Chapter 7 Load, Switch, and Commutation Considerations 252

CHAPTER 7 Vg

Load, Switch,
and Commutation Considerations
Figure 7.1. A typical IGBT transistor switching circuit incorporating a resistive load.

Figure 7.3 shows the safe operating area (SOA) characteristics for an IGBT, on logarithmic axes.
Power switching devices are employed for controlling inductive, resistive or capacitive loads. Inductive
Illustrated are the collector switch-on and switch-off trajectories, which are virtually coincident. In the off-
loads include electrical machines, transformers, solenoids, and relays. High-current in-rush occurs with
state, point A on figure 7.2b, the transistor supports the supply rail voltage Vs while in the fully on-state,
loads such as incandescent lamps, pulse-forming networks, snubbers, and motors. Incandescent lamps
point C on figure 7.2b, the collector current Im is Vs /RL, neglecting the low on-state voltage of the
are essentially resistive, but the cold resistive in-rush current during turn-on is 12 to 18 times the steady-
transistor. During switching the collector voltage and current traverse the I-V switching trajectory
state current. This turn-on surge presents special switch-on problems. Capacitive loads, such as
between the steady-state operating conditions on → Vs /RL and off → Vs, as shown in figure 7.3.
fluorescent lighting, also present high-current in-rush at turn-on. Electromechanical loads, such as
shakers, present loads that vary between capacitive and inductive over their operating frequency range.

The interaction of the load circuit on the switch arrangement and its commutation depends on three
inter-related factors. Vg
 The type of load, usually inductive, and rarely purely resistive. ON OFF
 Switching mechanism classification, how the load effects switching commutation,
namely hard switching, resonant, etc.
 The switch characteristics required to fulfil the supply and load I-V requirements,
such as a bidirectional current switch, an asymmetrical sustaining voltage
switch, etc.

Each of the three factors and their interdependence with the switching mechanisms are considered
separately.

7.1 Load types

The two principal load types of general interest in power electronics are
 the resistive load and
 the inductive load.

Turn-on and turn-off voltage and current switching waveforms, hence losses in a switch, depend on the
type of load. Purely capacitive loads are rare, and tend to be inductive at the switch transient frequency. turn-on on-state turn-off off-state

7.1.1 The resistive load

A purely resistive load is rarely encountered in power switching applications (other than at load
resonance). Figure 7.1 shows a simple resistive load being switched by a common emitter-connected
IGBT transistor, which could equally be another appropriate semiconductor switch, for example, a
MOSFET. When the gate is driven by the voltage waveform shown in figure 7.2a, the resultant collector
voltage and current waveforms are as shown in figures 7.2b and 7.2c. These figures show that at turn-
on, as the collector current increases, the voltage across the resistive load increases proportionally, as
the collector voltage vce decreases at the same rate. That is, at turn-on,v ce (t )  Vs  i c (t )RL , while at
turn-off the inverse process occurs. Figure 7.2d shows transistor instantaneous power loss during turn-
on and turn-off, which in each case has a peak value of ¼VsIm when the collector voltage and current
reach half their respective maximum values. The energy loss W during switching is given by
W  v ce (t ) i c (t )dt (J) (7.1)
Figure 7.2. Transistor switching waveforms for a resistive load: (a) on-off gate drive voltage; (b) collector-
where the integration is performed over the switching transition period. to-emitter voltage; (c) collector and load current waveform; and (d) instantaneous collector-emitter losses.

BWW
253 Power Electronics Chapter 7 Load, Switch, and Commutation Considerations 254

It is important that this trajectory does not exceed the shown SOA bounds set by the device voltage
and current limits, and that the SOA region be traversed rapidly. For slow transitions, greater than a few
microseconds, power dissipation considerations become the limiting design factor, which is a thermal
limitation.
In order to perform the required thermal design calculations (for heatsink determination) it is
necessary to be able to specify device-switching losses. To simplify analysis, the switching waveforms
shown in figure 7.2 are linearised as shown in figure 7.4. As indicated on these waveforms, the collector
voltage fall at turn-on is given by v ce (t )  Vs (1  t / t on ) while the collector current rise is i c (t )  I m t / t on ,
where I m  Vs / RL . Combining vce(t) and ic(t) by eliminating time t, gives off on on off
i c  Vs (1  v ce /Vs ) / RL (7.2)

As shown in figure 7.3, this describes the linear turn-on transition of slope -1/RL from the on-state
voltage with Vs / RL collector current, shown as C, to the off-state at A where no current flows and the
collector supports the supply Vs. Note figure 7.3 uses logarithmic axes, so the transition trajectory does
not appear as a straight line (the inset figure is for linear axes).

Using equation (7.1), the switch-on loss for a resistive load is given by
t on t t
Wonr   0
V s (1  t )I m
on t on
dt
(7.3)
V s2
  I mV s t on or  t (J)
RL on
where I m  Vs / RL and ton is the period of the switch-on interval, as shown in figure 7.4.

Ic C B

Figure 7.4. Linear approximations of switching intervals for a purely resistive load:
(a) collector voltage and current linear waveforms and (b) corresponding energy and power losses.

