Flyback transformer design

Flyback transformer design

considerations for efficiency

and EMI
Bernard Keogh
System and Solutions Engineer
High Voltage Power Systems
Isaac Cohen
Principal System Architect
Low Power Controller and Converter
Texas Instruments
Power Supply Design Seminar 2016/17
AC/DC power supplies widely use the flyback

converter given its simplicity and wide operating
range, and because it eliminates the output filter
inductor and free-wheeling rectifier required for
forward-mode topologies.
Three main topology components dominate flyback-converter performance: the
primary switch, secondary rectifier and transformer. This paper focuses on the
importance of transformer design, since this single component has a profound impact
on converter efficiency and electromagnetic interference (EMI). This paper will discuss
the various conflicting design requirements, the often-neglected subtleties of core loss
and snubber clamp level, and ways to improve transformer performance.
Introduction often-neglected absorption of magnetizing energy

Many AC/DC and DC/DC power supplies, from very by the snubber is also highlighted.
low power levels to as much as 150 W or more, For conducted EMI, we will outline the causes
use the flyback converter. Often maligned and not of common-mode (CM) EMI, suggesting various
always fully understood, the transformer is the winding structures and techniques to ensure good
heart of the flyback power supply and probably the CM balance.
most important component. When designed and Finally, through several examples we will show how
implemented well, the transformer can deliver the transformer construction can have a significant
required performance cost-effectively. When poorly impact on both efficiency and conducted EMI. In
designed, it can cause EMI issues, low efficiency these examples, we changed none of the other
and possible thermal overstress issues. components – only the transformer – in order to
This paper will discuss the causes of the major demonstrate how a well-designed transformer can
losses in the flyback transformer. In particular, we simultaneously improve both efficiency and EMI.
will review core loss in light of recent research
findings that highlight the significant impact of duty The flyback topology
cycle and DC bias. The significant contribution The flyback transformer is not really a transformer
of proximity effect on AC copper loss is also in the conventional sense; it is actually a coupled
discussed. inductor. Figure 1 is a simplified schematic of a
flyback converter. The flyback transformer in this
We will review wire-size selection and winding
example has three windings: primary, secondary
methods to reduce AC copper loss. The effect of
and bias (sometimes called the auxiliary winding).
snubber clamp voltage levels and the
Texas Instruments 2 September 2016

When the primary switch turns on, the input voltage required by the pulse-width-modulation (PWM)
is imposed across the primary winding. Since the controller for regulation, the primary switch is
dot-end of the primary winding is connected to turned off. The primary current then transfers to the
ground, the dot-end of both the secondary and secondary winding and the current decays at a rate
auxiliary windings will be negative and proportional proportional to VOUT. In this way, the energy stored in
to the input voltage. The respective rectifier diodes the transformer during the buildup of primary current
on those windings will thus be reverse-biased. gets released to the load and output capacitor
While the primary switch remains on, current builds during the flow of secondary current. This is, of
up in the primary winding at a rate dependent on course, a simplified explanation; for more detailed
the input voltage and the primary magnetizing descriptions of the flyback topology and modes of
inductance, LP. operation, see references [1], [2] and [3].

Based on this description, the flyback transformer
BIAS actually operates as a coupled inductor, where
SEC VOUT current builds up to a peak value in the primary
PRIM winding and then decays back down in the
+ secondary winding during the flyback interval.
VIN
Thus, when designing the flyback transformer and
assessing the losses, you must consider it more of
an inductor than a transformer.
DRV
PWM CS
VDD Controller Flyback operation
FB Figure 2 shows the different operating phases
of the flyback converter during a single switching
Figure 1. Simplified schematic for a typical flyback converter. cycle, with the corresponding voltages and currents
shown in Figure 3. During the primary switch
Once the current in the primary reaches the level on-time interval in Figure 2a, current flows from
PRIM SEC VOUT PRIM SEC VOUT

+ +
VIN VIN
ON OFF
(a) (c)
PRIM SEC VOUT

PRIM SEC VOUT
+
+ VIN
VIN
OFF
OFF
(b) (d)
Figure 2. Flyback converter operating intervals per switching cycle: primary switch on-time (a); primary switch turn-off, transition interval (b);
secondary rectifier clamping and conduction interval (flyback interval) (c); discontinuous conduction mode (DCM) ringing interval (d).

(a) (c)
(b) (d)
Figure 3. “Idealized” flyback converter voltages and currents, with highlighted operating intervals per switching cycle: primary switch on-time (a);
primary switch turn-off, transition interval (b); secondary-rectifier clamping and conduction interval (flyback interval) (c); DCM ringing interval (d).
the input-voltage source through the transformer’s capacitance on the switch node. The losses in the
magnetizing inductance, storing energy in the transformer core and the AC resistance (ACR) of the
inductor air gap. During the transition interval in windings dampen this ringing.
Figure 2b, the primary current transitions to the In continuous conduction mode (CCM), the interval
secondary, while the transformer’s primary voltage in Figure 2d does not occur because the primary
swings positive. When the transformer primary on-time commences before the secondary current
voltage swings sufficiently more positive than VIN, decays to zero. In CCM, not all of the energy
the output flyback diode becomes forward-biased stored in the transformer’s magnetizing inductance
and clamps the voltage. Subsequently, during transfers to the secondary during each switching
the interval in Figure 2c, the secondary current cycle.
will decay linearly (since the voltage across the
secondary winding is negative). During the interval Flyback transformer losses
in Figure 2c, some or all of the energy previously The flyback transformer is responsible for a large
stored in the transformer’s magnetizing inductance percentage of the total losses in a flyback power
will be released to the secondary-side storage stage. There are four categories of losses:
capacitor and to the load. • Core losses.
In discontinuous conduction mode (DCM), all of the • Copper (winding) losses.
energy stored in the inductance during the primary • Transition losses.
on-time interval is delivered to the secondary during
• External losses.
the flyback interval. In this mode, the secondary
Core losses occur in the transformer’s ferrite core
current decays to zero at the end of the flyback
and depend on the core’s flux density (amplitude,
interval. Subsequently, the interval in Figure 2d is
duty cycle and flux-density rate of change),
the DCM ringing interval, where the magnetizing
frequency of operation, core size or volume, and
inductance resonates with the total parasitic

properties of the chosen ferrite material. Different are significant causes of loss. As transition loss is
materials optimized for different frequency and peak beyond the scope of this topic, see reference [4] for
flux-density ranges will exhibit varying core-loss further details.
characteristics. We will describe core losses in more While the transformer itself incurs most of the
detail in the next section. losses, two significant external losses occur due to
The flow of current through the resistance of the parasitic elements of the transformer. First, leakage
windings causes copper or winding losses. Most inductance results in a loss incurred in the external
designers refer to it as copper loss because copper clamp or snubber circuit, which is necessary to
is by far the most commonly used wire material keep the voltage stress on the primary switch
given its low resistance, ease of manufacture and below its VDS maximum rating. Second, transformer
wide availability. capacitance contributes to the total parasitic
Copper loss breaks down further into DC loss and capacitance of the switch node. An increase in the
AC loss. DC loss is caused by DC or low-frequency switching node capacitance increases the switching
root-mean-square (rms) current flowing through the losses in the primary switch. We discuss the effects
DC resistance (DCR) of the winding. Maximizing the of leakage inductance and interwinding capacitance
wire cross-sectional area and minimizing the wire further in the section on EMI shielding.
length minimizes DCR.

Core loss in a flyback converter
AC loss is caused by high-frequency
Traditionally, designers assumed that DC flux did
electromagnetic effects from the magnetic field not affect the core losses in an inductor, and these
produced by the time-variant current flowing in the losses are largely independent of the flux-density
wires. AC loss can be very significant, especially for waveform. For example, as Figure 4 shows, the
large wire diameters. We will discuss AC losses in flyback flux-density waveform is nonsinusoidal, not
more detail later. necessarily 50 percent duty cycle and contains a
Transition losses refer to the losses associated significant DC component. Yet when calculating
with the transition or commutation of transformer core loss, most designers neglect the DC
current from the primary to secondary winding. component and duty cycle and consider only the
In this region, the rate of change of the currents peak-to-peak flux swing, as shown in Figure 5.
(di/dt) is very high, so the currents will have large Another common assumption is that all waveforms,
high-frequency harmonic content. Also in this regardless of duty cycle and DC bias, have the
region, since both primary and secondary currents same core loss because they have the same Bpk-pk
flow simultaneously, the flyback transformer flux-density swing. Thus, designers extracted the
behaves more like a conventional high-frequency core losses from the published sine-wave-specific
transformer, and so high-frequency effects and ACR loss curves using the flux-density amplitude and
frequency experienced by the converter.

Flyback Waveforms at CCM/DCM Boundary offer equations that relate manufacturer-published

specific loss data for sine-wave excitation only (no
VIN ΔBac Bpk-pk DC bias) to actual losses generated with rectangular
BDC waveforms and DC bias, and provide empirical
VREFLECTED support for their theories.
TSW = 1/FSW
Effect of rectangular waveforms with
variable duty cycle
Figure 4. Flyback transformer flux-density waveform at CCM/DCM
boundary. Reference [6] investigates the ratio of core loss
Flyback Waveforms, Neglect BDC and D under rectangular-wave excitation to that of a
sinusoidal-wave excitation of equal flux amplitude
for a number of magnetic materials (Figure 6
ΔBac Bpk-pk reproduced from [6]). It also introduces a curve-
fit equation for the core loss versus duty cycle
TSW = 1/FSW
(Equation 1):
�� (1)
Figure 5. Flyback transformer waveforms at CCM/DCM boundary,
� ��
�� ∙��∙��
neglecting the BDC component and duty-cycle variation.
where D is the duty cycle and γ is a correction factor
A closer qualitative examination reveals that these specific to the material, operating frequency and
assumptions must be incorrect. It should be
temperature, and has to be extracted from careful
apparent that the eddy currents induced in the core
measurements. Reference [6] tabulates measured
are higher when the rate of flux change is faster,
values of γ for several different ferrite materials.
since the induced voltage driving the eddy currents
will be higher. Thus, compared to the eddy current
loss generated by a sine wave of equal frequency,
a low duty-cycle rectangular voltage waveform PV_RECT
generating an equal peak-to-peak flux density must PV_SIN
generate higher eddy current loss in the core.
Additionally, the magnetic domains theory suggests
that the domain walls cause nonuniform flux density,
which results in eddy current losses in excess of
those related to the material’s conductivity.
References [5], [6], [7] and [8] discuss in more

Figure 6. Core-loss ratio for rectangular versus sinusoidal excitation
detail the mechanisms that relate the losses to as a function of duty cycle [6]. (Image: Courtesy of the Institute of
waveforms, duty cycle and DC bias, which are Electrical and Electronics Engineers [IEEE], © IEEE 2014)
beyond the scope of this paper. Those authors

3F35 at 500 kHz (various Bpk, D values) PC90 at 1 MHz (various Bpk, D values)
Curve fit: F(HDC) = 2.1875 x 10-4 (HDC)2+ 1 Curve fit: F(HDC) = (0.04 x HDC + 1)0.5
Figure 7. Core-loss ratio for DC bias versus no bias excitation. Source: Reference 5. (Images: Courtesy of Virginia Tech)
Effect of DC bias where PV_SINE is the Steinmetz equation loss for

