Sie sind auf Seite 1von 152

Inductance factor calculation

Inductance Calculations - Ferroxcube Standard Cores


A

F
G
H
I
J

H Click Datasheet to view the data sheet of


the selected core type.

A Select a Core Family from our standard range.


B Available sizes are displayed, make your selection.
C Available materials for the chosen size are
displayed. When you select one, its permeability
and the usual surface quality are shown.
D You can also define your own material by directly
entering its features.

Click User defined cores if you want to


design your own core.

Effective permeability is caculated based on


material permeability, total airgap and its position.

K Effective core parameters for the selected


core are shown.

E Flux density (Bpeak) is calculated based on a given


voltage and number of turns.

L Inductance factor (AL) is caculated based on


material permeability, total airgap and its position.

F When a bias current is entered, the


resulting magnetic bias field (Hbias) is calculated.
The bias flux density (Bbias) in the core is
estimated with the help of the effective
permeability (e).

M Inductance value (L) is caculated based on AL


and number of turns. These parameters can
be changed freely to optimize the design.
If required the number of turns can be fixed.

G Click Go to start a new calculation,

Inductance factor calculation

Using the program Inductance Factor Calculation


Functional description
Ferroxcube Standard Cores
The program consists of two parts. The first screen appears when you click the button Inductance factor
Calculations and can be used for calculations on standard Ferroxcube Cores.
The second program can be activated by clicking the button User Defined in the lower right corner of the
window.
User Defined Cores
This program calculates effective magnetic dimensions of user defined, magnetically closed core shapes.
With these data, the properties of the chosen ferrite material and airgap length, the inductance factor AL
of the core is derived. With the number of turns the inductance value of an inductor based on this core
is calculated. For magnetically open circuits like rods the so-called rod permeability is caculated. With
the number of turns and the position of the winding with respect to the rod the inductance of the coil is
predicted.
The program can handle a variety of core shapes. The principal shapes are : Toroid (ring core), U core, E
core, P core and rod. Several other core types are variations on the principal shapes so that the same
formulas can be used. The core dimensions are entered with the help of a dimensional drawing in the
core type window. In the case of cores with known effective dimensions, these dimensions can be entered
directly without the need to specify the whole geometry.
After the calculation of magnetic dimensions, a menu appears for the choice of the inductor calculation.
For a magnetically closed core type this can be AL value, inductance, flux level, field strength or inverse
calculations. Every core shape, including ring cores, can be gapped. The influence of fringing flux and parasitic
gaps is accounted for. In case of a rod the calculation can be inductance, or an inverse calculation. An AL
value is not appropriate here, because the inductance also depends on coil length and not only on the rod
and the number of turns.

Further remarks:
If entered values are out of range, a warning message box will displayed. After changing a value, a
recalculation can be triggered by clicking the Go button or by pressing <ENTER> in the updated text field.
Calculations can be saved in a file and later be opened again with the Save en Open options in the File
menu. Results can be printed with the Print option in File menu. Core sizes are always listed in ascending
order of effective volume Ve. In cases where the calculated gap would become smaller than the parasitic gap,
the parasitic gap length is taken as default.

Inductance factor calculations for User Defined Cores

E
F

H
I
J

K
A Select a Core Shape from the list according
to the design you want to make.

calculated. The bias flux density (Bbias)


in the core is estimated with the help
of the effective permeability (e).

B Enter its main dimensions using the outline


drawing as a guide.

H Click Go to start a new calculation,

C Use direct input when effective core


parameters le and Ae are known.
Choose between a closed core or 2 halves.
Enter the type of airgap and its position.
The core window height is required for the
estimation of the fringing flux.

Click Ferroxcube Standard Core to


switch back to the standard core ranges.

Effective permeability is caculated based on


material permeability, total airgap and its
position.

D Available materials for the chosen shape


are displayed. When you select one, its
permeability and the usual surface quality
are shown.

K Effective core parameters for the selected


core are shown.

E You can also define your own material by


directly entering its features.
F Flux density (Bpeak) is calculated based on,
core parameters, a given voltage and the
number of turns.
G When a bias current is entered, the
resulting magnetic bias field (Hbias) is

L Inductance factor (AL) is calculated based


on material permeability, total airgap and its
position.
M Inductance value (L) is calculated based on
AL and number of turns. These parameters
can be changed freely to optimize the
design. If required the number of turns can
be fixed.

Inductance factor calculation

Definition of terms
Permeability
When a magnetic field is applied to a soft magnetic material, the resulting flux density is composed of that
of free space plus the contribution of the aligned domains.

B = 0 H + J or B = 0 ( H + M )

[1]

where 0 = 4.10 -7 H/m, J is the magnetic polarization and M is the magnetization.


The ratio of flux density and applied field is called absolute permeability.

B
M
---- = 0 1 + ----- = absolute
H
H

[2]

It is usual to express this absolute permeability as the product of the magnetic constant of free space and
the relative permeability (r).

B
---- = 0 r
H

[3]

Since there are several versions of r depending on conditions the index r is generally removed and
replaced by the applicable symbol e.g. i, a, etc.

Initial permeability
The initial permeability is measured in a closed magnetic circuit (ring core) using a very low field strength.

B
1
i = ------ ------- 0 H

( H 0 )

Initial permeability is dependent on temperature and frequency.

[4]

Inductance factor calculation

Effective permeability
If an airgap is introduced in a closed magnetic circuit, magnetization becomes more difficult. As a result, the
flux density for a given magnetic field strength is lower.
Effective permeability is dependent on the initial permeability of the soft magnetic material and the
dimensions of airgaps and circuit.

i
e = -------------------------G i
1 + ----------------le

[5]

where G is the gap length and le is the effective length of magnetic circuit. This simple formula is a good
approximation only for small airgaps. For longer airgaps some flux will cross the gap outside its normal area
(fringing flux) causing an increase of the effective permeability.

Amplitude permeability
The relationship between higher field strength and flux densities without the presence of a bias field, is
given by the amplitude permeability (a).

B
1
a = ------ ---0 H

[6]

Since the BH loop is far from linear, values depend on the applied field strength.

Incremental permeability
The permeability observed when an alternating magnetic field is superimposed on a static bias field, is called
the incremental permeability.

1 B
= ------ ------- 0 H H DC

[7]

If the amplitude of the alternating field is negligibly small, the permeability is called the reversible
permeability (rev).

Inductance factor calculation

Effective core dimensions


To facilitate calculations on a non-uniform soft magnetic core, the effective dimensions are given on each
data sheet. These dimensions, effective area (Ae), effective length (le) and effective volume (Ve) define a
hypothetical ring core which would have the same magnetic properties as the non-uniform core.
The reluctance of the ideal ring core would be:

le
----------------- Ae

[9]

For the non-uniform core shapes, this is usually written as:

l
1
------ ---A
e

[10]

the core factor divided by the permeability.


The inductance of the core can now be calculated using this core factor:
2

9
2
0 N
1.257 10 N
L = ----------------------- = -------------------------------------------------- (in H)
l
1
l
1
------ --------- ---A
e
A
e

[11]

The effective area is used to calculate the flux density in a core, for sine wave:

U 2 10
2.25U 10
B = ------------------------------ = ---------------------------------- (in mT)
A e N
fNA e

[12]

for square wave:

0.25U 10
B = ---------------------------------- (in mT)
fNA e

[13]

Inductance factor calculation


where:
Ae is the effective area in mm2 .
U is the voltage in V
f is the frequency in Hz
N is the number of turns.
The magnetic field strength (H) is calculated using the effective length (Ie):

IN 2
H = -------------- (A/m)
le

[14]

If the cross-sectional area of a core is non-uniform, there will always be a point where the real cross-section
is minimal. This value is known as Amin and is used to calculate the maximum flux density in a core. In well
designed ferrite core a large difference between Ae and Amin is avoided. Narrow parts of the core could
saturate or cause much higher hysteresis losses.

Inductance factor (AL)

To facilitate inductance calculations, the inductance factor, known as the AL value (nH), is given in each data
sheet. The inductance factor of a core is defined as:

L = N A L (nH)

[15]

The value of AL is calculated from the core factor and the effective permeability:

1.257 e
0 e 10
A L = ------------------------------ = --------------------- (nH)
( l A)
( l A)

[16]

Inductance calculations on rods and tubes


Rods and tubes are generally used to increase the inductance of a coil. The magnetic circuit is very open and
therefore the mechanical dimensions have more influence on the inductance than the ferrites permeability
(see Fig.1) unless the rod is very slender. In order to establish the effect of a rod on the inductance of a coil,
the following procedure should be carried out:
Calculate the length to diameter ratio of the rod (l/d).
Find this value on the horizontal axis and draw a vertical line.
The intersection of this line with the curve of the material permeability gives the effective rod permeability
(rod).

Inductance factor calculation


The inductance of the coil, provided the winding covers the whole length of the rod is given by:

N A
L = 0 rod ----------- ( H )
l

[17]

where:
N = number of turns
A = cross sectional area of rod (mm2)
I = length of coil. (mm)

i = 10.000

103

5000
2000
1000
700
500

rod

400
300
200
150
102

100
70

40

20

10

10

10

Length / diameter ratio

100

Fig.1 Rod permeability (rod) as a function of length to


diameter ratio with material permeability as a parameter.

