This report documents the results of the performance evaluation of a 3.3KV, 6A CIGBT Clustered Insulated Gate Bipolar Transistor CIGBT is a three terminal MOS controlled thyristor semiconductor device. CIGBT was developed with the aim to replace IGBT (Insulated Gate Bipolar Transistor) in the power semiconductor market. IGBT is widely used in power semiconductor industry but it has been shown experimentally that CIGBT can provide superior on-state voltage drop and switching performance. IGBT's on-state voltage drop and switching losses increase with increase in voltage rating more significantly compared to CIGBT. This is due to thyristor conduction in CIGBT and a more efficient turnoff process.
The aim of the project is to evaluate quantitatively and qualitatively the performance of 3.3KV, 6A CIGBT for three different n-well doping concentrations. Change in the n-well doping concentration directly affects the static and dynamic performance parameters. Devices with three n-well doping concentrations of 6.5, 6.7 and 6.9e+13 were tested to determine the optimum doping concentration.
The report includes the literature review, explaining the evolution of power semiconductor devices from BJT up to CIGBT. Also included are simulated and practical results of electrical characterization. The report concludes with recommendations and suggested future work for future development of 3.3kV CIGBT.
Table of Contents
Introduction
Aim of the project is to evaluate quantitatively and qualitatively the performance of 3.3KV CIGBT. Project will determine the optimum n-well concentration of the n-well doping. CIGBT is designed to achieve a lower on state voltage drop and turnoff losses as compared to the state-of-the-art trench IGBT. Devices were tested to formulate a data sheet for the devices and the results were documented in the report. Device functionality was simulated and the results were compared with the real test readings to analyze the deviation between real and simulated test results for the future development of the CIGBT.
Evolution of Semiconductor devices
Bipolar junction transistor (BJT)
BJT was invented in 1947 by using insight gained by point-contact transistors [1]. .Multiple advantage of replacing vacuum tubes with a solid-state device publicized its use in industrial application. BJT is a three terminal current controlled bipolar device constructed from doped silicon. BJT is primarily used as a switch. BJT works by the injection of minority carriers from one junction and the collection across anther. The
Figure 1: cross section of a power BJT [2]
A basic structure of a NPN bipolar transistor is illustrated in figure 1.When compared to a conventional BJT, power BJT consists of a N-drift region that supports the high voltage during the forward blocking state. Forward blocking capability of the device is determined by the thickness and the doping concentration of the n-drift region. The device is put into the conduction state by forward biasing the emitter-base junction (J2) to start the injection of electrons. Electrons flow through the P-base region and are collected at collector-base junction. In a BJT, a small base current can produce a large collector current. Similarly, a small base voltage produces a large collector voltage thus making it a power amplifier. BJT can be turned off by reverse biasing the base region. Reverse biasing base region stops the injection of minority carriers from emitter-base junction and also removes the charge stored in the base region. Since the discovery of BJTs, they were widely used in switching applications such as air condition and discrete circuit design. For several decades, BJT's ruled the power semiconductor market and several successful efforts were made to improve the performance of BJTs. Research highlighted the one of the major shortcoming of the BJT's i.e. poor current gain. Consequently, BJT's were overshadowed by the invention and the rapid industrial acceptance of the IGBT's. IGBT's has a simpler gate control circuit because of its high input impedance, making the integration simpler. In addition, BJT's are not able to support high frequencies due to the charge in their drift region and fail under hard inductive switching. BJT's has a negative temperature coefficient, which makes it harder to characterize in a parallel configuration. CIGBT positive temperature coefficient and has a higher on-state current density when compared to a power BJT, making it favorable for numerous industrial applications.