The author of reference [5] measured the effect sinusoidal excitation.
on core losses when adding a DC bias, HDC, for References [5] and [6] contain the information
rectangular-waveform excitation and proposed necessary to use Equation 2 for several Ferroxcube
a curve-fitting factor, F(HDC), to account for the materials. We hope that magnetic materials
increase in loss due to the presence of DC bias in manufacturers will consider verifying the validity of
the core. the results reported and generate the information
Figure 7 shows curve-fitting equations for F(HDC) necessary to enable users to accurately calculate
for 3F35 and PC90 ferrite materials generated from core losses in PWM applications, which are far
measurement data of core losses with DC bias, and more common than sine-wave applications.
rectangular-waveform excitation at different duty In order to put Equation 2 to practical use,
cycles and flux-density amplitudes. manufacturers must make available the following
The F(HDC) function represents the increase in core information about magnetic materials:
loss caused by DC bias; it appears that it is relatively • Frequency and flux-density exponents to generate
insensitive to the amplitude and duty cycle of the the correct PV_SINE at the relevant flux density and
excitation voltage. frequency range. (Note: Ferroxcube provides an excellent
spreadsheet documenting their materials, available
Total core loss for arbitrary upon request).
waveforms • The γ parameter and an appropriate equation with which
Equation 2 combines the results presented to use it.
above with the core-loss equations provided by
• The equation for FDC(HDC ).
manufacturers for sine-wave excitation to calculate
core loss for the rectangular-waveform excitation
present in flyback (and many other PWM) converters:
𝑃𝑃!_!"!#$ = 𝑃𝑃!_!"#$ ∙ 𝐹𝐹!"#$%&'( (𝛾𝛾, 𝐷𝐷) ∙ 𝐹𝐹!" (𝐻𝐻!" ) (2)

We must emphasize a few points: will depend on the frequency, wire diameter and
• Duty cycle and DC-bias effects on core losses are overall layer structure. The eddy currents induced
significant, and should not be ignored. inside the wires (as a result of the magnetic field
• The substantial increase in core loss at extreme duty- inside the wires) are the main cause of AC loss and
cycle values is an often-neglected penalty of wide input/ increased ACR. These eddy currents lead to skin
Use original Excel file for
effect and proximity effect, which we will explain
output-voltage-range converters.
a better qualty image.
further in the next sections.
• The increase in loss due to DC bias reduces the benefits
expected from CCM operation. Current Harmonic Content
1.2 7.0
• The assumption of equal core loss in single- and double- IPRI 4 A peak
Np/Ns = 5 6.0
ended applications with equal AC flux excursions is 1 DPRI = 20%
DSEC = 35%
probably incorrect. 5.0
Harmonic Current (A)

0.8
4.0
Copper loss and AC effects in 0.6
flyback transformers 3.0
0.4
Current flowing through the resistance of copper IPRI 2.0
ISEC
windings causes copper loss in a transformer. 0.2
1.0
Losses arise because of the DC component of

0 0.0
the current and the DCR of the windings, but also 0 2 4 6 8 10 12 14 16 18 20
Harmonic Order
(and often more significantly) from high-frequency
AC effects. Figure 8. Harmonic content of typical flyback primary and secondary
current.
For flyback converters operating in DCM or
transition mode (TM), the current flowing in both Skin effect
the primary and secondary windings is triangular When DC current flows in a wire, the current density
in shape (Figure 3). Since the flyback converter is uniform throughout the wire’s cross-section; in
stores energy during the primary conduction interval other words, the current is distributed equally across
and then delivers energy to the load and the output the wire. But when a time-varying AC current flows,
capacitor during the secondary conduction interval, the changing current produces a changing magnetic
the duty cycle of each interval is typically less field around the wire. This changing magnetic field
than 50 percent. The primary duty cycle will often is also present inside the wire. Faraday’s law states
be much less than 50 percent at high-line input that whenever there is a changing magnetic field,
voltages, where the di/dt of the current ramp is a voltage (or electromotive force [EMF]) is induced,
much steeper, and so the high-frequency harmonic so as to oppose the changing magnetic field.
content will also be greater. Consequently, AC loss The induced voltage causes circulating eddy
mechanisms can become more significant. Note currents to flow, and since the conductivity of
that in Figure 8, the zeroth harmonic is actually the copper is high, these currents can be very
DC component of the current waveform. significant. The eddy currents reduce or cancel the
current flow in the center of the wire, and reinforce
Since the flyback’s primary and secondary currents
or increase the current flow in the outer regions of
have a significant DC component and significant
the wire cross-section, leading to current-density
high-frequency harmonic content (Figure 8), both
distribution as shown in Figure 9.
ACR and DCR are important. The ACR-to-DCR ratio

Use original Excel file for
a better qualty image.
Skin Depth vs. Frequency

0.8
Current density, J
0.7
0 1/e 1
0.6
0.5
Depth (mm)
7d 0.4
5d 0.3
0.2
0.1
Figure 9. Nonuniform AC current distribution due to induced eddy 0.0

0 200 400 600 800
currents. Source: Reference 9. Frequency (kHz)
As the frequency of the AC current increases, the Figure 10. Copper skin depth (or penetration depth) in millimeters vs.
current becomes more concentrated near the frequency in kilohertz at 100°C.
outer edges of the wire, and the central portion ACR to DCR, assume that all of the AC current
of the wire will carry almost none of the current. flows in an annular ring around the outside of the
The “skin depth” is defined as the depth inside wire, one penetration-depth wide. Thus, Equation 5
the wire where the current density has fallen to approximates the ratio of ACR to DCR by the ratio
approximately 37 percent (1⁄e) of the value at the of the total wire cross-section to the cross-section
surface. This depth is also where the penetrating of the 1-δ wide outer annulus:
magnetic field strength has fallen by the same 1⁄e
ratio – hence it is also sometimes referred to as 𝐴𝐴𝐴𝐴𝐴𝐴 (7𝛿𝛿)! 49 49
= ! !
= = ≅ 2 (5)
𝐷𝐷𝐷𝐷𝐷𝐷 (7𝛿𝛿) − (5𝛿𝛿) 49 − 25 24
“penetration depth.” Penetration depth, δ, depends
on the resistivity of the wire material, ρ, the relative
This illustrates the significance of skin effect when
magnetic permeability of the wire material, μr, and
using large diameter wires. Using the example
the frequency of interest, f. See Equation 3:
from Figure 9 and Equation 5, reducing the wire
diameter to 2-δ reduces the ACR to DCR ratio
𝜌𝜌
𝛿𝛿 = (3) to approximately 1; however, the DCR will have
(𝜋𝜋 ∙ 𝜇𝜇! ∙ 𝜇𝜇! ∙ 𝑓𝑓)
increased twelvefold due to the significantly smaller
wire diameter. Filling the space occupied by the
Since the wire used in transformers is almost
single 7-δ wire with multiple 2-δ wires reduces
exclusively copper, δ can be conveniently expressed
DCR and consequently ACR. Replacing the single
as a function of only frequency. At 100°C, Equation 4
large 7-δ wire with an array of nine paralleled 2-δ
gives the δ of copper, where f is in kilohertz (plotted
wires (to fit in approximately the same total area as
in Figure 10):
the original wire), DCR is now 136 percent of the
2.3 ∙ 10!! 1 2.4 𝑚𝑚𝑚𝑚 original value (72/(9*22)). Thus ACR is now 1.36 times
𝛿𝛿 = !!
∙ = (4) the original DCR, compared to twice the original
(𝜋𝜋 ∙ 4𝜋𝜋 ∙ 10 ∙ 1 ∙ 1𝑘𝑘) 𝑓𝑓 𝑓𝑓
DCR for the single large-diameter wire. Of course,
Looking back at Figure 9 as an example, the this improvement comes at the penalty of more
wire diameter is seven times larger than δ at the complicated multistranded wires – but these are kind
frequency of interest. To approximate the ratio of of trade-offs that you need to consider when weighing
cost/complexity against efficiency performance.

(a) (b) (c)
Figure 11. AC current distribution due to induced eddy currents for single wire (skin-effect only) (a); two adjacent wires with current in same
direction (b); and two adjacent wires with current in opposite directions (c). Source: Reference 9.
Proximity effect – single layer When current flows in two adjacent wires, the
In the previous section, we explained skin effect in the magnetic field from the AC current flow in each wire
context of a single isolated wire. But rarely will you affects the current distribution of the other.
encounter a single isolated wire in practice. Flyback- When currents flow in the same direction, the
transformer windings always consist of multiple current distribution will tend toward the farther-
turns, built up in multiple layers, including at least away outer surfaces, and the current density at the
one primary winding and one secondary winding. facing edges drops. When currents flow in opposite
They usually also include an auxiliary winding, and directions, the current density concentrates at the
sometimes multiple secondary windings. inner-facing surfaces.
Skin effect alone is actually not that significant. What
If you place multiple adjacent wires together in a
is far more important in the context of transformers
typical single-layer transformer winding, the current
is “proximity effect.” This is very similar to skin effect,
flow will be in the same direction in each wire,
but arises from the effect of the magnetic field that
assuming that they are connected in series. The
AC current flow in one wire causes on all adjacent
proximity effect will reduce the current density at the
wires. As you will see, proximity effect can build up
adjacent-facing edges of each wire (except for the
rapidly as you add more layers of wire – to the point
first and last wire in the layer), as shown in Figure 12.
where the inner layers are carrying significantly more
The current density is concentrated along the
eddy current than load current.
top and bottom surfaces of the wires in the layer,
We will first explain how proximity effect occurs in
with little current flow in a central strip along
a pair of wires and then in a single layer of multiple
the layer. This qualitatively highlights how much
wires. A common misconception is that proximity
more significant and important proximity effect is
effect only applies to multiple-layer windings and does
compared to skin effect alone. Even for a single
not occur in single-layer windings. But proximity effect
layer, if the wire diameter is too large compared to
does occur in single-layer windings, and its extent
the penetration depth, proximity effect will occur.
depends on the chosen wire diameter.
Figure 11 shows the AC current flow in a single

wire, the effect of two adjacent wires with current
flow in the same direction, and the effect of currents Figure 12. AC current distribution due to proximity effect for a
in opposite directions. Note that for ease of single-layer winding, with all currents flowing in the same direction.
illustration, the wire diameter is much greater than

the penetration depth at the frequency of interest.

adjacent layers and how the resulting current

concentration is worse than skin effect alone.
Proximity effect becomes progressively worse
with the addition of more winding layers, inducing
canceling eddy currents in each layer that
Figure 13. AC current distribution due to proximity effect for a
two-layer winding, with all currents flowing in the same direction. contribute significantly higher losses. Figure 15
illustrates a three-layer 24-turn winding, with eight
turns per layer. Current is flowing in the same
direction (out of the page surface) in each winding
+
+
+
+
+
layer. Once again, the wire diameter is much larger

than the penetration depth in order to highlight the
Figure 14. AC current distribution due to proximity effect for a t proximity effect.
wo-layer winding, with currents in each layer flowing in opposite
directions.
L3
Proximity effect – multiple layers
+
+
L2
If you extend and implement the transformer
+
+
+
+
+
+
+
+
L1
winding over two layers, the proximity effect will
impact the distribution within each layer as already
seen – but each layer will also impact the other. Figure 15. AC current proximity effect for a three-layer winding, with
currents in each layer flowing in the same direction.
Figure 13 illustrates how the current distribution
is concentrated only along the outer surface of the Assuming a normalized 1-A current in the winding,
wires in each layer. A two-layer flyback transformer with 24 turns the magnetomotive force (MMF) is
primary or secondary winding could typically have 24 At. Since the wires are so large compared to
this kind of winding structure. the penetration depth, the magnetic field cannot
Arranging a two-layer winding with currents flowing penetrate far enough into any of the winding layers.
in opposite directions in each layer causes the A corresponding 24-At MMF on the inner surface of
current density to concentrate along the inner- the first innermost winding layer (L1) cancels the 24
facing surfaces of the wires in each layer, as At of MMF of the air gap. Thus, the inner surface of
shown in Figure 14. A forward-mode transformer each wire in layer L1 must carry 3 A each in order to
would typically have this type of winding structure, generate 24 At of MMF across eight turns.
where both primary and secondary current flows Since they are all connected in series, the net
simultaneously in opposite directions. A flyback current in each wire must be 1 A. This means that
transformer with adjacent primary and secondary a canceling 2-A current must flow in the opposite
winding layers has this type of structure during direction on the outer faces of the wires in L1 in
the transition interval, when the primary current order to get 1 A net. The magnetic field from that
commutates to the secondary. opposing 2-A current on the outer face of L1 will
The illustrations in Figure 13 and Figure 14 are then force a canceling 2-A current to flow in the
of course grossly simplified, with very large wire opposite direction on the inner face of L2 as shown
diameters, to illustrate proximity effect between in Figure 15.