Example 7.1: Resistive load switching losses

I
SOA An IGBT switches a 10 ohms resistive load across a 100V dc supply. If the switch on-state duty cycle is
C
25%, (δ = ¼), calculate the average load voltage and current. Calculate the switch losses if the switch-on
time is ton =1μs, switch-off time is toff =2μs, and the on-state voltage is 2V.
-1/RL

Solution
A V

When the switch is on, the current in the resistor is IL =Vs /R = 100V/10Ω = 10A.
VCES
The average load voltage is
Figure 7.3. Transistor I-V characteristics showing safe operating area and Vo  V s
the switching trajectory with a resistive load, on logarithmic axes, and inset, on linear axes.
 0.25  100V  25V
The average load current is
Similarly, using the time dependant collector voltage and current equations shown on figure 7.4a, the I o  Vo / R = 25V/10Ω = 2.5A
turn-off switching loss is given by
t off t t
Woffr   0
Vs I (1 
t off m t off
) dt The total switch losses PT are made up of three components.
PT = on-state loss + loss at switch-on + loss at switch-off
(7.4) PT   v ce  I L  1
VIt f  1
VIt f
V2 6 s L on s 6 s L off s
  I mV s t off or  s t off (J) = ¼×2V×10A + 1
×100V×10A×1μs  10kHz + 1
×100V×10A×2μs  10kHz
RL 6 6

where toff is the turn-off period as shown in figure 7.4. = 5W + 5

W + 10
W
The average power loss due to switching, which is required for the thermal design outlined in 3 3
chapters 5 and 6, is obtained by multiplying energy loss W by the switching frequency fs. That is, the = 10W
turn-on switching loss is given by
Pon   I mVs t on fs (W) (7.5) Since the off-state leakage current and gate power losses are not specified, it is assumed
these are insignificant. Technically the load current should be calculated based on 98V
while the turn-off loss is given by
across the load since the switch supports 2V. Also the switching loss calculations should
Poff   I mVs t off fs (W) (7.6) use a voltage of 98V, rather than 100V and a load current of 9.8A rather that 10A. The
Because of IGBT current tailing and voltage overshoot at turn-off, the practical switching losses will percentage error is small, and becomes increasingly insignificant at higher voltages.
be larger than those given by the linear approximating methods outlined. ♣
255 Power Electronics Chapter 7 Load, Switch, and Commutation Considerations 256

Example 7.2: Transistor switching loss for non-linear electrical transitions

Assume the transistor collector current at turn-off falls according to

Df
i c  ½I m (1  cos  t / T 0 ) for 0  t  T 0 (7.7)
For a resistive load, RL

i. Calculate transistor loss at turn-off.

ii. Show that the switching trajectory across the SOA is as for the linear
current fall case, as given by equation (7.2) and shown in figure 7.3.
Vg Vg
iii. Calculate the peak power dissipation and the time when it occurs.

Solution

i. The collector voltage for a resistive load, on a dc supply Vs, is given by

v ce (t )  Vs  i c (t )RL
 Vs  ½I m (1  cos  t / T 0 )RL
and since V s  I m RL
v ce (t )  ½Vs (1  cos  t / T 0 ) R
The turn-off energy loss is given by
T0 T0
Woff   0
p (t ) dt   0
i c (t )v ce (t ) dt
D
T0
  0
½I m (1  cos  t / T 0 )  ½Vs (1  cos  t / T 0 ) dt
Vg Vg
 Vs I mT 0

ii. Combining vce(t) and ic(t) so as to eliminate the time variable, yields (d)
Vs v
ic 
(1  ce )
RL Vs Figure 7.5. Four methods of limiting inductive load turn-off voltage spike and of absorbing the
which is the same straight line expression as in equation (7.2) and shown in figure 7.3, associated energy: (a) freewheel clamping diode; (b) Zener diode clamp; (c) R-C snubber circuit;
for the linear switching transition case. and (d) capacitor soft voltage clamp.

iii. Instantaneous power dissipation is given by

Vs v The freewheel diode Df in figure 7.5a is used to clamp the maximum device voltage to the supply rail
P  v ce i c  v ce (1  ce ) voltage. The stored load energy is dissipated after turn-off as a result of the current that flows in the
RL Vs
diode and load. The low impedance of the diode causes the current to decay slowly, since the inductor
stored energy can only dissipate slowly in the freewheeling loop parasitic resistive components. A
Peak power Pˆ occurs when dP/dvce = 0, that is, when vce = ½Vs, whence on shorter current decay time can be achieved if series loop resistance R is added, as shown in figure 7.5a.
substitution into the power expression P, yields Now the peak off-state voltage experienced by the switch is increased from Vs in the case of only the
Pˆ  ¼Vs 2 / RL  ¼Vs I m at t  ½T 0 diode, to Vs + ImR because of the initial voltage drop across the optionally added resistor. This extra
♣ voltage drop, ImR, decreases exponentially to zero. The resistor in figure 7.5a can be replaced by a
Zener diode, thereby clamping the switch voltage at turn-off to Vs + VZ. The load now freewheels at a
Turn-on loss can be similarly analysed to yield virtually identical expressions, as required in problem 7.4. fixed voltage VZ thereby improving the rate of current decay, which is now constant. The inductive load
current will fall linearly from Im to zero in a time given by
t  LI m /Vz (s)
7.1.2 The inductive load

