IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 35, NO. 2, MARCH/APRIL 1999
487
Comparison of High-Power IGBT’s and Hard-Driven GTO’s for High-Power Inverters Steffen Bernet, Member, IEEE, Ralph Teichmann, Student Member, IEEE, Adrian Zuckerberger, and Peter K. Steimer, Member, IEEE
Abstract— This paper compares hard-driven gate-turn-off thyristors (IGCT’s) and high-power insulated gate bipolar transistor (IGBT) modules in a two-level pulsewidth modulation inverter. The structure, fundamental operation, and specific characteristics of the considered devices are shown. Simulations enable a loss comparison of IGCT’s and IGBT’s in a 1.14-MVA inverter at switching frequencies of fs = 250 Hz=500 Hz. The evaluation of device characteristics is the basis for a derivation of potential applications. Index Terms— Hard-driven gate-turn-off thyristor, insulated gate bipolar transistor, losses, pulsewidth modulation inverter.
I. INTRODUCTION
U
P-TO-DATE conventional gate-turn-off thyristors (GTO’s) are the gate-controlled semiconductors mostly V) and high power used at high voltages (i.e., MVA) in traction and industrial inverters. (i.e., Several manufacturers offer GTO’s up to a rated switch power of 36 MVA (6000 V, 6000 A) on the market. The tradeoff between conduction, turn-on, and turn-off behavior of conventional GTO’s leads to typical turn-off gains between 3–5. The inhomogeneous turn-off transient caused by the constriction of the turn-off current toward the center of the to about 500–1000 cathode islands limits the turn-off V/ s requiring bulky and expensive snubber circuits [1], [11]. The rather complex gate drive, as well as the relatively high power required to control the GTO, are other substantial disadvantages. However, the high on-state current density, withstand the high blocking voltages, the high off-state capability, and the possibility to integrate an inverse diode are considerable advantages of this device. Therefore, the GTO technology has found interesting applications in a power range of 0.5–20 MVA, mainly in adjustable-speed drives and railway interties. At present, there are two interesting alternatives for conventional GTO’s—the high-power insulated gate bipolar transistor (IGBT) and the hard-driven gate-turn-off thyristor (GTO), Paper IPCSD 98–61, presented at the 1998 IEEE Applied Power Electronics Conference and Exposition, Anaheim, CA, February 15–19, and approved for publication in the IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS by the Industrial Power Converter Committee of the IEEE Industry Applications Society. Manuscript released for publication August 3, 1998. S. Bernet is with Electrical Drive Systems, ABB Corporate Research, D69115 Heidelberg, Germany. R. Teichmann is with the Power Electronics Department, Dresden University, 01062 Dresden, Germany, A. Zuckerberger is at Youvalim 131, Misgav, 20142 Israel. P. K. Steimer is with ABB Industrie AG, CH-5300 Turgi, Switzerland. Publisher Item Identifier S 0093-9994(99)01130-5.
which shall be called the integrated gate commutated thyristor (IGCT) herein. The name IGCT stands for the essential technological improvements on the GTO and inverse diode, as well as the integrated gate drive. Both devices have the potential to decrease the costs and to increase the power density, as well as the inverter performance, because of snubberless operation at higher switching frequencies. There are numerous publications about the design, the characteristics, and the target applications of high-power IGBT’s ( V) [1]–[8] and IGCT’s [10]–[16]. However, up to now, a detailed comparison of both power devices has not existed. Thus, this paper derives specific characteristics on the basis of a description of the fundamental function and structure of both high-power switches. The design and simulation of a 1.14-MVA two-level pulsewidth modulation (PWM) inverter applying the first high-power IGBT module FZ1200R33KF1 (3300 V, 1200 A) [8] on the market and two reverseconducting low dc-link voltage IGCTs—5SGX08F4502 (4500 V) and 5SGX26L4502 (4500 V, 1560 A, V) [15]—enable a detailed V, 3120 A, comparison and evaluation of both switches. The active silicon area used, the semiconductor losses, the complexity of gate drives, protection, and reliability issues are addressed. II. HIGH-POWER IGBT’S The IGBT has gained more and more importance since its introduction in 1988. Today, there are 1200-V, 1700-V, 2500-V, and 3300-V IGBT’s on the market. A. Construction/Design While most of the IGBT manufacturers produce nonpunchthrough (NPT) IGBT’s for breakdown voltages of V (e.g., [3], [4], [5]), even punch-through (PT) IGBT’s have been introduced for 3300-V devices [6]. All power IGBT’s consist of many parallel chips due to the applied “MOS technology.” Today, the maximum chip size of IGBT’s is limited to 2.6 cm [12]. The considered 3300-V 1200-A IGBT module comprises 24 IGBT chips, 12 fast diode chips, and 24 gate resistance chips. Each chip is covered by a very thin (about 5 m) aluminum metallization. The connections of the chips are realized by aluminum wires which are bonded to the chip metallization by ultrasonic soldering [9]. In the 3300-V 1200-A IGBT module, 450 wires with 900 wedge bonds are required. To protect the wire bond soldering, the plastic box of the module is often filled with silicon gel. The IGBT and
0093–9994/99$10.00 1999 IEEE
488
IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 35, NO. 2, MARCH/APRIL 1999
Fig. 1. Physical arrangement of a 4500-V 1000-A IGBT module.