Once again, since the net current in each wire in L2 Proximity effect – passive layers
must be 1 A, yet another 1-A canceling current will Passive layers are layers of a winding structure that
flow on the outer faces of L2. The magnetic field do not carry any useful load current. In some cases
from the 1-A current on the outer face of L2 forces they never carry useful current (such as an EMI
a corresponding canceling current in the inner face shield), while in other cases they carry current only
of L3. Because the wire diameter is so large that the part of the time (such as a center-tapped secondary
magnetic field cannot penetrate far enough into the in a forward-mode push-pull converter – each
wire, these canceling currents develop to allow the half only carries current 50 percent of the cycle at
magnetic field to propagate through the multilayer most). During any interval when no load current
winding structure. flows, proximity effect-induced eddy currents can

flow in the nonconducting winding, contributing to
In the example in Figure 15, the initial expectation
conduction loss even when not conducting.
is that the conduction loss would be proportional
A flyback transformer with interleaved primary and
to (3 * I2), since each of the three layers carries the
secondary layers is another example of a passive
same net current, I. Using Equation 6 to sum the
layer, when the nonconducting secondary is
contribution of the currents on all of the faces results
sandwiched between conducting primary layers.
in a loss proportional to:
Even with a noninterleaved flyback transformer, the

𝑃𝑃!"#$ ∝ 𝐼𝐼 ! + 𝐼𝐼 ! + (2𝐼𝐼)! + (2𝐼𝐼)! + (3𝐼𝐼)! = 19𝐼𝐼 ! (6) primary or secondary layer that sits closest to the
core air gap will also be a passive layer when it is
not conducting.
The total losses are more than six times higher than
expected. Adding more layers with the same large Figure 16 illustrates this passive-layer proximity
wire diameter makes the situation progressively effect, where a single secondary layer, S, is
worse. For four layers, the loss would be (44 * I2) sandwiched between two inner primary layers
vs. (4 * I2), 11 times higher. For five layers, the loss (P1, P2) and two outer primary layers (P3, P4). As
would be (85 * I2) versus (5 * I2), 17 times worse; and before, assume that the wire diameter is much
so on for more layers. larger than the penetration depth and that a
normalized 1-A net current in each wire where the
Intuitively, you can see that by reducing the wire
total MMF is 32 At. Since the magnetic field cannot
diameter sufficiently, the 3-A current on the inner
penetrate the wires, all of the required current to
face of L1 will eventually merge with the canceling
balance the MMF flows on the inner face of layer
2-A current on the outer face of L1 to achieve both
P1. This results in a 4-A current on the inner face of
a net and an actual 1-A current flow in L1, reducing
each wire in L1, with a canceling 3-A current on the
proximity effect considerably. Of course, as we
outer face, and so on as before.
noted earlier when discussing skin effect, a narrower
wire diameter will have considerably higher DCR, As shown in Figure 16, the secondary layer will
which you must compensate for by using more have a 2 A of current induced along one face
paralleled strands of thinner wire. and a canceling 2 A of current induced on the
opposite face. So while the secondary layer is in
a nonconducting phase of the cycle and the net

current in the winding is zero, there are significant at the frequency of interest), and layer count.
eddy currents induced in the layer, contributing extra Figure 17 (reproduced from [14]) illustrates the
conduction losses. In this case, the nonconducting trade-off between wire size and layer count using
secondary will exhibit losses eight times higher Dowell’s equations.
than the outermost primary layer, P4. Intuitively, as
already noted, reducing the wire diameter sufficiently
will cause the canceling eddy currents to merge and
diminish greatly.
P4
+
+
P3
+
+
+
+
+
Figure 17. ACR factor RAC/RDC versus layer thickness and layer count.
+
+
+
+
+
P2
Source: Reference14.
+
+
+
+
+
P1
For low layer counts, you can use a larger wire size
without incurring a major increase in the ACR/DCR
ratio – a diameter twice the penetration depth will
Figure 16. AC current proximity effect for a four-layer flyback primary
winding, with a nonconducting secondary passive layer sandwiched result in an ACR/DCR ratio of 2 for a single layer.
in-between the primaries. However, even for a single-layer winding, ACR
Again, these examples are gross oversimplifications increases significantly if the wire size is much larger
to illustrate the fundamentals of proximity effect. than the penetration depth. This highlights the
But they do qualitatively highlight the impact of significance of proximity effect, even for single-layer
proximity effect, and the importance of carefully windings.
choosing the right wire diameter and winding If you need to use a large number of layers, you
construction. In many cases, adding more must keep the wire size to a smaller fraction of the
copper (either by increasing the wire size or penetration depth as the layer count increases.
adding more layers to fill the winding window) For example, a 10-layer winding would require the
can actually be counterproductive, leading to layer height to be approximately half the penetration
higher transformer losses.
depth to keep the ACR/DCR ratio at 2.
ACR factor Dowell’s equations apply to sinusoidal waveforms

Extensive analysis exists on the topic of ACR and at a single frequency. In flyback transformers, there
proximity effect, most notably by Dowell [10]. is significant high-frequency harmonic content,
Despite the large number of assumptions and the particularly at smaller duty cycles (Figure 8). So
fact that they deal only with sinusoidal currents, although the ACR/DCR ratio may be acceptable
Dowell’s equations have proven very useful for when based on the penetration depth at the
predicting the ACR factor, kP (the ratio of RAC to fundamental switching frequency, the ACR/DCR
RDC), as a function of layer ratio, Q (the ratio of layer ratio will increase significantly for higher-order
thickness to penetration depth [DPEN or δ] harmonics. This indicates that you may need to

Figure 18. Normalized effective resistance factor vs. layer thickness and layer count for triangular currents at a 50 percent duty cycle; reproduced
from [9]. (Image courtesy of Bruce Carsten)
choose a much smaller wire diameter to reduce the methods here based on work published by Carsten
losses associated with high-frequency harmonics. [9] and Hurley [12].
Note that the graphs in Figure 17 show the ACR/

Optimized wire size based on Carsten
DCR ratio. When using a smaller diameter wire,
DCR will increase rapidly, since it is inversely Figure 18 is reproduced from [9], where Carsten
applied Dowell’s equations to triangular currents
proportional to the square of the diameter. Although
at 50 percent duty cycle and evaluated the losses
the ACR/DCR ratio decreases with smaller wire
based on a Fourier expansion of the harmonics of
diameter, the absolute-value ACR will eventually
the waveform. The “effective resistance factor,” KR,
increase. Moreover, since the flyback current
is defined as the ratio of the effective resistance
waveforms have a significant DC component and a
to the DCR, with the layer height, h (equivalent to
fundamental switching-frequency component, the
the layer thickness in Figure 17), set equal to the
DCR is significant and requires minimizing.
penetration depth at the fundamental frequency, δ0.
Methods to choose the optimum Carsten generated curves for the KR factor versus
wire size the ratio of layer thickness to penetration depth
Given the conflicting requirements to minimize both for a range of layer counts. As you can see from
DCR and ACR, how can you choose the optimum the curves in Figure 18, for a given number of
wire diameter and strand count to minimize copper layers, there is a value of layer height where KR is a
loss for a given design? We will propose two minimum – this is the optimum layer height for that
layer count. Setting the layer height smaller than this TM), the ratio of the optimum wire diameter to
value will result in higher losses due to increased penetration depth is given by:
DCR; a larger layer height will have higher ACR.
! !! ∙!
Note that although there is an optimum KR value for Δ!"# =
!∙!
each layer count, the actual KR value will increase as (7)
!!! !!
the layer count increases – see the red diagonal line 𝑤𝑤ℎ𝑒𝑒𝑒𝑒𝑒𝑒: 𝜓𝜓 =
!"
in Figure 18.
Where Δ = d/δ0, d = wire diameter, D = duty cycle,
For a given transformer design and a target p = number of layers.
number of winding layers, you can use the curves
in Figure 18 to select h as a ratio of δ0 at the Equation 7 can be used to estimate the optimum
fundamental frequency. Knowing the optimum value layer height as before. However, where the Carsten
of h, you can calculate the optimum wire diameter, d. curves are available only for 50 percent duty cycle,
Depending on the transformer’s bobbin geometry, the Hurley equations can be used for different duty
choose the number of strands to fill the full layer cycles, and to plot the variation in optimum layer
widths as neatly as possible – it may be necessary height as a function of duty cycle.
to vary the chosen wire diameter up or down
Using Equation 7 for duty cycle D = 0.5, the ΔOPT
somewhat to achieve a good fill of the available
ratio (d/δ0) for one-layer winding is 1.57, and for
window width. We will provide real-world examples
two-layer it is 1.07. These figures agree reasonably
of winding optimization later.
closely with the red line that highlights the minimum
Optimized wire size based on Hurley loss points on the various curves of Figure 18,
Hurley et al [12] propose an alternative method to approximately 1.6 and 0.95, respectively, (note that
choose the optimum wire size where you determine the horizontal axis of Figure 18 is a log scale).
the optimum layer thickness for any arbitrary current
waveform simply by evaluating the rms value of the Wire type – solid core versus
current waveform and the rms of the derivative of multistrand versus Litz
the current waveform. This method gives reasonably As we have shown, the ACR factor can be very
accurate results – typically within 5 percent of the significant due to skin and proximity effects.
result calculable by using the first 30 harmonics However, since ACR is a multiple of DCR, and
of the waveform – but with significantly less the flyback currents contain a significant DC
computation and complexity. component, minimizing DCR is also important.
Hurley generated equations for rms values Figure 19 compares a single-strand wire of
of various common current waveforms, their diameter, d, and various combinations of smaller
derivatives and the corresponding optimum strand sizes that take up roughly the same space
ratio of layer height to fundamental-frequency as a large single strand. (In this example, the effect
penetration depth. From that table [12], for a of the small thickness of outer enamel insulation
variable duty cycle triangle-wave current wave- coating is ignored, although in practice you cannot
shape (closest to the current wave-shape in a always neglect this effect, especially for very small
flyback transformer when operating in DCM or wire diameters where it becomes increasingly more

