An Insulated Gate Bipolar Transistor (IGBT) is a more attractive device for high power and high voltage applications. The device combines the advantages of BJT which has fixed voltage drop in the on-state, high voltage, and high current ratings and a MOSFET which is a voltage controlled majority carrier device. The major problem of IGBT for operation at high frequency is the "current tailing". At turn-off the device current does not fall rapidly but a considerable portion of the current lingers or tails for a longer time. The co-existence of tail current and high collector-to-emitter voltage of IGBT cause high turn-off switching losses. This sets the upper limit on the switching frequency of an IGBT.

In order to increase the switching frequency of operation with IGBT, one must use the switch in soft-switching mode. That is, the switching transitions must take place either under zero-voltage-switching (ZVS) or zero-current-switching (ZCS) conditions. In ZVS, turn-on losses are eliminated but the turn-off losses are reduced by slowing down the rise of voltage across the device. With the presence of tail current, ZVS does not completely eliminate the turn-off losses. In ZCS, collector current is forced to zero before turn-off. Therefore the effect of tail current on the turn-off losses is eliminated. However, with ZCS the loss due to discharge of device's output capacitance is not prevented.

Resonant, quasi-resonant, and resonant-transition converters can create favorable ZVS or ZCS conditions for an IGBT in a circuit. This is achieved with the help of auxiliary switches and/or additional inductors and capacitors. Apart form the increased complexity, these techniques can increase the peak voltage and current stresses. A simple and effective method of reducing turn-off loss in an IGBT with the help of MOSFET is proposed in Reference 9 of Qian et al. In the said technique a MOSFET is either operated in series or parallel with the IGBT and with the proper sequencing of the turn-on and turn-off of the MOSFET, soft-switching of IGBT is achieved.

In this article the MOSFET-assisted soft-switching of IGBT is re-investigated. With simulation models, it was noticed that although the total power dissipation in MOSFET and IGBT is considerably reduced, the loss in MOSFET is almost equal to that in the IGBT under the most favorable operating conditions. This contradicts with the intention: since a MOSFET is used only to assist the soft-switching of the IGBT, we intend that the MOSFET ratings should be small as compared to the main switch—IGBT. If we have a MOSFET that can dissipate half of the total switch power dissipation, we would rather put two MOSFETs in parallel instead of using an IGBT.

An alternative switching method suggested in this article can reduce the ratings of MOSFET.

Emitter-Open Operation
The emitter-open operation of an IGBT is shown in Figure 1a. The series connection of an IGBT and a MOSFET allows zero-current-turn-off for the latter. When the gate-drive of the MOSFET is taken off, the IGBT current is forced to zero at the switching speed of the MOSFET. There are two ways of driving the gate signals to the series connected switches. As shown in Figure 1b, the IGBT and MOSFET are turned on and off simultaneously. Once the MOSFET goes into the blocking state after turn-off, minority carriers in the IGBT have to recombine. Until then the IGBT stays in the on-state and the forward voltage is blocked by the MOSFET. Thus the IGBT turn-off under ZCS condition. Alternatively, the MOSFET can be switched off before the IGBT as shown in Figure 1c. The MOSFET turns off first forcing the current in the series connected devices to zero. After a short delay time, T_d, the IGBT turns off. During T_d although the IGBT does not carry any current, the voltage across the device is zero since the forward voltage is blocked by the MOSFET. Thus the IGBT turns off under ZC-ZVS.

Figure 1:  Emitter-open operation of IGBT. (a) Circuit diagram. (b) Switching method for simultaneous switching of IGBT and MOSFET. (c) Switching method for delayed switching of IGBT and MOSFET.

Mixed Parallel Operation
In this mode, a MOSFET and an IGBT operate in parallel as shown in Figure 2a. In the on-state, MOSFET and IGBT conduct simultaneously. IGBT conducts most of the current while the MOSFET carries only a part because the on-state drop of the IGBT is smaller than that of the MOSFET. As shown in Figure 2b, the IGBT is turned off first and the load current is transferred to the MOSFET. The IGBT exhibits tail current but its terminal voltage is maintained low due to parallelly conducting the MOSFET. The turn-off losses in the IGBT are thus reduced. After a short delay, T_d, MOSFET is turned off. The turn-off loss in IGBT depends on T_d, longer the T_d lower will be the loss. However, upper limit on T_d is decided by the switching frequency.