the diode chips are soldered on a direct copper bonding (DCB) substrate consisting of a ceramic layer of Al O or AlN (which provides the internal insulation) plus two copper layers (one at each side) which are soldered on the copper base plate. Fig. 1 shows the basic structure of a 4500-V 1000-A IGBT module as an example. The design of the IGBT module has to guarantee a reliable insulation, an equal current sharing during on-state and switching transients, a homogeneous temperature distribution, and low collector-emitter and gate-emitter stray inductances. Thus, the applied chips of one module require a narrow distribution of the chip parameters and a positive temperature coefficient of the on-state voltages.
Fig. 2. Measured hard turn-on transient of a 3300-V 1200-A IGBT module 2:25 kV; Io = 1:05 kA; Tj = 25 C). (FZ1200R33KF1; Vdc
=
Fig. 3. Measured hard turn-off transient of a 3300-V 1200-A IGBT module (Vdc = 2:25 kV; Io = 1:05 kA; Tj = 25 C).
B. On-State Behavior High-power IGBT’s realize acceptable current densities due to the bipolar injection of charge carriers. The conductivity modulation of the considered 3300-V 1200-A NPT-IGBT module is only controlled by a special low -emitter efficiency [4]. Thus, the considered IGBT does not require lifetime control. The resulting long carrier lifetime leads to the desired positive temperature coefficient of the saturation voltage, the variations of different chips, and a small influence low of the temperature on the switching losses [3]. C. Switching Behavior In Fig. 2, the snubberless turn-on transient of a 3300-V 1200-A IGBT module at a dc-link voltage of V and a load current of A is depicted. Obviously, the load current of the inductive load commutates from the freewheeling diode (inverse diode) to the turning-on IGBT. Therefore, the IGBT has to take over the reverse-recovery current of the inverse diode during the turn-on transient in addition to the load current. In the measured switching transient, the collector current increases moderately with a A/ s. The entire turnrate of current rise of on transient takes about 1.2 s. Since hard turn-on transients are basically determined by the turn-on transient of the IGBT ’s, internal MOSFET, the occurring switching times, the ’s can be adjusted by the gate drive. The and the is limited by the safe maximum rate of current rise operating area (SOA) of the inverse diode which describes
the maximum peak reverse recovery current as a function of the reverse blocking voltage of the diode [5]. Therefore, the minimum gate resistances for turn-on transients depend essentially on the dc-link voltage and the stray inductances of the circuit. Fig. 3 shows the measured snubberless turn-off transient of the 3300-V 1200-A IGBT module. The small tail current is a typical characteristic of NPT-IGBT’s. The gate drive realizes a A/ s and a rate of voltage rate of current fall of V/ s. The turn-off transient takes rise of , as well as the resulting about 5 s. The occurring turn-off losses, can be adjusted in a wide range by the gate drive. However, the reverse bias safe operating area (RBSOA) of the IGBT limits the maximum turn-off current to twice the nominal IGBT current if a maximum collector-emitter voltage V is not exceeded [5]. of Since (1) is the total stray inductance of the circuit, where is the dc-link voltage, and is the rate of current fall of IGBT turn-off, the maximum collector current to be turned off depends essentially on the dc-link voltage, the stray inductance, and the gate drive. An extension of the nonrectangular RBSOA of high-power IGBT’s to a rectangle can be expected in the near future.