significant). As the number of strands of smaller wire frequency would need to be 575 kHz before the
increases, the DCR increases since the fill factor will penetration depth equals the 0.1-mm diameter.
worsen because of space lost to gaps in-between The required number of strands, N, would depend
the individual strands. The fill factor for very small on the rms current and required DCR.
strands will be even worse due to the enamel At such high frequencies, Litz construction has
insulation on each strand. However, if the frequency advantages if you choose the correct wire size
is high enough, the decrease in ACR factor of the and strand count for the frequency and current
waveform of interest. A bundle of N strands of Litz
Single-strand 15-strand bundle,
diameter d diameter d/5 wire is equivalent to √N layers. So a single physical
167% DCR
layer of Litz wire is actually equivalent to √N layers
(a) (c)
when using Figure 18 or the Hurley equations from
4-strand 65-strand bundle, reference [12] to determine the optimum strand
diameter d/2 diameter d/10
Same DCR 154% DCR diameter. A poorly chosen Litz bundle (wire diameter
and strand count for the frequency of interest) might
(b) (d)
actually make losses worse.
Figure 19. Comparison of single versus multistrand wire bundles.
Litz wire is rarely used and probably offers little ACR
stranded wire may be sufficient to justify an increase
advantage for more conventional flyback switching
in DCR. By using a sufficient number of wire
frequencies (<150 kHz typical for EMI reasons).
strands, an acceptable DCR may also be achieved.
Moreover, Litz wire costs more, has poorer window
For bundles with a high number of multiple strands,
utilization and comes with handling difficulties. Since
twisting and bundling is very important in order to
there are many strands of small wire diameters,
ensure that all strands equally occupy all positions
sometimes a small percentage of strands can break,
in the cross-section of the bundle along each turn
which can impact the effective resistance, and result
around the core center leg. Litz wire is usually
in induced eddy current losses in the nonconducting
woven together from a number of sub-bundles to
strands. Soldering the small wire diameters can
help achieve this goal. If a bundle is poorly twisted,
be difficult at the transformer terminations; high
such that some strands occupy central positions
soldering temperatures can actually vaporize
in the bundle for all or most of the time, the losses
some strands, again leading to a percentage of
can actually be dramatically worse. In this case,
nonconducting strands.
most of the AC current will flow in the outer strands
only, increasing the effective resistance. Even worse, Leakage inductance
besides not carrying any (or as much) useful current, Leakage inductance is caused by the magnetic flux
the central strands can actually suffer large losses from one winding in a transformer that does not
due to induced eddy currents from the current flow couple to other windings. It is due to the magnetic
in the outer strands. flux in the spaces and gaps between windings,
Litz wire is commonly used for very high frequencies which stores energy in those gaps, in the same way
where the ACR factor would be significant. For that energy is stored in the air gap of a ferrite core.
example, for a Litz bundle using 0.1-mm wires, the The energy in the leakage inductance is typically

dissipated external to the transformer. Several of the h1 h2 h3

references – [1], [11], [13], [14] – discuss the causes
and consequences of leakage inductance in
more detail.
For flyback transformers, minimizing leakage P S P
b
inductance is very important, since in most cases
the leakage energy dissipates in an external
snubber or clamp circuit. As you will see later,
it is also important to minimize the ratio of leakage
c c
to the magnetizing inductance in order to minimize
Figure 20. Typical interleaved flyback transformer winding layer
the amount of magnetizing energy lost to the
spacing.
clamp circuit.
Per the assumptions used in Dowell’s equations
How to estimate and minimize [10], each winding layer of circular wire diameter d
leakage inductance converts to an equivalent rectangular block of the

same cross-section. Thus, h is not quite the same
Several published methods explain how to
as d (Equation 9):
estimate leakage inductance based on the physical
geometry of the windings and the layer structure in 𝑑𝑑 ! 𝜋𝜋
a transformer. Figure 20 and Equation 8 show a ℎ! = 𝜋𝜋 ∙
(9) ⇒ ℎ = 𝑑𝑑 ∙ = 𝑑𝑑 ∙ 0.886
4 4
method used by Carsten [9]. This method assumes
that all winding layers are the same full width, b, As Figure 21 shows, you can use the wire diameter
and cannot be used with some partial-width layers. and Equation 9 to calculate h for each winding
Note, that the leakage inductance to any partial- layer. You can calculate the spacing, c, using the
width layers will be much higher, increasing as the thickness of an intervening tape or insulation,
width of the partial layer narrows. plus the thickness or enamel coating or any other
insulation on the wire, plus the extra gap due to the
𝜇𝜇! ∙ 𝑁𝑁 ! ∙ 𝑀𝑀𝑀𝑀𝑀𝑀 ∙ ( ℎ + 3 𝑐𝑐) 1 (8)
𝐿𝐿!"#$ = ∙ ! conversion of diameter d to an equivalent h.
3𝑏𝑏 𝑚𝑚
Following are the parameters of Equation 8: d h=d*0.88

c=d*0.12+
insulation
• N = number of turns on the winding to which the d h=d*0.88
leakage is referred
Figure 21. Converting each winding-layer pair into equivalent h
• MLT = Mean length per turn of the windings and c values.
• Σh = sum of the heights of all winding layers
• Σc = sum of the heights of the spacing gaps
between winding layers
• m = level of interleaving (number of winding
“portions”)
• b = bobbin winding breadth (width of the
winding layer)

m=1 m=2 m=3
P S P S P P S P S
MMF MMF MMF

0 0
0
P S P S P S P S P S
MMF MMF
0 0
Figure 22. Different levels of interleaving to reduce leakage inductance.
Leakage inductance reduction and to a flyback transformer during the transition

minimization interval, when primary and secondary current flow
simultaneously, and when leakage inductance
Leakage inductance depends largely on the physical
is relevant.
winding geometry; you can estimate the leakage
inductance from this geometry. An inspection of In the first example, the primary and secondary
Equation 8 shows that you can reduce the leakage are placed side by side with no interleaving. The
inductance by: corresponding MMF diagram underneath shows
• Interleaving – increasing the value of m. how the primary MMF builds up to a maximum
• Using a wider bobbin winding width – maximizing b. value, and is then canceled by the secondary, back
to zero. The MMF diagram shows just one portion
• Minimizing turns N and mean length per turn MLT.
or region of MMF build up and return to zero – so in
• Minimizing the thickness and quantity of spacing gaps, cn,
this example with one portion, m = 1.
between winding layers.
The primary and secondary are interleaved in the
• Minimizing winding layer thicknesses, hn, and the number
second examples. Both interleaving methods are
of layers.
different, but equivalent. That is because in both
Interleaving cases, the MMF builds to a peak and declines

back to zero over two separate portions, so in both
Figure 22 shows examples of different levels
cases m = 2. Similarly, the third example shows
of interleaving to reduce leakage inductance
two different interleaving methods, but as both are
in the context of a conventional forward-mode
equivalent in terms of interleaving portions, m = 3 in
transformer. But the same principles also apply
both cases.

Since leakage inductance is inversely proportional Leakage inductance estimation –

to m2, a first level of interleaving with two portions worked example
(i.e., m = 2) is expected to result in a four-fold
We now use Equation 8 to estimate the leakage
decrease in leakage inductance. By changing from
inductance of the typical flyback transformer
m = 1 to m = 2, as shown in Figure 22, the winding
winding structure illustrated in Figure 23. This
structure changes from a single primary-secondary
structure consists of a split primary with auxiliary
interface to two interfaces, one at each face of the
bias and secondary layers sandwiched in-between.
secondary. In the interleaved case, there are now
Each half primary is wound over two layers, and
two interface spacing gaps, c; both are likely to be
secondary and bias layers are wound over single
approximately the same dimension as the single
layers. There is also a shield layer inserted between
c in the noninterleaved case. So in reality, the factor
the secondary and outer half primary. Figure 23
Σc/m2 in Equation 8 only decreases by a factor of
lists all of the wire sizes.
two, and the leakage inductance will really only
We converted all round wire diameters to an
decrease by half. Nonetheless, this is a significant
equivalent rectangular block with the same cross-
reduction and is usually very beneficial.
section. The switching frequency of the target
However, the designer should be aware that
design was 60 kHz, with a δ ~0.31 mm. The shield
interleaving primary and secondary windings may
is sufficiently thin (0.05 mm, one-sixth of δ), so that
possibly increase transformer cost, complicate
for simplicity, we included it as part of the insulating
EMI and safety compliance (since there are now
space between the secondary and outer half
two primary-secondary interfaces), and increase
primary (Figure 24).
transformer interwinding capacitance.
All tapes 60 mm
Primary 0.25-mm enameled copper
Shield 50-mm copper foil
Secondary 0.55-mm triple-insulated copper
Bias/CM cancel 0.2-mm enameled copper
Primary 0.25-mm enameled copper
Figure 23. Interleaved flyback transformer winding structure.
c5 = (0.25*0.12) + (0.01*2)
c4 = (0.55*0.06) + (0.01*2) + (0.06*2)
+ (0.05 +0.105) + (0.25*0.06) h6 = 0.25 * 0.88
c3 = (0.20*0.06) + (0.01*2) h5 = 0.25 * 0.88
+ (0.06+0.105) + (0.55 * 0.06) h4 = 0.55 * 0.88
c2 = (0.25*0.06) + (0.01*2) h3 = 0.20 * 0.88
+ (0.06) + (0.20 * 0.06) h2 = 0.25 * 0.88
c1 = (0.25*0.12) + (0.01*2) h1 = 0.25 * 0.88
Figure 24. Conversion of sample interleaved flyback transformer winding structure to layer heights hn and spacing gaps cn.

When you know the heights, hn, of all of the diverts into the clamp circuit (represented in
conducting layers and the spacing gaps, cn, of Figure 25 by the voltage source, VCLAMP). A voltage
the insulating layers, you can estimate the leakage will develop across the leakage inductance, LLEAK,
inductance using Equation 8. In this case, the that is equal to the difference between VCLAMP
primary total turns N = 34; we used the RM10 and the reflected secondary voltage, VREFLECTED.
bobbin, with MLT = 52 mm. The primary is Consequently, the current in the leakage inductance
interleaved with two portions, m = 2; the winding (which is also the primary current) will decrease at a
breadth b = 9 mm. (The bobbin nominal dimension rate dependent on the difference between the two
is actually 10 mm, but due to wire insulation and voltages; thus, magnetizing current will flow into the
entry/exit wire routing, the layer fill is typically ~90 clamp circuit until the current in the primary decays
percent of the available width). to zero.
LLEAK D
𝜇𝜇! ∙ 34! ∙ 52 𝑚𝑚𝑚𝑚 ∙ (1.54 𝑚𝑚𝑚𝑚 + 3 ∙ 0.78 𝑚𝑚𝑚𝑚) 1
𝐿𝐿!"#$ = ∙ ! = 2.71 𝜇𝜇𝜇𝜇 LMAG
3 ∙ 9 𝑚𝑚𝑚𝑚 2 VOUT
+ VCLAMP
(10) VIN
In Equation 10, 2.71 µH is reasonably close to Q

Drive
the measured leakage inductance of 3.2 µH. The
difference comes down to practical nonidealities of
Figure 25. Simplified flyback schematic showing a clamp/snubber
the actual winding, such as tape creasing, entry/ network as an ideal voltage clamp.
exit thickness and return wires. Nevertheless, we
The rate at which the current decreases in the
have shown how to estimate leakage inductance
leakage inductance determines the rate at which
for a given winding structure and how to objectively
the current transfers to the secondary. Figure 26
compare different winding structures for leakage
illustrates this process.
inductance performance. We also have shown the
In Figure 26a, where the clamp voltage is only 10
impact of the chosen bobbin shape/geometry on
percent higher than the reflected secondary voltage,
leakage inductance.
a considerable amount of the magnetizing energy
Effect of the clamp voltage value is lost in the clamp. The current transferred to the
on transformer leakage inductance secondary is significantly lower compared to
losses Figure 26b, where the clamp voltage is 50 percent

higher than the reflected secondary voltage.
In a single-switch flyback converter, you typically
need a clamp circuit to limit the drain voltage of the In an extreme case, where the reflected secondary
switch; the clamp circuit absorbs energy stored voltage is equal to the clamp voltage, no voltage
in the transformer’s leakage inductance and, is available to force the current in the leakage
depending on the value of the clamp voltage, will inductance to decay faster than the current in the
also absorb a fraction of the magnetizing energy. magnetizing inductance; thus, no magnetizing
In Figure 25, you can see that immediately after current can be transferred to the secondary.
transistor Q turns off, all of the primary current The clamp circuit will absorb all magnetizing
energy and leakage energy.