Transformer Core Selection

Transformer Core Selection


B

E
G
H
C

F
A

E
Enter the design parameters. Take a reasonable
value for the maximum throughput power, for
instance 2 times the minimum required power, to
limit the choice of suitable core sizes.

Click Go to start a new calculation.


F
The core types capable of handling the required
throughput power are displayed in this window.

B
Make a choice of converter type. Enter the
required creepage distance between windings,
expected copper fill factor and duty cycle, or
accept the default values.

G
Click Graph to view a graph of throughput
power versus frequency for the selected core
type.

H
Check the core families you wish to consider for
your design.

Click Datasheet to view a data sheet of the


selected core type.

D
Select the material you wish to apply or all to
leave the choice up to the program.

Transformer Core Selection

Transformer Core Selection


Introduction
This program can be used to select ferrite cores from the actual product range of FERROXCUBE for the
design of a forward or flyback transformer for use in a switched mode power converter. The user is asked
to fill in the required power level, expected operating conditions and a number of other variables. Then he
chooses the core families and the ferrite materials he wants to consider.
The program calculates the expected throughput power for all chosen cores and displays the cores that
fulfil the criteria in the result window. The background of these calculations and the formulas used are
explained in the following paragraphs.
For the core set(s) a graph showing the losses as a function of frequency can be generated. This can be
useful for the optimization of the converter design. When the choice is made, the data sheet of the product
is displayed on request.

Input data
The requested input data, grouped in 3 sets, can be inserted arbritrary order.

1. Application conditions
Description

Symbols (used in calculations)

Minimum throughput power


Maximum throughput power
Converter operating frequency
Ambient temperature
Allowed temperature rise
Converter type
Creepage distance
Copper fill factor
Effective duty factor
Preferred airgap

Pthrmin
Pthrmax
f
Tamb
T
Forward, Flyback or Push-Pull
Cr *)
Fw

lg

*) For all core families, except toroids (T, TN, TX, TL, TC), the total creepage distance is the sum of the creepage
distances between winding and the ferrite core. If one choses e.g. a value of 8 mm it is assumed that the distance
between the primary and secundary winding package and the ferrite are 4 mm each. For toroids the creepage distance is the sum of both insulation distances between the primary and secondary winding.

2. Selection of core family(ies)


The search should be limited to those families one wants to consider to speed up the proces.

3. Selection of power ferrite(s)


The search can be limited to those materials one wants to consider to speed up the proces or the option
all can be checked to leave it up to the program.

Transformer Core Selection

Background of the calculations


1.Calculation of througput power
(example forward converter)
In general the throughput power can be calculated as a time average of the product of voltage and current.
Applying this to the pulse shaped input voltage Vp and current Ip with duty cycle (active part of period)
gives :
T

Pthr =

1
T

(Vp Ip)dt = Vp Ip

[1]

The relation of voltage Vp and flux density B is given by Faradays law of induction. Using this for the applied
signals and keeping in mind that 1/T equals the frequency f, and Np stands for the number of primary turns,
this results in:

Vp

= Np Ae

fB
B
dB
= Np Ae
= Np Ae

dt

[2]

Note that B in eq.[2] is referring to the total flux excursion, So, from its minimum to its maximum value.
Substituting eq.[2] for Vp in formula [1] shows that throughput power is proportional to the product of
(f.B), which is interpreted as the material performance factor.

[3]

Pthr = Np Ae fB Ip

Next the current Ip will be worked out in terms of winding losses. For this purpose it is necessary to
consider the effective (or rms) current. This current (Ieffective) is related to power losses in a resistance R by
P = I2R. The effective current for a time varying current I(t) with period T is defined in general as:
T

I2

effective

1
T

I 2 (t) dt

[3A]

Transformer Core Selection

Applying this definition to the appropriate current signals we get the following results:

Ipeffective = Ip

[4]

Relating the current Ip to the primary winding losses Pw.pr and primary resistance Racpr gives:

Ip =

Ipeffective

Pwpr

[5]

Racpr

The a.c. resistance of the windings is related to the d.c. resistance by a constant factor FR which is equal
to 1 for a d.c. current. The increase of the a.c. resistance value is caused by the skin and proximity effects.
Normally these effects strongly depend on the arrangements of the windings in the available winding space
and on the frequency of the signals. In literature, methods to calculate the FR value are published (e.g. based
on the theory of Dowell). For these calculations computer programs can be helpfully.
The d.c. resistance Rdc is equal to: Rdc = .Np.Lav/Acopper.
is the copper resistivity in m
Lav is average winding turn length
Acopper is the cross-sectional area of the copper wire
Acopper can be expressed in terms of a copper fill factor Fw, which is defined as total area of
copper divided by total winding area, or in symbols : Fw = N.Acopper / Aw
This gives : Acopper = Fw.Aw / N.
Using the quantity FR and the expression for Acopper one can write for the the a.c. resistance in:

Racprim = FR Rdc =

FR Np lav
Acopper

FR Np lav
Fw Awprim

[6]

Substituting formula [6] in formula [5] for the current Ip gives:

Ip =

Np

Pwprim Fw Awprim

lav FR

[7]

Transformer Core Selection

In transformers the following rules apply :


Np.Ip = Ns.Is
Pw.pr (Np.Ip)2 and Pw.s (Ns.Is)2.
Also it is assumed that :
Pw.pr Pw.s Pw, where Pw is the total winding losses.
Fw.Aw.pr Fw.Aw.s Fw.Aw
Applying this to [7] gives the primary current in terms of the total winding losses and total area:

Ip =

1
2

Np

Pw Fw Aw

lav FR

[8]

If equation [8] for Ip is substituted in [3] the following expression for the throughput power is obtained:

Pthr =

1
2

Ae (fB)

Pw Fw Aw

lav FR

[9]

The same formula is valid for a flyback transformer, however the derivation is somewhat more complicated
due to the triangular current shapes. Due to the influence of the airgap (fringing flux) the throughput power
is about 90 % compared to a forward concept.
For a standard push-pull transformer it can be derived that the throughput power is a factor 2 higher
compared to the forward transformer. For transformers used in other converter topologies only the factor
might change. So in general the througput power factor is equal to:

Pthr Ae (fB)

Pw Fw Aw

lav FR

[10]

Transformer Core Selection

2. Determination of the optimum flux density.


The total loss in a transformer and its temperature rise are related by an analogon of Ohms law:

P trafo =

T
R th

[11]

Measurements on wire wound transformers resulted in an empirical formula for the thermal resistance
Rth = 1/ (Cth.Ve 0.54 ).
The constant Cth depends on core shape and winding arrangement. For standard wound cores like RM or
ETD and EFD it is about 17. For planar cores with windings integrated in a PCB (epoxy) it is about 24, and
for flat frame and bar cores used for LCD backlighting the value is about 20.
The total transformer losses (in mW) can be related directly to its allowed temperature rise T and
effective volume Ve:

0.54

P trafo = T Cth Ve

[12]

The power loss density (mW/cm3) in a core can be determined with the following empirical fit formula:
y

x
Pcore = C m f Bac ct ct 1 T + ct 2 T = C m C (T) f Bac

[13]

Note that Bac is here the peak flux density, so half the total flux excursion .(Bmax Bmin).
The parameters Cm, x, y, ct, ct1 and ct2 are specific for each power ferrite and often only applicable in a
defined frequency range. Total transformer loss consists of winding loss plus core loss. So, winding loss can
be written as:

0.54

Pw = P trafo Pcore = T Cth Ve

Cm C (T) f B Ve

[14]

If eq. [14] is substituted in equation [10] the result is a long formula with many terms. Denoting some terms
in advance makes it easier to calculate:

Transformer Core Selection

C1 =

Ae f

Fw Aw
FR lav

0.54

Ve Fw Aw
FR lav

C2 = T Cth Ve

C3 = Cm C (T) f

[15]

Substituting formula [14] in [10] and using the denoted terms [15] gives:

Pthr (Bac) = C1 Bac

(C2 C3 Bac)

[16]

The throughput power is now expressed as a function of Bac only because the other terms are considered
as constants.
Remark:
Bac is the peak flux density which is half the value of B as used in equation [10].
The value of Bac for which Pthr is maximum, called the optimum flux density, can be calculated by solving:
d[Pthr(Bac)]/dBac = 0.
Using formula [16], the result for which this differential quotient equals zero is given by:
y

x
0.54
2
T Cth Ve = Cm C (T) f Bac Ve
2+y

[17]

From expression [17] it can be seen that:

Pcore =

2
Ptrafo
2+y

and/or

Pw =

y P
trafo
2+y

[18]

Transformer Core Selection

By isolating Bac from [17] and noting that B = 2.Bac we find for the optimum B:

Bopt = 2

2
2+y

) (
1/y

x/y
0.46/y
T Cth 1/y
f
Ve
Cm C(T)

Bopt is expressed as a function of :


6 fit parameters for the ferrite Cm, x, y, ct, ct1 and ct2
Allowed temperature rise T
Constant Cth (17 for wire-wound and 24 for planar transformers)
The effective core volume Ve
The switching frequency f

[19]

Transformer Core Selection

3. Calculation of throughput power for forward and


push-pull transformers
The calculations are done for all the combinations of cores and ferrites which are given as input by the
selection on core family(ies) and type of ferrite(s).
For each core the required information as e.g. Ae,Ve, lav, Aw is read from a data base, which is also done for
each ferrite with the constants Cm, x, y, ct, ct1 and ct2.
The first step is the calculation of the optimum flux density Bopt with [19] for all the core / ferrite
combinations. The required parameters can be read from the data bases or are known because of the input
on the application condition.
The second step is the calculation of Ptrafo with [12] followed by the calculation of Pw with equation [18].
The third step is the inserting of the calculated values of Bopt and Pw in [10] to calculate the throughput
power factor.The copper resistivity is used as temperature dependent (T) in the following way:

T = 393 . 10-5

T =

1 + T ( T-20)
1 + T 76

(T) = (T=100) T

[20]

For (T=100) the value of 2.24 .10-8 m is used


Dependent on whether it is a forward or a push-pull transformer a multiplying factor of 2/ or 22/
is used respectively to calculate the throughput power. If the value for a ferrite core is outside the range of
the inserted values for minimum or maximum throughput power it will not be mentioned in the range of
the determined available cores.

