1

Lecture 15
The Forward PWM Converter Circuit Topology
and Illustrative Examples

I. Erickson Problem 6.4 A DCM Two

Transistor Flyback Converter

II. Forward Converter

A. Overview
B. Forward Converter with a Three Winding
Transformer Drive: Two Case Studies
1. One Transistor Implementation
2. Two Transistor Implementation
C. Forward Converter Transformer

D. External Reset of a Transformer:

Erickson Problem 6.9
 2
Forward PWM Converter Circuit Topologies

I. EricksonProblem 6.4 A DCM Two

Transistor Flyback Converter
b) Erickson Problem 6.10: One transistor Flyback implementation
in the DCM mode of operation.

We will consider only DCM operation below:

np:n1
ip . i1
+15V
Vg . 1A

165 Vdc
:n2
Q1 i2
-15V
. 0.5A

:n3
i3
+5V
. 4A

Note the ability to provide ±15 V as well as +5 easily. If we

employed feedback for Vo stabilization we could use just one of
the three Vo to monitor and slave the other two. WHAT OTHER
WAY IS POSSIBLE??

Neglect:
1. All losses in the electronic circuits.
2. No magnetic core losses.

Also for this problem we arbitrarily set the specification Vg is fixed

 3
and iload is fixed for the DCM operation.

We expect given 3 DCM time intervals: D1, D2, D3 but we

arbitrarily fix D3 = 0.1 ⇒ D2 = 0.9 - D1.

Goal: Find steady state design, where the V(off) blocking voltages
are minimum for both the transistors (< 300 V) and diode (< 30 V)
solid state switches.
Below we give circuit conditions for each time interval.
Interval D1Ts: Control signal puts Q1 on and using current flow
dot’s on the transformer we see that all three diodes in the
secondary are off.
ig np:ns
i . +

Lm vp C R Vo
Vg .
-

↓ ↓
VLm = Vg ic + iR = 0
V
i = ig ic = - o
R

Interval D2Ts: Control signal puts Q1 off, all secondary diodes

conduct using current flow dots on the transformer.
ig
+
i . i/n +

vLm Lm vp C R Vo
Vg - .
-
1:n
V(off Tr)

ns/np

↓ ↓
 4
i
ig = 0 ic = - Vo
n R
V o = V(primary)
VL =
n
Period D3Ts: Q1 off, all secondary diodes off V(off diode)
1:n v(off diode)

+ . +

vLm Lm vp C R Vo
Vg - .
-
V(off Tr)

↓ ↓
VLm = 0 ir + ic = 0
V
ig = 0 ic = - o
R

Draw ic waveform versus time using the above circuits to

determine values of ic for each interval.
Vg
ic
D1 D2 0.1
t

-Vo/n=-V(primary)

Remember - V o = - V(primary) for all ac conditions

n
encountered.

Applying volt-sec balance to the Lm inductor:

< VLm > Ts = 0 D1V g + D2 (- Vp ) + 0 = 0

↓
 5
D1 V g = (0.9 - D1) Vp
D1( V g + Vp) = .9 Vp
.9 Vp
D1 ≤ , setting an upper bound on D1.
V g + Vp

For cost considerations we limit maximum voltages to the switches

as follows:

Choose: Transistor: Voff < 300 V

Diode: Voff < 30 V

300 - Vp+ Vg from primary circuit and transistor maximum standoff

voltage spec.

Vp = 300 - Vg(165) = 135

0.9(135)
D1 < = 0.405
300

Pick D1 = 0.4 ⇒ D2 = 0.5 Try this 1st choice and see the effect.

Diode currents for all three secondaries which are similar in time.

Three diode currents each have a unique charge passed through

them during D2Ts.

iD charge
1/2 ipeakD2Ts

D1 D2 0.1

Current flow in LM is on for 0.9 Ts, given D3 fixed at 0.1.

 6

iLm

0.9
t
diodes 0.1
on

Notice the fact that for the buck-boost and flyback:

iLm = i(diode) + i(transistor)

The corresponding current in the transistor and the primary

winding is:

ip=iTR

transistor t
on for D1Ts
Remember iD is secondary current, whereas iLm is the primary
current.