Power MOSFET
Power metal-oxide-semiconductor field effect transistor (MOSFET) was developed in mid-1970 with an aim to capture the bipolar transistors market. MOSFET is a single carrier, three terminal voltage controlled device constructed from doped silicon. A voltage control device instead of current controlled bipolar transistors was an exciting idea for the industry from the application point of view. More over, high input impedance of the metal-oxide gate structure further simplified the gate circuitry. MOSFET has a higher switching frequency, which opened the new application streams for power semiconductor industry. Figure 2 shows the cross section of a single cell structure of a MOSFET. To put the device in conduction sate, a gate voltage is applied along with the drain voltage. Gate voltage induces an inversion in p-base under the gate electrode. When the positive drain voltage is applied, the electrons flow from the source through the inversion layer and enter the narrow JFET (Junction gate Field Effect Transistor) region located between the two P-base regions. After passing through the JFET region, electrons disperse in the entire drift region and eventually get collected at the drain terminal. When the gate voltage is removed, the device goes in the forward blocking state. The junction J1 between the P-base and the N-Drift region becomes reverse biased and the voltage is supported across the drift region. The doping concentration and the length of the drift region determine the forward breakdown rating of the device. Both the drift region and the narrow JFET region add internal resistance to the device.
Figure 2: cross section of a power MOSFET [3]
The large internal resistance of the MOSFET was the reason behind the further development of the different types of MOSFETs leading to IGBT
IGBT
Insulated Gate Bipolar Transistor (IGBT) is a three terminal MOS controlled bipolar transistor made from doped silicon. IGBT was developed in early 1980s with an aim to capture the bipolar transistors market. BJTs have a low current gain at high voltages, which increases the cost of a gate drive circuitry. Moreover, slow switching speeds of BJT's made researchers to look for an alternate hybrid solutions combining both MOSFET and BJT. Equivalent circuit of this hybrid MOS-Bipolar configuration is shows in figure 3.
Figure 3: equivalent circuit of the IGBT [4]
The gate signal for the IGBT is applied to the MOSFET utilizing the high input impedance, thussimplifying the gate drive circuitry and providing the voltage controlled operation. MOSFET source current is used to power the base drive current for the bipolar transistor. This hybrid structureis very beneficial. The low resistance of the power MOSFET allows an increased base drive current for the bipolar transistor. IGBT has no internal diode as MOSFET.
Figure 4: cross section of IGBT
Absence of diode in the internal structure gives a user an open choice to choose an external recovery diode for a particular application. Figure 4 shows a cross section of a planar gate IGBT. IGBT structure is capable of supporting high voltages in a forward blocking state. When the positive bias is applied to the anode while grounding cathode and gate terminals, junction between p-anode and the n-base becomes forward biased and the junction between n-base and p-base becomes reverse biased. In this forward blocking condition, the applied voltage is supported across the reverse biased n-base/p-base junction with the depletion region extending towards the p anode with an increase in the anode voltage. The breakdown voltage of the device is determined by the doping concentration and the thickness of the n-base region. When the positive bias is applied to the gate terminal, an inversion layer is created under the gate in the p-base region. Inversion layer connects the n+ cathode region to the n-base allowing the transportation of electrons. This electron current is used to power the base drive current for the P-N-P transistor. Device can be switched from the off state to the on state by changing the gate voltage from 0 to the gate voltage (Vg) greater than the threshold voltage. Performance of the device is improved by using a trench gate structure as shown in figure 5.
Figure 5: cross section of a trench IGBT [5]
To put the device in conduction state, a gate voltage is applied along with the drain voltage. Gate voltage induces an inversion in p-base along the sides of gate electrode. When the positive drain voltage is applied, the electrons flow from the source through the inversion layer and disperse through the entire drift region and eventually get collected at the drain terminal. When the gate voltage is removed, the device goes in the forward blocking state. The junction J1 between the P-base and the N-Drift region becomes reverse biased and the voltage is supported across the drift region. As the device doesn't have a JFET region, the internal resistance is significantly reduced. To protect the gate, the gate oxide edges are rounded as the gate oxide is exposed to high voltage. Trench structure has many benefits, for e.g. increases the channel density in the MOSFET region and increase in free carrier concentration at the N-Base/emitter junction.
The turn-off losses and the on-state losses are inversely proportional so it's important to make a trade-off between them while designing the IGBT. Researchers have been constantly working to improve the design for the better performance of IGBT by changing the gate structure, adding the buffer layers to the n-base/p-anode junction and changing the doping concentrations until CIGBT's were introduced in the power semi conductor market. CIGBT's offered a lower On-state and the switching losses thus making it a possible replacement of IGBT for the future industrial applications.