The voltage spikes generated by inductive loads at turn-off may have high energy content, and the An alternative Zener diode clamping circuit, as shown in figure 7.5b, can be employed in low power
power generated may cause excessive device temperature, voltage stressing, and device failure. applications. The Zener breakdown voltage Vz is selected between the rail voltage Vs, and the switch
breakdown voltage Vs  V z  VBR  . At turn-off, the Zener diode clamps the switch voltage to a safe level
At turn-off, the switch decreases the inductive load current from Im to zero at a high di/dt and the VZ and absorbs the stored inductive load energy. The higher the clamping voltage level, the faster the
resultant inductive voltage spike is given by energy is dissipated. The inductive load current decays linearly to zero in a time given by
di t  LI m / (Vz - Vs ) (s) (7.8)
v (t )  L (V)
dt
where L is the load inductance. The spike energy to be absorbed by the switch is given by The two different Zener diode approaches perform the same switch clamping function in the same
W  ½LI m2 (J) current decay time, if the voltage experienced by the switch is the same, but with different Zener diode
losses. The desirable feature in the case of the Zener diode in parallel to the switch as in figure 7.5b, is
that the protection component is directly across the element to be voltage protected. When placed in
Both the voltage spike and its associated energy may be well outside the capabilities of the switching parallel with the load as in figure 7.5a, the switch is indirectly voltage protected, relying on the supply
device. The peak voltage induced must be limited to a value below the breakdown rating of the device. decoupling being a low inductance path. A reverse blocking diode Df in figure 7.5a is mandatory.
The parallel-switch Zener diode approach in figure 7.5b has a number of disadvantages
Four commonly employed voltage limiting techniques are shown in figure 7.5.
257 Power Electronics Chapter 7 Load, Switch, and Commutation Considerations 258

 The Zener diode voltage rating must be in excess of the supply rail, Vs, while any Zener value 100V 100V 100V

can be used when the Zener diode is in parallel with the load.
 At higher voltages, >280V, Zener diodes will have to be series connected, thus the low
inductance advantage of clamping with just one component is diminished.
 Assuming no resistance in the load, the energy dissipated with the two Zener diode Df Df
approaches differs. When in parallel with the load, the load energy ½LI m2 is dissipated while
in the second case, load and supply energy are dissipated in the clamping Zener diode. The
extra supply energy, in addition to ½LI m2 , dissipated in the Zener diode, is ½LIm2 Vs / (Vz Vs ) .
This is derived by recognising that, assuming a purely inductive load, the dc supply Vs Z1
delivers a current Im which linearly falls to zero over the period given by equation (7.8). 25V

The R-C snubbing circuit shown in figure 7.5c is commonly used in power conversion circuits to limit
spikes caused by transformer leakage inductance, diode recovery, and interconnection wire inductance.
The stored load energy is resonated to the snubber capacitor at switch turn-off. The reset resistor R
Z2
(non-inductive) must overdamp the L-C-R oscillation by absorbing the transferred energy. The resistor
also limits the snubber capacitor discharging current to a maximum of Vs /R at switch turn-on. For a
purely inductive load, the snubber resistor power losses are given by the sum of the turn-off and turn-on
losses, that is
P  ½LI m2  ½CVs2  f s (W) Figure 7.6a. Three inductive load clamping circuits.
Icoil Icoil
Figure 7.5d shows a capacitive voltage clamp used to soft clamp the switch voltage overshoot caused 1A 1A

by the inductive energy stored in the load. The capacitor retains a charge of at least Vs. At switch turn-
off, when the switch collector voltage reaches the capacitor (supply Vs) voltage level, the inductive
stored load energy is transferred to the capacitor and concurrently, the capacitor discharges the energy
in excess of Vs into the supply. When the capacitor is over charging, energy is taken from both the load Iswitch IZener Iswitch IResistor
inductance and the supply. When the capacitor discharges through the resistor back into the supply, the
earlier energy taken from the supply is returned. The net effect is that only the energy ½LI m2 is dissipated 0 10 50 t (μs) 0 10 50 t (μs)
in the resistor. A reset resistor of low inductance is not necessary - a wirewound resistor can be used.
This capacitive soft voltage clamp is analysed in detail in chapter 9.2. switch switch
voltage 175V voltage
125V

Example 7.3: Zener diode, switch voltage clamping 100V 100V

A reed relay coil of 1 mH inductance is switched at 20 kHz with a 20 per cent on-time duty cycle, across Coil voltage Coil voltage
a 100 V dc rail. The energy stored in the coil at turn-off is dissipated in a 25 V Zener diode connected as equal areas equal areas
shown in figure 7.5a.
0V 0V
i. Sketch the coil current and voltage, and the switch voltage waveforms. t t
ii. What is the average coil voltage? -25V -25V
iii. What Zener diode voltage is required for the circuit in figure 7.5b so as to produce -75V
the same coil current waveform as in figure 7.5a when using a 25 V Zener diode? Figure 7.6b. Coil voltage and current waveforms.
iv. For each circuit, calculate the power requirement of the Zener diode and the
average power delivered from the 100 V supply.
v. Calculate the minimum resistance that replaces the Zener diode in figure 7.5a if the ii. From V=Ldi/dt, for a steady-state continuous waveform, VL (t )dt  0 , i.e. 1 / T v (t )dt  Vave  0 ,
coil is to be switched on with almost zero current. Draw the coil current and switch as shown on the coil voltage waveform (the coil voltage areas cancel).
voltage waveform, showing the switch peak voltage at turn-off.
vi. Discuss the relative features of each voltage clamping approach. iii. The parallel Zener diode requirement is VZ2 = Vs+VZ1 = 100V+25V = 125V.