Values of Diode Peak Currents: (Given chosen maximum off

voltages)
2 iD1 (DC) (2)(1)
iD1(peak) = = = 4A
D2 0.5
1
2 iD2 (DC) (2)( )
iD2 (peak) = = 2 = 2A
D2 0 .5
2 iD3 (peak) (2)(4)
iD3 (peak) = = = 16 A
D2 0.5
We must check that each diode can carry these current levels
from the diode manufacturer spec. sheets.

Peak ILm is worth calculating because hysteresis core loss varies

 7
with peak current, not rms current.
iLm(peak) = Σ iDx*1/nx
↓ ↓
primary secondary
current diode currents

15 V 15 5
n1 = = , n2 = = V 2 , n3 =
135 Vp 135 Vp 135
4 2 16
iLm (peak) = + + = 1.26 A
9 9 27

We must check that this peak current does not saturate the
magnetic core.

Transistor peak current

iTR(peak) = iLm(peak) = 1.26A
di
VLm = Lm Lm
dt
165 .5 10-5
↓ ↓ ↓
VLm Vg
iLm (peak) = D1T s = D1Ts
Lm Lm

iLm depends on Lm values. The Lm value of the chosen transformer

is set by the peak transistor current.

Vg (165)(.4)
Lm = D2 Ts = 10 -5
ipeak 1.26
= 0.52 mH

Calculate IRMS using the above current waveforms and Appendix 1

on pg. 705 of Erickson’s text.
 8

[Ix (output) ]RMS = ∑ iDx (peak) * D2

3
x
x = 1-3 for the three secondary windings.
I1(rms) = 4 * .41 = 1.63
I2(rms) = 2 * .41 = 0.81
I3(rms) = 16 * 41 = 6.53
D1 + D2 = 0.69 A
ILm(rms) = iLm (peak)
3

II. Forward Converters

A. Overview
Here we choose in the circuit topology the dots on the transformer
coils and the primary / secondary diode placement so that when
primary current flows so will secondary current unlike the flyback
converter. Hence, the name forward converter. Three secondary
arrangements for the forward converter are given below:
•Simple center tapped transformers are the key elements to
one approach
•Full wave rectification without isolation of the secondary
voltages and without center taps is a second approach to the
forward converter topology
•Fully isolated secondaries without center taps with full wave
rectification is the third topology approach
Note that in all cases the transformer has the role of dielectric
isolation, which is accomplished by the choice of isolation material
between wire windings on the transformer. The winding turns ratio
provides the step up or step down ratio desired. There is no air
gap in the core so the forward transformer stores no energy, it
merely transfers energy form the primary coil to the secondary coil.
Of, course we must never allow the peak flux in the transformer
core to EXCEED THE SATURATION VALUE of the chosen core.
Forward converters posses both a transformer and an output
choke and this distinguishes them from flybacks of lecture 14.
 9

On the next page we will outline the voltage and current

waveforms in a simple half-wave forward converter to give a clear
picture of the unipolar drive that occurs in each portion of the
transformer secondary that is synchronous with the primary drive
sequence. In full wave center tapped operation each half of the
transformer sees similar waveforms. The full wave rectification in
the secondary insures that the current waveforms are unipolar as
does half-wave rectification.
 10

Notice that the output circuit of an L-C filter directly after the power
switch or output rectifier is the CALLING CARD of the forward
converter. This is clearly a BUCK-LIKE TOPOLOGY with Vout
being proportional to D x Vin.

B. DCM Forward Converter Operation: Two Case Studies

On the following pages we will outline the operation of

froward converters operating in the DCM of operation via two
illustrative examples. One will employ a single transistor
switch while the second will employ two temporally
synchronized switches. DCM operation is more complex
than CCM operation due to the third time interval that is
introduced into the switching period, Tsw.
 11
1. DCM operation using only one transistor Iout is non-
pulsating for a fixed load and the L-R-C filtering.

For the Forward Converter notice that: buck- like operation occurs
in the secondary circuit to the right of diodes “D2”and “D3”.

To start with, for the dotted transformer windings always assume

all currents flow into the dots and n1i1+ n2i2 + n3i3 = 0.
D2 L
n 1 : n2 : n3
. . +

D3 C R V
Vg .
-

Q1 D1

If the input, or primary, current flow direction is known by circuit

means to flow the opposite to that initially assumed then this
reverse flow occurs also at other coils. The dotted side specifies
which way current flows.

DCM operation has three intervals in Ts of duration D1, D2 & D3.

iLm is reset to zero so that saturation of the magnetic core does not
occur as follows, in DCM operation.

