaN FET can be obtained by observing the variance.
The D-mode GaN FET signal is depicted in Figure 4. When the gate-to-source terminal voltage of the D-mode GaN FET is 0 V, because it is typically turned on, its drain and source terminals are short. In this case, the source-to-ground Gnd voltage should be increased to the power supply voltage , and the GaN FET gate-to-ground Gnd voltage should be . When the gate-to-source terminal voltage is , the transistor turns off. Concurrently, the gate-to-ground Gnd becomes . The voltage discharges through the circuit stray capacitance [22] and the load resistance , as shown in Figure 2 (), until the next PWM to the gate-to-source terminal voltage is 0 V, which turns the GaN FET on again. The drain terminal voltage provides voltage to . When the gate terminal PWM signal ceases and the GaN FET remains off for a sufficient period, the gate-to-ground Gnd voltage discharges through the circuit stray capacitance and resistance to 0 V. The condition of the D-mode GaN FET is the same as that of the E-mode GaN FET. The turn-off impedance of D-mode GaN FET is smaller than that of Si MOSFET; therefore, the leakage current offers a load resistance to produce voltage . The relationship between the turn-off impedance , leakage current, and voltage is discussed later.
2.2.3. Isolated Gate Driver Circuit Model
To make the device model adaptable to and suitable for a system-level simulation, a subcircuit model was developed using the measured characterization results as the parameters of the model [23]. A simplified isolated gate driver detection circuit architecture is shown in Figure 5(a). A voltage probe is used to measure the voltage between the gate terminal and the ground. When is 0 V (to turn the E-mode GaN FET off) or (to turn the D-mode GaN FET off), the E- or D-mode GaN FETs are equivalent to a turn-off resistance , which can be used to evaluate the turn-off capacity of the GaN FET. In this test, with a known probe resistance value and by applying the Kirchhoff laws, the current through the power supply voltage at resistances and is obtained as the leakage current , which can be described as follows:The leakage current flows through the voltage probe and produces as follows:By substituting (1a) into (1b), the impedance can be derived as follows:A simplified simulation of the isolated gate drive detection circuit is shown in Figure 5(b). The simplified simulation circuit consists of the controlled signal source, E- or D-mode GaN FET current source, the isolated power supply ground , ground Gnd, the parasitic capacitances , , and , turn-off impedance , and voltage probe resistance . The turn-off resistance and voltage probe resistance connect together and, with the applied voltage and ground Gnd, form a loop that can use the Kirchhoff voltage law to estimate the leakage current . The current source is extracted from the characteristics. The characteristic curve of the GaN FET follows the Level 1 MOSFET model characterized by (3a), (3b), and (3c) and is divided into cut-off (3a), linear (3b), and saturation (3c) regions.
(a)
(b)
Cut-off region:
Linear region:
Saturation region:where is the transduced value and is the short-channel width-modulation slope coefficient in the saturated region, which is initially set to 0. The sign of determines the mode: positive is for E-mode and negative is for D-mode. Using the established Level 1 MOSFET characteristics model equations to describe the GaN FET current value in the saturation region reveals a large difference between experimentally measured and simulated data. Therefore, referring to a smoothing equation, the coefficient 1/2 in (3b) is replaced with 1/3 and that in (3c) is replaced with 2/3. The smoothing equation (4) is used to smoothen the characteristic curve in the linear and saturation regions; the voltage value is modified to and substituted in (5a) and (5b). The + and − signs denote the D- and E-modes of the GaN FET, respectively, which complies with the characteristics of the GaN FET model. The value impacts the degree of smoothness between the linear and saturated regions; the higher the value is, the smoother the curve is [24]. The smoothing equation is as follows:After smoothening, the GaN FET characteristic equations for the linear and saturation regions are depicted as follows:
Linear region ():
Saturation region ():SPICE simulation software [25] was used to simulate the electrical characteristics and to verify the measured gate drive signals. The Shenai model [23] was used and the built-in Level 1 MOSFET capacitance model was replaced with external capacitances. The measured gate-source capacitance was relatively independent of voltage, and a constant measured capacitance was used in the circuit model. The and can be described using the following equations:where is the zero-bias gate-to-drain capacitance, is the zero-bias drain-to-source capacitance, is the gate-to-drain voltage, is the drain-to-source voltage, is the junction built-in potential, and is the junction grading coefficient. The parameters and were adjusted to obtain the optimal fit with the measured capacitance data. Moreover, the effect of external couple capacitances on during turn-off is considered.
