200 degrees C operation of semiconductor power devices - IEEE Xplore

4 downloads 25991 Views 543KB Size Report
Using the results of these measurements, power supplies have been built using IGBT's and MOSFET's and operated at an ambient temperature of 200°C for up ...
159

IEEE TRANSACTIONS ON COMPONENTS, HYBRIDS, AND MANUFACTURING TECHNOLOGY, VOL. 16, NO. 7, NOVEMBER 1993

2OOOC Operation of Semiconductor Power Devices R. Wayne Johnson, Member, ZEEE, James R. Bromstead, and G. Bennett Weir

Abstract-There is a growing need for commercial and military power electronics to operate above 175°C. Changes in operating parameters at 200°C have been measured for four devices, an N-P-N bipolar junction transistor (BJT), an insulated gate bipolar transistor (IGBT), an N-channel metal-oxide-semiconductor field effect transistor (MOSFET), and a P-type MOS controlled thyristor (MCT). Using the results of these measurements, power supplies have been built using IGBT's and MOSFET's and operated at an ambient temperature of 200°C for up to 72 h.

I. INTRODUCTION

S

EMICONDUCTOR power devices are typically rated for operation below 150°C. While digital and low power analog devices have been characterized at elevated temperatures [ 11-[4], little data are available for power semiconductors over 150OC [5]. In most cases, the device is derated to zero operating power at 150-175°C. Typical space-based power sources generate high currents at low voltages. To reduce resistive losses and the weight associated with high current buses, it would be advantageous to locate the power conversion and conditioning electronics as close to the primary source as possible. Location, however, is limited by the local temperature generated by the power source. Electronics capable of operating at higher temperatures could be placed nearer the power source. In addition, conversion and conditioning electronics produce significant heat which must ultimately be removed. Enhanced temperature electronics would result in a higher heat rejection temperature for the thermal management system resulting in a reduction in the size of the thermal radiators and decreasing launch weight. Higher operating temperatures are also beneficial in terrestrial applications. The maximum temperature for automotive underhood electronics is increasing due to changes in styling and airflow patterns for improved fuel efficiency. Recently Texas Instruments discussed an automotive fuel injector power IC for operation to 200OC [6]. In computer and office equipment, higher operating temperatures reduce or eliminate the need for forced air, reducing noise levels. Temperature variations in the device parameters must be quantified in order to implement extended temperature range operation. The goal of this program was to characterize four power transistor types: a bipolar junction transistor (BJT), an n-channel metalhxide-semiconductor field effect transistor Manuscript received January 20, 1993; revised July 22, 1993. This work was supported by the National Aeronautics and Space Administration, Lewis Kesearch Center under Cooperative Agreement NCC3-I 75 and the Center for Commercial Development of Space Power with funds from NASA Grant NAGW-1 192-CCDS-AL Auburn University and the Center's Industrial partners. The authors arc with the Electrical Engineering Department, Auburn University, Auburn, AL 36849-5201. IEEE Log Number 9212411.

TABLE I DEVICES USED IN Device Type NPN N-MOSFET IGBT MCT

Device Number 2N6032 RFH7SN0SE TA9196 MCTA60P60

STUDY

Current Rating 50 A 15 A 34 A 60 A

Voltage Rating 150 V

so v

1000 v 600 V

(N-MOSFET), an insulated gate bipolar transistor (IGBT), and a MOS-controlled thyristor (MCT), and to demonstrate device operation in power conversion applications.