BERNET et al.: COMPARISON OF HIGH-POWER IGBT’S AND HARD-DRIVEN GTO’S
489
D. Protection The IGBT is able to limit its maximum collector current, which depends on the gate emitter voltage and the junction temperature. For the considered 3300-V 1200-A IGBT V limits the current module, a gate voltage of to about three times the nominal current. If a short circuit appears, the IGBT has to be turned off within 10 s from the active region. The rectangular short-circuit safe operating area (SCSOA) limits the maximum dc-link voltage to V and the maximum short-circuit current to 3.5 times the rated module current [5]. However, if a short-circuit current cannot be turned off, the IGBT module will be destroyed by the occurring large overcurrent. After destruction, the IGBT will act as an open circuit. Both the heavy destruction of the IGBT module and the open circuit afterwards are serious drawbacks of the considered module. To prevent large surge currents of up to several hundred kiloamperes, a special short-circuit current limiter, which consists basically of a resistor with an abrupt positive temperature coefficient, can be used [7].
Fig. 4. Mechanical parts of a GCT.
E. Gate Drive Besides the recharge of the input capacitance of the IGBT during turn-on and turn-off transients, the control of and during the switching transients, the supply under voltage protection, the short-circuit protection, the adjustment of switching times and protection, as well as the generation of error signals, require sophisticated gate drives with a substantial part count of electronic devices. F. Reliability Since IGBT modules provide insulation between the power circuit and cooling system, the durability of the insulator is one major aspect of reliability. By changing the ceramics from Al O to AlN, the considered IGBT module now meets the partial discharge test defined in the international standard IEC 1287 [2]. The reliability of the aluminum wires, including the bonds, is another critical issue. After extensive work, it was reported in [2] that bond wire lift-offs and corrosion are not the main failure mechanism anymore. Instead, intercrystalline cracks along grains of the current leads were the limiting points in a K. 9 000 000-cycle test at Another serious source of failure is the increase of the thermal resistance after power cycling tests. Both the migration of the thermal contact grease between module and heat sink and the degradation of the internal thermal contacts caused by the thermomechanical stress in the solder between the DCB substrate and the copper base plate can lead to an inhomogeneous thermal contact of the module [9]. III. INTEGRATED GATE COMMUTATED THYRISTORS Substantial improvements of the conventional GTO structure, the gate drive, the packaging, and the inverse diode, as well as the change of the turn-off process, resulted in a drastically improved GTO which is considered as a new component—the IGCT. 4.5-kV (1.9 kV/2.7 kV dc-link) and
Fig. 5. Physical arrangement of an IGCT.
5.5-kV (3.3 kV dc-link) IGCT’s with currents of 275 A 3120 A have been developed [13]. A. Construction/Design The key idea of the IGCT is the hybridization of an improved GTO structure and an extremely low inductive gate drive. In contrast to the IGBT and its many parts (e.g., 60 450 bond wires for the 3300-V 1200-A IGBT), the chips applied gate commutated thyristor (GCT) consists of only a few mechanical parts (Fig. 4): 1) silicon waver which is divided into a GCT part and a diode part for reverse-conducting IGCT’s; 2) gate ring which permits a low inductive contact from the gate terminal to the gate segments on the waver; 3) molybdenum plates; 4) copper cases of anode and cathode; 5) gate ring terminal. Fig. 5 shows the GCT with integrated gate drive (IGCT). The distance of 15 cm between gate driver and GCT guarantees that this arrangement will fit into different types of stacks. A new IGCT design allows maximum compactness since gate drive, GCT, and cooler form one robust unit [14]. A “snap-in assembly” eases the assembly and maintenance of both types of IGCT’s [12].
