Today's IGBTs are efficient and simply cost less. They're displacing MOSFETs in many SMPS applications up to and sometimes exceeding 200 kHz. Let's look at some of the application-related differences between the various types of devices to effectively specify and employ the latest technology IGBTs and MOSFETs.
Losses in Semiconductor Switches
In an N-channel MOSFET, only electrons flow, so current is unipolar — composed only of majority carriers. IGBTs are bipolar — electrons flow just as in a MOSFET, and minority carriers (called holes) also flow. This greatly increases the current density, consequently reducing the on-state voltage. This fundamental difference not only affects conduction loss and switching loss, it also becomes the basis for advantages and drawbacks in MOSFETs and IGBTs.
Conduction loss is simply the on-state voltage multiplied by the current and the switching duty factor. Since current and duty factor depend on the load requirement, the question is the value of the on-state voltage. The answer to this question depends on current, temperature, gate voltage, and device technology.
Current — A fully enhanced MOSFET in the on state can be modeled as a voltage controlled resistor called RDS(on), because drain-source voltage increases from zero almost linearly with current. RDS(on) is the slope of the on-state voltage versus current curve, and it increases slightly with current as seen by the slight curvature in the MOSFET on-state voltage curves in Fig. 1.
The MOSFET and punch-through (PT) IGBT referenced in Fig. 1 are based on the latest Power MOS7® technology from Advanced Power Technology. The MOSFET's 25°C current rating is 46A, and the IGBT's 110°C current rating is 96A. Each is in a T-MAX™ package (TO-247 without a mounting hole) and has about the same die size.
At very low current, IGBTs behave like a diode rather than a resistor. In fact, the voltage across an IGBT, called VCE(on), is always at least a diode voltage drop because of the intrinsic diode in series with the drain of the MOSFET structure in the simple IGBT model shown in Fig. 2a. The diode is the source of minority carriers in the IGBT and also the reason for the diode voltage drop at low current . The IGBT can also be thought of as a MOSFET and PNP transistor in a Darlington type configuration, as shown in Fig. 2b. In contrast to a MOSFET, the IGBT on-state voltage increases very little with current because of the wide-base PNP transistor. As current increases, more minority carriers are injected in the IGBT, which causes a decrease in resistance to current flow.
The main disadvantage for conventional MOSFETs is conduction loss, which Fig. 1 illustrates. At 46A and 25°C, the MOSFET on-state voltage is 4.6V, whereas the IGBT on-state voltage at 45A is about 1.9V — the IGBT has about 2.4 times lower conduction loss at room temperature.
Temperature — At 46A and 125°C, the MOSFET on-state voltage in Fig. 2 approaches 10V, whereas at the same current the IGBT on-state voltage is only about 1.8V. The IGBT has more than five times lower conduction loss at 125°C. Only below about 6A at 125°C does the MOSFET have lower conduction loss than the IGBT. The dramatic difference in conduction loss at higher current means you can use a smaller die size IGBT compared to a MOSFET. Of course, a smaller die size gives IGBTs a significant cost advantage over the MOSFETs.
In contrast to a MOSFET, the IGBT on-state voltage changes very little with temperature. The IGBT conduction loss being relatively insensitive to current and temperature creates an overload advantage for the IGBT, making it much more immune to thermal runaway than a MOSFET.
GateVoltage — On-state voltage depends on gate voltage for MOSFETs and IGBTs, with a higher gate voltage corresponding to a lower on-state voltage at any given current. This effect is small for high-voltage MOSFETs and IGBTs with the gate voltage significantly higher than the threshold voltage, because the voltage sensitive channel resistance is so small compared to the drift region resistance. The resistance of the drift region decreases as the device voltage rating decreases, making the channel resistance a more significant portion of total the resistance. Fig. 3, on page 36, is a graph of the normalized RDS(on) copied from the APT5010B2LL datasheet. Because the APT5010B2LL is a high voltage MOSFET, the effect of gate voltage on RDS(on) is small. For example, choosing 30A as a reference point, there is about a 5% increase in RDS(on) at room temperature as the gate bias drops from 20V to 10V. The variation in RDS(on) with gate voltage is even less at higher temperature because the threshold voltage decreases.
IGBTs have higher transconductance than high-voltage MOSFETs, due to minority carrier injection. With this, they benefit more from higher gate voltage in terms of minimizing conduction loss. Fig. 4 shows the increase in VCE(on) for the APT30GP60B is about 4.6% at 30A, as the gate voltage drops from 15V to 10V. As the current increases, the VCE(on) sensitivity to gate voltage increases slightly. In Fig.4, the temperature is 125°C; however, the change in VCE(on) with gate voltage would be about the same at any given current — regardless of temperature.
It would seem that 10V to 20V gate drive would be suitable for the MOSFET and IGBT in terms of conduction loss. In many cases this is true; however, there's another important parameter to check before settling on a gate drive voltage — especially when squeezing the maximum performance out of the lowest cost device.