4. Calculation of throughput power for flyback transformers


The operating principle of a flyback transformer differs from that of a forward or push-pull type. In a flyback
transformer magnetic energy is stored during the active part of the duty cycle. The energy is transferred to
the secondary side during the off period of the duty cycle.
The energy density (J/m3) of a magnetic field is: E = .B.H
The amount of energy Eind ( J ) stored in an inductor is:

Eind = B H Ve =

B2
Ve
2

[21]

Transformer Core Selection

Writing out eq.[21] for the ferrite part (Ve = Ae.le ) and the airgap (VLg = Ae.lg) gives:

Eind =

B2 Ae le + B2 Ae lg = B2 Ve
2 0 a
2 0
2 0

( 1

lg
le

[22]

The throughput power Pthr = Eind/T = Eind . f becomes with [22]:

Pthr(lg) =

B2 f V e 1 +
a
2 0

lg
le

[23]

The fringing flux has the effect of increasing the airgap area. The reluctance Rg, which is equal to lg/.Ag,
decreases by the fringing flux factor F. Then Rg becomes lg/(.Ag.F). For the calculations it is more
convenient to keep a constant cross-section Ae for the total circuit. Then the effect of the fringing flux on
the effective airgap lg is accounted for by deviding lg by a factor F.
A formula for the fringing flux factor F can be derived with conformal transformations and is equal to:

F=1+

lg
Ae

In

( 2 l Wh )

[24]

where Wh is the height of the winding window.


The throughput power for a flyback inductor choke with fringing is equal to:

Pthr(lg) =

B2 f Ve 1 + lg/F
a
le
2 0

[25]

The throughput power capability of a ferrite core is (for a flyback) equal to:

Pthr =

1
2

Ae (fB)

Pw Fw Aw

lav FR

Expression [9] can be written out in three steps by:


Substituting for Pwinding in eq.[18] : Pwinding = [y/(2+y)].Ptrafo
Substituting for Ptrafo in [18] : Ptrafo = T/ Rth
Substituting for B in [9] the optimum value from eq. [19]

10

[9]

Transformer Core Selection

The result of these substitutions gives for the throughput power capability:

Pthr =

( C 2C )
m (T)

1/y

( )

( ) ( )

(1x/y) (1/y+)
(0.270.46/y) Aw
Cth (1/y+)
Ae Ve
f
T

2+y
Lav

Fw
[26]
FR

Remark :
For a better understanding of this long formula it is useful to have a closer look at its different parts.
1. Circuit dependent part
2. Ferrite dependent part, important is to see how Pthr depends on:
- frequency : Pthr f (1 x/y)
- temperature : Pthr T(1/y + )
3 Core dependent part, known as the core performance factor
4 Winding dependent part
The throughput power in eq. [9] or [26] can, in theory, be equal to the value from eq.[25] when the airgap lg
is given its maximum size. The maximum airgap can be calculated by reworking [25] and using Pthr from eq.
[9] or [26] denoted now as Pthrmax:

lgmax =

( A 2 B f )
0

Pthrmax

le
a

( )

[27]

Remarks :
1. Fringing factor F is ignored
2. The maximum airgap is in principle strongly dominated by the allowed temperature rise T, and is
almost lineair with T.
3. If a larger airgap than lgmax is used the allowed temp. rise T will be exceeded. Magnetically this
means that the transformer is not operating with the optimum flux density as stated in [19]. Larger
airgaps lead to a higher field strength H, which means higher currents and winding losses. In this case
the optimum equilibrium between core and winding losses is disturbed.
4. Even if lgmax is used for the airgap, Pthr will always be somewhat lower (10 to 15 %) than calculated
with eq. [26], due to fringing effects, and additional losses in the windings due to fringing flux.
The sequence to calculate Pthr for a flyback transformer is:
The steps as mentioned in chapter 3; The calculation of Pthr for forward and push-pull
transformers. The values obtained from eq. [10] are interpreted as Pthrmax.
The calculated values for Bopt are used to calculate values for a. For this purpose the graphs a(B) from
our Data Handbook, are approximated by 4 straight lines for each power ferrite.
For each selected ferrite core Pthrmax and other data is used in eq.[27] to calculate lgmax. These are the
maximum values that can be used for the airgap.
The program uses eq.[25] to calculate Pthr for a certain airgap lg. If the desired value for the airgap is too
large for certain cores, automatically the value lgmax is used.

11

Transformer Core Selection

5. Option of Graph Pthr(f) for forward and push-pull transformers


Select one of the displayed cores and click on the button Pthr(f) for a graph of througput power as a
function of frequency.
The graph is generated by stepping through the following procedure at several frequencies in the operating
range of the applied ferrite.
Optimum flux density Bopt with eq. [19]
Calculation of Ptrafo with eq.[12]
Calculation of Pwinding with eq.[18]
Inserting the calculated values of Bopt and Pwinding in eq.[10] to calculate the throughput power
and finally multiplying it with 2/ for forward or 22/ for push-pull transformers.

6. Option of Graph Pthr(f) for flyback transformers


For this calculation it is assumed that for the selected core:
Airgap lg is constant
Number of turns is fixed
f and can change (to f and ) but, they have to be balanced that new values B and I are
ensuring still Pcore + Pwinding = Ptrafo. So the control loop has to ensure that f and are
kept in balance.
Suppose a transformer was optimized for nominal frequency fn which fixed Np, Lg and n (usually 0.5).
This can be written as:

Pcore (fn, n) + Pw (fn, n) = Ptrafo

[28]

Consider the following period T for the flux density:


From t = 0 to prim the flux increases from 0 to Bmax.
From t = prim to sec the flux decreases from Bmax to 0.
From t = ( prim + sec) to T, the flux remains 0.
The following assumptions are than usually justified:

prim sec

and

1
=
2T

f
2

feff =

1 2

[29]

[30]

12

Transformer Core Selection

at fn:

[31]

2n 1

Using equations [29], [30] and [31] leads to:

Pcore (f, ) = const. feff By (1-d) = const.

f x By 2
2

( )

Pw (f, ) = const. I2 (1-d) = const. I2 2

[32]

[33]

Airgap lg and Np are already fixed. And because V is fixed and since V = LdI/dt ~ dB/dt the slopes I(t) and
B(t) are fixed as well:
For f, the slope B(t) = B(f,)/T
For fn, n the slope B(t) =B(fn,n)/nTn

B (f, ) = B (fn, n) T
= B (fn, n)
nTn

( f nfn)

[34]

Using expression [34] for B(f,) in eq.[32] to find the core losses Pcore(f,) gives:

Pcore (f, ) = const.

f x B (fn, n) fn y 2
2
f n

( ) [

)]

[35]

Equation [35] can be expanded to eq.[36]:

x
y
x
x
y
Pcore(f,) = const. fn B (fn,n) 2n f 2n .fn 2
fn
n.f
2n
2
2n

( )

( ) ( ) ( ) ( )

Pcore (fn, n)
By using the substitution for Pcore(fn ,n) in [36] it can be written as:

13

[36]

Transformer Core Selection

Pcore(f,) = Pcore (fn, n)

y-x

fn
f

( )

1-x+y

( )

[37]

The winding losses can be expressed in the same way. Since dI/dt dB/dt the function for I(f ,) is similar to
B(f,) in equation [34]:

I (f,) = I (fn, n)

( f

fn
n

[38]

Using expression [38] for I(f,) in eq. [33] to find the winding losses Pw (f,) gives:

Pwe(f,) = const. I (fn, n)

fn
n
f

)]

[39]

Equation [39] can be reworked to eq. [40] and [41]:

Pwe(f,) = const. I (fn, n)2 2n

Pwe(f,) = Pw (fn, n)

( n ) ( fnf )
3

fn 2
f

( ) ( )

[40]

[41]

For the nominal case (fn ,n) as well as for another (f ,) this yields:

Pcore(fn,n) + Pw(fn,n) =

2 P
( 2+y
)

(y )

trafo + 2+y Ptrafo = Ptrafo

Pcore(f,) + Pw(f,) = Ptrafo

14

[42]

[43]

Transformer Core Selection

Substituting equations [41] and [37] in eq. [43] and comparing with eq. [42] gives:

fn y-x 1-x+y P (fn, n) 3


+ w
n
n
f

fn 2
=
f

( ) ( )
( ) ( )
[44]
3
2
y-x
1-x+y
y
fn
2
( 2+y
) ( fnf ) ( ) Ptrafo + ( 2+y) ( n ) ( f ) Ptrafo= Ptrafo
Pcore (fn, n)