3. Results and Discussion
3.1. Characteristic Curve
E- and D-mode GaN FET devices under test are shown in Figure 6. The tested D-mode GaN FET chip is 80 mm in size and is packaged in the TO-3P form. Figure 7 depicts the measured characteristics of the four power transistors. Solid lines represent the waveforms specified in the datasheets, dotted lines represent the measured waveforms, and solid lines with circles represent the SPICE-simulated waveforms.
(a)
(b)
(a)
(b)
(c)
(d)
The measured characteristics of the E- and D-mode MOSFET waveforms are similar to those specified in the datasheet. The on-resistance at a specific turn-on gate voltage and drain current can be extracted directly from the output characteristic curves.
Figure 8 plots the of the four power transistors. In the enhanced MOSFET, at = 10 V, = 1.44 V, and = 4.5 A is approximately 1.44/4.5 = 0.32 Ω, which is under the maximum value specified in the datasheet (0.4 Ω). In the D-mode MOSFET, at = 0 V, = 1.66 V, and = 3 A is approximately 1.66/3 = 0.55 Ω, which is close to the value specified in the datasheet (0.5 Ω). Although the D-mode MOSFET is 0 V, it conducts current but not at full conduction. When is 5 V, is 30 V and the output current is 35 A. When is −2 V, the power transistor turns off and the output current is close to zero. GaN FET output characteristic variation is considerably large compared with that of the MOSFET. The output current value exhibits drift phenomena in different E-mode GaN FET samples when the inputs and are the same. The experimental results show that the linear region of the on-resistance is approximately 0.025–0.03 Ω. When is 5 V and the average output current is 6 A, the average voltage is 0.18 V; therefore, the average on-state resistance is 0.18/6 = 0.03 Ω, which exceeds the datasheet value of 0.025 Ω. The D-mode GaN FET on-resistance is approximately 0.25–0.30 Ω. When is 0 V and is 1 V, is 3.87 A; therefore, is 1/3.87 = 0.26 Ω.
Equations (5a) and (5b) are used to establish the current source model of a transistor. The output voltage of the E-mode GaN FET is increased from 0 to 3 V at intervals of 0.1 V, and the gate input voltage is increased from 0 to 5 V at intervals of 1 V; the output voltage of the D-mode GaN FET is increased from 0 to 10 V at intervals of 0.5 V, and the gate voltage is increased from −5 to 0 V at intervals of 1 V.
The waveforms are shown in Figures 7(c) and 7(d) as solid lines with circles; the simulated characteristics are similar to the measured characteristics (dashed lines). Drain current (-axis) at a particular voltage (-axis) can be obtained from the simulated waveform. In addition to the on-resistance characteristics, the saturation voltage of the D-mode GaN FETs exhibits variance. The experimental results show that the average saturation voltage is at = 7 V and that the maximum saturation current is between 16 and 18 A. The conduction resistance of the D-mode GaN FET is 0.26 Ω, which is smaller than those of the two MOSFET power transistor (0.32 and 0.55 Ω) but much larger than that of the E-mode GaN FET (0.03 Ω). In the future, of the D-mode GaN FET can be improved using internal or external parallel methods [6, 7]; therefore, uniform performance should be sorted.
3.2. Threshold Voltage
The measured threshold voltage is plotted in Figure 9. The conduction threshold current of the MOSFET power transistor is defined as 250 μA. From the experimental results of the enhanced MOSFET, the threshold voltage is 3.21 V, which is in the range specified in the datasheet (2–4 V). Furthermore, the conduction threshold current of the D-mode MOSFET is 250 μA; therefore, the gate threshold voltage is −2.98 V, which is in the range specified in the datasheet (−4 to −2 V). From the information in the manual, E-mode GaN FET conduction threshold current is defined as 3 mA.