Fig. 5 shows gate voltage versus gate charge curves at three currents for the APT30GP60B. The plateau voltage increases by one volt as the current goes from 15A to 60A. There's a similar relationship between plateau voltage and current for a MOSFET. The plateau voltage also varies directly with threshold voltage. This variation of plateau voltage with current and threshold voltage impacts the available gate voltage overdrive. The threshold voltage range must be taken into account when choosing the gate drive voltage. For example, the part used for Fig. 5 measurements has a threshold voltage of 4.6V, which means the plateau voltage could shift down 1.6V or up 1.4V since the threshold voltage range for the APT30GP60B is 3V to 6V. Also, gate voltage must increase if operating at high current to maintain the same gate voltage overdrive margin. Although a 10V gate drive can be used with Power MOS7® IGBTs, it is recommended to use a 15V gate drive with IGBTs in general, for the full current capability.
Device Technology — Transistors use a P-N diode junction to block voltage. To increase the blocking voltage, the resistance of the silicon material used to create this diode junction, called the drift region, must increase. In high voltage MOSFETs, more than 80% of the conduction loss is in the resistance of the drift region. The classic approach to decrease conduction loss is to decrease feature size on the die, but this does not affect the conduction loss associated with the drift region. The IGBT solution is to use conductivity modulation by injecting minority carriers into the drift region, greatly reducing the resistance to current flow as already discussed.
Another technique uses a compensation structure in a MOSFET, which is commonly known as a super-junction MOSFET. The resistance of the drift region is greatly reduced, which would normally result in a corresponding reduction of blocking voltage. However, the geometry of the P-N diode junction is altered, distributing the electric field such that blocking voltage is not reduced. The advantage of super-junction MOSFETs is significantly reduced RDS(on) without the “side effect” of a turn-off tail current. Similar to their conventional MOSFET counterparts, super-junction MOSFETs are unipolar devices and so have characteristics similar to conventional MOSFETs such as a positive RDS(on) temperature coefficient. One disadvantage of super-junction MOSFETs is much higher output capacitance compared to the Power MOS7® MOSFETs and IGBTs, and the turn-off delay time for some devices is much longer.
In terms of current density (on-state voltage at a given current, temperature, and die size), super-junction MOSFETs fall between conventional MOSFETs and IGBTs. Consequently, the super-junction MOSFET die size for similar overall efficiency in most applications falls between conventional MOSFETs and IGBTs. So the improvement in RDS(on) of a super-junction MOSFET results in a cost advantage compared to a conventional MOSFET, and the lowest on-state voltage of the IGBT results in the greatest cost advantage overall.
Similar to conduction loss, switching losses depend on current, temperature, gate voltage, and device technology. Other factors also affect switching losses, which we'll separately discussed for turn-on and turn-off.
The turn-on characteristics of an IGBT are very similar to a power MOSFET. The normal part-to-part variation in IGBT VCE(on) has very little impact on turn-on switching energy, Eon. Threshold voltage has almost no effect on Eon for MOSFETs and IGBTs, and threshold voltage has a negative temperature coefficient for both.
Temperature has practically no effect on MOSFET and IGBT turn-on switching speed and loss. However, increasing temperature does increase the reverse recovery current of a silicon P-N diode clamp (freewheeling diode). This diode recovery current flows through a hard-switched MOSFET or IGBT during turn-on, significantly increasing the turn-on switching loss and making it temperature sensitive.
When replacing a MOSFET with a smaller die size IGBT or super-junction MOSFET, the gate resistance may need to be increased. The smaller size combined with differences in technology usually result in lower capacitance and gate charge and consequently faster switching speed.
As with Eon, threshold voltage has almost no effect on MOSFET and IGBT Eoff. MOSFET turn-off speed is independent of temperature (like the turn-on speed), and there's no tail current since there are no minority carriers.
IGBT turn-off differs from a MOSFET because of tail current. Trapped minority carriers that can only be removed by being swept out and by internal recombination cause this IGBT tail current. The tail current persists until recombination of minority carriers is complete. Thus the turn-off switching energy, Eoff in a hard-switched clamped inductive circuit gives an indication of IGBT switching speed and tail current characteristic.
NPT IGBTs control the amount of tail current by limiting the amount of minority carriers that enter the drift region. PT IGBTs also limit the amount of minority carriers, yet they also utilize lifetime control to greatly accelerate recombination. An advantage of lifetime control is recombination of minority carriers takes place even with little voltage across the IGBT, making PT IGBTs well suited for soft turn-off switching applications. The tail current of a NPT IGBT tends to be low in amplitude but long in duration — especially with high bus voltage.
Tail current in a PT IGBT is more temperature dependent than in a NPT IGBT, higher in magnitude with increasing temperature but still short in duration resulting in low Eoff. Minority carrier lifetime control in Power MOS 7® IGBTs results in turn-off energy much lower than previous generation IGBTs.