Figure 27 shows the percentage of magnetizing

energy lost in the clamp, γ, as a function of the ratio
of clamp voltage to the primary reflected voltage, α,
and the percentage ratio of the leakage to the
magnetizing inductance.
The energy lost in the clamp increases rapidly as
the ratio of the clamp to the reflected voltage drops
below 1.2. This loss is also very sensitive to the
leakage-to-magnetizing inductance ratio.

Although selecting a lower clamp voltage may allow
(a) VCLAMP/VREFLECTED = 1.1 use of a lower-voltage switching transistor with
a lower RDS(on), the added clamp loss may easily
outweigh the expected benefits.
EMI shielding and cancellation

techniques
Causes of EMI
Reference [15] gives a very good and thorough
explanation of EMI causes and solutions. This
section will focus solely on the CM interference that
occurs from flyback transformer capacitance. We
will also discuss techniques to minimize CM EMI
(b) VCLAMP/VREFLECTED = 1.5 through transformer design and construction.
You may want to use the TIF file for a sharper quality. Figure 28 shows the CM current, ICM, that flows
Figure 26. IComparison
Note that of in
overlaid text current
Visio transfer to the
to change secondary
all headers winding
from
(a)serif
versus the voltage-clamp
to Arial font. level (b). between the primary and secondary transformer
windings caused by the voltage waveform imposed
% Magnetizing Energy Lost to Clamp across the interwinding capacitance. Since the
primary voltage swing is typically much greater
than the secondary voltage (for offline AC/DC
applications), ICM will flow from the primary to the
Energy Lost
secondary. ICM will then flow to earth through the

impedance from the output circuit to earth, causing
a potential CM interference issue.
Vclamp/N*VOUT
Figure 27. Graph of percentage energy lost to the clamp versus the
clamp level and leakage/magnetizing inductance ratio.

The shield typically connects back to local primary

NP NS ground as shown in Figure 29, or sometimes
connects to the input DC bulk capacitor’s positive
terminal – this is possible because that point is also
ICM
a quiet AC ground. With the shield in place, ICM will
flow into the shield and back to local primary ground
instead of flowing to output and from there back
Figure 28. CM current flow from the primary to the secondary due
to earth.
to the switching voltage waveform across the primary-secondary
interwinding capacitance. Shield
In many applications, the output return (negative

VOUT) terminal is often directly connected to earth
anyway – resulting in the worst-case potential
CM EMI result. This CM EMI needs filtering at the
power-supply input using a combination of CM filter
chokes and Y-capacitors. Any steps to reduce CM
EMI at the source can result in significant cost, size Figure 29. Flyback transformer with shield layer between the primary
and secondary.
and power-loss reductions for the EMI filter.
Even though ICM is trapped by the shield and
In order to minimize CM EMI, you can construct
returned to local primary ground, capacitance still
the transformer to minimize the interwinding
exists between the shield and secondary winding.
capacitance, which in turn will minimize ICM.
Since the voltage induced in the one-turn shield is
However, reducing the capacitance typically
not the same as in the secondary winding (unless
involves increasing the thickness of dielectric
you are using a one-turn secondary), there is
spacing between the primary and secondary
still some CM current flow between shield and
windings (moving them further apart), and/or
secondary. Therefore, although the shield helps
reducing the surface area of overlap between
greatly attenuate CM EMI, the shield cannot
them. Both of these changes will lead to poorer
eliminate it completely. A disadvantage with
primary to secondary coupling and increased
interleaved windings is that the number of primary-
leakage inductance. Remember that increased
secondary interfaces increases – more shield layers
leakage inductance will increase losses. So there
are usually required, one at each primary-secondary
is usually a trade-off between low-leakage
interface.
inductance construction for efficiency, versus
low capacitance for EMI.
EMI mitigation methods II – CM
EMI mitigation methods I – cancellation windings and CM
transformer shields balance
Electrostatic shields help reduce CM EMI when As an alternative to shields, you can introduce a
added to the transformer [15]. As discussed earlier, separate auxiliary cancellation winding, NAUX, as
the shield should be as thin as possible in order shown in Figure 30. The polarity of the auxiliary
to minimize eddy current loss in the shield due to winding is oriented to produce a canceling current
proximity effect. of opposite polarity to the current injected from the

NP NS NP NS NAUX
CP_S CS_AUX
ICM
ICM1 ICM2
Figure 30. Flyback transformer with additional auxiliary cancellation winding.
primary. By adjusting the number of turns NAUX and deployed between primary and secondary windings
the secondary-auxiliary capacitance, CS_AUX, you can as before – but rather than connect the shield to
make the magnitude of ICM2 equal to ICM1. Thus, ICM1 an AC ground, it is instead driven by an auxiliary
from the primary is canceled, and close to zero net winding, with NAUX = ½ NSEC. This ensures that the
CM current will flow to the output, and from there average voltage on the shield is the same as the
to earth. average voltage on the secondary winding. Since
This approach depends on tight manufacturing the average voltage at both ends of CSEC_SHIELD is
controls over the value of CS_AUX. If CS_AUX varies, the same, it is balanced for CM, so there will be
then the CM nulling will not be perfect. Even if the zero average CM current flow from the shield to the
manufacturer tightly controls CS_AUX capacitance, secondary winding.
its value will potentially vary as the ambient Shield
temperature changes and as the transformer’s
temperature changes due to internal self-heating. NAUX NSEC
Over the power supply’s lifetime, the capacitance
CSEC_SHLD
will also change as multiple heat/cool cycles cause CPRI_SHLD
the insulating tapes to expand, compress and
harden over time – such that the capacitor dielectric
thickness can change. Figure 31. Flyback transformer with shield plus auxiliary winding for
CM balance.
As an alternative to CM cancellation, you can
arrange the transformer windings to achieve CM CM balance – design example
balance. With this method, the average voltage
Figure 32 is an example implementation of CM
at both ends of the interwinding capacitances are
balance in a flyback transformer winding structure.
arranged to be the same amplitude and polarity,
Figure 33 shows the equivalent schematic, with
thus minimizing or nulling the CM current through
the windings colored to match the equivalent
the parasitic capacitance in a way that does not rely
physical winding. The structure is interleaved, with
on the value of the capacitance itself.
the primary split into two half primaries and all other
Figure 31 shows an example where a shield is windings sandwiched in-between.

Primary NP2
Shield
Secondary NS
Bias NB1
CM NB2
Primary NP1
Figure 32. Flyback transformer winding structure with CM balance.
Between the inner half primary and the secondary Using the schematic in Figure 33, the arrangement
is an auxiliary winding that serves multiple functions. is explained using the voltages in the circuit. The
First, as the primary-side bias winding (pink strands), output voltage is approximately 20 V, so the average
it supplies bias power for the primary controller. voltage for the flyback phase across the secondary
Second, it is wound with multiple parallel strands will be approximately 10 V. With a 6T secondary,
(pink and gray strands) that fill the full bobbin width this is equivalent to 1.67 V/T. NB1 is a 4T winding
and serve as a shield layer between the inner half to generate the required bias rail, and NB2 consists
primary and the secondary winding. Third, there are of a further 2T, for a total 6T of auxiliary winding to
extra turns (gray strands) such that the total turns match the 6T secondary winding. Consequently, the
in that layer match the secondary winding turns in capacitance between auxiliary and secondary has
order to achieve CM balance between the layers. the same voltage waveform and amplitude at both
A shield layer (purple) is placed at the other interface sides, so it is CM-balanced.
between secondary and outer half primary. Again, There is a tap on the NB1 winding after 3T to drive
the shield in this case is not tied to an AC ground, the shield. Thus, the average shield voltage will also
but is instead driven by a tap on the auxiliary be 10 V. Once again, the same average voltage on
winding, such that the average voltage on the shield both the secondary winding and the shield layer
matches the average voltage on the secondary, (10 V) achieves CM balance. The major advantage
again achieving CM balance. of this arrangement is that it does not rely on the
VDD
value of the parasitic capacitance, or controlling it to
a required value. CM balance is assured regardless
NP1 NB2 Shield
of capacitance value or variations.
NP2
In practice, you will need to adapt the
NB1 NS
implementation and structure used to achieve CM
balance depending on your specific application’s
NB1 + NB2 = NS
⇒ zero current in circumstances. Very often, figuring out the best
parasitic caps to NS
implementation for a given case involves trial
Figure 33. Equivalent schematic of the flyback transformer shown in and error.
Figure 32.