(a)
(b)
In Figure 9(a), when the output current of the E-mode GaN FET is 3000 μA, the gate threshold voltage is 1 V, which is in the range specified in the datasheet (0.7–2.5 V). The output current of the D-mode GaN FET is plotted logarithmically in Figure 9(b). Near = −3.9 V, the output current rises rapidly. Hence, this voltage is defined as the threshold voltage of the D-mode GaN FET. The threshold voltage of the E-mode GaN FET is much lower than that of the MOSFET.
3.3. Parasitic Capacitance
The datasheet provides power transistor parasitic capacitance characteristics, including the input capacitance (), output capacitance (), and transpose capacitance (), where , , and . Figure 10 plots the measured parasitic capacitance values of the four power transistors. From Figure 10(a), for the E-mode MOSFET, at = 0 V and the bias voltage = 25 V, the parasitic capacitances , , and are 553.8, 91.5, and 27.3 pF, respectively, which are close to the datasheet values (540 (typ.), 90 (typ.), and 35 pF (typ.), resp.). From Figure 10(b), the E-mode GaN FET, at = 0 V, and a bias voltage = 100 V, parasitic capacitances , , and are 473.7, 301, and 16.7 pF, respectively, which are close to the datasheet values (540 (max.), 350 (max.), and 12 pF (max.)). From Figure 10(b), the D-mode MOSFET, at = −10 V, and a bias voltage = 25 V, parasitic capacitances , , and are 2361.7, 243.7, and 61.7 pF, respectively, and is under the value specified in the datasheet (2800 pF (typ.)), whereas and are close to the datasheet values (255 (typ.), 64 pF (typ.)). For the E-mode GaN FET, at = 0 V and = 100 V, parasitic capacitances , , are 473.7, 301, and 16.7 pF, respectively, which are close to the datasheet values (540 (max.), 350 (max.), and 12 pF (max.)). For the D-mode GaN FET, at = −5 V and = 50 V, the parasitic capacitances , , and are 72.7, 64.4, and 10.2 pF, respectively. Comparison of datasheet and measured characteristics are listed in Table 1, and Table 2 lists the important - and capacitance model parameters for E- and D-mode used in this study.
(a)
(b)
(c)
(d)
According to (6a) and (6b), the parameters listed in Table 2 are used. The relationships , , and = are used. Plots of , , and are shown in Figures 10(c) and 10(d). The simulated and experimental curves are similar.
3.4. Isolated Gate Drive Detection
A turn-off voltage of 0 V is used for the E-mode GaN FET; therefore, the full-conduction voltage is limited to 5.5 V. For the D-mode GaN FET, the used turn-off voltage is −5 V, and full-conduction voltage is limited to 2 V. Hence, the driving voltage for the E-mode GaN FET gate-to-source voltage is set to 0−5 V; in other words, is set to 5 V, and the D-mode GaN FET gate source driving voltage is set to −5 to 0 V. At driving voltages of 0–5 V and −5 to 0 V, the E- and D-mode MOSFET waveforms can be contrasted. Regardless of the MOSFET mode, the voltage probe was used to measure the voltage between the gate terminal and the ground terminal. The E-mode MOSFET waveforms are the same as the ideal isolated gate drive circuit detection signal, as depicted in Figures 3 and 4; the gate voltage when turned on is +29 V and decreases to 0 V when turned off (Figure 11(a)). The D-mode MOSFET gate voltage waveform is +24 V, which decreases to 0 when turned off (Figure 11(b)). When measuring the E-mode GaN FET, the gate voltage is +29 V when turned on, but a difference in voltage level exists between the gate and the ground. The large change in the voltage level is in the 0–24 V range, as shown in Figure 12(a). D-mode GaN FETs exhibit the same phenomenon, as shown in Figure 12(b). The differences are caused by the turn-off impedance . The larger the turn-off impedance is, the smaller the leakage current is; the across voltage is small, and the difference between source-to-ground voltage value is close to 0. Conversely, when the turn-off impedance is small, the leakage current is large, and the source-to-ground voltage approaches +24 V. Therefore, the turn-off ability of GaN FETs can indirectly screen device uniformity. Moreover, the impedance value can be quantified.