Fewer minority carriers, due to a lower injection rate or more aggressive recombination, result in higher VCE(on) because fewer carriers are available to support the current. Normal manufacturing variations result in part-to-part variation in VCE(on) and corresponding Eoff. Fig. 6 shows the Eoff versus VCE(on) relationship for a typical manufacturing range of VCE(on) in 600V Power MOS 7® IGBTs.
Gate drive voltage and impedance have an impact on switching speed and switching delay times. Turn-off delay time can be shortened by reducing the gate drive voltage, which is an advantage for very high frequency applications. However, to maintain the same turn-on switching delay and speed, the gate resistance for turn-on must be reduced by the same ratio as the reduction in gate drive voltage. Due to very low gate charge and extremely short turn-off delay times, total switching times are very short for Power MOS7® MOSFETs and IGBTs, enabling very high frequency operation with a high gate drive voltage.
Based on inductive switching energy measurements, it is easy to simulate the performance of different parts in a particular application before testing in a circuit. The following example simulates a hard-switched 3.5kW boost converter operating at 150 kHz. An APT5010B2LL Power MOS7® latest technology conventional MOSFET and an APT30GP60B Power MOS7® PT IGBT are evaluated along with a 600V, 47A super-junction MOSFET. The same test conditions were used for each device in an inductive switching test circuit, namely 400V output, TJ = 125°C, RG = 5Σ, VGG = 15V, and the same 30A high-speed silicon P-N output diode in the circuit.
The first case to consider is switching an average input current of 13A, corresponding to boosting 280V in to 400V out with a switch duty factor of 30%. Fig. 7 shows the losses for each, where Poff and Pon are the turn-off and turn-on switching losses respectively, and Pcond is the conduction loss.
The total losses are very similar for each part, which is not surprising since the IGBT has the smallest die size and highest current density and the conventional MOSFET has the largest die and the lowest current density. The conventional and super-junction MOSFET ID and IGBT IC2 current ratings are about the same.
Distribution of losses is interesting. The conduction loss is the smallest portion of the total loss, even for the conventional MOSFET. The turn-on loss dominates the total loss for each part, which is due to the diode reverse recovery current adding to the load current during turn-on. The tail current penalty of the IGBT is minimal, due to its minority carrier lifetime control; its turn-off loss isn't much higher than the super-junction MOSFET. The very fast switching speed of a Power MOS7® MOSFET is evident by the low turn-off loss of the APT5010B2LL. In this case, the Power MOS7® MOSFET is best suited for very high frequency operation, since switching losses scale with frequency.
Another case to consider is doubling the duty factor to 60%. At the same output power, this corresponds to 160V at 23A input with the same 400V output. The results of this case are shown in Fig. 8.
At higher duty factor and higher current, the difference in conduction loss becomes more evident. Under these conditions, using a larger die size part would result in a noticeable reduction in total loss.
The IGBT tail current is very controlled, and its high gain results in fast switching with low losses. In fact, the IGBT has the lowest total loss in spite of its smallest die size. Earlier generations of SMPS IGBTs could not really keep up with a MOSFET above 100 kHz, but in this case a PT IGBT performs better at 150 kHz. In fact, it has the highest efficiency of all the three devices at higher current.
What's the Best Choice?
It's a close race for each type of part in this application. In terms of total loss, any of the three devices work. The IGBT has efficiency and overload capability advantages at higher current. Operating temperature tends to be lower for the largest part, the conventional MOSFET, but junction temperature differences are modest when total thermal resistance are considered.
A simple rule is that MOSFETs perform best at low current and high frequency where their conduction loss is still fairly low and full advantage can be taken of their unrivalled switching speed. Super-junction MOSFETs extend the range of maximum usable current for a given die size compared to conventional MOSFETs due to their lower RDS(on). Depending on operating conditions, latest technology PT IGBTs can operate up to 200 kHz or even faster in hard switched applications, which means these IGBTs can now displace MOSFETs with comparable performance and lower cost in a wide variety of SMPS applications, including soft switched applications.
It's a good idea to compare inductive switching delay, rise, and fall times to make sure switching transients finish within an acceptable amount of time — especially for very high frequency applications. These parameters can vary widely between part types, and resistive switching data isn't very helpful in highlighting these differences. Inductive switching data in datasheets for all types of devices make it easier to predict total device losses, overall performance, and the amount of each type of loss.
J. Dodge, J. Hess, “IGBT Tutorial,” Application Note APT0201, Advanced Power Technology.
R. Severns, E. Oxner, “Parallel Operation of Power MOSFETs,” Technical Article TA 84-5, Siliconix Inc.
“Application Characterization of IGBTs,” Application Note INT990, International Rectifier.
For more information on this article, CIRCLE 332 on Reader Service Card