From equation [44] follows that:

2+y

y 3
fn y-x 1-x+y
+
n
2+y
n
f

( ) ( ) ( )

fn 2= 1
f

( ) ( ) ( )

[45]

This equation can only be solved numerically. The parameters y, fn and n are known. Frequency f has to be
chosen and can be calculated (or approximated).
Using eq. [23] and [34], the throughput power as a function of f and can be written as:

2
2
2
Pthr(f,) = const f B(f,) = const f B(fn,n) fn
f n
2
= Pthr(fn,n) fn
n
f

[46]

( )( )

The duty cycle is always 0.5. It is assumed that the nominal duty factor n has the optimum and
maximum value of 0.5 at the nominal frequency fn. Two situations can be distinquished for the new
frequency f:
f > fn
In this case the new duty cycle cannot increase above n but will stay on the maximum value of n = 0.5
Using eq. [23] and [46] shows that in this case we can write the new flux density B(f,) and Pthr(f,) as:

( fnf )
Pthr(f,) = Pthr(fn,n) ( fn )
f
B(f,) = B(fn,n)

[47]

15

Transformer Core Selection

f < fn
For this situation the new duty cycle can be lowered by the control circuit with respect to the maximum
nominal value of n = 0.5. The new duty cycle can be calculated by solving numerically equation [45]. By
using eq. [23] and [46] again with n = 0.5 the new flux density B(f,) and Pthr(f,) can be written as:

B(f,) = B(fn,n) 2 fn
f

( )

[48]

Pthr(f,) = Pthr(fn,n) (2)2 fn


f

( )

Equations [47] and [48] are used to calculate Pthr(f) for a flyback transfomer.
A rough estimation can be made to give some insight in the curve of Pthr as a function of frequency.
The left term in eq [45] is constant if:

fn
f

( ) ( )
=

yx
1x+y

fn (0.45...0.60)
f

( )

[49]

The right term in eq. [45] is constant if :

fn
f

( ) ( )
=

2
3

fn
f

( )

0.6

[50]

So, the solution is:

( n ) ( fnf )

0.6

[51]

For a given f, the is adapted to:

n = n

fn
f

( )

0.6

[52]

Using [52] in eq. [46] the approximation for Pthr in the situation : f > fn can be seen as follows:
cannot increase above n. So, Pthr which is a linear function of (f.B2) decreases. Because B decreases with
1/f also Pthr(f) will decrease with 1/f.

16

Transformer Core Selection

Using [52] in [48] the approximation for Pthr in the situation: f < fn is:

Pthr(fn,n) = Pthr(fn,n)

fn
f

( )

0.2

[53]

So, for frequencies lower than nominal, Pthr(f) increases with about approximately f 0.2 up to the maximum
for Pthr(f) at fn. For frequencies higher than nominal, Pthr(f) decreases with 1/f.

17

Inductor design

Inductor Design
D

A
B
G

J
I
L

I
Choose the type of inductor to design.

Press Go button to start the calculation.

J
Choose the type of coil former to be used.

Shows the minimum core size suitable for the


present design complete with effective core
parameters.

C
Choose the type of magnetic circuit to be used for
the design.

Overview of all relevant properties of the nished


design.

Fill in the design parameters and limitations.


E

L
Make a choice from the available core shapes.

Press Datasheet button to open PDF le of the


data sheet of the resulting core type.

F
Choose a ferrite material from the list.
G
Shows the effective initial permeability for this
core / ferrite combination.
H
Shows the usual surface nish for this core / ferrite
combination.

Inductor design

Inductor design
Introduction
The design of an inductor is often done by trial and error. There are many parameters to play with. Current
and inductance value are usually xed, but there is a choice of core types, materials and airgap lengths to
realize the design with.
This program greatly simplies the design procedure by checking all cores in a range for magnetic saturation
and overheating. Then it proposes the smallest possible core size tting all boundary conditions for the
design.
In common-mode chokes, the 2 windings generate counteracting uxes of the same magnitude. The
resulting ux is practically zero, there is no danger of saturation so airgaps are not required.
However, because of the safety isolation requirements there is a certain distance between both windings.
This allows some stray ux to escape from the core before passing through the other winding. This stray
ux is not compensated and will cause saturation at very high current levels. The program accounts for this
effect and proposes the smallest core for the design.

Design procedure
Inductor design
The program starts by checking the cores from the chosen range for energy storage capability using the
following relations:
2

Energy stored: E = Imax L

[1]

e 0 E
Ve min = -------------------------2
Bsat

The saturation ux density Bsat depends on the ferrite material and the temperature of the core. Initially,
the ambient temperature is taken. The effective permeability e is equal to the material permeability i in
case of a closed core without mating faces or airgaps. In a gapped core it is a variable, mainly controlled by
the length of the airgap.
The iteration start is different for closed and gapped cores.
Closed core
e = i, so Ve.min follows from E and denes the rst core of the range to check.
Gapped core
The iteration is started with the smallest core of the range, so e follows from E and the corresponding
core volume Ve.
When a core with the required Ve is found, the winding is designed by rst calculating the number of turns
N from :

Inductor design
2
L = e 0 N Ae
le

[2]

Then the conductor length (lcopper) and cross-section (Acopper) are derived from the available winding
window (Aw), copper ll factor (Fw) and the average turn length (lav).

Acopper = Fw Aw
N

[3]

lcopper = lav N

This gives the total winding resistance :

R =

lcopper
Acopper

[4]

Since the resistance of copper is a function of temperature, the program takes that into account with the
following formula :

= 20C [ 1 + ( Toperating 20 )]

[5]

where = 0.0039 / C.
The winding losses can now be calculated with :

Pwinding = Ieff R

[6]

Usually this loss is the predominant cause of heat dissipation in the inductor. The temperature rise of the
inductor design is calculated from :

T = Toperating - Tamb = Pwinding


Rth

[7]

Rth is the thermal resistance of the inductor which is a function of Ve of the ferrite core and is calculated
with the following empirical equation:

Inductor design

Rth =

1
Cth . Ve

[8]
0.54

The constant Cth depends on core shape and winding arrangement. For standard wound cores like RM or
ETD and EFD it is about 17. For planar cores with windings integrated in a PCB (epoxy) it is about 24.
The final operating temperature of the proposed design can be found by combining eq. [4] to [7], resulting
in a linear equation for Toperating.
With this new temperature, the core is checked for saturation again, using the Bsat value at this
temperature. Bsat(T) is approximated by a piecewise linear function. Also the other boundary conditions
like Tmax and RDCmax are checked. If one of these conditions is not fulfilled the core is found not suitable
and the core with the next larger Ve is taken for a new design cycle. The first core size fulfilling all boundary
conditions is proposed for the the inductor design. All necessary core parameters such as airgap length and
e are given in the output window. For the calculation of this airgap the effect of fringing flux is taken into
account. Also the resulting winding design is presented together with power dissipation and temperature
rise.
In order to fine-tune the design, the core size can be fixed to find out what happens when a parameter is
adapted manually.

Inductor design

Common-mode Choke
Design procedure
The rst step in the design procedure is to estimate the non-compensated proportion of the ux. The
amount of stray ux is controlled by the geometry of the windings and the number of turns only. The total
generated ux is controlled by the same parameters but also by the core permeability. This means that the
fraction of non-compensated ux only depends on the core permeability. It is approximated by a multi-step
function.
This non-compensated ux fraction s is used as input for the design procedure using the same steps and
largely the same formulas as for the normal inductor design.
The stored energy is reduced

Energy stored: E = s Imax L

[1a]

The winding space is divided

Acoil = (Aw Aiso) / 2


Acopper = Fw Acoil
N

[3a]

The area Aiso follows directly from the safety isolation distance diso.
The losses double
2

Pcoil = Ieff R

[6a]

Pwinding = 2 x Pcoil

Magnetic Regulator Design

Magnetic Regulator
A

A
Make a choice from the available toroid sizes

D
A list with results for different numbers of
turns is presented in the window

B
Insert design parameters
C
Graph explains the meaning of the
abbreviations used in the output window

Magnetic Regulator Design

The magnetic regulator calculation formulas


Bs
td

Bsr=Vintd/Ae
Br
B=Vintbl/AeN

Bda=Vintb/AeN

Hc

tb

tbl

-Hc
Ireset=BdaHcle/(BrN)

The picture above shows the origin of the terms used in the equations that describe the fields and delay
times in the regulator. A summary of the formulas that are valid for the different parts of the BH-loop are
quoted below :
Vout T = Vin (t on t bl )

[1]

B Ae N = Vin tbl
Bda Ae N = Vin tb

[2]
[3]

Bsr Ae N = Vin td

Bsr = 11 N

0.235

(I out l e )

[4]

0.235

[mT]

[5]

(1 0.0048 T )/ 3250 [mW/cc], B in mT, f in kHz.


12
Pw = 8 10 I out f (t on t bl ) Cu N l av [mW], f in kHz

[6]

Ptotal = T 17 Ve

[8]

Pcore = B

1.7

1.3

[7]

[mW] for Ve in cc, [W] for V e in m .