490
IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 35, NO. 2, MARCH/APRIL 1999
A substantial improvement of the GCT has recently been achieved by the introduction of a buffer layer at the anode side. Buffer-layer power semiconductors generate distinctly fewer on-state losses and switching losses than conventional NPT elements due to their up to 30% reduced device thickness for the same forward breakdown voltage [13], [11]. In the new IGCT’s, the buffer layer is combined with a transparent anode which is basically a pn-junction with current-dependent emitter efficiency. Trigger current and on-state gate current (or backporch current) are very small, since the emitter efficiency of the transparent anode is high at low current. On the other hand, electrons can be extracted as efficiently as through conventional anode shorts during turn-off because the transparent emitter is designed for low injection efficiency at high current density in the latching state [13]. In the past, the advantages of monolithic NPT-GTO and diode combinations were always diminished by the fact that the NPT-GTO required a thicker silicon chip than its corresponding freewheeling diode. Thus, reverse-conducting GTO devices suffered from excessive diode losses. However, in the new buffer layer concept, the minimum thickness of a PTGCT and of the inverse diode are essentially the same, which makes the monolithic GCT/inverse diode configuration very attractive [13]. B. On-State Behavior Compared to IGBT’s, IGCT’s have the important advantage of substantially lower on-state voltages in the considered V). GCT’s (GTO’s) blocking voltage range ( enable high current densities (second only to thyristors) at low on-state voltages, even at high blocking voltages due to the latching state of the thyristor structure. Compared to conventional GTO’s, the on-state gate current of GCT’s is reduced by a factor of ten due to the applied transparent anode technology. C. Switching Behavior The active turn-on transient of the IGCT at inductive loads nH), is improved by the low-inductance gate drive ( as well. The fast impression of the positive gate current leads to a more homogenous turn-on transient. In experiments, no inhomogeneties have been observed at a rate of current rise of A/ s [12]. However, to keep the turning-off must be limited diode within its safe operating area, the during the turn-on transient of the IGCT. Due to the latching during the turn-on transient, the IGCT cannot provide (or ) control. Instead, a small concentrated turn-on snubber consisting of an inductor, a freewheeling diode, and of the turning-off diodes for a a resistor limits the two-level inverter (Fig. 8). clamp circuit relieves the turn-on Additionally, the transient of the IGCT’s and transfers losses to the clamp resistor, which accepts higher temperatures and requires less cooling than semiconductors. Fig. 6 shows the measured hard turn-off transient of a 4.5V kV 3-kA IGCT. Using a gate voltage of during the turn-off transient, the negative gate current rises
Fig. 6. Hard turn-off transient of a 4500-V 3000-A IGCT (Vdc Io = 3 kA; Tj = 125 C, ts = 1:6 s).
= 3:5 kV;
with kA/ s, thereby commutating the complete cathode current to the gate before the main GCT blocking junction takes over voltage. Thus, the GCT changes from its pnpn latching state to the rugged transistor pnp mode within 1 s, enlarging the SOA to full dynamic avalanche. Therefore, the IGCT does not require any turn-off snubbers. If the load is purely inductive, the anode current remains unchanged until the IGCT voltage reaches the dc-link voltage. The main part of the losses is generated during this interval and only determined by the rate of voltage rise. As soon as the dc-link voltage is reached, the current commutates into the clamp. The occurring IGCT tail current is short due to the buffer layer technology. The hard gate drive causes a storage time of only 1.6 s in the investigated operating point. In contrast to conventional GTO’s, where a fairly long minimum time between consecutive turn-off transients is defined to return to a uniform junction temperature, the homogenous turnoff transient of IGCT’s overcomes this drawback. Therefore, only the thermal impedance limits the maximum switching frequency of IGCT’s. As an example, [13] presents a test pulse pattern where an IGCT is stressed with ten 25-kHz pulses (10 s on, 30 s off). D. Protection The output short-circuit protection profits directly from the of an external shortfast switching of IGCT’s. If the circuit current is limited by a filter or a cable inductance, the IGCT’s can turn off before the maximum turn-off current of the semiconductors is reached [12]. clamp In the case of an internal shoot-through, the of the inverter limits the maximum peak current. Of course, protection firing of all elements is possible in order to reduce the stress of the defect phase. A shoot-through will safely discharge the dc-link capacitance, since the IGCT’s will safely short circuit under all worst case failure conditions. This is especially advantageous in converters with series-connected devices. E. Gate Drive The IGCT gate drive delivers the required gate current for the switching transients and the on state. Despite the increase
BERNET et al.: COMPARISON OF HIGH-POWER IGBT’S AND HARD-DRIVEN GTO’S
Fig. 7. Circuit configuration of a two-level IGBT inverter.