CIGBT
CLUSTERED INSULATED GATE BIPOLAR TRANSISROR (CIGBT) is a three terminal MOS controlled thryster made from doped silicon. CIGBT was developed in late 1990s with an aim to capture the IGBT market. Although IGBT combines the current carrying capability of a bipolar transistor with a high input impedance of MOSFET but it is limited to medium power applications because the On-state losses increase for the high power rating devices [4]. To tackle this problem, a MOS controlled thryster like structure is created, which has a high current carrying capability and a simplified gate control circuitry. Figure 6 shows the application range of Power semiconductor devices. CIGBT aims to replace the IGBT market circled red on figure 6.
Figure 6: Application range of Power semiconductor devices [6]
Figure 7 shows a cross section of a CIGBT. The difference between the IGBT and the CIGBT structure is the addition of an n-well and a p-well. This additional n-well and p-well makes an internal thryster comprising p-anode/n-base/p-well/n-well. CIGBT has two gates. First one is used to turn on the device by connecting n-well to the n-drift and the second gate is used to control the MOSFET feeding current to the thyristor.
Figure 7: cross section of CIGBT
When the positive bias is applied to the gate terminal, inversion layer is induced under the gate, which connects the n-well to the n-base and provides a path for the transportation of the electrons. When the anode and the gate voltage is ramped up, p-anode/n-base junction becomes forward biased and holes start conducting from anode to the cathode terminal. On the other hand, the p-base/n-well junction becomes reversed biased and the electrons flow from cathode cell, through the n-base to the anode. P-base in the cathode cell has a higher doping concentration as compared to n-well, thus the depletion region moves faster in the n-well when the anode voltage is increased until the punch through occurs. When the depletion region in n-well touches the p-well, punch through occurs which clamps the cathode cell voltage. This "self-clamping" feature is unique to the IGBT and protects the cathode cell from high voltages. Moreover, self-clamping controls the saturation current of the device. CIGBT offers a lower On-state and the lower turnoff losses at higher power ratings as compared to the IGBTS, hence an attractive alternative of IGBT a potential for an economical green future. Further research has been carried out to optimize the CIGBT structure by using trench gates and by changing the doping concentrations of p-well and n-well to minimize the on-state and switching losses.
Experimental Evaluation
Under the scope of this project, devices with three n-well concentrations i.e. 6.5, 6.7 and 6.9 e+13 will be tested for DC and dynamic characterization. Results will determine the optimum n-well concentration to be used for the future development of the CIGBT's. These value of n-well doing concentration are chosen after a careful analysis because n-well change has an effect on the overall performance of the device. Increase in n-well doping decreases the threshold voltage and breakdown voltage because of counter doing and decrease in n-well doping decreases the saturation current. The effects of n-well doping on the individual performance parameters are explained later on in detail.Curve traces obtained were analysed using ORIGIN 6 software and the results were documented in the table 5
DC characterisation
DC characterization was carried out to monitor the effects of change in n-well doping concentration over following DC parameters; I(sat), Vce(sat), BV and Vth. Tektronix W71B curve tracer has been used to perform the testing.
On-State [ Vce(sat) ]
On state is the voltage drop across the device when operating at the rated current. On of the benefits CIGBT has over IGBT is the reduced On-state, which reduces the operational losses. Figure 8 shows a typical I-V curve of CIGBT.
Figure 8: On-State variation at different n-well concentrations, Vg= 15v, Id=5mA
Devices were tested to find out theoptimum n-well concentration. Theory suggests that an increase in n-well doping concentration increases the carriers in the drift region hence reduction in On-state. Devices were tested at the gate voltage (Vg) = 15 volts, device current (Id) = 6 amps. For testing, cathode was grounded and the anode and the gate voltage were slowly ramped up together. Figure 8 shows the I-V curve of three devices with a same threshold voltage. Devices with same threshold voltage were chosen to ensure devices turn on at the same gate voltage.On-state is calculated by reading the anode voltage at device current of 6 Amps on Figure 8. Results show a reduction in the On-state with an increase in n-well doping concentration. Results were summarised in the table 1 with average, maximum and the minimum values for a direct comparison between the three n-well doping concentrations.
To analyse the effect of temperature on the On-state, devices were tested at 25, 60 and 125 degree Celsius. Figure 9 shows the results of the high temperature testing. Results show an increase in On-state with an increase in the n-well doping concentration hence demonstrating a positive on-state temperature coefficient. Positive on-state temperature coefficient is required in parallel device applications.