Solution iv. Zener diode VZ1 in the parallel-load reset circuit:

The energy ½LI 2 is transferred from the coil to the Zener diode when the switch is turned off. The
The three voltage clamping circuits being considered are shown in figure 7.6a. power dissipated in the Zener diode at 20kHz is therefore ½LI 2f s  ½×1mH×1A2×20kHz =10W.
The total power drawn from the supply is the power stored by the coil at the end of the 10μs on-
i. With a 20kHz switching frequency, the coil current rises and falls every 50μs, with an on-state duty time, namely 10W.
cycle representing 10μs for the current to increase in the coil and 40μs for the current reset decay to
reach zero. Zener diode VZ2 in the parallel-switch reset circuit:
From V=Ldi/dt, in steady-state, with zero coil resistance and zero initial current, the peak coil current When the coil releases its stored energy (10W) into the Zener, current is also drawn from the
is I = Vs t /L = 100Vx10μs/1mH = 1A. Thus the coil current rises linearly from zero to 1A in 10 μs. supply. The total average power delivered by the supply over the 50μs period is given by
During reset, the coil current waveform depends on the reset circuit. For Zener diode (constant Vs I ave  ½  100V×1A = 50W . This comprises ½LI 2 (10W) from the supply into the coil when the
voltage) reset, the current falls linearly, while with a resistor the reset current decays with an L / R switch is on for 10 μs, and the remainder (40W) into the Zener diode (plus the coil energy, 10W),
exponential time constant, as shown in figure 7.6b, for each case. when the switch is off for 40 μs. The Zener diode losses are 50W during the switch off period.

The various circuit voltage and current waveforms are shown in figure 7.6b, where data derived from v. When a resistor is used in the reset circuit, the current decays exponentially from 1A to 0A. The
the rest of this example has been incorporated. resistance determines the peak switch voltage. The resistance does not affect the amount of
energy dissipated, only the period over which the coil energy is released, dissipated as heat.
259 Power Electronics Chapter 7 Load, Switch, and Commutation Considerations 260

Assume the coil current to be near zero after three L/R time constants, that is 3L/R = 40μs = toff. It will be seen in figure 7.7 that during both turn-on and turn-off the switch must support instantaneously
For L = 1mH, this gives R = 75Ω, with a power dissipation rating of 10W from part iv. At switch a maximum voltage, Vs, and full load current, Im, condition. These severe electrical conditions are shown
turn-off the collector voltage rises to (100V+1A×75Ω) 175V and then decays to 100V. Use an 82Ω on the SOA characteristics in figure 7.8. In switching on from the operating point A to C, a maximum
(preferred value, exceeding 75Ω which reduces the time constant), 15W metal oxide resistor for voltage and current condition (Vs, Im) occurs at point D. Because of freewheel diode current reverse
low inductance. recovery effects at turn-on, an SOA trajectory point B is reached. At turn-off, due to stray inductance,
vi. A Zener diode approach gives a fixed over-voltage on the switch, independent of current or voltage over shoot occurs and the point E is reached. By comparison with figure 7.2, it is seen that
stored energy. When clamping is in parallel with the switch, only one clamping element is needed, power losses during the switching intervals are higher for an inductive load than a resistive load.
but its power requirement is significantly higher than when the clamp (Zener plus diode) is in
parallel to the load. Any resistive element must have low inductance. This is restrictive given the diode
power levels involved, and may result in only less effective wire wound elements being viable. Ic
recovery

♣ Icmax

By far the most common technique used to limit inductive switch-off voltage spikes in power circuits Voltage
involves the use of a freewheel diode without Ropt, as shown in figure 7.5a and 7.7a. Typical switching on-state tfv overshoot
E
waveforms for an inductive load clamped by a freewheel diode are shown in figure 7.7.
trv
 At turn-off, the switching device conducts the full load current as the collector voltage
rises to the supply rail. When the collector voltage reaches the supply rail level the
freewheel diode becomes forward-biased and begins to conduct. Only then can the
switch current fall to zero. The freewheel diode conducts the load current. Rds(on) tfi
limit
 At switch turn-on, assuming the diode is still freewheeling load current, the switch
tri
current increases, displacing freewheeling diode current, while the load is clamped to the
rail voltage by the conducting freewheel diode. Only when the switch conducts the full
load current can the freewheel diode recovered (and block), so that the switch voltage
can fall to the low on-state level.
Vs off-state
A

VCES

Figure 7.8. I-V characteristics for an IGBT

Vg showing its safe operating area and switching trajectory for an inductive load (linear axes).