Interval D1Ts: Q1 and diode “D2”conduct creating simultaneous

primary and secondary transformer currents while diodes “D1”and
“D3”are off again by current flows.

All currents are assumed into the dots of all windings.

a) When Q1 is on +Vg appears across n1
b) A voltage of polarity -Vg n2/n1 appears across turn n2 ⇒ D1
is off. This is also seen by current flow. i1 flows into dot of
 12
coil #1 ⇒ i3 flows out of the dot of coil #3 and D2 is on.
c) +Vg n3/n1 appears across turn n3 ⇒ diode D2 is on drawing
a secondary current which appears in the primary as i1’.
Then the current drawn from Vg is i1 = im + i1'.

Note that the primary is composed of the magnetizing inductance

in parallel with the n1 winding as per the standard transformer
model. Coil #2 is open because of diode D1 being off.
L
n 1 : n2 : n3 D2 on
im .
+
i1' . +
+
- +
Lm v1 v2 v3 vD3 C R V
Vg - .+ -
- -
i1 i2 i3

Q1 on D1 off

Interval D2Ts: Q1 is put off actively by the control signal forcing a

current loop in Lm and the n1 coil. This causes current to flow out
of the n1 coil at the dotted end, and current flows into the dotted
end of the n2 coil.

n 1 : n2 : n3 L
im . i1'
-
. +
+
+ +
Lm v1 v2 v3 D3 on v C R V
Vg - .+ -
D3

- -
i1 i2 i3

Q1 off D1 on

im now flows in the n1 winding as shown exiting the dot of the n1

coil. The n3 coil has current flow into the dotted end.
 13

. n1 .
Lm im im

In total then the conservation of mmf gives:

-n1i1 + n2i2 - n3i3 = 0
+ sign is into dot
- sign is out of dot

i2 and i3 coil current directions are as assumed but i1 is known to

flow out of the dot of the n1 coil due to Q1 being off during the i1
time interval D2Ts. Due to the current flow direction we can say
that during D2Ts:

•Q1 is off and diodes D1 and D3 are on whereas diode D2 is off as

shown above.
•Diode D1 is on because of current flow into the dot of winding n2
which turns D1 on.
•Diode D2 is off because of current flow into the dot of winding n3
which turns diode D2 off.
•Diode D3 is on because of current flow into the dot of winding n3
equals i3 out of undotted n3 turning D3 on.

With diode D1 on we have:

INPUT OUTPUT
+Vg across n2 coil so Current i3 as shown flows
by the dot convention from the n3 coil to the output
n1
VLm = - V g
n2
and this negative voltage across Lm causes the magnetizing
n1 V g
current to decrease with a slope di/dt:
n2 L m
 14
Note also that in the primary we have the sum of two voltages
across the transistor. [ V Tr (off) ]max = V g [1 + n1 ]
n2

Interval D3Ts: iLm will try to reverse sign after the time period D1Ts
going up and the time period D2Ts going down as shown below in
the iLm plot versus time.
Vg n1 / n2 (downslope)
(upslope) Vg
Lm Lm
Vg(n1/n2)/Lm
Vg/Lm

iLm

First iLm hits zero then iLm tries to go negative. At this point a new
circuit topology arises as diode D1 goes off and diode D3 goes on.
n 1 : n2 : n3 L
im
=0
. i1'
-
. +
+
+ +
Lm v1 v2 v3 D3 on v C R V
Vg - .+ -
D3

- -
i1 i2 i3

Q1 off D1 off

If iLm now goes negative with Q1 off then the dot convention tells
that the iLm current loop now enters the n1 winding from the dotted
side in the current loop seen below.

n1 . This current flow direction for i1

causes i2 to also try to flow into
the dotted side of coil #2 as
Q1 off shown below:
 15

.n But i2 tries to flow out but diode

2
D1 does not allow this direction
of current flow. So diode D1
goes off.
D1

Likewise iLm flow into n1 dot ⇒ current flow into the n3 dot and D3
is turned on when iLm hits zero.
Since both transistor Q1 and diode D1 are off iLm remains zero for
the whole period D3Ts.
A tradeoff must be made in the forward converter since:
n2 n2 ↓ allows the interval D ↑
D2 = D1 moving the turns ratio 1
n1 n1
for fixed interval D2
BUT n1 ↑ implies that the standoff voltage across the transistor
n2
increases since:
n1 ] So we trade off decreased Tr on time
V Tr (off) = V g [1 +
n2
D1 for increased voltage stress.
This transistor switch voltage stress may be too much for one
transistor to work in its safe operating area (SOA). Below we
show a way to solve this by utilizing two rather than one switch
and dividing the switch stress between them.
 16
2. Two transistor implementation of the Forward Converter
operating in DCM operation.
What do you guess the phasing of Q1 and Q2 gate control is?