These are the basic equations for a magnetic regulator in which we shall now step by step insert the known
input variables.
List of input variables :
core
magnetic regulator core size
Vin
Input voltage
Vout
Output voltage
Iout
Output current
f ,T
Frequency or cycle period
ton
on-time of the input voltage, related to the duty factor of the input cycle
Ta
Ambient temperature
T
Maximum temperature rise

Magnetic Regulator Design

Some of the input variables are related to each other:


ton = . T and f = 1 / T
The first step in the magnetic regulator program is to calculate the blocking time tbl.
From equation [1] and the relation above it can be shown that :

V
t bl = out T
Vin

[9]

The blocking time is needed when we want to calculate the minimum number of windings Nmin.
First we introduce some constants for convenience :

= 17 Ve
= T
= 8 1012 I out f (t on t bl ) Cu l av
Ve f 1.3
(1 0 . 0048 Ta )
3250
V f 1.3
0.0048
= e
3250
= Ve f 1.3 (1 0.0048 T )/ 3250 = T
=

1.7

V t 10 3

= in bl
Ae

We can now express the formulas for the power losses in a more compact form :

Ptotal =

Pw = N

Pcore = B 1.7

from :
Ptotal = Pw + Pcore
Where Pw represents the winding losses,
we find for the total flux swing :

B =
'

1
1.7

[10]

By inserting [10] in equation [2] we can derive an expression for Nmin :

Magnetic Regulator Design

1.7
2.7
N min
N min
= 0

[11]

The computer iterates Nmin until [11] > 0.


Inserting Nmin in [10] gives the maximum flux swing. The maximum flux swing is limited to 2 Br.
The following step is to calculate Nmax. How many turns can we wrap around the chosen core ?
For the calculation of Nmax we assume that the current density J = 8 A/mm2

J=

1
4

I out
=8
2
d Cu

d Cu =

I out
2

[12]

The effective diameter is calculated with an empirical formula for winding cores. A 5% isolation and pitch of
the winding is incorporated in this equation :

0.921
d eff = 1.05 1.14 d Cu

[13 a]

For a core with inner diameter di it can be shown that :

N max = i 1
d

eff

[14]

Note that dCu from [12] or [13] is a minimum value. The maximum value is calculated with :

d eff ,max = d i (N + )

[13b]

The maximum diameter is limited to one quarter of the inner diameter.


There is one more restriction to Nmax. Increasing N means that Bsr is increasing. We are applying a larger
H field. Bsr is causing the delay time td. We do not want td to exceed tbl. The number of turns for which this
happens is when td=tbl. With [4] and [5] we can calculate Nmax :

N max

Vin t bl

=
0.235
0.011 Ilou t
Ae
e

( )

1
1.235

[15]

For Nmax we choose the minimum of [14] and [15]

Magnetic Regulator Design

Finally the program iterates N from Nmin to Nmax and calculates B from [2], td from [4] and [5], tb from
tbl-td, Ireset and the temperature rise. Ireset is calculated by :

I reset =

H c le
Bda
Br N

[16]

Iteration from Nmin to Nmax gives various values for B from [2] which result in a specific temperature rise.
The temperature rise can be calculated by rewriting [10].

T N

T
B 1.7 + N
T =
+ B 1.7

B 1.7 =

[17]

List of input variables :


Nmin
minimum number of turns
Nmax
maximum number of turns
dCu,min
minimum wire diameter
Bmax
maximum flux swing

[from eq. 11]


[from eq. 14 & 15]
[from eq. 13a]
[from eq. 10]

Output results from iteration Nmin to Nmax


B
flux swing
dCu,max
maximum wire diameter
Ireset
reset current
td
time Br to Bs
tb
time Bda
tbl
total blocking time
T
temperature rise

[from eq. 2]
[from eq.13b]
[from eq.16]
[from eq. 4 & 5]
[tb = tbl-td]
[from eq. 9]
[from eq. 17]

Power Loss Calculations

Power Loss Calculation

A
B
C
D
E
F
G

E
Make a choice between sine wave and
customized waveform

A
Enter the switching frequency
B
Make a choice of ferrite material

F
Press to draw a graph (up to 4 per screen)

C
Enter the AC peak flux density

G
Press to clear the graphs

D
Enter the DC bias flux density

Power Loss Calculations

Power Loss Calculation


Introduction
With this program it is possible to calculate the power loss density for Ferroxcube power ferrites under
various conditions. The flux density signal can be an arbitrary wave shape or a sine wave, with or without
a bias flux. The calculations are based on fit formulas describing the power loss density as a function of
frequency, flux density and temperature. In the program the desired operating frequency (in kHz) or cycle
period (in s) and magnetic flux densities (in mT) can be given as input. The program calculates the losses in
the temperature range from 0 to 140 C for the selected power ferrite.
The program protects the user for wrong choices. Once a frequency is given as input, only those ferrites
can be selected which are intentended for use at that frequency.
The flux density cannot be chosen higher than the saturation level of the selected ferrite.
Theory behind the fit formulas and part of the fit parameters can be found in the references.

1.Basic formula
The calculation is based on an empirical fit formula for the core loss density of power ferrites:

Remarks
The constants are based on measurements with symmetric sine waveforms (no bias).
Bac is the peak flux density, so half of the total flux excursion : .(Bmax Bmin)
The parameters Cm, x, y, ct, ct1 and ct2 are specific for each power ferrite and often applicable
only in a limited frequency range.
The temperature dependent coefficients ct, ct1 and ct2 are dimensioned in such a way that at
T = 100 C the temperature term (ct - ct1.T + ct2.T 2) is equal to 1.

Power Loss Calculations

2. Calculation for sine wave


If there is no DC flux density (bias), the basic formula [1] can be applied directly to calculate the power loss
density in kW/m3 which is identical to mW/cm3.
With a DC flux density a correction is required. For this purpose measuring data from reference [2] is used.
For several power ferrites the power loss density at defined values of Bac and Bdc is given there.
The relative increase of the losses at Bac+ Bdc levels with respect to corresponding levels without bias is
calculated from this data.
The maxtrix obtained in this way consists of the elements: Pcore(Bac,j ,Bdc,i)/Pcore(Bac,j ,Bdc = 0).
The determination of the loss density on a arbritrary point of (Bac, Bdc) is done in four steps:
1. First the Bdc level under consideration is rounded to the nearest Bdc value available in the matrix.
2. The Bac level under consideration is rounded to the nearest lower and higher Bac value available
in the matrix . A lineair interpolation between these 2 Bac values is made to arrive at the desired Bac
level.
3. Finally with this rounding of Bdc and interpolation of Bac the approximated value of
PV(Bac,Bdc)/PV(BacBdc = 0) for the levels of Bac, and Bdc under consideration is obtained.
4. The obtained correction factor [1 + PV(Bac,Bdc)/PV(BacBdc = 0)] is multiplied by the power loss
density for the same Bac but without bias.

3. Calculation for custom waveform


For an arbritrary waveform the power loss density can be approximated as described in the following
chapters:

3.1. Sine wave equivalent of an arbritrary waveform


The arbitrary waveform is approximated by a maximum of ten pieces of straight lines. Each interval is
transformed to an sinus frequency equivalent
The method used is partly described in reference [3] and makes use of the sum of the weighted time
derivative of B which is defined as:

In words: the sum of the time derivative weight factor.

Power Loss Calculations

Example :
For a symmetric (no bias) triangle waveform with its maximum Bm at T, zero at T, Bm at T and zero
again at T, Bw can be calculated with eq. [2] as:

------------------------------------------------------------------------------------------

For a symmetric sine wave with the same period and values for Bm the calculation of can Bw can be done by
writing it in a integral representation:

Comparing the end result of equations [3] and [4] shows that the sine wave equivalent frequency of the
triangle form can be given as:

A piece of line representing part of a complete triangle period T is denoted as Ti:

Inserting eq. [6] in eq. [5] gives the expression for the sinus equivalent frequency of a line piece with period
Ti :

Example :
Suppose Ti = 2.5 s. From eq. [7] follows that fsin.eq = 81 kHz.
Ti = 2.5 s corresponds to T = 10 s, so by using eq. [5] we also find fsin.eq. = 81 kHz.
A sine wave with T = 10 s corresponds with fsin = 100 kHz

Power Loss Calculations

3.2 Division of a waveform into pieces of straight lines


The division of flux levels into periods of Ti is done in the following steps:

1.First the DC level of the flux waveform Bdc is determined. This is done by calculating numerically:

2.The begin and end points of the pieces of flux lines are now (numerically) placed at the :
Corners, where the inserted wave form changes from increasing B to decreasing B and vice versa.
Intersections of the inserted wave form with the calculated value for Bdc.

3.3. Calculation of total power loss

For all periods Ti, which are constructed as explained in chapter 3.2, the equivalent sine wave frequency is
calculated with eq. [7]: fsin.eq = 2/(BTi).
The corresponding fit parameters (Cm, x, y, ct, ct1, ct2) valid for the calculated frequency fsin.eq are now read
from the data base.
The power loss density contribution for each period is now calculated by inserting eq. [7] in eq.[1] and
multiplying it by its weighting factor Ti / T

The total power loss Ptotal( J ) is calculated by taking the sum of the obtained values with formula [9]:

If there is a DC flux density the same correction will be made as explained for the sine waveform in chapter
3.1 . The DC flux density is already calculated as in chapter 3.2. As AC flux density the maximum value of
the waveform with respect to the DC flux Bdc is used. The total losses are in this case:

Power Loss Calculations

References
[1] Mulder S.A., 1993, Loss formulas for Power Ferrites and their use in transformer design, Application
note Philips Components.
[2] Brockmeyer A., 1995 Experimental Evaluation of the influence of DC Premagnetization on the
properties of Power electronic ferrites,
Aachen University of Technology.
[3] Durbaum, Th, Albach M, 1995, Core losses in transformers with an arbitrary shape of the Magnetizing
current, 1995 EPE Sevilla.