Figure 9: On-State variation at different temperatures, Vg= 15v, Id=5mA
Saturation Current [ I(sat) ]
Saturation current is the maximum amount of current that can flow through a device at a fixed gate voltage. CIGBT has a unique self-clamping feature, which limits the potential of the cathode cell and is used to control the saturation current. When the CIGBT is switched on, the junction between p-base and n-well become reversed biased. P-base has a higher doping concentration than the n-well, because of which the depletion region in n-well moves faster towards p-well as compared in p-base as the anode voltage is increased. When the depletion region touches the p-well, the cathode cell voltage cannot increase any further hence gets clamped. When the voltage across the cathode cell gets clamped, saturation current saturates. Theoretically, Increase in the n-well doping concentration increases the cathode cell potential because it takes longer for the depletion region in n-well to touch p-well hence increases the saturation current. This unique feature is very useful in designing the CIGBT for a specific application with a strict saturation current requirement.
Devices were tested at the gate voltage (Vg) = 15 volts, device current (Id) = 50 amps and anode voltage (Vdd) = 20 volts. For testing, cathode was grounded and the anode and the gate voltage were slowly ramped up together.
Figure 10: I-V characteristics at different n-well concentration, Vg= 15v, Id=50A, Vdd= 20V
Figure 10 shows the typical IV curve for three CIGBT's with different n-well concentrations. Test results show an increase in saturation current with an increase in n-well doping concentration. Results were summarised in the table 1 with average, maximum and the minimum values for a direct comparison between the three n-well doping concentrations. Testing was repeated at 125 degree Celsius to monitor the effects of high temperature on saturation current. Figure 11 shows the change in saturation current with an increase in temperature. Test results show a decrease in I(sat). . Test results show a decrease in I(sat) because of increase in channel and drift region resistance.
Figure 11: I-V characteristics at different n-well concentration, Vg= 15v, Id=50A, Vdd= 20V
Forward Blocking voltage [ BV ]
Forward blocking voltage is the maximum voltage a device can withstand without breakdown when forward biased. This parameter is significant in industrial switching applications. Theory suggests that Increase in the n-well doping concentration decreases the forward blocking capability because of counter doping. Devices were tested at the gate voltage (Vg) = 0 volts, device current (Id) = 0 amps and anode voltage (Vdd) = 3k volts.For testing, cathode and gate was grounded and the anode voltage was slowly ramped up to 3KV. Figure 12 shows a I-V curve of a CIGBT in a forward blocking state. Test results shows that the change in n-well concentration does not affect the BV and the devices does withstand the blocking voltage of 3KV. Due to test apparatus limitations, the device was not tested up to 3.3KV.
Figure 12: Forward Blocking Breakdown Voltage, Vg= 15v, Vdd= 3kV
Threshold Voltage [ Vth ]
Threshold voltage is the minimum voltage required to induce an inversion layer for the device to go in the conduction state. Theory suggests that an increase in n-well concentration will result in decrease in threshold voltage due to counter doping of p-base. Devices were tested at the device current (Id) = 5mA and anode voltage (Vdd) = 20 volts. For testing, Cathode was grounded; gate and the anode voltage was slowly ramped up. Figure 13 shows a gate voltage (Vg) versus device current (Id) curve of a CIGBT. Threshold voltage is calculated by reading the voltage at 5mA on Figure 13. Results show a reduction in the threshold voltage with an increase in n-well doping concentration. The change is not significant because the n-well doping concentration change doesn't alter the gate oxide. Results were summarised in the table(X) with average, maximum and the minimum values for a direct comparison between the three n-well doping concentrations.
Figure 13: Threshold voltage variation vs. n-well doping concentration, Vg= 15v, Id=5mA, Vdd= 20A
Summary: DC Characterization
Table 1 provides the summary of the DC characterization of the CIGBT. Total of 89 deviceswere test tested and the results were displayed in the table. Results show that the change in n-well doping has a direct impact on the gate threshold voltage(Vg), On-state and saturation current density(Id) but has no effect on the forward breakdown voltage. Table shows the minimum, average and the maximum values of the test results, which is important to determine the outliers. Table also provides the direct comparison between the results when the same tests are performed at higher temperature. Results showthat the On-state reduces with an increase in n-well doping concentration because of increased carriers in the drift region.Increase in the n-well doping concentration increases the cathode cell potential because it takes longer for the depletion region in n-well to touch p-well hence saturation current increases. Increase in n-well doping concentrationresults in reduction of the threshold voltage because of counter doping.