Switching losses can be calculated by using linear approximations to the switching transitions. It can
V
be assumed that a silicon carbide Schottky freewheel diode is employed so as to allow reverse recovery
Vg
g
effects to be neglected. Figure 7.9 shows the linearised switching waveforms for an inductive load,
(V)
ON OFF where maximum voltage Vs and current Im occur simultaneously during both turn-on and turn-off. The
equations for the collector voltage and current at turn-on and turn-off are also shown in figure 7.9.
The turn-on switching interval loss is given by the time integral over the current rise period plus the
voltage fall period,
Won   Vs  I m t dt   Vs (1 - t )  I m dt
t ri t fv
di
L  c 0 t ri 0 t fv (7.9)
stray dt
= ½Vs I m t on (J)
where ton = tri + tfv, as shown in figure 7.9. The current rise time at turn-on is termed tri, while the switch
voltage fall time at turn-on is termed tfv.
B
A Similarly, from figure 7.9c, the turn-off loss is given by
t
Vs t  I m dt 
D t rv t fi
Wof f   0 t ri  0
Vs  I m (1 -
t fv
) dt
(7.10)
= ½Vs I m t off (J)
where toff = trv + tfi, as shown in figure 7.9c. The switch voltage rise time at turn-off is termed trv, while the
switch current fall time is termed tfi.
Comparison of switching losses for a resistive load, equations (7.3) and (7.4), and an inductive load,
equations (7.9) and (7.10), shows that inductive switching losses are three times those for the resistive
load case. The peak power experienced by the switch during switching of an inductive load, Vs Im, is four
Won Woff times greater than that experienced with a resistive load, ¼VsIm. As for the resistive load switching
circuit, actual switch losses with an inductive load are also higher than those predicted by equations
(7.9) and (7.10). The effects of current tailing, voltage over-shoot, and freewheel diode reverse recovery
Figure 7.7. Inductive load switching waveforms: can together produce losses of the same order as those predicted for theoretical switching by equations
(a) the circuit including the freewheel diode Df; (b) on-off gate drive voltage; (c) collector-to-emitter (7.3), (7.4), (7.9), and (7.10).
voltage; (d) collector and freewheel diode current; and (e) switch instantaneous power losses.
261 Power Electronics Chapter 7 Load, Switch, and Commutation Considerations 262

Im = iDf + ic
iDf ic KCL
Kirchhoff’s
ic iDf 7.1.3 Diode reverse recovery with an inductive load
current law
t When a silicon bipolar diode conducts the pn junction scl region accumulates charges. When the diode
turns off and the current falls to zero, the junction retains charge that must recovery before diode
(a) reverse voltage can be supported. Negative diode current flows. This phenomenon was considered in
Vs = vDf + vce chapter 4.2.2 and is shown in figure 7.10a. The maximum collector current at turn-on is increased above
vce vDf vDf vce the load current level Im by the reverse recovery current Irr, to Im+Irr. The diode begins to support
Kirchhoff’s Vs KVL
voltage
reverse voltage once the peak reverse recovery current is reached. As a consequence the switch turn-
law vDf Im on losses are increased as shown in figure 7.10c.
iDf The circuit current at peak recovery has a discontinuous derivative, and as a consequence, high circuit
voltages are induced across circuit stray inductance due to v = Ldi/dt. High-frequency voltage ringing
ic t occurs as the stored energy in the stray inductance is dissipated and reverse voltages far in excess of
vce Vs are experienced by the recovering diode.
(b) Im+Irr

Im = iDf + ic
iDf ic iDf ic
Kirchhoff’s
current law
t
off on on off Schottky bipolar
diode diode
(a)
Irr
(c)
Vs = vDf + vce
vce vDf vce vDf
Kirchhoff’s
voltage law

off on off on

(b)
(d)
Vs (Im+Irr)Vs
Figure 7.9. Linear approximations of transistor switching intervals for an inductive load: ImVs
(a) Kirchhoff’s current law Im = iDf + ic; (b) Kirchhoff’s voltage law Vs = vDf + vce; (c) collector voltage PD
and current waveforms with switching parameters defined; and (d) corresponding switching losses. vDf = Im
Vload iDf

Example 7.4: Inductive load switching losses ic t

A power n-channel MOSFET switches a 10A, 100V dc, highly inductive load at 10kHz. vce
ton PD
Calculate the worst case switch losses if the switch turn-on time is ton = 1μs, switch turn-off time is toff = (c)
2μs, and the MOSFET channel on-state resistance is 0.2Ω at 10A.
Calculate the maximum instantaneous power dissipation in the switch, and determine when it occurs.

Solution Figure 7.10. Linear approximations of transistor switching turn-on interval for an inductive load
showing freewheel diode reverse recovery effects on the right: (a) Kirchhoff’s current law
Maximum switch losses occur when the duty cycle approaches one (δ →1) such the both turn-on
Im = iDf + ic; (b) Kirchhoff’s voltage law Vs = vDf + vce; and (c) corresponding switching losses.
and turn-off still occur (that is conduction is not continuous).
The total switch losses PT are made up of three components
PT = on-state loss + loss at switch-on + loss at switch-off
PT    I L2  Rds (on )  ½V s I Lt on f s  ½V s I Lt off f s Example 7.5: Inductive load switching losses with device models
= 1×102×0.2 + ½×100V×10A×1μs  10kHz + ½×100V×10A×2μs  10kHz
A MOSFET 340V dc chopper feeds an inductive dc motor load at 50kHz. In steady state the load
= 20W + 5W + 10W current rises from 10A to 25A when the switch is on with a 75% on-state duty cycle (δ = ¾).
= 25W The MOSFET switch turn-on time is ton = 100ns, switch turn-off time is toff = 200ns, and the
Since the off-state leakage current and gate power losses are not specified, it is assumed these channel on-state resistance is Rds on = 0.025Ω.
are insignificant. The switching loss calculations should use a voltage of 98V, rather than 100V, The freewheel diode is modelled by a 1V on-state voltage and on-state resistance of 0.05 Ω.
since (10A×0.2Ω) 2V is dropped across the channel resistance of the MOSFET. The percentage Neglecting diode recovery and diode turn-on loss, for this continuous load current case, calculate
error is small, and becomes insignificant at higher voltages.
i. the MOSFET total losses
Maximum switch loss occurs when during the switching transitions, the drain current is 10A and ii. diode losses
the drain voltage is 100V. The maximum instantaneous loss is 10A×100V=1000W, (IL ×Vs). iii. power delivered to the motor load, if the armature resistance is 1 Ω and back emf is 170V
♣ iv. electromagnetic energy conversion efficiency and total circuit efficiency
263 Power Electronics Chapter 7 Load, Switch, and Commutation Considerations 264