Figure 2:  Mixed parallel operation of IGBT. (a) Circuit diagram. (b) Conventional switching method.

Discussion
The emitter-open mode has an advantage over mixed parallel operation since the turn-off loss of the latter is a function of T_d which is limited by switching frequency. The voltage rating of the MOSFET in both the modes of operation must be equal to that of the IGBT. However, the MOSFET in emitter open additionally is rated for full load current which not only increases the conduction loss but contradicts the intension. For a high power application, if we have a MOSFET that is required to be rated as much as the IGBT, then we can use the MOSFET and eliminate the IGBT. Recall that the IGBT is chosen as a switching device due to its high voltage and current rating and we intend to reduce the turn-off loss and increase the switching frequency with the help of a fractionally rated MOSFET. Therefore, emitter-open mode is not practically suitable for high power applications.

The mixed parallel operation of an IGBT and a MOSFET can be advantageous since the structure reduces the turn-off losses in the IGBT while the current in the MOSFET during the on-state is low. The total conduction losses are not increased significantly. The current rating of an auxiliary MOSFET is expected to be low and thus this mode of operation can fulfil the intention. In the next section the mixed parallel mode is treated in detail.

V_ce : Collector-to-emitter on-state voltage drop of the IGBT
I₀ : Load current
R_d : Drain-source on-state resistance of the MOSFET
F_s : Switching frequency
T_s : Switching period (1/F_s)
T₁ : On-time of the MOSFET
T₂ : On-time of the IGBT
T_d : Delay time (T₁-T₂)
D : Effective output duty cycle (T₁/T_s)
DD : Delay duty cycle (T_d/T_s)

During period T₂ when both the IGBT and MOSFET are on, the current in the devices are given by,

and    (1)

During the delay time T_d, the current through the MOSFET is the difference between the load current and the tail current of the IGBT. For simplified calculations you can approximate that during delay time T_d,

   (2)

The MOSFET conduction loss in periods T₂ and T_d is not same since the device carries different currents and it depends on the device's on-resistance R_d in a different manner. From Equation 1 it can be seen that during period T₂, conduction loss in the MOSFET is given by,

   (3)

Similarly during T_d we get,

   (4)

Equations 3 and 4 show that it's tricky to choose on-resistance of the MOSFET since conduction loss is proportional to the resistance during some part of the switching cycle and inversely proportional to the resistance during the other part. You can derive the total conduction loss of the MOSFET as,

   (5)

Similarly, you can derive the IGBT conduction loss as,

   (6)

Figure 3:  Simulation model of mixed parallel operation used in the study

Thus we see that the conduction losses in the IGBT and MOSFET depend on R_d and DD, in addition to V_ce, I_O and D. Also, as mentioned in the previous section, the turn-off power loss also depends on DD: the larger the DD, the lower the losses. To validate this, we constructed a simulation model of a buck converter in PSpice as shown in Figure 3. The output filter and load was replaced by a current source assuming that the buck converter operates in continuous conduction mode. The real device models for the IGBT (BSM50GB120D) and the MOSFET (IRFPE50) were used whereas an ideal diode model is used for the free-wheeling diode to avoid the effect of reverse-recovery. Figure 4 shows the simulated waveforms of IGBT voltage and current with and without proposed soft-switching method. The plots show that the turn-off energy loss in IGBT can greatly be reduced with MOSFET since the overlap of high collector-to-emitter voltage and tail current is eliminated.

With D=0.8 and F_s=50 kHz, repeated simulations were performed with different values of T_d. Figure 5 shows the effect of T_d on the losses. As expected, the IGBT loss reduces but the MOSFET loss increases with increasing T_d. The total loss (IGBT plus MOSFET) thus first decreases with T_d, reaches a minimum and then increases slightly with further increase in T_d. It is important at this point to note also that when the total loss is near minimum, the IGBT and MOSFET losses are equal.