Table 1: Summary of DC characterisation
Dynamic Characterisation
DC characterization was carried out to monitor the effects of change in n-well doping concentration on inductive switching losses and the gate charge.
Inductive switching
One industrial application of the CIGBT is switch inductive loads such as traction motors. One advantage CIGBTs have over IGBT is the low turn off loses. To demonstrate the superiority of CIGBT, Inductive switching test were carried out to calculate the turn off losses.The circuit in figure 14 was used to switch the devices at gate resistance (Rg) of 22, 47 and 1000.
Figure 14: Inductive Switching circuit [7]
Theory suggests that increase in n-well doping concentration increases the carriers in the drift region. Turn off process includes removal of carriers from drift region. More carriermeans longer time to remove the charge from the drift region.Devices were tested at the gate voltage (Vg) = 15 volts, device current (Id) = 6 amps and anode voltage (Vdd) = 1.8k volts. For testing, CIGBT was connected to the circuit and the anode voltage was slowly ramped up to 1.8kV. Current was ramped up to the rated 6Amps and the turn off voltage and current curves were analysed.Turns off losses were calculated by integrating the overlapping area of the voltage and current curves shown in figure 15.
Figure 15: Turn off curves , W37K9(6.9e+13), Rg=1k, Vdd=1.8k , Vgs=13.5 v, Itail= 0.1A
Results were documented into table 2 and 3 for a comparison. Results show an increase in turn-off losses with an increase in n-well doping concentration. Further more, turnoff losses increase with an increase in gate resistance.
Table 2: Summary of switching data
Table 2 shows which n-well was able to switch at what Rg and temperature. N-well with a doping concentration of 6.9e+13 was unable to switch at lower Rg because of gate polysilicon problem. Devices were unable to switch at higher temperature because of the packaging problems.
Table 3: Summary of switching data
Table 3 shows the effect of change in n-well doping concentration on turnoff losses. At Rg 22, there is no significant difference between the 6.5 and 6.9 n-well but the increased losses at 6.7 n-well are unexpected. Further work needs to be done to understand the reason behind this unexpected rise in losses. Figure 16 shows a trade-off curve between On-state and the turnoff losses. Figure 16 shows that the increase in n-well doping concentration increases the turnoff losses but decreases the On-state. Figure 16 can be used to determine the optimum doping concentration for a specific industrial application where lower On-state is more important than the turnoff losses or vice versa.
Figure 16: Switching Off Power loss vs. Vce(sat) tradeoff graph; Vdd = 1.8kv, Vgs= 15v, Rg= 1k/22/47, T=25C, Id= 6A
Gate charge measurement
Gate charge is the total gate capacitance that needs to be charged before the device turns on. Gate charge comprises of two capacitances i.e. gate-source capacitance (CGS) and gate-drain capacitance (CGD). When the gate voltage is applied, the capacitances charges up to turn the device on. Figure 17 shows the circuit used to calculate the gate charge of CIGBT.
Figure 17: Inductive Switching circuit [8]
Gate charge is calculated by integrating gate current over the switching on period. Gate current is obtained by recording the voltage immediately before and after the gate resistance (Rg). Voltage difference is then divided by the Rg to get the gate current. Gate current is integrated to obtain the gate charge. Figure 18 shows a variation of gate charge with a change in n-well doping concentration. N-well doping concentration has no effect on gate charge because the change in n-well doping concentration doesn't alter the gate oxide.
Gate-source charge
Figure 18: Gate voltage vs Gate charge , Vdd = 1.8kV, Vgs= 15v, Rg=1k, T=25C, Id=6Amps
Figure of merit (FOM)
Figure of merit (FOM) is a way of evaluating the performance of the device. FOM is the product of device resistance while operating at rated current (Rdson) multiplied by gate charge. Gate charge and Rdson are inversely proportional thus result of multiplication is always a constant.