11. DEVICES UNDER TEST

The four devices used for testing were supplied by Harris Semiconductor. The devices were chosen as typical representatives of the four transistor types (See Table I). The bipolar transistor is available commercially in a hermetic metal package, while the other three devices are in plastic packages. The glass transition temperature for semiconductor molding compounds is typically less than 200°C and plastic packages are not suitable for 200°C operation. The N-MOSFET, IGBT, and MCT were custom packaged by Harris Semiconductor in hermetic metal packages for testing. 111. DEVICEMEASUREMENTS

During the testing, each device was mounted to a heat sink in a Delta Design 9032 test chamber. Under software control the test chamber was brought to within +/ - 0.2"C of the desired test temperature and held for 5 min. This allowed the heat sink and device to achieve the ambient chamber temperature. The individual test was then performed and the temperature raised. The test temperatures ranged from 20 to 200°C in 10°C increments. Tests performed using the TEKTRONIX 371 high power curve tracer utilized pulse current measurements. The pulse (half amplitude pulsewidth) employed by the curve tracer is 250 p s / - lo%, with a rise/fall time of 40-120 ps with a repetition rate of 0.25 times the line frequency at power levels of 3 kW and 0.5 times the line frequency at power levels of 300 W.

+

IV. RESULTS A. Saturation Voltage and On-Resistance

In current switching applications, the saturation voltage,

Vc,5a,, plays a major role in determining the bipolar transis-

01484411/93$03.00 0 1993 IEEE

IEEE TRANSACTIONS ON COMPONENTS, HYBRIDS, AND MANUFACTURING TECHNOLOGY, VOL. 16, NO. 7, NOVEMBER 1993

760

.=...==....., ....

4-

IGBT Ice = 30A. Vge=IOV

L

-P

9

Y s

2 *

3-

'

0.0

MCT la = 60A, Vga = -1OV 1-

50.0

100.0

150.0

200.0

250.0

Temperature ( "C)

2-

Fig. 2. On-resistance of N-MOSFET versus temperature.

E E . . . . " . m o . . . . m .

BJTIb=5A,Ic=SOA

0

100

* * *

200

I

10 -3-

Fig. 1. On-state voltage of the BJT, IGBT, and MCT versus temperature. h

d

10 47

B

. .. B

.

.

..:..** D

I

:I

IO I O -z. 1

Temperature ("C)

.

BJTVce= 80V ' tor losses. Under normal switching operation, the saturation IO -s e o * * * 0 voltage drop across the transistor occurs in the nonconducMCTVak=UK)V D . 00 0 tivity modulated drift region of the collector. The increased m . E m resistivity of the drift region with increasing temperature will result in increasing saturation voltage. V,, was measured at a forced Beta of 10 (using a base current It, = 5 A and limiting the collector current to I, = 50 A). Fig. 1 shows the measured collector-emitter saturation voltage for the BJT. MOSFETVds=MV There is a slight increase in saturation voltage with temperature 0 100 200 (1.7 mV/"C). Temperature CC) To measure the VcesU,of the IGBT, a gate-to-emitter voltage, \'&, of 10 V was applied. The collector-emitter voltage was Fig. 3. Collector (drain, anode) leakage current versus temperature. then acquired at a collector current of I, = 30 A, Fig. 1. Because of the MOS gate of the IGBT, the base current of the bipolar structure could not be controlled as in the BJT, so B. Leakage Currents Device leakage is often considered a limiting factor in unlike the bipolar devices studied, the IGBT shows a negative temperature dependence above 80°C. silicon devices at elevated temperature. The leakage current of The MCT's on-state voltage drop was measured with the a p-n junction approximately doubles for every 12°C increase gate-to-anode voltage (V,,) set to -10 V, and the anode- in temperature. Increased leakage results in increased off-state to-cathode current set to 60 A. Fig. 1 shows the results of losses in the power switch. Data for the BJT collector-emitter the on-state voltage versus temperature measurements. As leakage was generated by applying a collector-emitter voltage reported in [7] and [8], the forward voltage drop of the of 80 V with a HP6030A power supply, shorting the base MCT decreases with temperature at these current levels. In and emitter leads and reading the collector current with a the forward conducting state, the transistors in the thyristor TEKTRONIX DM.5120 programmable digital multimeter. The structure are in saturation and the excess carrier concentrations collector-emitter leakage current, I,,,, of the BJT as a function in the base regions reach high level injection. The doping of temperature is plotted in Fig. 3. The leakage current concentrations in these two layers becomes unimportant and increases with increasing temperature after an initial plateau. the structure behaves like a p-i-n diode with the transistor base This plateau is attributed to other parasitic leakage paths in the device which are temperature independent and dominate at the layers forming the intrinsic region. On-resistance (Rds,,) and rms drain current determine the lower temperatures. In the portion of the curve not dominated power loss in a MOSFET. Using the TEKTRONIX 371 curve by the parasitic leakage paths (100-200°C) the leakage current tracer with a pulsed drain current, I d , of 75 A and a constant was found to double every 12°C as expected. Collector-emitter leakage current for the IGBT was perVps - C$ (using the temperature dependent threshold voltage I+ shown later) of 6.5 V, the data in Fig. 2 were measured. formed using the same test circuitry as the BJT with a V,, Rds increases approximately 50 pR/OC. The increase in Rds voltage of 400 V. The MOSFET measurements were made was expected due to the decreasing carrier mobility with with V d , = 50 V and the gate connected to the source. The IGBT and MOSFET test results are plotted in Fig. 3. An increasing temperatures.