of the amplitude of the gate current , the gate turn-off is reduced to about 40% of that of a conventional charge GTO, since the storage time is decreased by a factor of 1/15. This, and the 90% reduction of the on-state gate current, leads to a 50% reduction of the gate drive power in comparison to a GTO. The required gate drive power of the 4500-V 1500-A IGCT is about five times the gate drive power of the 3300-V 1200-A IGBT module in the considered PWM inverter. The fast switching transients of the IGCT’s do not require a control of the switching times on the gate drive itself. It is interesting to note that the part count of the IGCT gate drive is only slightly higher than that of a standard IGBT gate drive. F. Reliability Due to a low total part count and established technology, a high reliability is guaranteed. A number of qualification tests, field experience on reliability of key components (up to 400 000 000 device operations hours), and recent data from a 100-MVA railway intertie indicate a failure in time (FIT) , where of a full 3-MVA inverter of failure in 1 000 000 000 000 h. The contribution of the gate drivers is not significantly larger than with standard 600–1200V IGBT inverters, since fiber optics and logics are similar, and the power devices, including the pulse capacitors, behave extremely well. IV. LOSSES OF A TWO-LEVEL PWM INVERTER WITH IGCT’S AND IGBT’S A. Data of the PWM Inverter To compare the losses of two-level PWM inverters using IGCT’s and IGBT’s, the lower part of the power range of IGCT converters was chosen. The considered inverter ( V, V, A, MVA) features sinusoidal modulation with added third harmonics. The IGBT inverter was assumed to operate totally snubberless to achieve a minimum part count (Fig. 7). In contrast to that, a small clamp was assumed in the IGCT inverter to limit the A/ s (Fig. 8). rate of current rise to about The following devices were chosen: IGBT module: V, A) [8]; FZ1200R33KF1 ( V, reverse-conducting IGCT’s: 5SGX08F4502 (
491
Fig. 8. Circuit configuration of a two-level IGCT inverter.
V, A) [15], and 5SGX26L4502 V, V, A) [15]. Both IGBT and IGCT can be operated up to a dc-link V. The IGBT module has a voltage of about proportion of about 2:1 between IGBT and diode chip size area. It should be noted that the active area of the 4500-V 3120-A IGCT is only 69% of the active area of the 3300V 1200-A IGBT. The active area of the integrated inverse diode of the 4500-V 3120-A reverse-conducting IGCT is only about 58% of the active area of the inverse diode of the IGBT module. Considering the 4500-V 1560-A IGCT, the active areas of this IGCT and of the integrated inverse diode are only 33% of the active area of the IGBT and the inverse diode, respectively. -
(
B. Loss Model The losses of the voltage-source inverter were simulated using a previously developed power semiconductor loss model [16]. The basis of the program is the loss approximation of typical turn-on and turn-off losses, as well as on-state voltages given in the data sheet. In addition to these fitted functions, the model contains an algorithm which determines the type of each commutation during the simulation of the inverter and distributes the losses to the active semiconductors. The developed loss model describes the semiconductor losses very accurately, since, on the one hand, the switching and on-state losses of each semiconductor are calculated at each integration s) and, on the other hand, the fitted functions step ( closely approximate the measured switching losses and onstate voltages. To simplify the simulated system, the load was represented by three ideal sinusoidal current sources. Losses clamp of the IGCT inverter and of a short-circuit of the current limiter of the IGBT inverter were not considered. C. Loss Simulations Fig. 9 shows the on-state voltages of the considered IGBT and IGCT’s as a function of the collector/anode current. As expected, the on-state voltages of the IGCT’s are substantially lower at higher current densities than the on-state voltage of the IGBT. The turn-on and turn-off loss energies of the IGBT, IGCT’s, and inverse diodes at a dc-link voltage of
492
IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 35, NO. 2, MARCH/APRIL 1999
Fig. 9. On-state voltages of IGBT’s and IGCT’s (Tj
= 125 C).
V and a junction temperature of are approximated by the following functions:
-
[J/A] [J/A] [J/A] [J/A] [J/A] [J/A]
-
[J/A]
-
C
Fig. 10. Conduction, switching, and total losses of a PWM inverter using 3300-V 1200-A IGBT modules and 4500-V 3120-A IGCT’s as a function of the modulation index (Vdc = 1500 V; Io = 600 A; o = 25 C; fs = 500 Hz; Tj = 125 C).