Q1
D1 D3 L

. 1:n . +

D4 C R V
Vg
-

Q2
D2

The tandem transistors Q1 and Q2 have the same phase gate

control. So they are both in the same state.

Here our Lm is considered large as the core has no slotted

opening and the core magnetic reluctance is low. Although iLm is
small we still we need to consider Lm effects.
Now the primary voltage when Q1 and Q2 are both off can be
dropped across the two transistors as both are in series, reducing
switch stress to more allowable levels.

Interval D1Ts: Control signal forces both Q1 and Q2 on, primary

diodes D1 and D2 are both off. Current flow is such that i1(primary)
flows into the dot end of the one turn coil so i2 flow is out of the
secondary winding of n turns. The secondary current is 1/n of the
primary current.
 17
in
i1
. 1:n .

when i1 flows into the dot on winding 1 a current i1/n flows out of
the dot on the n turn winding as shown above. This insures that
diode D4 is off and diode D3 is on.
L
1:n
. . +

Vg ig +nVg C R

im

Vg/Lm

The current in Lm increases linearly during switch interval D1 with a

slope Vg/Lm.

Interval D2Ts: The external control signal forces both Q1 and Q2

off. Both primary diodes D1 and D2 are on as shown below during
the interval D2Ts driving the magnetizing current down during time
D2Ts.
gnd

D1 L
1:n D3 off

-
. .- +

iLm Lm nVg C R Vo
+ + D4 on
-

D2

+Vg
 18

Note that iLm flow is always positive with respect to zero during
intervals D1 and D2. It first rises during D1Ts and then falls during
D2Ts as shown below.
What occurs if iLm does not reach zero?
for core
iLm reset
D1<1/2
Vg/Lm -Vg/Lm
D1 D2
t
Ts
im
-Vg/Lm

The iLm current period as shown in DCM of operation could lead

over a few switch cycles to core saturation if iLm did not return to
zero. DCM operation, however, guarantees that this occurs. If iLm
tries to go negative, then diodes “D1”and “D2”go off and prevent it
from doing so.

Setting/Fixing iLm = 0 so the core doesn’t saturate

We will find below in part 3 for iLm to reset the transistor on time D1
< ½ is required. Is this clearly why?

Period D3Ts: Q1, Q2, as well as diodes D1 and D2 are all off in the
primary circuit
iLm = 0, Vprimary = 0
C. Forward Transformer Overview
As stated earlier we must never allow the flux density, B, in the
forward converter core to exceed the saturation flux for that core.
If we do, the transformer will look like a short circuit and no doubt
fry the power switches. We will employ Faraday’s law to see the
trends between transformer core size and the choice for the
 19
number of turns in the windings.
V(out) = N(# turns of wire) x ω(radian frequency)x φ(flux)

Where φ(flux) given by is given by B(flux density)x Acore(core area)

Now each core material has a maximum allowable flux density that
it can handle before the onset of core saturation. Hence, we can
say that there is a required number of turns on the transformer
windings. Consider first the primary windings. We will find below
the result that N varies inversely with Bmax(saturation value).

Nprimary = Vprimary /(ω x Bmax x Acore)

That is the MINIMUM number of primary winding turns varies as:
•the input voltage level Vprimary
•Inversely with operating frequency
•Inversely with the saturation flux of the core
•Inversely with the core size
Clearly, there is lots of design tradeoffs to be considered.
Regardless, once the number of primary winding turns is set this
acts as the reference for all other secondary winding turn choices
as described below.
The secondary turns are unique for each secondary voltage
desired. We start with the secondary that requires the most output
power. The voltage across the output rectifers should also be
accounted for. The starting point for all such calculations is the
fact that the voltage in a specific winding divided by the number of
turns in that winding must be equal to the output voltage of
another winding divided by its number of turns. Finally, we have to
consider worst case when the input voltage is minimum, Vin(min),
and the duty cycle is maximum, Dmax.. That is we find the
relationship:
Nsecondary = Nprimary x( Vout +Vdiode)
Vin(max) x Dmax
We solve for Nsec at lowest Vin and highest D we expect to occur.
 20
The result is always a non-integer number and one must round off
to the nearest number of turns. This could result in the output
voltage being not what we desire. Again an iterative process is
required to meet all desires. Note that adding secondary turns on
the primary winding will always move you in the safe direction of a
lower flux density, below the dangerous B(saturation)