Solution 7.2 Switch characteristics

i. The MOSFET losses comprise turn-on, turn-off, and conduction losses. Having considered the switching of inductive and resistive loads, the following are the stable electrical
The rms current in the MOSFET is given by and thermal characteristics desirable of commutable switching devices (as well as low cost and weight):
  2    2  off-state (ideally open circuit → Ioff = 0 and Pdis = 0 for -∞ ≤ Voff ≤ ∞):
I M rms  I  I  I  I 
3   Low, temperature independent leakage current in the off-state, to minimise off-state
power loss, and to simplify resistive networks for device series connection.
  25A 2 +25A×10A+10A 2  = 15.6A
0.75 
 High forward and reverse voltage blocking ratings to reduce the need for series device
3 connection, which would otherwise complication control and protection circuitry
The MOSFET conduction losses are therefore requirements. Series connection increases the on-state voltage, hence on-state loss.
Pc  I rms
2
Rds on  324.75  0.025  8.1W When a diode is used in antiparallel across the switch to allow reverse principal
The switching losses are current flow, the switch does not require a significant reverse voltage blocking rating.
  High static off-state avalanche capability to absorb transient overvoltage stresses.
at turn-on Pt -on = ½Vs I t on f s  ½×340V×10A×100ns×50kHz = 8.5W  High static and re-applied dv/dt capability to withstand high applied off-state voltages
 without avalanche or false turn-on, with minimal displacement current.
at turn-off Pt -off  ½Vs I t off f s  ½×340V×25A×200ns×50kHz = 42.5W
on-state (ideally short circuit → Von = 0 and Pdis = 0 for -∞ ≤ Ioff ≤ ∞):
Total MOSFET losses are PMOSFET  Pc  Pt on  Pt off  8.1W+8.5W+42.5W = 59.1W  Low on-state conducting voltage or low on-state resistance, in order to minimise on-
state conduction power loss: with a slight positive (but stable) temperature co-
efficient at high current densities, to allow reliable parallel device connection.
ii. The diode RMS current is
 High on-state current density capability so as to avoid need for and problems
1    2    2  associated with parallel device current sharing and differential thermal coefficients.
I D rms  I  I  I  I 
3    Safe controlled switch off from a short circuit current condition.
switching (ideally instantaneous → ton = 0, toff = 0 with Pdis = 0):
  25A 2 +25A×10A+10A 2  = 9A
0.25
  Low control power to produce switching between states, with no ‘Miller’ interaction.
3
 Short, temperature independent, turn-on and turn-off times to result in low switching
The average diode current is
losses (due to instantaneous simultaneous high voltage and current) which will allow
I d  ½ 1     I  I   ½× 1 - ¾ × 25A+10A   4.375A

high frequency switching.
   High initial di/dt capability at turn-on to allow rapid low loss build-up of the turn-on
The total diode losses are principal current.
Pdiode  I D2 rms RD on  I VD on  High dv/dt capability to avoid false turn-on and the need for snubbers.
 High surge current capability to withstand transient over current fault conditions,
=81.25×0.05Ω + 4.375A×1V = 8.4W
resulting in better fault tolerance and nuisance tripping ride through.
 Large switching safe operating area, being able to simultaneously, but instantaneously,
iii. The power delivered to the load comprises losses in the 1Ω armature resistance and the support rated voltage and rated current, without the need for switch snubber circuits.
power delivered into the 170V dc back emf.
The rms load current is given by thermal/mechanical:
 Easy to electrically connect and mechanically mount, with low thermal resistance and
1  2   2 impedance for efficient heat removal.
I M rms  I  I  I  I 
3   Mechanically, electrically, and thermally robust, with the ability to operate at high (and
low) junction temperatures in high (and low) ambient, pressure, humidity conditions.
  25A 2 +25A×10A+10A 2  = 18A
1 
 Matching substrate structure and thermal properties to minimise stressing due to
3 thermal, mechanical, and power stressing.
The load resistor loss is PRa  I a2rms Ra  325  1  325W  Infinite storage and operational lifetime, at any temperature, atmosphere, and pressure.