FOM = RDSON â‹…QG
Table 4 shows the FOM of the CIGBT for three n-well doping concentrations. Results show that the doping concentration of 6.9e+13 has the highest FOM hence fastest switching capability.
N-well doping Concentration
Rds on/Ω
Gate charge/nC
FOM/ΩnC
6.5e+13
0.31
99
3.07e-8
6.7e+13
0.24
103
2.47e-8
6.9e+13
0.29
110
3.19e-8
Table 4: Figure of merit (FOM) calculation
SUMMARY: Electrical Characteristics
Table 5 provides the summary of the Electrical characteristics of the CIGBT. Total of 89 deviceswere test tested and the results were displayed in the table. Results show that the change in n-well doping has a direct impact on the gate threshold voltage (Vg), On-state and saturation current density (Id) and inductive switching, but has no effect on the forward breakdown voltage. Table shows the minimum, average and the maximum values of the test results, which is important to determine the outliers. Table also provides the direct comparison between the results when the same tests are performed at higher temperature. Results show that the On-state reduces with an increase in n-well doping concentration because of increased carriers in the drift region.Increase in the n-well doping concentration increases the cathode cell potential because it takes longer for the depletion region in n-well to touch p-well hence saturation current increases. Increase in n-well doping concentrationresults in reduction of the threshold voltage because of counter doping.Increase in n-well doping concentration increases the carriers in the drift region, which increases the turnoff losses.
Table 5: Summary of Electrical Characteristics of CIGBT
Simulations
The CIGBT structure was simulated in a virtual environment called TSUPREME. Figure 19 shows the device structure, which was simulated.
Figure 19: CIGBT structure for TSUPREME
Simulations were performed to analyze the difference between the experimental data and simulation data. Results will be use to optimize the simulation parameters to match the simulation data as closely as possible to the experimental data. This will increase the confidence in the simulation results for future development. Table 6 shows a comparison of typical values of test parameters.
Table 6: Summary of Electrical Characteristics of CIGBT
Conclusions
Results show that the n-well doping concentration of CIGBT has a direct impact on the On-State[Vce(sat)] and the turnoff energy loss. On-state reduces with an increase in n-well doping concentration because of increased carriers in the drift region. Turn-off losses increase with an increase in n-well doping because increase in the doping concentration increases the carriers in drift region and it takes longer to remove the carriers.
Results show that the gate charge and the forward Breakdown voltage (BV) are unaffected by the change in n-well doping concentration. N-well doping concentration has no effect on gate charge because the change in n-well doping concentration doesn't alter the gate oxide. BV is unchanged because the thickness and the doping concentration of n-drift region are unchanged.
Simulation data for DC correlates with the experimental results which increases the confidence in simulation for future work. However switching data doesn't correlates with the simulation data. One of the possible reasons could be the inclusion of parasitic while simulations.
Based on analysis of the data collected 6.9e+13 n-well doping concentration gives the best trade off between the Vce(sat) and the switching off losses
Future Work
Analyze the packaging problems because of which devices broke down at high temperature. After solving the problem, repeat the test again at high temperature.
Analyze the gate silicon problem and repeat the switching for the devices with n-well doping concentrationof 6.9e+13 at lower Rg
Switch all the devices at higher temperature and record the results in table 4 for a comparison.
3.3kV CIGBT is new to the industry. Work done in this project should be publicized through publishing the research papers in leading science and technology journals to attract the industry's attention.
Test the CIGBT performance in the convertor applications and compare it to the IGBT performance to demonstrate the performance superiority.
Repeat testing at higher frequency. All the testing performed was at 1Hz. It will be useful to add to the data sheet, by how much does the losses change when you change the switching frequency.
To replace IGBT in industrial application, CIGBT doesn't just have be efficient but also reliable. CIGBT should be put under a fatigue test to analyze the effects of continuous use and the mean lifetime of the device. If CIGBT fails to meet the industrial reliability standards, all efforts should be directed towards analyzing the failure reasons and improving the device design
Change the n-well doping concentration in the trench gate IGBT and analyze the effects of change of n-well doping concentration on the turn-off losses.
Results will be use to optimize the simulation parameters to match the simulation data as closely as possible to the experimental data. This will increase the confidence in the simulation results for future development.
Data Sheet