It is difficult to estimate the cost/performance However, if you are also using the auxiliary/bias
trade-offs between applying CM mitigation inside winding for primary-side regulation (PSR), it is not
the transformer versus external CM EMI filtering. possible to put the bias rectifying diode in the return
Shielding and CM balance layers inside the leg. In this case, the shield drive winding NB2 must
transformer will add material and labor cost to wind in the opposite direction, with appropriate
the transformer – but they can be very effective turns, to provide the necessary balance versus the
at reducing CM noise in a way that does not add secondary layer. In such cases with a secondary
significantly to the overall size or power loss. rectifier in the return leg, it can be simpler and more
On the other hand, adding an external CM filter cost-effective to use two shields on either side of
choke, or increasing its size/inductance is relatively the secondary driven by an anti-phase winding
straightforward. But this will incur extra cost as the NB2 = ½ NSEC.
external CM choke will occupy more space and
incur extra losses. Real-world examples of EMI and
efficiency improvements
One further important point: the secondary-side
rectifier in Figure 33 is shown in the conventional Transformer No. 1
high-side location. But locating the rectifier (diode or
Using the TI UCC28630EVM572 (EVM572) [16]
synchronous rectifier [SR]) on the low side (for gate-
as a starting point, we reviewed the structure
drive ease), will change the polarity of the secondary
of its transformer. We identified possible further
winding with respect to the auxiliary/bias layer.
improvements to the structure, and implemented
This means that a winding with opposite polarity is
and tested several alternative versions against
required to drive the shield – ideally with an inverted
the original transformer for EMI and efficiency
auxiliary winding and a bias rectifying diode placed
performance. The evaluation module (EVM)
in the return leg. This preserves all relative winding
demonstrates a 19.5-V output reference design
polarities and maintains CM balance.
All tapes 60 mm
W5 primary 15 x 0.1-mm enameled
copper, twisted bundle
W4 shield 50-mm copper foil
W3 secondary 7 x 0.2-mm twisted
bundle, triple-insulated
W2 bias/CM cancel 0.2-mm
enameled copper
copper, twisted bundle
Figure 34. UCC28630EVM572 transformer winding construction (transformer No. 1).

using the UCC28630 DCM/CCM PSR controller, We designated this standard EVM572 transformer
and is rated for an average output power of 65 W, as transformer No. 1.
or up to 130 W of intermittent peak power. Analysis of transformer No. 1 performance
Figure 34 illustrates the structure of the original We measured transformer No. 1 for leakage
transformer, while Figure 35 is the equivalent inductance, EMI and efficiency performance in
electrical schematic. The transformer uses a the EVM572 board. Since the EVM efficiency
standard RM10/I core set and bobbin. The windings was typically worse at low line, the efficiency
are interleaved – the primary split into two half performance with all transformer variants was at
primaries, with all other windings sandwiched 115 VAC. We tested conducted EMI at both
in-between. A twisted bundle is used for each half 115 VAC and 230 VAC, and found it worse at low
primary, comprising 15 strands of 0.1-mm wire, line, 115 VAC.
with each half primary wound over a single layer. We measured the leakage inductance at 4.51 µH
4 across the full primary terminals, with the main
W1 secondary shorted (all other windings open).
8
5 9 The full load efficiency at 65-W load, 115-VAC/
W5 60-Hz input was 87.7 percent, or 9.11 W of total
6
W3 power loss (not including the output cable drop).
1
W2 Figure 36 shows the typical conducted EMI
2 10
1 performance at 115 VAC, 65 W. The results show
11
W4 very good pass margin due to the balanced CM
shielding structure of this transformer, as previously
Figure 35. UCC28630EVM572 transformer schematic
(transformer No. 1). discussed.
Between the inner half primary and the secondary is

an auxiliary winding.
The secondary uses a twisted bundle of seven

strands of 0.2-mm wire; the outside of this bundle
is triple-insulated for safety compliance. Using two
triple-insulated bundles in parallel fills the full layer
and reduces the total resistance of the secondary.
A shield layer is placed between the secondary
interface to the outer half primary. As noted
previously, the shield in this case is not tied to an
AC ground but is instead driven by the bias-rail tap
on the auxiliary winding so that the average voltage
Figure 36. Conducted EMI plot for the UCC28630EVM572 with
on the shield is close to the average voltage on the transformer No. 1. Conditions: input voltage 115 VRMS, output power
secondary, to get close to CM balance. 65 W, output return connected to earth; quasi-peak (QP) result in blue,
average (AVG) result in green.

Reviewing the construction from a loss perspective, The RM10/I bobbin winding window width is
the primary wire strand diameter is 0.1 mm. Since 10 mm, which allows for a target of 9 mm (90
the EVM’s full-load switching frequency is 60 kHz, percent) for the wires. We assumed an enamel
the fundamental frequency δ is approximately coating thickness of 0.02 mm for all standard
0.31 mm. Thus, the 0.1-mm strand h/δ ratio is only wires and this is added to get the actual wire outer
0.29 – this is very low. Assuming that the 15-strand diameter (OD). We assumed that the triple-insulation
bundle is sufficiently well twisted and woven to give thickness was 0.2 mm, again added to the bare-
Litz-like performance, this construction is equivalent copper diameter to get the actual wire OD. To
to almost a four-layer winding (N-strand bundle/Litz choose the best wire diameter and strand count
⇒ √N layers). Even with four layers, going back to requires several iterations to find the combination
Figure 18, the optimum h/δ ratio is approximately that fits the available width comfortably, using
0.7. Clearly, there is room for further optimization. complete full winding layers. Full-width layers are
The secondary has seven strands of 0.2-mm important to minimize leakage inductance.
diameter wire and is thus equivalent to a ~2.6-layer Using the optimum wire diameter – 0.315 mm
winding. Again looking at Figure 18, the optimum (SWG30) wire with OD = 0.34 mm – four strands
h/δ ratio is approximately 0.9, but the actual h/δ would not fit in two layers, while three strands
ratio is 0.57, so again there appears to be room for would not fully fill two layers. Since the curves in
further optimization. Figure 18 show a reasonably flat characteristic near
The next sections will investigate various the optimum value, there should be little penalty in
alternate transformer constructions and compare adjusting the wire diameter up or down slightly to
performance. As we mentioned in the introduction, get a better full-layer-width fill factor. For this reason,
we changed only the transformer winding structure, we decreased the wire size to 0.25 mm (0.27-mm
keeping all other components on the board OD), using four parallel strands to completely fill two
unchanged for all tests, including the transformer layers with 17T for each half primary.
core material (Ferroxcube 3C95) and the turns Regarding the secondary layer, for a one-layer
count for all windings. winding, the optimum h/δ ratio is approximately
1.6, again according to Figure 18. At a 60-kHz
Transformer No. 2 fundamental switching frequency, δ is 0.31 mm,
For transformer No. 2 a structure similar to giving h = 0.5 mm. Converting from rectangular
transformer No. 1 is used, but with two layers layer back to round wire, the optimum diameter is
for each half primary and a single layer for the then approximately 0.56 mm. Considering 0.55-mm
secondary. For a two-layer half-primary winding, triple-insulated wire, with OD = 0.75 mm, six turns
the optimum h/δ ratio is approximately 0.9 of two strands will neatly fill the full available
according to Figure 18 At a 60-kHz switching layer width.
frequency, δ is 0.31 mm, giving h = 0.28 mm.
As Figure 34 showed, the original auxiliary bias
Converting from rectangular layer back to
layer is not completely full, so the shielding will not
round wire, the optimum diameter is then
be as effective as it could be. For transformer No.
approximately 0.32 mm.
2, we added an extra strand to the auxiliary bias

All tapes 60 mm
copper, laid flat (quad-filar)
W3 secondary 1 x 0.55-mm
single-core, triple-insulated
enameled copper
Figure 37. Transformer No. 2 winding construction.
layer for a total of eight strands. Transformer

No. 2’s schematic is the same as Figure 35.
Figure 37 illustrates the final winding cross-section
and construction of transformer No. 2.
Analysis of transformer No. 2 performance

We checked transformer No. 2 for leakage
inductance, EMI and efficiency performance in the
EVM572 board.
We measured the leakage inductance at 3.24 µH
across the full primary terminals, with the main
Figure 38. Conducted EMI plot for the UCC28630EVM572 with
secondary shorted (all other windings open). transformer No. 2. Conditions: input voltage 115 VRMS, output power
This shows a significant reduction of 30 percent 65 W, output return connected to earth; QP result in blue, AVG result
in green.
compared to transformer No. 1.
The full load efficiency at 65-W load, 115-VAC/ Transformer No. 3

60-Hz input was 89.03 percent, or 8.01 W of total The leakage inductance of transformer No. 2 was
power loss (not including the output cable drop). 30 percent lower than transformer No. 1, and the
This is a 1.3 percent efficiency improvement or a overall losses reduced by 12 percent, or 1.1 W.
loss reduction of 1.10 W (12 percent lower total How much of this improvement was due to the
losses in the whole board) – with no other change leakage inductance reduction (less energy lost in
but the transformer. the external clamp), and how much was due to the
Figure 38 shows typical conducted EMI expected improvement in winding ACR?
performance at 115 VAC and 65 W. This result is To answer this question, we implemented
similar to transformer No. 1, as expected, since we transformer No. 3; see Figure 39. This is an almost
used a similar CM balance approach. identical construction to transformer No. 2, except
that we added several more layers of tape to the
construction on top of the first half primary and
below the second half primary. This increase in
separation increases the leakage inductance. We

All tapes 60 mm
W3 secondary 1 x 0.55-mm single-
core, triple-insulated
enameled copper
used the process described earlier for estimating This demonstrates that transformer No. 2’s
leakage inductance and Equation 8 to estimate improvements were roughly 50 percent due to the
the required number of extra layers of tape to leakage inductance reduction and 50 percent due
increase the leakage inductance to the same level to the improved ACR factor, thus emphasizing the
as transformer No. 1. Based on our calculations, importance of minimizing both leakage inductance
an extra 12 layers of tape would increase the and ACR for a flyback transformer.
leakage inductance by approximately 40 percent. In order to highlight the benefit of the CM-balanced
In practice, we actually needed an extra 14 layers winding structure, we added a jumper to allow
of tape to get the desired 40 percent increase in connection or disconnection of the shield and CM
leakage inductance, which is very close to the cancellation winding. As Figures 40-41 show,
estimate using Equation 8. there is a considerable difference. With the shield/
Analysis of transformer No. 3 performance CM-balanced winding connected, the conducted
We checked transformer No. 3 for leakage EMI drops as much as 26 dB on QP emissions and
inductance and efficiency performance in the 20 dB on AVG.
EVM572 board. For the conducted EMI plot of Figure 41, the pass
We measured the leakage inductance at 4.46 µH margin is at least 17 dB below the Class B QP limit
across the full primary terminals, with the main line (at 900-950 kHz) and at least 10 dB below the
secondary shorted (all other windings open). This is AVG limit line. This indicates that you can decrease
approximately the same result as transformer No. 1, the EMI filter size to achieve the more typical 6-8 dB
as targeted. pass margin, and yield significant savings in size,
The full load efficiency at 65-W load, 115-VAC/ cost or losses.
60-Hz input was 88.39 percent, or 8.54 W of total
power loss (not including the output cable drop).

the EMI filter size to achieve the more typical 6-8 dB

pass margin, and yield significant savings in size,
cost or losses.
Transformer No. 4
Revisiting our initial assumptions, what if each
primary was wound over one layer instead of two?
We calculated that the optimum wire diameter for
one layer is approximately 0.55 mm. However,
Figure 40. Conducted EMI plot for UCC28630EVM572 with
17T does not quite fit in one layer in the available
transformer No. 3 – shield/CM cancellation winding disconnected.
Conditions: input voltage 230 VAC, output power 65 W, output return window width. To accommodate the full 17T in one
connected to earth; QP result in blue, AVG result in green.
layer, we decreased the wire diameter slightly to
0.5 mm, which almost fully fills one layer (0.53-mm
wire would be ideal but was not readily available
during prototyping).
Thus, we implemented transformer No. 4 with each

half primary in a single layer, using a single 0.5-mm
strand. All other layers were identical to transformer
No. 2 (including the minimum number of tape
layers). Figure 42 illustrates the final winding cross-
section and construction of transformer No. 4.
Figure 41. Conducted EMI plot for UCC28630EVM572 with Compared to transformer No. 2, the primary
transformer No. 3 – shield/CM cancellation winding connected.
Conditions: input voltage 230 VAC, output power 65 W, output return winding cross-sectional area is unchanged
connected to earth; QP result in blue, AVG result in green. (4 x (0.25 mm)2 vs. 1 x (0.5 mm)2), so the DCR
should be the same. The one-layer half-primary
For the conducted EMI plot of Figure 41, the pass
winding height of 0.5 mm is also the same as the
margin is at least 17 dB below the Class B QP limit
two-layer 0.25-mm version, so given Equation 8,
line (at 900-950 kHz) and at least 10 dB below the
we would expect the leakage inductance to also
AVG limit line. This indicates that you can decrease
be very similar.
All tapes 60 mm
W5 primary 1 x 0.50-mm
enameled copper
enameled copper
W1 primary 1 x 0.50-mm
enameled copper