Power Inductor Properties


C

F
Choose Ferroxcube cores or user
dened cores.

Fill in core parameters for a customized


core design
G

Make a choice from the available core


shapes.

Choose an effective permeability,


inductance fator or gap length.
After pressing enter the other
windows are updated and show the
calculated results.

C
Choose a core size.
H
D

Shows the resulting plots


(after pressing enter).

Choose a ferrite material from the list.


I
E

The graph can be exported or stored.


Choose an operating temperature

Power Inductor Properties

The power inductor properties program is a tool to explore the properties of a power inductor in more
detail. One can select various materials for a core shape and it is possible to define an effective permeability,
inductance factor or gap length. All properties are then calculated and displayed in graphs. The plots give
more insight in how the inductor will behave as a function of field strength as well as the energy storage as
a function of gap length.

BH-loop
The BH-loop of our power materials were measured at three temperatures at field strengths from
250 A/m to 250 A/m. The measurements were carried out on cores without gap. With this design tool one
can calculate the gap length required to get the given effective inductance or permeability or visa versa,
using the relations:

AL = 0 e

Ae
le

[1]

l g Ae
1
1
=
+
e i l e Ag F

F = 1+

2 lleg
ln
l
Ae
g

lg

[2]

[3]

The factor F is the contribution of the fringing flux to the effective permeability.
When a gap is introduced, one can recalculate the BH-loop by re-scaling the applied field H.
Amperes law for a gapped core can be written as:

NI = Hap l e = H f (l e l g ) + H g l g H f l e + H g l g

H g lg =

Bg
0

lg = B f

l g Ae
0 Ag

l g Ae
NI
= H ap = H f + B f
= H f + H
le
0 Ag l e
Hf = field strength in the ferrite,
Hg = field strength in the gap,
Bf = flux density in the ferrite
Bg = flux density in the gap
H = extra field strength to compensate for reluctance of gap
Hap = applied field strength,

[4]

Power Inductor properties


So, for the gapped core an extra field H is needed to reach the same level of B compared to a non gapped
core.
original BH loop without gap

sheared BH loop with gap

Br

point (H, Br)

Br(gap)

-Hc

Using equation [2] we can rewrite equation [4] as:

Hap = H f + Bf

1
0

1
1


e i

[5]

This enables the program to draw a so-called sheared BH-loop for any gap.

Incremental permeability versus H


The incremental permeability is defined as the slope of the BH-loop at given field strength H.
The BH-loop can be divided into an upper loop and a lower loop. For each value of H there are 2 values of
B. In the calculation of H according to [5] the average of these two B values is used.

Hap = H f + 12 Bf ,upper

1
0

1
1
1 1
1

+ 12 Bf ,lower

0 e i
e i

[6]

The value of B of the upper and lower loop is also averaged:

Bf ' = 12 ( Bf ,lower + Bf ,upper )

[7]

The slope of Bf(Hap) gives the incremental permeability :

Power Inductor properties


=

B f
Hap

[8]

The plot produced by the program gives an indication of the field strength for which the permeability of the
selected ferrite material collapses. This happens when the gapped core is saturated at the given operating
temperature.

AL of a gapped core versus NI


The behaviour of the AL of the gapped core set can now be calculated from the incremental permeability:

AL (gapped) = 0

Ae
le

[9]

From the average field strength H [6] the program calculates NI:

NI = Hap l e

[10]

IL versus NI
Important design parameters for a power inductor are inductance value (L) and the maximum current (I).
This is related to the required energy storage per cycle in the inductor, which can be expressed as IL.
Following the information given in the application section of the FERROXCUBE data book, the program
plots IL as a function of NI.

I 2 L = Hap2 l e2 AL(gapped)

[11]

Important design information can be taken from these graphs. The peak in the curve gives an indication of
the maximum energy storage in the core. If this is too low for the design a larger ferrite volume or another
material is obviously required. The value of NI below the peak gives the optimal number of turns.

Transformer Design
In switching power supplies, problems often arise with the design of the magnetic components, the ferrite-cored transformers and chokes.The interaction
between electronic and magnetic design deserves particular attention. Firstly,
in switched-mode power supplies (SMPS), the power transformer and choke,
and the electronic circuitry are so interdependent that design is hardly possible
without the magnetic aspects being constantly taken into account. Secondly, by
combining magnetic and electronic design, a far better insight is gained into the
operation of the circuit, with a consequent improvement in the design itself.
In the following sections these problems are tackled. Design procedures are
explained with the help of many useful formulas and graphs.

Part 1 - Magnetic components functional requirements


Part 1 covers most aspects of switching power supply design, with emphasis on the magnetic aspects. Here,
the basic electrical relationships for SMPS are given for forward, push-pull and fly back converters. Practical
formulae are given for inductance and effective-current values; auxiliary outputs and other special features
are also covered. Some aspects of control are treated. All treatments are related to the magnetic design.

Part 2 - Selection of suitable cores for a transformer design


Part 2 deals with the design of the magnetic components themselves, especially the selection of the
appropriate core.

Part 3 - Transformer winding design


Part 3 deals with the design of transformer windings

Part 4 - Power inductor design


Part 4 concentrates on the design of power chokes

Each Part contributes to an overall, step-by-step design procedure. Use of the procedure requires only a general electronic engineering background. Before entering
intothe procedure, a choice of converter type must have been made.

Part 1

Part 1 - Magnetic components functional requirements


1.The forward converter
1.1 Non-isolated
Principle of operation
Figure 1 shows the outline circuit of a forward converter. The basic operation of an ideal converter is
described by: Vo = Vi

+
D

Vi

Io

Vo

Fig.1 Outline circuit of a forward converter.

Figure 2 shows the voltage waveform across the inductance and the associated current waveforms.

VL

Vi - Vo

time
Vo
IL
Io
2Iac
time
/f

(1)/f

1/f

Fig. 2 Voltage waveform across the power choke in a forward


converter and the associated current waveform.

Part 1
Minimum choke inductance
For continuous-mode operation (uninterrupted current flow through the choke) - otherwise regulation
deteriorates-output current Io must always be greater than half the choke ripple current 2Iac. This is ensured
by using a minimum value for the choke inductance

Vo
Lmin = 1
2f Io min

( 1- max

Vi min
Vi max

[2]

An increase in as a result of a sudden load increase will cause a temporary increase in choke ripple
current. As long as the ripple component is much smaller than the d.c. component, which is usually the case,
this does not affect the design of the choke.

Choke design
Output choke L carries a direct current equal to the d.c. current in the load. Thus, to avoid saturation, an air
gap is required in the core. The design steps are:
- determine IM = Io max + Iac
- calculate I2MLmin
- proceed to Part 4 of this series.
For core loss, see Part 2.

Deriving auxiliary power from the choke


During the flywheel period (1-)/f, that is, while the power switch is not conducting, the voltage across
the choke is stabilized. By adding secondary windings to the choke, auxiliary, stabilized, low voltages can be
obtained as shown in Fig. 3. These auxiliary voltages are rectified by diodes that conduct during the flywheel
period. Since auxiliary loads decrease the amount of energy recovered by flywheel diode D, the amount of
auxiliary power that can be derived is limited to 20 to 30% of the total output power.

V1 aux
Vo
D

V2 aux

Fig. 3 Auxilliary low-voltage outputs can be obtained by adding


windings to the choke of a forward converter.

Part 1
In order to store sufficient energy in the choke to supply an auxiliary load, inductance Lmin calculated from
Eq. 2 must be increased to Laux. This might lead to the use of a larger core: see Part 4. The relationship
between Lmin and Laux is

Laux

Lmin

>
1

[4]

(0.3 to 0.4)
1 max

The factor (0.3 to 0.4) corresponds to an auxiliary load being 20 to 30% of the total. If output ripple is not
important, a higher proportion of auxiliary load can be drawn.
The turns ratio

Vo
Vaux

ND
Naux

[5]

During forward conversion, during the period /f the input power is: Pin = ViIo
Similarly, the throughput power is: Pth = VoIo
The difference is the power stored in and then removed from the choke: PL = Pin P = Io ( Vi Vo )
From Eq.1: PL = Io ( Vi Vi ) = ViIo ( 1 Vi ) = Pin ( 1 )
If, during flywheel period (1- )/f, the choke current is divided between output and flywheel diodes as
indicated in Fig.4 by line a, Paux is maximum, Paux max . The current below a is that through the flywheel diode
of the primary output. Fig.5 shows some waveforms. Due to leakage induction, flywheel diode current does
not begin at A, but at Imax as the auxiliary output currents are then zero. This decreases the value of Paux
rnax . If the current is shared according to line b in Fig.4, Paux = kPaux max, where k = 0.5. In practice, k will be
somewhat higher: a value of k = 0.7 is reasonable.

L = Lmin

Imax
L = Laux
a (k=0.5)

k = 0.7

A
a (k=1)

time
(1)/f

Fig.4 Current through a forward converter choke with


auxiliary windings.