D. DCM Forward Converter with External

Reset to actively avoid core saturation.
Next we explain how to actively achieve Lm core reset by use of an
external voltage source, with the secondary goal of causing less
standoff voltage stress for both Q1 and diode D2. Also the duration
D1Ts can be increased so for a fixed D2Ts:
n3 n3
D 2 = D1 ⇒ can be reduced if D1Ts increases. Why is this good?
n1 n1
↓ ↓ ↑
fixed
If n3 ↓ ⇒ Lower peak transistor current occurs
n1

Erickson Problem 6.9: External Vr resets the core with less stress
than reset with - Vg as shown below:

Note that the auxiliary secondary winding n2 is not an output

circuit, rather it is a reset voltage entry point.
The full schematic is shown on the following page.
 21

n1:n3 D2
. . +

D3 R v
Vg
-

Q1
D1
reset Vr
.winding
:n2

In DCM operation there will be 3 periods as shown below:

D1Ts D2Ts D3Ts

Q1 on Q1 off Q1 off
Diode D2 on Diode D2 off Diode D2 off
Diode D3 off Diode D3 off Diode D3 off
Diode D1 off Diode D1 on Diode D1 off

Interval D1Ts: Diode D2 on and diode D3 off in the winding n3.

n1:n3 L v/R

+
i . . ic +

vL Vgn3/n1 C R V
Vg -
-

VR
.
:n2

During interval D1Ts an effective DC voltage of magnitude

D(n3/n1)Vg appears across winding n3. As diode D1 is off, VR does
not affect the core flux levels.
 22
Interval D2Ts: Diode D1 in winding #3 is on activating the reset
current on the magnetic core. The current iLm flows out of the n1
winding dot implying current flows into both n2 and n3 windings.
Hence, diode D2 is off and diode D1 is on.

L v/R
. . iC +

iLm n1 n3 C R V
Vg
-

iR
n2 VR
.
Reset of core via integrating VR for the interval D2Ts.

Interval D3Ts: All circuits are open and no volt-sec drive to the
magnetic core flux occurs.

L v/R
. . iC +

C R V
Vg
-

VR
.
Below we calculate the applied VR to the windings required for
duration D2Ts to cancel Vg applied for duration D1Ts to winding #1.

Magnetizing Inductor: LM

Volt-sec balance to LM has only two components.

 23

< VLm > Ts = 0 V g D1 Ts + D2 [- Vr N1 ] Ts = 0

N2
Vg n2
< VLm > Ts = 0 sets the relation D2 = D1
Vr n1
Setting D1 and D2 still must leave a non-zero interval D3Ts.
D3 = (1-D1-D2) > 0
1- D1 n1 1
⇒ >
V g D1 n2 VR
V g D1 n2
or VR > Sets minimum value of VR to be applied
(1- D1) n1
during interval D2Ts for Lm core reset

Consider Switch Stress for Interval D2Ts.

n1
V Tr (off) = V g + VR → substitute for via
n2
↑ ↑
+Vg from appears on n1 coil due
source to VR on n2 coil
D1 ]
V Tr (off) = V g [1 +
1- D1
V Tr (off) = V g / (1- D1)

Output Inductor of “Buck”: L

<D3> = Vo ⇒ <VL>Ts = 0
↑ voltage equal either side
n n V0
D1 3 V g = V o ⇒ D1 = 1 * sets D1Ts time so L has
n1 n3 Vg
<VL>Ts=0 in steady state.
 24

The above summarizes a comparison between flyback and forward

converter topologies as regards the switch stress and a crude
estimate of the required SOA for the switches.

Finally, For HW#3 Due in 1 week:

1. Answer any Questions asked throughout lectures 12-15.
2. Chapter 6 Problems 7, 8, and 11(the forward converter).