All tapes 60 mm
enameled copper
W1 primary 0.5-mm
enameled copper (2 layers)
Figure 43. Transformer No. 5A winding construction.
Analysis of transformer No. 4 performance its effectiveness. Transformer No. 5A had no

We checked transformer No. 4 for leakage interleaving, and was similar in construction to
inductance and efficiency performance in the transformer No. 4. As Figure 43 shows, the
EVM572 board. 0.5-mm half primaries are now wound together in
We measured the leakage inductance at 3.29 µH two layers, without being interleaved as they were
in the case of transformer No. 4. The auxiliary bias
across the full primary terminals, with the main
layer is unchanged, providing shielding and CM
secondary short circuit (all other windings open).
balance, and the secondary layer is also similarly
This is very close, as expected, to the 3.24-µH
unchanged. Since this version has no interleaving,
value for transformer No. 2.
there is no need for a shield layer on the outside of
The full load efficiency at 65-W load, 115-VAC/ the secondary layer.
60-Hz input was 88.58 percent, or 8.37 W of total
In order to show the difference between interleaved
power loss (not including the output cable drop).
and noninterleaved implementations, we constructed
This demonstrates that although you can save transformer No. 5B as shown in Figure 44. This
~0.5 W through the reduction in leakage inductance is very similar to transformer No. 5A, except that
of both transformer No. 2 and transformer No. 4 we added a shield layer outside the secondary,
compared to transformer No. 1, the net power loss and another identical two-layer 0.5-mm full primary
savings of transformer No. 4 is only 0.74 W. The outside the shield. The shield is driven from the bias
change in primary winding structure (to use a single, winding just as in transformer No. 2. Connecting
larger-diameter wire) actually increased losses by the inner and outer full primaries in parallel achieves
0.34 W compared to transformer No. 2. Despite interleaving, just as in previous examples where two
the same DCR, the ACR factor of transformer No. 4 half primaries are connected in series. In the parallel
is considerably worse than transformer No. 2. configuration, the current will split approximately
50/50 between the two sets of primaries, so each
Transformer Nos. 5A and 5B
one will produce half the MMF compared to the
We wound transformer No. 5 in two versions, noninterleaved case.
with and without interleaving, to demonstrate

All tapes 60 mm
W5 primary 0.5-mm
enameled copper
W1 primary 0.5-mm
Figure 44.Transformer No. 5B winding construction.
Compared to transformer No. 4, the primary The full load efficiency at 65-W load, 115-VAC/
winding cross-sectional area of transformer No. 5A 60-Hz input for transformer No. 5A was 86.94
is unchanged, so the DCR should be the same. The percent, or 9.76 W of total power loss (not including
one-layer primary winding height of 1 mm is also the the output cable drop). This is a huge drop in
same as the two-layer 0.5-mm for transformer efficiency, even compared to transformer No. 1,
No. 4, so given Equation 7 we would expect that predominantly due to the leakage inductance
the leakage inductance would approximately increase from the lack of interleaving. For
double, due to the noninterleaving. transformer No. 5B, efficiency was 88.26 percent, or
Transformer No. 5B by comparison should have half 8.65 W of total loss. Despite a leakage inductance
the DCR as transformer No. 4, and slightly more similar to transformer No. 4, the total loss was
than half the leakage inductance due to the increase almost 0.3 W worse. Using large-diameter two-layer
in total winding layer heights, Σh, partly offsetting the primaries made the ACR factor worse, despite a
50 percent reduction due to interleaving. 50 percent reduction in DCR.
Analysis of transformer Nos. 5A and 5B Transformer No. 6

performance
In the previous transformer iterations, we varied the
We checked transformer Nos. 5A and 5B for primary between one- and two-layer constructions,
leakage inductance and efficiency performance in with 0.5-mm and 0.25-mm wire stands. The
the EVM572 board. 0.25-mm implementations resulted in better ACR
We measured the leakage inductance for factor and lower loss. But in all cases (except for
transformer No. 5A at 6.41 µH, approximately transformer No. 1), the secondary implementation
double the value of transformer No. 4, in keeping did not vary from the two-strand, triple-insulated,
with expectations due to noninterleaving. With 0.55-mm one-layer implementation. Based on the
transformer No. 5B, the leakage inductance findings for primary loss, maybe thinner secondary
dropped to 3.7 µH – again in keeping with wire with more strands would also
expectations, almost half compared to transformer yield improvement.
No. 5A due to the change to interleaving, but slightly To meet safety requirements, the secondary wires
higher than transformer No. 4 due to the increased must be triple-insulated, or you would need a
total winding layer height.

All tapes 60 mm
W3 secondary 7 x 0.2-mm twisted
bundle, triple-insulated
enameled copper
margin-wound construction. Given the significant We measured the leakage inductance at 3.24 µH,
reduction in winding-window widths for margin- pretty much identical to transformer No. 2,
wound constructions, triple-insulated wire is more as expected.
common. Using a multistranded secondary winding The full load efficiency at 65-W load, 115-VAC/
would result in far too much window lost to the 60-Hz input was 89.0 percent, or 8.03 W of total
0.2-mm triple-insulation thickness for each individual power loss (not including the output cable drop). This
strand. Triple-insulated pre-stranded bundles (such was somewhat disappointing because compared to
as that used on transformer No. 1) are the most transformer No. 2, the results are almost identical.
suitable option for a stranded secondary. The multistrand triple-insulated secondary did not
We implemented transformer No. 6 with the same deliver the hoped-for conduction-loss improvement.
primary, auxiliary and shield structure as transformer The total cross-section, and therefore the DCR, of
No. 2, combined with the stranded secondary transformer No. 6’s secondary is 8 percent worse
winding of transformer No. 1. Figure 45 shows than transformer No. 2’s secondary. Since the overall
the structure. efficiency result is similar, this tells us that despite the
Analysis of transformer No. 6 performance DCR dis-improvement, there is a similar magnitude of
You would expect transformer No. 6’s structure ACR improvement – but the two effects cancel out.
to have similar leakage inductance as transformer If we had implemented the multistrand secondary
No. 2. However, we hoped that transformer with the same total cross-section (for the same
No. 6 would exhibit improved secondary-winding DCR), then we should have seen an improvement.
conduction loss. However, in this case, there is not sufficient space
We checked transformer No. 6 for leakage to accommodate a triple-insulated secondary
inductance and efficiency performance in the with more strands.
EVM572 board.

Summary of transformer Efficiency Comparison by Transformer at 230 VAC

91%
optimizations 90%
TR#1
TR#2
TR#3
Table 2 summarizes the various results discussed 89%

TR#4
TR#5B
Efficiency
TR#5A
in this paper. Figures 47-48 graph the variation in 88%

TR#6
efficiency of each transformer across line voltage 87%
and load. 86%
85%
Leakage Full load Pdiss full 5 15 25 35 45 55 65
Transformer inductance efficiency at load at Output Power (W)
version (μH) 115 VAC (percent) 115 VAC (W)

1 4.51 87.71 9.11 Figure 47. Efficiency performance comparison of each transformer at
high line (230 VAC).
2 3.24 89.03 8.01
3 4.46 88.39 8.54 For the transformer windings, we showed the
4 3.29 88.58 8.37 importance of minimizing leakage inductance and
5A 6.41 86.94 9.76
ACR, and gave recommendations and approaches
5B 3.70 88.26 8.65
to improve performance. It is particularly important
6 3.24 89.00 8.03
Table 2. Performance comparison of various transformer to choose bobbin and core styles with large
constructions. breadth-to-height ratios appropriate for the required
turns counts in order to minimize total layer count.
Conclusions
Avoid partial layers and ensure that all layers are full
This paper highlighted many important design
and neatly wound.
considerations to achieve high-efficiency flyback
transformers for offline AC/DC applications. The work of Hurley [12] and Carsten [9] serve as
guides for choosing the best wire size and strand
In particular, we highlighted the importance of the
count, but multiple iterations will still be required in
effects of duty cycle and DC bias on core losses.
most cases to find a suitable compromise between
There has been (and still is) much ongoing research
cost/complexity and neat, full layers.
in this area. However, the available data from ferrite
core manufacturers continues to be inadequate, Finally, we showed that you can build very effective
and must improve so that designers can make more CM EMI reduction into transformer construction at
accurate and realistic estimates of core losses and the design stage, reducing system-level EMI filtering
see the genuine impact of operation at very small requirements and time-consuming EMI debugging
duty cycles. later.
Efficiency Comparison by Transformer at 115 VAC
91% Acknowledgements
90%
Thanks to our colleagues at Texas Instruments in
TR#1
89%
Efficiency
88%
TR#2 Cork, Ireland, for their invaluable contributions to
TR#3
87% TR#4 this paper: Joe Leisten, Peter Meaney, Billy Long,
86% John Griffin, Colin Gillmor and Seamus O’Driscoll.
85%
5 15 25 35 45 55 65
Output Power (W)
Figure 46. Efficiency performance comparison of each transformer at

low line (115 VAC).

References 9. Carsten, B. “High Frequency Conductor Losses in
1. Picard, J. “Under the Hood of Flyback SMPS Designs.” Switchmode Magnetics.” Technical Papers of the
Texas Instruments Power Supply Design Seminar First International High Frequency Power Conversion
SEM1900, 2010. 1986 Conference (1986) 155-176.
2. Maksimovic, D., and Erickson, R.W. Fundamentals of 10. Dowell, P.L. “Effect of eddy currents in transformer
Power Electronics. Springer, 2004. windings.” Proceedings of the Institution of Electrical

Engineers 113, No. 8 (1966) 1387-1394.
3. Mohan, N., Undeland, T.M., and Robbins, W.P. Power
Electronics: Converters, Applications, and Design. 11. Hurley, W.G., and Wölfle, W.H. Transformers and
John Wiley & Sons, 1995. Inductors for Power Electronics: Theory, Design and
Applications. John Wiley & Sons, 2013.
4. Sullivan, C.R., Abdallah, T., and Fujiwara,
T. “Optimization of a Flyback Transformer Winding 12. Hurley, W.G., Gath, E., and Breslin, J.G. “Optimizing
Considering Two-Dimensional Field Effects, Cost and the AC Resistance of Multilayer Transformer Windings
Loss.” APEC 2001 Sixteenth Annual IEEE Applied with Arbitrary Current Waveforms.” IEEE Transactions
Power Electronics Conference and Exposition (2001) on Power Electronics 15, No. 2 (2000) 369-376.
116-122. 13. Dixon, L.H. “The Magnetics Design Handbook for
5. Mu, M. “High Frequency Magnetic Core Loss Study.” Switching Power Supplies.” 2001.
PhD diss. Virginia Tech, 2013. 14. Dixon, L.H. “Transformer and Inductor Design for
6. Mu, M., and Lee, F.C. “A New Core Loss Model Optimum Circuit Performance.” Texas Instruments
for Rectangular AC Voltages.” 2014 IEEE Energy Power Supply Design Seminar SEM1500, 2002.
Conversion Congress and Exposition (ECCE) (2014) 15. Mammano, B., and Carsten, B. “Understanding
5214-5220. and Optimizing Electromagnetic Compatibility in
7. Mühlethaler, J., Biela, J., Kolar, J.W., and Switchmode Power Supplies.” Texas Instruments
Ecklebe, A. “Core Losses Under DC Bias Condition Power Supply Design Seminar SEM1500, 2002.
Based on Steinmetz Parameters.” 2010 International 16. Long, B. “Using the UCC28630EVM-572 User Guide
Power Electronics Conference (IPEC) (2010) 2430- (SLUUAX9).” TI User’s Guide, 2015.
2437.
8. Roshen, W.A. “A Practical, Accurate and Very General
Core Loss Model for Nonsinusoidal Waveforms.” IEEE
Transactions on Power Electronics 22, No. 1 (2007)
30-40.