JOHNSON et a / . ; 20O0C OPERATION OF SEMICONDUCTOR POWER DEVICES

-4-

761

b

Veb-IV Vebc3V Veb5V VeMV

1200

-

b

b

.

b

.

*

*

b

.

b

b

.

b

.

*

b

*

b

b

*

b

IGBT

**,Mi3

* * * .

5 " 10-71. - - = I

0.0

50.0

J

n

.

.

100.0

. .

I

. . .

150.0

I

.

200.0

.

. . I

250.0

Temperature("C)

0

2409

me

200

Fig. 4. Emitter-base leakage current versus temperature.

*

b

b

*

*

b

b

b

b

b

b

b

b

b

0 0

attempt to reduce the MOSFET drain-source leakage current by applying a negative 5-V gate-source voltage reduced the leakage current by a factor of two over the entire temperature range. Measurements of the off-state leakage current of the MCT versus temperature were made with anode-to-cathode voltages of 500 and 300 V and a gate to anode voltage of 10 V. The 500-V measurement lead to breakdown and destruction of the device at temperatures near 200°C. (See next section.) The MCT leakage data for an anode-to-cathode voltage of 300 V is plotted in Fig. 3. The data are within the expected range, doubling every 10-12°C. The emitter-base leakage current of the BJT was measured by applying emitter-base voltages of 1-7 V using a PS5010 programmable power supply while measuring the leakage current using a DM5 120 programmable digital multimeter. The device again demonstrated a saturation of leakage current in the lower temperature range, Fig. 4, with a doubling of leakage current every 10-12°C. Breakdown of the emitterbase junction occurred at Veb = 7 V, but was soon overcome by the increase in the junction breakdown voltage with temperature. Analysis of the gate-source and gate-emitter leakage, for the MOSFET and IGBT respectively, indicated currents of less than 10 nA (these data were lost in the noise resolution of the measurement equipment) at all temperatures for all gate voltages. While the leakage currents did increase with increasing temperature, the leakage currents at 200°C were still 3-5 orders of magnitude lower than the typical on-state currents (milliamperes versus 30-75 A). The lower leakage currents of the IGBT and N-MOSFET coupled with the very low drive current leakage indicate their potential for high temperature application. C. Breakdown Voltage

The breakdown voltage determines the maximum voltage the device can block. Changes in breakdown voltage must be considered when specifying maximum operating conditions. The breakdown voltage of a p-n junction increases with increasing temperature. The increase is dependent on the doping concentrations. The variation in drain-source breakdown voltage for the devices were obtained using- a TEKTRONIX 371 curve tracer. For the MOSFET, the gate was shorted to

100

200

Temperature ("C) Fig. 5.