Part 1
If L = Lmin, no auxiliary power can be drawn. From these considerations it follows that the auxiliary power
that can be obtained is limited and depends on k, and L, such that

Paux

k(1-)(1-

Lmin
Laux

) Paux

The required value of Laux is obtained from

Laux

Lmin

>
1

Paux
k Ptot (1 max )

Substituting k = 0.7 and Paux / Ptot = 0.2 to 0.3 yields Eq.4.


Ptot = Pth - I0 ( VF + VR + Vo )
Further discussion may be found in Ref. 1.

V
I

Fig.5 Oscillogram of choke waveforms (a) L> Laux and (b) L < Laux>

Part 1
1.2 Single forward converter with isolating transformer
Principle of operation
The outline circuit of a forward converter with mains isolation is given in Fig.6. The magnetic energy stored
in the transformer while S is ON must be removed while S is OFF, otherwise the energy stored and
removed during a complete switching cycle would not be zero and the transformer core would rapidly
saturate. A solution involving minimum power loss is to add a winding, closely coupled to the primary, and a
diode D3 such that a flow of magnetizing current is ensured while S is OFF.

D1

+
Vi
C

D2

Vo
-

D3

Fig.6 Outline circuit of a mains-isolated forward-converter


SMPS.

The operation of the transformer-isolated forward converter is described by the same basic expression, Eq.
1, as was used for the non-isolated version. The transformer also adds an extra degree of freedom of choice
of output voltage for practical values of . This output voltage becomes Vo = Vi / r
Voltage and current waveforms for a transformer-isolated forward converter are given in Fig. 7

VL

Vi /r - Vo

time

Vo
IL
Io

2Iac

time
/f

(1)/f

1/f

Fig.7 Output circuit waveforms for the mains-isolated forward


converter.

Part 1
Duty factor
The maximum allowable duty factor at which the core will not saturate due to flux staircasing depends
on r and m:

max < 1

m
m+r

[7]

The maximum voltage across the power switch (here, a transistor) is then

VCEM = Vi max m + r
m

[8]

When using a transistor with VCESM = 850 V in a forward converter with a maximum (rectified) input voltage
of 375 V, it is usually adequate to limit VCE to 2 x 375 = 750 V. Thus, with m = r, there is 100 V to spare for
ringing and integrated supply-voltage surges.
The effective duty factor depends on frequency, turns ratio r, rated load current, the leakage inductance of
the transformer and the inductance of the leads to the output diodes. As a guide for mains-operated SMPS,
decrease the conduction time so that

e = - rI
o
f
f

10-9

1.2

[9]

The inductance of the leads to the output diodes is reflected into the primary as the square of the turns
ratio. With large turns ratios, combined with high switching currents, the loss due to commutation delay
becomes substantial. The effect is as if the available duty factor is decreased. This problem is discussed
in greate detail in Ref. 3.
An example is given in Fig8a. The commutation delay is: tc IorLs / Vi .
Now, Ls is about 1 nH/mm of leads and, for a 220 V mains supply,Vi min 200 V.
With the shortest possible leads to the output diodes, experiment shows that the commutation delay is
tc 1.2Ior 10-9
This is an important reason for not operating low-voltage high-current SMPS at high frequencies, but, rather,
to use a frequency just above the audible range.
Io
Ls

I2Ls

Io

Io /r
0

time

tc
e/f

I Ls

(a)

(b)

Fig.8 Effect of stray inductances in the output circuit of an isolated forward converter: (a)
inductance in the secondary is reflected back into the primary as the square of the turns ratio;
(b) there is a commutation delay tc during which neither output diode conducts.

Part 1
Preliminary turns ratio and core selection
The preliminary turns ratio is

r' =

e max Vi min
Vo max + VF + VR

[10]

For transformer core selection, see Part 2.

Multiple-output transformers
Additional outputs at any d.c. Ievel can be obtained simply by adding further secondaries of the appropriate
number of turns to the output transformer. The regulation of the additional outputs will be better than
with a flyback converter. However, each output needs two diodes and a power choke, against the single
diode needed with a flyback converter.
Warning: the flywheel diode must always conduct when the forward diode does not. Otherwise, peak
forward conversion rectification occurs and the output voltage could rise to the peak value of the forwardconversion voltage, which might be much higher than the nominal voltage, with disastrous results. So, ensure
that the appropriate minimum load is always present at each output.
With several different outputs, it is necessary to find that value of volts-per-turn for the transformer that
allows each output voltage to be obtained within the permitted tolerance with an integral number of turns.
The procedure for this is described in Part 3.

Control method
The function of the control circuit is to stabilize the output against variations in input voltage and load
by adjusting the duty factor of the switching device. However, the effect of step load changes cannot be
corrected immediately because some time is needed for the current through the choke to assume the new
value of the load current. A momentary change in output voltage is thus inevitable. The time required for
resumption of the desired level of output voltage after the sudden load change depends greatly on the
properties of the control system. Two basic control characteristics can be distinguished: Fig.9; these are
discussed in greater detail in Ref.2.

Vi
max Vi max

max

Vi min

10% feed
forward

min

Vi

1.1Vi

Vi bo

Vi

0
Vi bo Vi min

Vi max

time

Fig. 9 Input voltage and duty factor combinations under transient


load conditions for feedback and feedforward control.

Part 1
Feedback:
A step increase in load current causes the powerswitch duty factor to increase instantly to its maximum
value max regardless of the level of input voltage. It is, thus, possible for Vi max = max Vi max.

Feedforward:
A step rise in load current causes the powerswitch duty factor to increase to a value x % higher than its
steady-state value for a constant load. The Vi max product in response to a step load increase will be
higher with feedback control. This results in a shorter delay in adjusting to the new load because the choke
current is forced to increase at a maximum rate. However, the output transformer must be so designed that
it is able to cope with the product max Vi max without saturating.
With feed forward control,

Vi max

= (1+

x
100

) minVi max

The corresponding value of for the transformer design is

1+

x
100

Difficulties may be encountered with converter starting at full load with minimum mains voltage (15 %
below nominal). If the duty factor is close to maximum, all the output current will flow into the load
and little or no current will be available to charge the output capacitor. Starting will be improved if the
steady-state value of the duty factor is, say, 10 % below its maximum allowable value.
One disadvantage of feedback control is that a larger transformer core is generally required to avoid
saturation. The transient response time resulting from the method of control is discussed at the end of
Section 1.2.

Primary inductance and de-saturation


The primary inductance is given by

L1 = 0 a n1 Ae
le

[11]

where the value of a is obtained from the core data. The maximum, peak, primary magnetizing current is

I magM =

( Vi) max

[12]

L1 f

The peak primary current is

I magM = 1
f

( Io +

I o min
2

+ I magM

[13]

Part 1

L1
+

Vi

Fig. 10 Forward converter transformer with demagnetizing (energy


recovery) winding, and a slow rise capacitor in the primary circuit.

Primary inductance L1 together with slow-rise dV/dt capacitor C (Fig. 10) across the switch forms a
resonant circuit with a natural frequency

fr =

[14]

2 (L1 C)
The value of C used should satisfy the relationship

I1M tf = C > I1M


2VCESM
dV/dt
It is recommended that fr > f. At a lower value of fr, the core might fail to de-saturate and flux staircasing
could occur, with disastrous results. A higher fr offers sufficient safety margin under all conditions; a ratio fr/f
of about 1.2 is a good compromise. An electronic solution to the problem can be found in some control ICs
(e.g. TDA1060) where there is a facility for reducing the duty factor when core saturation is possible.
The value and spread of L1 may be reduced by introducing an air gap in the core; the penalty is higher
magnetizing current.

10

Part 1
For these small values of air gaps, the value of e ( and, thus, L1 ) can be obtained for cores of constant
cross-sectional area from

s = le
2

1 - 1
e a

[16]

If it is not possible to decrease L1 sufficiently to make fr > f, the addition of a sensor winding on the
transformer to generate a signal to prevent premature switch-on should be considered. (20)

Transformer currents
The ripple in the output choke is, in general, only a few percent of the d.c. load current. For this reason, the
transformer current can be regarded as a square wave for the purposes of winding loss calculation.
The maximum r.m.s. current values are given, approximately, for the primary and secondary by

I
Ie 1 = o
r

o max = Ie 2
r

[17]

where

o max = max

Vi

[18]

Vi av min

Duty factor omax is used in these calculations for the following reason.
For 220 V mains, the minimum line voltage is roughly 185 V r.m.s. Using an input filter capacitor of about
2 uF/W (to cope with a mains drop-out of 10 ms), the peak-to-peak ripple at 185 V mains input is about 20 V.
Under these conditions the minimum average (steady-state) input voltage

Vi av min =

2Vi min Vir

[19]

= 252 V
However, max is set so that the converter can handle a 10 ms drop out. But mains drop out is not a
steady-state condition, so that omax should be used to calculate Ie and, thus, the loss and consequent
temperature rise of the transformer.

Choke design
To determine the minimum required choke inductance and for the choke design, see Section 1.1.