(a)
(b)
(a)
(b)
When = 5 V, the E-mode GaN FET turns on. The gate-to-ground voltage is +29 V; when = 0 V, the E-mode GaN FET turns off, which is equivalent to the turn-off resistance ; the voltage probe resistance is . When the Kirchhoff circuit laws are applied, the power supply voltage through and generate the leakage drain current . Through the isolated gate drive circuit architecture, the voltage probe resistance is 10 MΩ and power supply voltage is 24 V. The values for the two modes are 13.6 V and 23.0 V, as shown in Figure 12(a). Substituting these values into (2), is obtained as 7.647 and 0.435 MΩ. Using a digital multimeter in series with the source terminal and the voltage probes () to measure the GaN FET device during the turn-off state, the leakage currents are obtained as 1.340 and 2.365 μA. By substituting in (1a), leakage currents are obtained as 1.36 and 2.30 μA, which are similar to the measured values.
Next, the gate-to-ground voltage waveform of the D-mode GaN FET is measured. When is 0 V, the D-mode GaN FET drain and source conducts and shorts, and the source terminal voltage is +24 V. Because = 0 V, the gate terminal voltage is +24 V; when = −5 V, the GaN FET is off, because the resistance of the D-mode GaN FET is not large enough; therefore, leakage current flows, and the source terminal voltage relative to ground cannot be reduced to 0 V. Next, the turn-off voltage of the D-mode GaN FET is measured using the 10 MΩ voltage probe; the value is always 19 V. The turn-off impedance is much lower than 10 MΩ; therefore, the voltage probe is adjusted to 1 MΩ to repeat the experiments. is in the 0–19 V range. and voltage probe = 1 MΩ divide the voltage, assuming that the probe is measured as voltage. From (1a), (1b), and (2), the leakage current and turn-off impedance can be obtained.
The waveform variability of the D-mode GaN FET is similar to that of the E-mode GaN FET, as shown in Figure 12(b). The D-mode GaN FET under a voltage probe = 1 MΩ varies in the 16–20 V range. From (2), of the D-mode GaN FET is 463.4 kΩ when is 16.4 V and 904.8 kΩ when is 12.6 V. However, when exceeds = 19 V, the turn-off impedance is insufficient and the transistor does not turn off.
In the E-mode GaN FET, the threshold voltage of the output characteristic curve model parameter is set to 1 V and is set to 24. An external capacitor is used as listed in Table 2. In the isolated gate driver circuit architecture, the voltage probe resistance = 10 MΩ, power supply voltage = 24 V, and measuring voltage = 13.6 V. From (2), impedance is derived as 7.647 MΩ when is 13.6 V; therefore, when E-mode GaN FET turns off, the drain-to-source is equivalent to a 7.647 MΩ resistor. is the internal voltage probe resistance, which is 10 MΩ at 10x magnification. In the D-mode GaN FET, the threshold voltage of its output characteristic curve model is −3.9 V and is 2.1. The external capacitor is used as parasitic capacitance. The turn-off impedance of the experimental device using the 10x magnification voltage probe is much lower than 10 MΩ. Hence, a 1x (1 MΩ) probe is used to measure ( = 1 MΩ); the power supply voltage = 24 V and the measured voltage = 12.6 V. From (2), when voltage is 12.6 V, is 904.8 kΩ. The equivalent turn-off resistance between the drain-source is equivalent to 904.8 kΩ.