Breakdown voltage versus temperature.

the source and the I d s versus vds transfer characteristic were acquired and used to determine the breakdown voltage plotted in Fig. 5. The breakdown voltage shows an increase of about 7.6% per 100°C. The IGBT breakdown voltage data in Fig. 5 also shows an increase of about 7.6% per 100OC. Breakdown voltage tests for the bipolar device, Fig. 5, display an unexpected drop in V,,, breakdown voltage in the temperature region above 140°C. These curves are typical of several devices tested. Because the device base and emitter are shorted externally for this test, it is believed that the increases in the internal base resistance with temperature and the increased leakage with temperature could allow the device to begin to turn-on (internal transistor action) without an applied base current. The MCT requires an applied voltage to hold the device off. A 10-V gate voltage was applied for the breakdown measurements. Fig. 5 shows the results of the breakdown versus temperature measurement. When forward biased and held in the off state by a positive gate-to-anode voltage, the collectorbase junction of the n-p-n transistor in the thyristor structure is the only reverse biased junction blocking the anode-to-cathode voltage. The transistor action of the n-p-n transistor multiplies the current through this junction. Therefore, it is expected that the forward breakdown voltage of the MCT would decrease as temperature increases. This decrease is evident in the results of the measurement. Arther et al. have recently reported similar results for P-MCT's [9].

D.DC Beta in the BJT The dc current gain of the bipolar transistor was measured using a TEKTRONIX 371 curve tracer at a constant V& of 2.6 V. Fig. 6 shows a typical set of curves for this test. These curves show the peak in Beta at low current levels as the generation and diffusion currents dominate the emitter-base junction, and the Beta "fall-off at high current levels due to the Webster effect. At lower current levels the dc Beta increases with increasing temperature, while at high current levels the temperature dependency is less evident.

IEEE TRANSACTIONS ON COMPONENTS, HYBRIDS, AND MANUFACTURING TECHNOLOGY, VOL. 16, NO. 7, NOVEMBER 1993

762

1

140.0

Fig. 6 . DC Beta versus temperature

Temperature ("C) Fig. 7. MOSFET threshold voltage versus temperature

(I= ,.5 V).

E. Threshold Voltage in the MOSFET Using a TEKTRONIX 371 curve tracer at a V& of 5 V, the transfer characteristics for the MOSFET were obtained. The data were then plotted as Id'2 versus V,, and the threshold voltage was determined as the straight line intercept with the 1, = 0 A (12.) axis. Fig. 7 presents the threshold voltages obtained as a function of temperature. The data has a slope of -4.5 mV/"C. The change in threshold voltage can vary over the range from -2 to -100 mV/"C depending on the doping concentration and oxide thickness. At 200"C, the threshold voltage is still sufficiently high for easy control of the transistor. V. CIRCUITDEMONSTRATION From the results of the parameter testing, the IGBT and the N-MOSFET were chosen for further testing. The BJT was rejected because of its high base current needs and the complexity of the base drive circuitry. The MCT was obtained late in the program, too late to be included in circuit design. For demonstration of the devices at 200"C, a circuit was needed that would repetitively stress the devices. The circuit would also need to show a practical application for the devices at 200°C. It was decided that an H-Bridge inverter [lo] with a switching frequency of 20 kHz and a peak device current of 20 A would meet these requirements. In the search for inverter designs, "hard switching" designs were not considered in order to avoid the stress of both high current and high voltage in the switching device, instead a resonant inverter was built. The H-bridge circuit is shown in Fig. 8. Using a Unitrode UC3860 with a International Rectifier