The power of this technique in reducing overall loss is illustrated by the graphs of Figure 5. The graph compares the total device loss as a function of switching frequency in the hard switching buck converter (that is, without MOSFET in Figure 5) and in the MOSFET-IGBT mixed parallel mode. At 25 kHz, the device loss is reduced to almost half and for operation beyond 75 kHz it reduced to one third.

Figure 4:  Simulated waveforms of IGBT collector-to-emitter voltage and collector current. (a) Hard-switching. (b) MOSFET-assisted soft-switching.

Figure 5:  Simulation results: the effect of Td on the losses

In another set of simulations the MOSFET's on resistance was changed to see the effect on overall loss. Figure 6 shows the plots of total loss versus the T_d for different values of R_d. It is seen in general that the total losses reduce with reducing on-resistance of the MOSFET.

Equations 5 and 6 can be used to study the effect of R_d on the IGBT and MOSFET loss. Following parameters are assumed:

V_ce = 3 V
I_o = 10 A
D = 0.8
F_s = 50 kHz
T_d = 2 µs

Figure 6:  Simulation results: comparison of total losses in IGBT-MOSFET composite switch under hard and soft switching at different operating frequencies

Figure 7:  Simulation results: effect of Td and Rd on the total losses in IGBT-MOSFET composite switch

Thus for DD = 0.1, Figure 8 shows the MOSFET, IGBT and total loss as a function of R_d. It's interesting to see that the MOSFET loss is in fact higher than the IGBT loss. As mentioned earlier this contradicts with the very purpose of MOSFET-assisted soft-switching of IGBT. It is intended to use a relatively "small" MOSFET to cause soft turn-off of the IGBT. In doing so, we do not intend to dissipate more power in a MOSFET than an IGBT. If we have a capable MOSFET, then we would use it instead of an IGBT.

Figure 8:  Analytical results: effect of Rd on the individual and total losses in IGBT-MOSFET composite switch

During period T₁ when only IGBT is on, the current in the devices are given by,

and    (7)

During period T_d1 when both the IGBT and MOSFET are on, the current in the devices are given by,

and    (8)

During the delay time T_d2, the current through MOSFET is the difference between the load current and the tail current of IGBT. For simplified calculations it can be approximated that during delay time T_d2,

   (9)

If T_d1 = T_d2 = T_d, the conduction loss of the MOSFET can be derived as,

   (10)

Similarly, the IGBT conduction loss can be derived as,

   (11)

Figure 9:  Alternative switching method for mixed parallel operation of IGBT-MOSFET composite switch

For the parameters listed in the previous section, Figure 10 shows the MOSFET, IGBT, and total loss as a function of R_d. It is clearly seen that the MOSFET loss with proposed alternative switching method is significantly less than that the IGBT loss. Figure 11 shows the total conduction losses (IGBT plus MOSFET) in the composite switch as a function of R_d with proposed and pre-proposed switching method. It is seen that the proposed switching method marginally increases the total conduction losses (by 3 W in this case) but drastically reduces the conduction loss in the MOSFET as compared to the IGBT. Therefore, it is possible to achieve required soft-switching of IGBT with the help of fractionally rated MOSFET.

Figure 10:  Analytical results: effect of Rd on the individual and total losses in IGBT-MOSFET composite switch with proposed switching method

Figure 11:  Analytical results: Comparison of total losses in IGBT-MOSFET composite switch with conventional and proposed switching method

A simulation model previously described in this article and shown in Figure 1 is simulated with proposed switching method. The simulated waveform at the turn-off of IGBT are shown in Figure 12. The results are close to the idealized waveform drawn in Figure 8. For the simulation we use, T_d1= T_d2 = T_d2= 2 µs and effective output duty of the composite switch is equal to 0.8.

Figure 12:  Simulation waveforms during turn-off of IGBT in IGBT-MOSFET composite switch with proposed switching method

To reduce the power dissipation in MOSFET, an alternative switching method is proposed. With simplified mathematical analysis it is shown that while achieving soft-switching of IGBT, power dissipation in the MOSFET is considerably reduced. This enables use of a MOSFET of smaller power rating. Simulation result confirms this.