SPICE circuits are established through the equivalent model described in Figure 5. The gate resistor uses 100 Ω , and the E-mode GaN FET gate terminal wave signal voltage is 0–5 V, whereas the D-mode GaN FET is −5 to 0 V. Gate pulse width, period, and frequency of the PWM signal are 100 μs, 500 μs, and 2 kHz, respectively. The numerical analysis software predicts that the gate drive circuit board has external drain-source stray capacitance and that the actual measurements of waveform segments have a slower falling slope. Because of stray capacitance parallel to drain-to-source and gate-to-drain, the turn-off slope falls slowly in the waveform. The estimates of the stray capacitance value are 2 nF. The GaN FET gate detection simulation parameters are shown in Table 2.
Figures 13 and 14 present the E- and D-mode GaN FET gate detection simulation circuit waveform as shown in black line and measurement waveform as shown in orange line superimposing they are matching each other. When the isolated gate detection circuit stops sending the PWM signal to the gate terminal and GaN FET is set to close long enough, the falling slope waveform segment of the discharge through circuit stray capacitance and load resistor from +24 V floating voltage discharge to the voltage is in accordance with the measurement. The influence of the parasitic capacitance of the input capacitance on voltage switching waveform can be observed by enlarging the voltage signal timeline, as shown in the inset of Figures 13(a), 13(b) and 14(a), 14(b); the enlarged voltage signal charge and discharge waveforms are shown in Figures 13(c), 13(d), 14(c), and 14(d). The simulation and experimental gate voltage waveforms of the charge and discharge match perfectly.
(a)
(b)
(c)
(d)
(a)
(b)
(c)
(d)
The results of the screening and recording of the turn-off voltage and the corresponding of the E-mode GaN FET voltages are shown in Figure 15. The off-resistance of E-mode GaN FETs is larger than 1 MΩ, whereas those of D-mode GaN FETs are approximately in the 0.5–1.5 MΩ range.
4. Conclusions
The on-resistance of E-mode GaN FET and NCTU D-mode GaN FET is 0.025–0.03 Ω and 0.25–0.3 Ω; both of these values are lower than that of MOSFET. Nevertheless, NCTU’s D-mode GaN FET can be further improved using the parallel method to reduce on-resistance. Regarding parasitic capacitance, the of the E-mode GaN FET is far lower than that of the enhanced MOSFET; a smaller capacitance value indicates that the Miller plain area is relatively short and that the switching time is shorter. Compared with the turn-off resistance of different samples, the electrical characteristics of each MOSFET device are highly consistent, whereas those of GaN FET exhibit less uniformity. GaN FETs are currently under development, and the electrical characteristics of each component are relatively unstable; the variability is larger than that in MOSFET. This study established a standardized electrical measurement procedure that provides necessary information for designers. In addition, a simple and accurate GaN FET model was established on the basis of the measured electrical characteristics. The simulation waveforms can be used to obtain information on GaN FET’s internal parasitic capacitance, turn-off impedance , and stray capacitance in the inverter circuit board. The proposed GaN FET isolated gate drive circuit screening method by detection provides a simple uniformity sorting method. The results show that the higher the off-state voltage is, the smaller the turn-off voltage is; in other words, the device has a lower . The leakage current in GaN devices is much larger than that in MOSFET devices in the turn-off state. The off-resistance of MOSFET is generally larger than 10 MΩ. By contrast, the off-resistance of E-mode GaN FETs is larger than 1 MΩ, whereas those of D-mode GaN FETs are approximately in the 0.5–1.5 MΩ range. Devices with the same off-state voltage perform similarly. Moreover, the larger the turn-off resistance is, the closer the characteristics are to those specified in the datasheet.
Conflict of Interests
The authors declare that there is no conflict of interests regarding the publication of this paper.
Acknowledgments
This work was supported by the CSIST Project CSIST-0101-V108(104), Taiwan. The authors would like to thank Professor Edward Yi Chang of NCTU for supporting GaN devices, Professor Stone Cheng of NCTU for supporting package technology, and National Nano Device Laboratories, Hsinchu, Taiwan, for their very helpful suggestions and technical support.