The average load current is 7.3 Switching classification

I a  ½  I  I   ½×  25A+10A   17.5A


  There are four principal I-V switching conditions during the commutation (turn-on or turn-off) of a switch,
The power delivered to the back emf is PE a  I a E a  17.5A×170V = 2975W viz.:
 hard switching;
The total power delivered to the dc motor is Pmotor  PRa  PE a  325W+2975W = 3300W  soft switching;
 resonant switching; and
iv. The dc motor efficiency is  naturally-commutated switching.
power output 2975W These four possibilities are classified in terms of the switching time ts and the commutation time tq,
dc    100  90.2%
power input 3300W where tq ≤ ts. Figure 7.11 shows the four electrical cases and specifies the switching and commutation
Including switch and diode losses yields total circuit efficiency, that is times for each.
power output power output
circuit    Switching time ts is the time for a switch to change from fully on (v=0, i = IL) to fully off (v =Vs, i = 0),
dc supply power input chopper circuit losses + dc motor power input
such that no further change occurs in the switch voltage and current due to the change of state.
2975W 
  100  88.3% Commutation time tq is associated with the external circuitry and is defined as the time the switch
59.1W+8.4W  +3300W takes to reach zero current at turn-off or to reach zero volts at turn-on. Alternatively, commutation
♣ time is the period of switch power loss at turn-on or turn-off, due to the switch changing states.
265 Power Electronics Chapter 7 Load, Switch, and Commutation Considerations 266

7.3.2 Soft switching: tq < ts

Vs Vs Figure 7.11b shows typical soft-switching waveforms for turn-on and turn-off. The switching losses are
vc vc complete before the switch has reached its final steady-state condition. That is, ts > tq such that the
IL IL periods ts and tq both commence at the same time.
ic ic At turn-on, the switch voltage reaches zero before the switch current reaches the steady-state full-load
W on W off
value IL. Once the switch voltage reaches zero, the rising current no longer results in a power loss.
This I-V characteristic at turn-on (usually involving inductance in series with the switch) is a form of
t on (a) off t quasi zero current switching, ZCS.
tq Hard Switching
tq The inverse occurs at turn-off. The switch current reaches zero before the switch voltage has settled at
ts tq = t s ts the supply voltage level Vs. (Usually involving capacitance in parallel with the switch.) This is a form of
quasi zero voltage switching, ZVS.
Soft-switching results when auxiliary stress diverting circuits, called snubber circuits, are used, as will be
considered in chapters nine and ten.

Vs vc Vs vc 7.3.3 Resonant switching: tq << ts

Resonant-switching waveforms at turn-on and turn-off are shown in figure 7.11c, with switching periods
IL IL
ic ic ts shown. Resonant switching occurs if the switching period is associated with either the switch voltage
W on W off or current being zero, due to external load circuit conditions which have diverted the load current or
voltage. That is ts > tq.
Switching of the voltage when the switch current is zero, usually at turn-on, is called zero current
t on (b) off t
tq tq resonant switching, ZCRS, while commutating the current while the switch voltage is zero, usually at
Soft Switching turn-off, is called zero voltage resonant switching, ZVRS. Because the exact instant of zero may vary,
ts tq < t s ts being load circuit dependant, some control restriction is inevitable. Zero voltage or current switching can
be readily attained with ac mains converter circuits since switching can be synchronised with supply
zero voltage crossing, or zero current when the load current reverses due to the supply voltage reversal.

Vs vc Vs vc 7.3.4 Naturally-commutated switching: W = 0, tq = 0

Figure 7.11d shows switching when the voltage and current are both zero, called naturally-commutated
IL IL
ic ic switching. This was a commonly used technique for force turn-off of thyristors before the emergence of
W on W off the GTO thyristor. Current from an auxiliary commutation circuit displaces (exceeds) the device principal
current and reverse biases the device, at turn-off. The method was not used at turn-on. Commutated
t t
turn-on and turn-off occurs in inverter circuits where the switch has an anti-parallel connected diode.
on (c) off When the diode conducts and the switch is on but not conducting, if the load power factor causes the
tq tq
Resonant Switching current to reverse, then the main switch automatically starts conducting with the switch voltage at zero
ts tq < t s ts because the diode was previously conducting, clamping the switch voltage slightly negative.
Naturally-commutated switching occurs for ac mains zero crossing switching, with a purely resistive load
such that the load V and I are in phase. Switching losses are virtually zero.

It should be noted from figure 7.11 that at turn-on the current change is modified, while at switch turn-off
Vs vc Vs vc the voltage is modified. Specifically, at switch turn-on, the switch voltage must fall from a supporting
IL W on = 0 IL W off =0 voltage level, to zero: only its current increase can be modified, Conversely at switch turn-off, the current
ic ic is conducting current which must reduce to zero: only the voltage increase across the switch can be
tq tq modified as the current reduces to zero.
t t