[A] (2) [A] (3) [A] (4) [A] (5) [A] (6) [A] (3120-A IGCT) (7) [A] (1560-A IGCT) (8)
Obviously, the IGCT’s also realize clearly less turn on losses clamp in the than the considered IGBT due to the applied IGCT inverter which relieves the turn-on transient distinctly. In contrast, the IGBT generates the lowest turn-off losses using . The turn-off losses of the a gate resistance of inverse diode of the IGCT are higher than those of the IGBT inverse diode due to the higher not optimum voltage rating of V) for this inverter. the IGCT ( Fig. 10 shows the sum of conduction losses , switching losses , and total losses of IGBT’s, IGCT’s, and inverse diodes of the PWM inverter as a function of the . In the considered working modulation index range, the reverse-conducting IGCT’s generate between 16% and 33% fewer total losses than the IGBT modules due to the lower total on-state losses. The loss reduction increases with increasing modulation index and rising conduction times of the active semiconductors, since the on-state losses of the IGCT inverse diodes are higher than those of the IGBT inverse diodes. The inverter on-state and switching losses are depicted in Fig. 11 as a function of the rms value of the output phase current. The PWM inverter applying 3120-A IGCT’s also realizes in this operating range up to 33% fewer total losses than the IGBT inverter due to the dominating on-state losses of the active semiconductors. Fig. 12 shows the total inverter losses and the distribution of IGBT, IGCT, and inverse diode losses as a function of
Fig. 11. Conduction, switching, and total losses of a PWM inverter using 3300-V 1200-A IGBT modules and 4500-V 3120-A IGCT’s as a function of the output current (Vo ph = 520 V; o = 25 C; fs = 500 Hz; Tj = 125 C).
Fig. 12. Losses of a PWM inverter using 3300-V 1200-A IGBT modules and 4500-V 3120-A IGCT’s as a function of the output phase shift (Vo ph = 520 V; Io = 600 A; fs = 500 Hz; Tj = 125 C).
the phase shift between output voltage and output current. The total losses of the IGBT inverter reach a minimum at where the diode conduction predominates, because the switching losses are basically independent of the phase
BERNET et al.: COMPARISON OF HIGH-POWER IGBT’S AND HARD-DRIVEN GTO’S
493
TABLE I COMPARISON OF CHARACTERISTICS OF AN IGBT MODULE AND IGCT’S.
Fig. 13. Losses of a PWM -inverter using 3300-V 1200-A IGBT modules, 4500-V 1560-A IGCT’s, and 4500-V 3120-A IGCT’s at low output voltage 35 V; Io = 600 A; o = 17 C; Tj = 125 C). (Vo ph
=
shift at the output in the chosen modulation scheme, and the on-state voltages of the diode are clearly lower than the IGBT on-state voltages. However, in the IGCT inverter, the majority of the losses are generated in the regenerative , since the on-state voltages of the mode at integrated inverse diode are higher than the IGCT on-state voltages. It should be noted that a substantial further loss reduction of the IGCT inverter could be achieved by the use of optimized 3300-V IGCT’s (including inverse diodes) in this application. ; where : IGCT/IGBT; The switching ( : diode) and on-state losses at very low output voltage, high output current, and low output phase shift are seen in Fig. 13. In this operating point, the inverse of the diodes conduct a substantial part of the period switching frequency. The high on-state voltages of the IGCT inverse diodes cause slightly lower total losses of the 1560-A Hz and a reduction of the total losses of IGCT at 15%–16% in the case of the 3120-A IGCT in comparison to the IGBT module. D. Comparison of IGCT’s and High-Power IGBTs Table I summarizes important characteristics of the considered IGCT’s and IGBT’s in the investigated 1.14-MVA inverter, as well as general features of the compared devices. Most of the numbers are normalized to the respective base value of the 4500-V 3120-A IGCT. Comparing the devices, it is important to note that the silicon area of the IGBT is about 50% larger than that of the 3120-A IGCT. The 1560-A IGCT has only one-third of the active silicon area of the IGBT module. Obviously, the IGCT’s realize a distinctly higher silicon utilization than the IGBT module in the considered inverter. Taking this silicon utilization and the high yield into consideration, the IGCT has a substantial cost advantage compared to the IGBT module. While the IGBT module consists of 60 chips which are connected by 450 bond wires, the IGCT’s have just one silicon waver in the proven reliable press pack. The IGCT’s realize just 50% (3120-A IGCT) and 67.5% (1560-A IGCT) of the on-
