The insulated gate bipolar transistor or IGBT is a three-terminal power semiconductor device primarily used as an electronic switch and in newer devices is noted for combining high efficiency and fast switching. It switches electric power in many modern appliances: electric cars, trains, variable speed refrigerators, air-conditioners and even stereo systems with switching amplifiers. Since it is designed to turn on and off rapidly, amplifiers that use it often synthesize complex waveforms with pulse width modulation and low-pass filters. In switching applications modern devices boast pulse repetition rates well into the ultrasonic range— frequencies which are at least ten times the highest audio frequency handled by the device when used as an analog audio amplifier.
The IGBT combines the simple gate-drive characteristics of the MOSFETs with the high-current and low–saturation-voltage capability of bipolar transistors by combining an isolated gate FET for the control input, and a bipolar power transistor as a switch, in a single device. The IGBT is used in medium- to high-power applications such as switched-mode power supplies, traction motor control and induction heating. Large IGBT modules typically consist of many devices in parallel and can have very high current handling capabilities in the order of hundreds of amperes with blocking voltages of 6000 V, equating to hundreds of kilowatts.
The IGBT is a fairly recent invention. The first-generation devices of the 1980s and early 1990s were relatively slow in switching, and prone to failure through such modes as latchup (in which the device won't turn off as long as current is flowing) and secondary breakdown (in which a localized hotspot in the device goes into thermal runaway and burns the device out at high currents). Second-generation devices were much improved, and the current third-generation ones are even better, with speed rivaling MOSFETs, and excellent ruggedness and tolerance of overloads.[1]
The extremely high pulse ratings of second- and third-generation devices also make them useful for generating large power pulses in areas like particle and plasma physics, where they are starting to supersede older devices like thyratrons and triggered spark gaps.
Their high pulse ratings, and low prices on the surplus market, also make them attractive to the high-voltage hobbyist for controlling large amounts of power to drive devices such as solid-state Tesla coils and coilguns.
Availability of affordable, reliable IGBTs is an important enabler for electric vehicles and hybrid cars.
Contents |
The IGBT is a semiconductor device with four alternating layers (P-N-P-N) that are controlled by a metal-oxide-semiconductor (MOS) gate structure without regenerative action. This mode of operation was first proposed by Yamagami in his Japanese patent S47-21739, which was filed in 1968.[2] This mode of operation was first experimentally discovered by B. Jayant Baliga in vertical device structures with a V-groove gate region and reported in the literature in 1979.[3] The device structure was referred to as a ‘V-groove MOSFET device with the drain region replaced by a p-type Anode Region’ in this paper and subsequently as the insulated gate rectifier (IGR),[4] the insulated-gate transistor (IGT),[5] the conductivity-modulated field-effect transistor (COMFET)[6] and "bipolar-mode MOSFET".[7]
Plummer found the same IGBT mode of operation in the four layer device (SCR) and he first filed a patent application for the device structure in 1978. USP No.4199774 was issued in 1980 and B1 Re33209[8] was reissued in 1995 for the IGBT mode operation in the four layer device (SCR).
Hans W. Becke and Carl F. Wheatley invented a similar device for which they filed a patent application in 1980, and which they referred to as "power MOSFET with an anode region".[9] This patent has been called "the seminal patent of the Insulated Gate Bipolar Transistor."[10] The patent claimed "no thyristor action occurs under any device operating conditions." This substantially means that the device exhibits non-latch-up IGBT operation over the entire device operation range.
Practical devices capable of operating over an extended current range were first reported by Baliga et al. in 1982.[4] A similar paper was also submitted by J.P. Russel et al. to IEEE Electron Device Letter in 1982.[11] The applications for the device were initially regarded by the power electronics community to be severely restricted by its slow switching speed and latch-up of the parasitic thyristor structure inherent within the device. However, it was demonstrated by Baliga and also by A.M. Goodman et al. in 1983 that the switching speed could be adjusted over a broad range by using electron irradiation.[5][12] This was followed by demonstration of operation of the device at elevated temperatures by Baliga in 1985.[13] Successful efforts to suppress the latch-up of the parasitic thyristor and the scaling of the voltage rating of the devices at GE allowed the introduction of commercial devices in 1983,[14] which could be utilized for a wide variety of applications.
Complete suppression of the parasitic thyristor action and the resultant non-latch-up IGBT operation for the entire device operation range was achieved by A. Nakagawa et al. in 1984.[15] The non-latch-up design concept was filed for US patents.[16] To test the lack of latchup, the prototype 1200V IGBTs were directly connected without any loads across a 600V constant voltage source and were switched on for 25 microseconds. The entire 600V was dropped across the device and a large short circuit current flowed. The devices successfully withstood this severe condition. This was the first demonstration of so-called "short-circuit-withstanding-capability" in IGBTs. Non-latch-up IGBT operation was ensured, for the first time, for the entire device operation range.[17] In this sense, the non-latch-up IGBT proposed by Hans W. Becke and Carl F. Wheatley was realized by A. Nakagawa et al. in 1984. Products of non-latch-up IGBTs were first commercialized by Toshiba in 1985.
Once the non-latch-up capability was achieved in IGBTs, it was found that IGBTs exhibited very rugged and a very large safe operating area. It was demonstrated that the product of the operating current density and the collector voltage exceeded the theoretical limit of bipolar transistors, 2x105W/cm2, and reached 5x105W/cm2.[1][17]
The insulating material is typically made of solid polymers which have issues with degradation. There are developments that use an ion gel to improve manufacturing and reduce the voltage required.[18]
An IGBT cell is constructed similarly to a n-channel vertical construction power MOSFET except the n+ drain is replaced with a p+ collector layer, thus forming a vertical PNP bipolar junction transistor.
This additional p+ region creates a cascade connection of a PNP bipolar junction transistor with the surface n-channel MOSFET.
An IGBT has a significantly lower forward voltage drop compared to a conventional MOSFET in higher blocking voltage rated devices. As the blocking voltage rating of both MOSFET and IGBT devices increases, the depth of the n- drift region must increase and the doping must decrease, resulting in roughly square relationship increase in forward conduction loss compared to blocking voltage capability of the device. By injecting minority carriers (holes) from the collector p+ region into the n- drift region during forward conduction, the resistance of the n- drift region is considerably reduced. However, this resultant reduction in on-state forward voltage comes with several penalties:
In general high voltage, high current and low switching frequencies favor IGBTs while low voltage, low current and high switching frequencies are the domain of the MOSFET.
Rather than using a device physics-based model, SPICE simulates IGBTs using Macromodels, a method that combines an ensemble of components such as FETs and BJTs in a Darlington configuration. An alternative physics-based model is the Hefner model, introduced by Allen Hefner of the NIST. It is a fairly complex model that has shown very good results. Hefner's model is described in a 1988 paper and was later extended to a thermo-electrical model and a version using SABER.[19]
|
||||||||||||||||||||