-

N

IR2110 to drive the opto-coupled bipolar drivers, the circuit was constructed open loop to reduce circuit complexity. Split inductor design of the inverter produces half-wave resonant current in the switches. 1N3891 fast recovery diodes were used as anti-parallel diodes to commutate the second half of this resonant current. Only the power devices and diodes were placed in the oven mounted on an aluminum heat sink. The driver circuitry remained outside the chamber, connected with twisted pairs of 12 gauge silver plated Teflon coated wire. Fig. 9 shows the drain-source voltage and drain current for a low-side N-MOSFET in the inverter at 200°C. Likewise, Fig. 10 shows the collector-emitter voltage and collector current for a low-side device in the IGBT inverter at 200°C. The transistor case temperatures for the IGBT and MOSFET were measured at 214 and 208"C, respectively. No degradation in performance was observed at elevated temperature. The IGBT inverter was successfully operated at 40-A peak current for 72 h at 200°C. A second demonstration circuit was fabricated - a 28 to 42 V, 100-W switch-mode converter. The converter was designed to operate from 30-200°C. Fig. 11 shows a schematic of this regulated power supply which utilized an IGBT. The control circuitry was based on a Unitrode 3860 resonant mode control integrated circuit. A zero-current switching scheme was selected for the power supply to reduce the switching stress on the IGBT/diode pair. The IGBT, with its 1000-V breakdown voltage is well suited to zero-current switching. Zero-current switching keep losses through the IGBT to a minimum. Losses in the converter circuit have been analyzed between 30 and 200°C. In this testing, all converter components except the two capacitors and the control circuitry were placed in the oven. Input power, output power, power loss in the input inductor, and power loss in the IGBT/diode pair were measured while the circuit was delivering approximately 100 W. Fig. 12 shows the overall circuit efficiency as a function of temperature. The efficiency of the converter was 79.7% at room temperature and dropped to 71.4% at 200°C. The input power and the output power were measured using a DM5110 multimeter to record the currents and voltages. Component current and voltage waveforms were measured using a Tektronix 2440 oscilloscope and used to calculate losses in the components. Fig. 13 shows the losses in the input inductor. The losses through this inductor nearly doubled over the temperature range and represent a major portion of the losses in the circuit. At 2OO0C, the total circuit losses are 41.45 W and the inductor losses account for 19.5 W. These losses are primarily attributed to the increased resistance of the copper wire with temperature. The inductance value of the inductors was measured as a function of temperature and found to be stable as a function of temperature. The second largest losses were associated with the IGBT/diode pair. These losses are plotted versus temperature in Fig. 14. Losses in the IGBT/diode went from 6.588 W at room temperature to 14.71 W at 200°C. The majority of this power was dissipated in the IGBT.

163

IOHNSON er al.; 200’C OPERATION OF SEMICONDUCTOR POWER DEVICES

NOTE OMY TRANmORS AND DIODES ARE MOUNTED OH HEATSINY , INTESTOVEN.

0-6OVDC

{ TA9896

-

-I-

Fig. 8. H-Bridge inverter.

4.SmH

N

+

0 t

6.OpH

Fig. 11. Zero current switching 28-42-V boost converter.

Fig. 9.

Switching waveforms for low-side MOSFET in switching inverter at 200OC.

16 14 12

Temperature (“C) Fig. 10. Switching waveforms for low-side IGBT in switching inverter at 200oc.

The 5-pF capacitor dissipated 3.39 W and the 45-pF capacitor dissipated 1.94 W. Since the capacitors were not in the oven, the waveforms across them changed very little as the ambient temperature of the IGBT’s and the inductors was changed. The capacitor losses were essentially constant. The 39.36 W lost between the capacitors, the IGBT/diode pair, and the input inductor accounts for most (95%) of the 41.45 W lost in the circuit at 200°C. The remaining loss of 2.09 W in the circuit includes losses in the resonant inductor and the output diode.

Fig. 12. Overall converter efficiency versus temperature

VI. CONCLUSION Four power devices have been characterized from 30 to 200°C. No abnormal behavior, with the exception of the BJT and MCT breakdown voltage, was observed. As expected, the leakage currents increase with increasing temperature, but remain in the milliampere range at 200°C. This will increase off-state losses at elevated temperatures. Switching power converters have been built using the MOSFET and IGBT. The feasibility of operating silicon power devices at 2OOOC has been demonstrated in “soft” switching applications.

164

IEEE TRANSACTIONS ON COMPONENTS, HYBRIDS, AND MANUFACTURING TECHNOLOGY, VOL. 16, NO. 1, NOVEMBER 1993

.