state voltage of the IGBT at 45% (3120-A IGCT) and 189% (1560-A IGCT) increased current density, respectively. The clamp of the IGCT inverter causes clearly lower turnon losses of the IGCT’s in comparison to the investigated IGBT. However, the IGBT generates lower turn-off losses is assumed. than IGCT’s if a gate resistance of Despite the substantially reduced area of active silicon, the 1560-A IGCT’s realize lower losses than the IGBT’s in the considered PWM inverter at low and medium modulation Hz. The 3120indices and a switching frequency of A IGCT’s generate about 16%–33% fewer losses than the IGBT modules. Comparing state-of-the-art gate drives of IGBT’s and IGCT’s, the complexity of a high-power IGBT module gate drive was assessed to be as complex as the gate drive of the 1560-A IGCT. The part count and complexity of the 3120-A IGCT was evaluated to be slightly higher. The IGBT gate drives require about 10%–20% of the IGCT gate drive power due to the MOS control of the high-power IGBT. However, the absolute value of the gate drive power is very small for all ’s and ’s semiconductors. The possibility to adjust during switching transients using the gate drive, the possibility of active clamping, and the limitation of short-circuit currents by the device combined with the possibility to turn off the
494
IEEE TRANSACTIONS ON INDUSTRY APPLICATIONS, VOL. 35, NO. 2, MARCH/APRIL 1999
short-circuit current within 10 s actively are advantageous features of the high-power IGBT’s. Both IGCT’s and IGBT’s have no problems with high switching frequency stress of worst case pulse patterns. The risk of a shoot-through is always present in a dcvoltage-link inverter. Of course, this situation has to be handled safely. In the IGCT inverter, the surge current is limited clamp. The IGCT’s will safely short circuit under by the all worst case failure conditions, and the control will stop the operation of the inverter immediately in this case. A special short-circuit current limiter can limit the destruction of IGBT modules caused by a shoot-through. However, the open circuit of an IGBT module after destruction is a serious drawback of this device in several applications, for instance, in converters with series connection. At the moment, there are not sufficient data about the reliability of high-power IGBT’s in industrial or traction converters available. In contrast to that, there are excessive data of field experience of IGCT’s. Recent data from qualification tests on a 100-MVA railway intertie in Germany indicate that an FIT is to be expected for a full 3-MVA IGCT of inverter. The proven outstanding reliability of IGCT inverters is another substantial advantage of this technology.
[7] G. Hilpert and T. Z¨ullig, “Integrated power module in IGBT technology for modular power traction converters,” in Conf. Rec. EPE’97, 1997, pp. 1106–1111. [8] Technical Information IGBT Module FZ12000R33KF1, EUPEC, Warstein, Germany, 1996. [9] A. Hamidi, G. Coquery, and R. Lallemand, “Reliability of high power IGBT modules testing on thermal fatigue effects due to traction cycles,” in Conf. Rec. EPE’97, 1997, pp. 3118–3122. [10] H. Gr¨uning, “High power integration,” EPE J., vol. 7, p. 5, 1997. [11] H. E. Gr¨uning, B. Ødegard, J. Rees, A. Weber, E. Carroll, and S. Eicher, “High power hard driven GTO module for 4.5 kV/3 kA snubberless operation,” in Conf. Rec. PCIM’96, 1996, pp. 169–183. [12] H. E. Gr¨uning and B. Ødegard, “High performance low cost MVA inverters realized with integrated gate commutated thyristors (IGCT),” in Conf. Rec. EPE’97, 1997, pp. 2060–2065. [13] S. Lindner, S. Klaka, M. Frecker, E. Caroll, and H. Zeller, “A new range of reverse conducting gate commutated thyristors for high voltage, medium power application,” in Conf. Rec. EPE’97, 1997, pp. 1117–1124. [14] P. K. Steimer, H. E. Gr¨uning, J. Werninger, E. Carroll, S. Klaka, and S. Lindner, “IGCT—A new emerging technology for high power, low cost inverters,” in Conf. Rec. IEEE-IAS Annu. Meeting, 1997, pp. 1592–1599. [15] Data Sheet—Reverse Conducting IGCT’s 5SGX26L4502, 5SGX08F4502, ABB Semiconductors, Lenzburg, Switzerland, 1997. [16] S. Bernet, T. Matsuo, and T. A. Lipo, “A matrix converter using reverse blocking NPT-IGBT’s and optimized pulse patterns,” in Conf. Rec. IEEE-PESC’96, 1996, pp. 107–113.
V. CONCLUSION Low losses at small active silicon area, fast switching, a small part count, and the reliable press pack in a compact mechanical arrangement which can be easily assembled enable the design of low-cost, compact, reliable, highly efficient, and 100% explosion-free IGCT inverters. A 300-kVA–10-MVA IGCT converter can be achieved without series or parallel connection of devices. The simple and robust series connection of IGCT’s will extend the power range of IGCT converters up to several hundred megavoltamperes for the power system market. High-power IGBT’s offer interesting features, like and , active clamping, shortactive control of circuit current limitation, and active protection. However, higher on-state losses, a substantially smaller utilization of the active silicon area, an open circuit after destruction, and reliability concerns limit the application of conventional highpower IGBT modules.