on (d) off 7.4 Switch configurations

Naturally-commutated Switching
tq = 0 Switch configurations can comprise any combination of one or more of three different switch types:
ts ts Uncontrolled switches (diodes);
Semi-controlled switches (phase controlling thyristors); and
Figure 7.11. Switch voltage (vc), current (ic), and power loss (Won and Woff) of four switching Controlled switches (MOSFET, IGBT, BJT, GTO, IGCT).
classifications: (a) hard switching; (b) soft switching; (c) resonant switching; and
(d) naturally-commutated switching.
Most semiconductor switches are unipolar, that is, allow current and/or voltage to be supported in one
direction. The MOSFET allows uncontrolled reverse current flow; hence cannot support reverse voltage
Generally, the switch loss magnitude (stress) for a given set of electrical and thermal operating because of its parasitic body diode. Some structures, like the RCT considered in chapter 3.3.3, integrate
conditions, decreases when progressing from severe hard switching through to virtually lossless an anti-parallel diode with a thyristor. Generally, such integrated approaches sacrifice some electrical
naturally-commutated switching. characteristics. Many applications require a bi-directional current and/or bi-directional supporting voltage
switches, so the basic switches can be configured as shown in figure 7.12, to give the necessary I-V
7.3.1 Hard switching: tq = ts characteristics. The net effect of the bi-directional voltage arrangements is good dynamic electrical
The turn-on and turn-off switching waveforms in figure 7.11a for an inductive load show that hard characteristics but poor static characteristics. Specifically, the switching performance is as for the
switching is characterised by tq = ts. The resistive and inductive switching considered in sections 7.1.1 principal switch but the on-state loss is that of two series connected devices. In the case of the
and 7.1.2 are examples of hard switching. In figure 7.4 for a resistive load, the switching periods ton and bidirectional blocking thyristor, the on-state voltage is increased slightly because an n-buffer can not be
toff (ts) correspond to the period of switch losses (tq) during each state transition. In figure 7.9 for the used in its fabrication. But the symmetrical GCT offers lower on-state voltage and significantly higher
inductive load, the tq periods correspond to the power loss periods at switching (trv + tfi and tfv + tri). current ratings than an equivalent a high voltage IGBT with a series connected diode. The bi-directional
267 Power Electronics Chapter 7 Load, Switch, and Commutation Considerations 268

conducting thyristor discussed in chapter 3.3.4 attempts to minimise the sacrificed on-state voltage 7.5 Power converter configuration classification
limitation. A reverse blocking IGBT can also be realised. Die edge passivation of the diode region by a
through the die p+ diffusion, plus guard rings, increase processing complexity, and hamper voltage
ratings. A punch through IGBT version with reverse voltage blocking properties, is therefore problematic. A power electronic converter processes the available energy form to another form having a different
On-state voltages are increased for a given switching speed and, as with the MOSFET body diode, the frequency and/or voltage or current magnitude. There are four basic types of electrical energy
non-optimal diode recovery characteristics are a compromise because of the overriding n-substrate low converters depending upon the conversion function performed:
resistivity requirements. See chapter 3.2.4.
 Controllable switching devices with reverse blocking capability are usually required for ac to ac
converters, half-wave resonant converters, and current fed inverters. Table 7.1 Four basic conversion processes
 Voltage source inverters, full-wave resonant converters, and dc to dc converters usually do not
require switching devices with reverse blocking properties, but may use an antiparallel Conversion
Topology Function Symbol
connected diode. from-to

chopper constant or variable dc

I Principal Current dc to dc to
smps variable or constant dc
V unidirectional bidirectional

dc source
RCT
to
dc to ac inverter
ac of desired voltage and
frequency
unidirectional

rectifier
ASCR
ac to dc ac to unipolar dc current
converter

IGBT ASCR
cycloconverter ac of desired frequency
ac to ac and/or magnitude from an ac
Voltage

matrix converter source

triac
bidirectional

reversible

adjustable magnitude and/or frequency

Constant magnitude and frequency
SSCR

on
¥

si
er

on
nv

si
co
SSCR RB-IGBT

er
SSCR

OUTPUT
in
INPUT

co
nv
er
in

si
v

on
er
si
¥ Can be arranged so that emitters are at the same potential. Switches may be reverse blocking IGBTs.

on
Figure 7.12. Switch configurations for uni-directional and bi-directional I-V characteristics.
269 Power Electronics Chapter 7 Load, Switch, and Commutation Considerations 270

Reading list blank

Peter, J. M., The Power Transistor in its Environment,

Thomson-CSF, Sescosem, 1978.

Problems

7.1. During turn-on and turn-off of a power transistor the current-voltage relationships are as shown
in figure 7.13. Calculate the energy loss during both turn-on and turn-off periods and the mean
power loss if the transistor is being switched at a frequency of 10 kHz. What is the maximum
instantaneous power dissipated?
[1.66 mJ, 16.6 mJ, 183 W, 5kW]

Figure 7.13.

7.2. The equivalent circuit in figure 2.4a involving parameters Eo and Ro can be extended to model a
thyristor by replacing the ideal diode by an ideal thyristor. Derive general expressions for the
thyristor mean power loss Pd and rms current io with a constant load current Io and switch on-
time duty cycle δ.
If Eo = 1 V and Ro = 0.01 Ohms, for Io = 50 A and a 25 per cent on-time duty cycle, calculate the
thyristor:
i. On-state voltage, VF
ii. Mean power, Pd
iii. rms current, io.
[See example 2.1: 1.5 V, 18.75 W, 25 A]

7.3. If the collector voltage at turn-on falls according to

v c  ½Vs (1  cos  t / T o ) for 0  t  T o
i. For a resistive load, RL, calculate transistor loss at turn-off.
ii. Show that the switching trajectory across the SOA is as for the linear current fall case.
iii. Calculate the peak power dissipation and when it occurs.

7.4. A transistor is used to switch an inductive load with a current of Im.

At transistor turn-off, the collector voltage rises to the supply rail Vs according to
vce = ½Vs (1 - cos π t /Tov) for t ≤ Tov,
then the collector current falls according to
ic = ½Im(1 + cos π t /Toi) for t ≤ Toi.

Using the same integration form as in equation (7.10), show that the turn-off loss is
P = ½VsIm To where To = Tov + Toi.