90 80

I

I

I

70 60

50

40 30 20 10

I

150

100

50

200

250

Temperature ( Cf

Fig. 13. Input inductor losses versus temperature. 20. 18

I

-

16: 14: 12

.

-

10-

m

m

Q 0

I

L. J Palkuti, J. L. Prince, and A. S. Glista, Jr., “Integrated circuit characteristics at 26OOC for aircraft engine-control applications,” IEEE Trans. Comp., Hybrids, Man$ Technol., vol. CHMT-2, pp. 405412, Dec. 1979. B. L. Draper and D. W. Palmer, “Extension of high-temperature electronics,” IEEE Trans. Comp., Hybrids, Manuf Technol., vol. CHMT2, pp. 399404, Dec. 1979. J. W. Swonger, S. J. Gaul, and P. L. Heedley, “An evaluation of op amp performance up to 30OOC using dielectric isolation and bonded wafer material technologies,” in Proc. First Int. High Temperature Electronics Conf, June 1 6 2 0 , 1991, pp. 281-290. J. R. Bromstead, G. B. Weir, R. W. Johnson, R. C. Jaeger, and E. D. Baumann, “Performance of power semiconductor devices at high temperature,” in Proc. First Int. High Temperature Electronics Conf, June 1 6 2 0 , 1991, pp. 27-35. A. Marshall, “Operating power IC’s at 200 degrees,” in Proc. I992 IEEE Power Electronics Specialists Con& Mar. 1992, pp. 1033-1039. V. A. K. Temple, “Power device evolution and the MOS controlled thyristor,’’ Power Conversion and Intelligent Motion, pp. 23-29, Nov. 1987. T. M. Jahns, R. W. A. A. DeDoncker, J. W. A. Wilson, V. A. K. Temple, and D. L. Watrous, “Circuit utilization characteristics of MOScontrolled thyristors,” IEEE Trans. Industr. Appl., vol. 27, no. 3, pp. 589-596, MayiJune 1991. S . D. Auther, V. A. K. Temple, and D.L. Watrous, “Forward blocking comparison of P and N MCTs,” Conf Record 1992 IEEE IAS Ann. Meeting, Oct. 1992. N. Mapham, “An SCR inverter with good regulation and sine-wave output,” IEEE Trans. Industly General Applications, pp. 1 7 6 1 8 7 , Mar.-Apr. 1967.

87 6 4 -

R. Wayne Johnson (S’85-M’87) received the B.E.

2 -

Improvements in the circuit design can be made to increase the high temperature efficiencies of power converters. ACKNOWLEDGMENT This research was performed in the Electrical Engineering Department of Auburn University. Special thanks are expressed to Eric Baumann of the NASA Lewis Research Center, Cleveland? OH, for his support.

and M.Sc. degrees in 1979 and 1982 from Vanderbilt University, Nashville, TN, and the Ph.D. degree in 1987 from Auburn University, Auburn, AL,all in electrical engineering. He is currently an Associate Professor in Electrical Engineering at Auburn University. At Auburn, he has established teaching and research laboratories for advanced packaging. His research efforts are focused on materials, processing, and modeling for multichb modules. Dower hvbrids. and hieh temperature electronics. He has published and presented numerous papers at workshops and conferences and in technical journals. He has also worked in the hybrid microelectronics industry for DuPont, Eaton, and Amperex. Dr. Johnson is a member of IEPS and was the 1991 President of ISHM. I

1

,

I

Y

James R. Bramstead, photograph and biography not available at time of publication.

REFERENCES J. L. Prince, B. L. Draper, J. N. Kronberg, and L. T. Fitch, “Performance of digital integrated circuit technologies at very high temperatures,” IEEE Trans. Comp., Hybrids, Manuf Technol., vol. CHMT-3, pp. 571-579, Dec. 1980.

G. Bennett Weir, photograph and biography not available at time of publication.