Steffen Bernet (M’97) was born in Ilmenau, Germany, in 1963. He received the Master’s degree from Dresden University, Dresden, Germany, in 1990 and the Ph.D. degree from the Technical University of Ilmenau, Ilmenau, Germany, in 1995, both in electrical engineering. He was a Development Engineer in the Department of Private Communication Systems, Siemens AG, from 1994 to 1995. During 1995–1996, he did postdoctoral work at the University of Wisconsin, Madison. In 1996, he joined ABB Corporate Research, Heidelberg, Germany, where he led several power electronics research projects and developed various novel high-power converters, including resonant snubber-based current-source converters, matrix converters, and three-level voltage-source inverters. Since 1998, he has led the Electrical Drive Systems group. His main research areas are high-power converter topologies, power semiconductors, and motor drives.
REFERENCES [1] J. M. Peter, “Power components: Which evolution? Consequences of the power converter design,” in Conf. Rec. AES’97, 1997, pp. 101–108. [2] K. Sommer, J. G¨ottert, G. Lefranc, and R. Spanke, “Multichip high power IGBT-modules for traction and industrial application,” in Conf. Rec. EPE’97, 1997, pp. 1112–1116. [3] H. Brunner, M. Hierholzer, T. Laska, A. Porst, and R. Spanke, “3300 V IGBT module for traction application,” in Conf. Rec. EPE’95, 1995, pp. 1056–1059. [4] H. Brunner, M. Bruckmann, M. Hierholzer, T. Laska, and A. Porst, “Improved 3.5 kV IGBT-diode chipset and 800 A module applications,” in Conf. Rec. IEEE-PESC’96, 1996, pp. 1748–1753. [5] M. Hierholzer, R. Bayerer, A. Porst, and H. Brunner, “Improved characteristic of 3.3 kV IGBT modules,” in Conf. Rec. PCIM’97, 1997, pp. 201–204. [6] K. Ishii, Y. Konishi, M. Takeda, E. Thal, and G. Debled, “A new high power, high voltage IGBT,” in Conf. Rec. PCIM’97, 1997, pp. 185–190.
Ralph Teichmann (S’96) was born in Dresden, Germany, in 1972. He received the Master’s degree in electrical engineering in 1997 from Dresden University, Dresden, Germany, where he is currently working towards the Ph.D. degree. From 1995 to 1996, he was a Guest Student at the University of Wisconsin, Madison. He was also a Research Assistant at Nagasaki University, Nagasaki, Japan. He gained industrial experience during several phases of project work for ABB Corporate Research, Heidelberg, Germany. His research interests include high-power conversion, hard- and soft-switching ac/ac converter topologies, and converter controls.
BERNET et al.: COMPARISON OF HIGH-POWER IGBT’S AND HARD-DRIVEN GTO’S
Adrian Zuckerberger received the B.Sc., M.Sc., and D.Sc. degrees in electrical engineering from the Technion—Israel Institute of Technology (IIT), Haifa, Israel, in 1977, 1983, and 1988, respectively. Between 1988–1989, he was a Lecturer and Researcher in the Electrical Engineering Department, Technion—ITT. From 1989 to 1991, he was a Researcher in the ABB Research Center, Baden, Switzerland, dealing with fast bipolar power devices. From 1991 to 1996, he was an Assistant Professor in the Department of Electrical Engineering, Technion—IIT. His research interests included matrix converters and resonant topologies and simulation of power electronics components and systems. During 1997–1998, he was with ABB Industrie AG, Turgi, Switzerland, dealing with IGCT technologies and MV systems design. Presently, he is a Consultant in the field of power electronics and drives.
495
Peter K. Steimer (S’88–M’89) was born in Zurich, Switzerland, in 1956. He received the Master’s and Ph.D. degrees in electrical engineering from the Swiss Federal Institute of Technology (ETH Zurich), Zurich, Switzerland, in 1981 and 1990, respectively. In 1981, he joined Brown, Boveri & Cie., as a Research Engineer, working in the field of control of drives and static VAR compensators. From 1991 to 1996, he was General Manager of the Research and Development Department, ABB Industrie AG, Turgi, Switzerland, where his focus was on high-power electronics for drives, high-power VSC interties, static VAR compensators, and high-current rectifier and excitation systems. From 1994 to 1997, he was responsible within ABB Industrie AG for the development of the new IGCT technology. Since 1996, he has been the Manager of Technology and Innovation. Dr. Steimer is a member of the International Steering Committee of the European Power Electronics and Drives Association.