TI Worldwide Technical Support
Internet
TI Semiconductor Product Information Center
Home Page
support.ti.com
TI E2E™ Community Home Page

e2e.ti.com
Product Information Centers

Americas Phone +1(512) 434-1560 Asia
Phone Toll-Free Number
Brazil Phone 0800-891-2616
Note: Toll-free numbers may not support
Mexico Phone 0800-670-7544 mobile and IP phones.
Australia 1-800-999-084
Fax +1(972) 927-6377 China 800-820-8682
Internet/Email support.ti.com/sc/pic/americas.htm Hong Kong 800-96-5941
India 000-800-100-8888
Europe, Middle East, and Africa Indonesia 001-803-8861-1006
Phone Korea 080-551-2804
European Free Call 00800-ASK-TEXAS Malaysia 1-800-80-3973
(00800 275 83927)
New Zealand 0800-446-934
International +49 (0) 8161 80 2121
Philippines 1-800-765-7404
Russian Support +7 (4) 95 98 10 701
Singapore 800-886-1028
Taiwan 0800-006800
Note: The European Free Call (Toll Free) number is not active in
all countries. If you have technical difficulty calling the free call Thailand 001-800-886-0010
number, please use the international number above. International +86-21-23073444
Fax +86-21-23073686
Fax +(49) (0) 8161 80 2045 Email [email protected] or [email protected]
Internet www.ti.com/asktexas Internet support.ti.com/sc/pic/asia.htm
Direct Email [email protected]
Important Notice: The products and services of Texas Instruments
Incorporated and its subsidiaries described herein are sold subject to TI’s
Japan standard terms and conditions of sale. Customers are advised to obtain the
most current and complete information about TI products and services
Fax International +81-3-3344-5317 before placing orders. TI assumes no liability for applications assistance,
Domestic 0120-81-0036 customer’s applications or product designs, software performance, or
infringement of patents. The publication of information regarding any other
Internet/Email International support.ti.com/sc/pic/japan.htm company’s products or services does not constitute TI’s approval, warranty
or endorsement thereof.
Domestic www.tij.co.jp/pic
A021014
The platform bar and E2E are trademarks of Texas Instruments. All other trademarks
are the property of their respective owners.
© 2016 Texas Instruments Incorporated

Printed in U.S.A. by (Printer, City, State) SLUP338
IMPORTANT NOTICE
Texas Instruments Incorporated and its subsidiaries (TI) reserve the right to make corrections, enhancements, improvements and other
changes to its semiconductor products and services per JESD46, latest issue, and to discontinue any product or service per JESD48, latest
issue. Buyers should obtain the latest relevant information before placing orders and should verify that such information is current and
complete. All semiconductor products (also referred to herein as “components”) are sold subject to TI’s terms and conditions of sale
supplied at the time of order acknowledgment.
TI warrants performance of its components to the specifications applicable at the time of sale, in accordance with the warranty in TI’s terms
and conditions of sale of semiconductor products. Testing and other quality control techniques are used to the extent TI deems necessary
to support this warranty. Except where mandated by applicable law, testing of all parameters of each component is not necessarily
performed.
TI assumes no liability for applications assistance or the design of Buyers’ products. Buyers are responsible for their products and
applications using TI components. To minimize the risks associated with Buyers’ products and applications, Buyers should provide
adequate design and operating safeguards.
TI does not warrant or represent that any license, either express or implied, is granted under any patent right, copyright, mask work right, or
other intellectual property right relating to any combination, machine, or process in which TI components or services are used. Information
published by TI regarding third-party products or services does not constitute a license to use such products or services or a warranty or
endorsement thereof. Use of such information may require a license from a third party under the patents or other intellectual property of the
third party, or a license from TI under the patents or other intellectual property of TI.
Reproduction of significant portions of TI information in TI data books or data sheets is permissible only if reproduction is without alteration
and is accompanied by all associated warranties, conditions, limitations, and notices. TI is not responsible or liable for such altered
documentation. Information of third parties may be subject to additional restrictions.
Resale of TI components or services with statements different from or beyond the parameters stated by TI for that component or service
voids all express and any implied warranties for the associated TI component or service and is an unfair and deceptive business practice.
TI is not responsible or liable for any such statements.
Buyer acknowledges and agrees that it is solely responsible for compliance with all legal, regulatory and safety-related requirements
concerning its products, and any use of TI components in its applications, notwithstanding any applications-related information or support
that may be provided by TI. Buyer represents and agrees that it has all the necessary expertise to create and implement safeguards which
anticipate dangerous consequences of failures, monitor failures and their consequences, lessen the likelihood of failures that might cause
harm and take appropriate remedial actions. Buyer will fully indemnify TI and its representatives against any damages arising out of the use
of any TI components in safety-critical applications.
In some cases, TI components may be promoted specifically to facilitate safety-related applications. With such components, TI’s goal is to
help enable customers to design and create their own end-product solutions that meet applicable functional safety standards and
requirements. Nonetheless, such components are subject to these terms.
No TI components are authorized for use in FDA Class III (or similar life-critical medical equipment) unless authorized officers of the parties
have executed a special agreement specifically governing such use.
Only those TI components which TI has specifically designated as military grade or “enhanced plastic” are designed and intended for use in
military/aerospace applications or environments. Buyer acknowledges and agrees that any military or aerospace use of TI components
which have not been so designated is solely at the Buyer's risk, and that Buyer is solely responsible for compliance with all legal and
regulatory requirements in connection with such use.
TI has specifically designated certain components as meeting ISO/TS16949 requirements, mainly for automotive use. In any case of use of
non-designated products, TI will not be responsible for any failure to meet ISO/TS16949.
Products Applications
Audio www.ti.com/audio Automotive and Transportation www.ti.com/automotive
Amplifiers amplifier.ti.com Communications and Telecom www.ti.com/communications
Data Converters dataconverter.ti.com Computers and Peripherals www.ti.com/computers
DLP® Products www.dlp.com Consumer Electronics www.ti.com/consumer-apps
DSP dsp.ti.com Energy and Lighting www.ti.com/energy
Clocks and Timers www.ti.com/clocks Industrial www.ti.com/industrial
Interface interface.ti.com Medical www.ti.com/medical
Logic logic.ti.com Security www.ti.com/security
Power Mgmt power.ti.com Space, Avionics and Defense www.ti.com/space-avionics-defense
Microcontrollers microcontroller.ti.com Video and Imaging www.ti.com/video
RFID www.ti-rfid.com
OMAP Applications Processors www.ti.com/omap TI E2E Community e2e.ti.com
Wireless Connectivity www.ti.com/wirelessconnectivity
Mailing Address: Texas Instruments, Post Office Box 655303, Dallas, Texas 75265
Copyright © 2016, Texas Instruments Incorporated

Uploaded by

Uploaded by

Flyback transformer design

considerations for efficiency

AC/DC power supplies widely use the flyback

Introduction often-neglected absorption of magnetizing energy

Texas Instruments 2 September 2016

inductance, LP. operation, see references [1], [2] and [3].

PRIM SEC VOUT PRIM SEC VOUT

PRIM SEC VOUT

Texas Instruments 3 September 2016

Texas Instruments 4 September 2016

length minimizes DCR.

Texas Instruments 5 September 2016

Flyback Waveforms at CCM/DCM Boundary offer equations that relate manufacturer-published

References [5], [6], [7] and [8] discuss in more

beyond the scope of this paper. Those authors

Texas Instruments 6 September 2016

Effect of DC bias where PV_SINE is the Steinmetz equation loss for

𝑃𝑃!_!"!#$ = 𝑃𝑃!_!"#$ ∙ 𝐹𝐹!"#$%&'( (𝛾𝛾, 𝐷𝐷) ∙ 𝐹𝐹!" (𝐻𝐻!" ) (2)

Texas Instruments 7 September 2016

probably incorrect. 5.0

Harmonic Current (A)

flyback transformers 3.0

Losses arise because of the DC component of

Texas Instruments 8 September 2016

Skin Depth vs. Frequency

Figure 9. Nonuniform AC current distribution due to induced eddy 0.0

Texas Instruments 9 September 2016

(a) (b) (c)

Figure 11 shows the AC current flow in a single

illustration, the wire diameter is much greater than

Texas Instruments 10 September 2016

adjacent layers and how the resulting current

layer. Once again, the wire diameter is much larger

Proximity effect – multiple layers

diameters, to illustrate proximity effect between in Figure 15.

Texas Instruments 11 September 2016

winding structure. flows, proximity effect-induced eddy currents can

Texas Instruments 12 September 2016

ACR factor Dowell’s equations apply to sinusoidal waveforms

Texas Instruments 13 September 2016

Note that the graphs in Figure 17 show the ACR/

Texas Instruments 15 September 2016

Texas Instruments 16 September 2016

dissipated external to the transformer. Several of the h1 h2 h3

leakage inductance converts to an equivalent rectangular block of the

Following are the parameters of Equation 8: d h=d*0.88

Texas Instruments 17 September 2016

m=1 m=2 m=3

MMF MMF MMF

Figure 22. Different levels of interleaving to reduce leakage inductance.

Leakage inductance reduction and to a flyback transformer during the transition

Interleaving cases, the MMF builds to a peak and declines

Texas Instruments 18 September 2016

Since leakage inductance is inversely proportional Leakage inductance estimation –

Figure 23. Interleaved flyback transformer winding structure.

Texas Instruments 19 September 2016

In Equation 10, 2.71 µH is reasonably close to Q

losses Figure 26b, where the clamp voltage is 50 percent

Texas Instruments 20 September 2016

Figure 27 shows the percentage of magnetizing

EMI shielding and cancellation

secondary. ICM will then flow to earth through the

Texas Instruments 21 September 2016

The shield typically connects back to local primary

In many applications, the output return (negative

Texas Instruments 22 September 2016

Figure 30. Flyback transformer with additional auxiliary cancellation winding.

Texas Instruments 23 September 2016

Figure 32. Flyback transformer winding structure with CM balance.

A shield layer (purple) is placed at the other interface sides, so it is CM-balanced.

Texas Instruments 24 September 2016

Figure 34. UCC28630EVM572 transformer winding construction (transformer No. 1).

Texas Instruments 25 September 2016

Between the inner half primary and the secondary is

The secondary uses a twisted bundle of seven

Texas Instruments 26 September 2016

Texas Instruments 27 September 2016

Figure 37. Transformer No. 2 winding construction.