Transient response time


The transient response time required for a forward con- verter to adjust to a step in the load current
of Io is

tr =

Io L
( Vo + VF + VR ) (tr/ - 1 )

11

[20]

Part 1
1.3 The double forward converter
Principle of operation
The equivalent circuit diagram of a double forward converter is shown in Fig. 11. It comprises two forward
converters in parallel, with flywheel diode D and filter LC common to both. Switches S1 and S2 operate
alternately, which doubles the ripple frequency of the choke current. Since energy is pumped twice per
converter period, the output voltage is Vo = 2 Vi

S1
L
+

+
S2

Vi

Io

Vo

Fig. 11 Outline circuit of a double-forward SMPS converter.


Switches S1 and S2 operate alternately.

Choke inductance
The minimum choke inductance is calculated in a similar way to that for the single forward converter.
However, since there are now two charges and two discharges per converter period

L>

Vo
4fIo min

2max Vi min

[22]

Vi max

Further choke design proceeds according to Section 1.1.

Preliminary transformer turns ratio


The transformer is designed in a similar way to that for the single forward converter, except that now the
turns ratio is twice as great. The preliminary turns ratio of a double forward converter transformer is

r' =

e max Vi min
Vo max + VF + VR

[23]

Further design proceeds according to Part 2 of this series.

Transient response time


The transient response time required for a double forward converter to adapt to a step in the load
current of Io is

tr =

Io L
2(Vo max + VF + VR) (tr/ -1)

12

[24]

Part 1

THE FLYBACK CONVERTER


2.1 Non-isolated inverting converter
The outline circuit of an inverting flyback converter is given in Fig. 12. The basic operation of an ideal
inverting flyback converter is described by

Vo =

[25]

Vi

D
-

Vi

Vo

Io

Fig. 12 Outline circuit of a non-isolated, inverting flyback


converter.

The voltage waveform across inductance L and the associated current waveforms under steady-state
conditions are given in Fig. 13.
VL

Vi

time

Vo
IL

2Iac

D
Io

time
/f

(1)/f

1/f

Fig. 13 Voltage and current waveforms for the choke of a nonisolated, inverting flyback converter.

13

Part 1
Power inductor
The minimum inductance required to ensure continuous mode operation at minimum load Po min is

L >

( min Vi max )2
2fPo min

[26]

Note: use of Po allows output-diode losses to be neglected.


The maximum peak current through the inductor is IM = Idc max + 2 Iac, and

IM =

Po max

V
+ max i min
max Vi min
2fL

[27]

For inductor design, see Part 4.


It is evident that the value of the inductor is inversely proportional to the minimum load. On the other hand,
the choke must not saturate at maximum load if the supply voltage is minimum. Thus, continuous choke
current with large output-power variations is not always practical. The alternatives are:
- Pre-load the supply to decrease the ratio P o max /P o min
- Change the operating frequency, as in the hiccup mode described in Ref. 2.
- Accept a discontinuous choke current at the expense of higher peak current through switch
and output capacitor.
In the last case, the choke current waveform will be triangular rather than trapezoidal. For a triangular
current waveform. Eqs (26) and (27) become

( min Vi max )2
L >
2fPo min

[28]

and

IM = 2

Po max
max Vi min

[29]

14

Part 1

Non-inverting boost converter


Principle of operation
The outline circuit of a boost converter is shown in Fig. 14. Its operation is described by

Vi
1

Vo =

The voltage waveform across inductor L and the associated current waveforms under steady-state conditions are shown in Fig. 15

D
+

Vi

Vo

Io

Fig. 14 Outline circuit of a boost flyback converter.

VL

Vi

time

Vo Vi
IL
2Iac
Io
time
/f

(1)/f

1/f

Fig. 15 Voltage and current waveforms for the choke of a boost flyback
converter.

15

Part 1

Power inductor
The minimum inductance required to ensure continuous- mode operation of the choke at load Po min is

L >

2
27

Vo2

[30]

fPo min

The maximum peak inductor current is

I maxM =

Po max
Vi min

max Vi min
2fL

Proceed now with Part 4.

16

[31]

Part 1

2.2 Transformer-isolated flyback converter


Principle of operation
For high-voltage low-current supplies, a useful variant of the inverting fly back converter is obtained by
adding secondary windings to the choke to form a transformer, Fig. 16. A further benefit of this arrangement
is mains isolation.
S

r:1

Vo

Vi
-

Io

Fig. 16 Outline circuit diagram of a transformer-isolated flyback


converter.

Turns ratio
In order to protect the switching devices, the turns ratio

r <

VCESM - ( Vi max + Vr )
Vo + VF + VR

[32]

At high voltages, such as rectified mains, a good compromise between inductor size, switching-transistor
peak current, and diode peak current can usually be obtained by one of the following procedures.
At moderate input-voltage range, say

Vi max
Vi min

<2

take min = 0.3. This yields

max =

1
1 + 7/3 ( Vi min / Vi max )

[33]

so that

r' =

3/7 Vi max

Vo + VF + VR

At large input-voltage range, that is

Vi max
> 2
Vi min

17

[34]

Part 1
take

r' =

(Vi max Vi min )

Vo + VF + VR

[35]

This yields

1
Vi max

min =
1+

[36]

r'(Vo + VF + VR)

Note: If dmin > 0.3, take dmin = 0.3 and proceed as for a small input-voltage range, otherwise

1 = 1 + (1 - ) Vi min
min
max
minVi max
A voltage limiting winding with turns ratio between primary and limiting winding r/m limits switching-device
voltage to Vi max The maximum duty factor must then be such that

max < 1

[37]

whence

min =
1+

(1 - max) Vi max

[38]

max Vi min

and

r' =

min Vi max
( 1 min) (Vo + VF + VR)

[39]

with

m =

r
1

[40]

18

Part 1
Power inductor
To ensure continuous-mode operation, design the choke primary for minimum load. (See also Section 2.1.)

L <

( min Vi max )2
2 f Po min

[41]

The maximum peak current through the primary is

I maxM =

Po max
max Vi min

max Vi min
2fL

Proceed with Part 4. The number of turns on the choke secondary, together with r, are found in Part 3.

Multiple output
Additional output voltages at any d.c. level can be obtained by simply adding additional secondaries of the
appropriate numbers of turns. Note that, if the range of d.c. output voltages is large, leakage inductance will
increase and regulation deteriorate. The design procedure is given in Part 3.

Effective currents
The maximum value of the r.m.s. current through the primary winding is

I e1 =

Po max

0 max

1+ 1
3

0 max Vi min

Po min
Po max

[42]

and through secundary x,

I ex =

rxPox max (1 0 max)


0 max Vi min

1+ 1
3

P
)
( Poo min
max

where

0 max =
1+

Vi av min
r (Vo + VF + VR)

19

[43]

Part 1

S1

L1

C1
T1
Nt
D1

Nt

C2

S2

S1

D3

C3

D2

D4

S3

L1

T1

Nt

Nt

S4

S2

D1

D3

S1

C3

D2

D4

L1

C1
T1
1/3Nt

1/3Nt

C2

C3

1/3Nt

S2

S1

S3

L1

T1
1/3Nt

Nt

S2

C3

1/3Nt

S4

S1

L1

T1
1/3Nt

1/3Nt

1/3Nt

D1

S2

D3

S1

D2

C3

Fig. 18 Fig. 17 The basic configurations of


push-pull converters. (a) Two-winding (1 +
1) transformer in a single-ended push-pull
converter with a bridge-rectified output. (b) Twowinding (1 + 1) transformer in a bridge converter
with a bridge-rectified output. (c) Three- winding
(1 + 2) transformer in a single-ended push-pull
converter with a bi-phase output. (d) Threewinding (1 + 2) transformer in a bridge converter
with a bi-phase output. (e) Three-winding (1 +
2) transformer in a push-pull converter with a
bridge- rectified output. (f)

D4

L1

T1
1/4Nt

1/4Nt

1/4Nt

1/4Nt

C3

S2

20

Part 1

3 THE PUSH-PULL CONVERTER


3.1 Principle of operation
Outline circuits of the various basic configurations of push-pull converters are shown in Fig. 17.
Since energy is pumped twice per converter period, the basic operation of a push-pull converter is
described by

Vo =

z Vi
2

[44]

For full bridge and conventional push-pull converters, z = 4; for half bridge push-pull converters z = 2.
The voltage waveform across the choke and the associated current waveforms are shown in Fig. 18.
VL

Vi

r - Vo

time

IL

Vo
Io

2Iac

time
/f

1/2f - f
1/f

Fig. 18 Voltage waveform across the choke of a push-pull


converter together with the associated current waveforms.

Duty factor
With the equal conduction times per converter cycle, the maximum allowable duty factor is 0.5, but a more
practical value is 0.45. In practice, wiring and transformer stray inductances result in a finite commutation
time between output diodes. As a result, the interval during which energy is supplied to the output is
shorter, and the effective duty factor smaller. This effective duty factor depends on operating frequency,
transformer ratio, load current, and the stray inductance of the leads to the rectifier diodes.
To compensate for the increased commutation time, the energy transfer period /f should be decreased
to about

e = - rI
o
f
f

1.2

10-9

Further discussion will be found in Section 1.2.

21

[45]

Part 1
Transformer turns ratio
The preliminary turns ratio is

r' =

2e max Vi min
Vo + VF + VR

[46]

For further transformer design, proceed to Part 2.

3.2 Power choke inductance


The minimum choke inductance that ensures continuous choke current, and, thus, continuous-mode
operation is

L >

Vo
4fIac

( 1 - 2max

Vi max
Vi min

[47]

here,

I ac =

I o min

I magM

[48]

and

I maxM =

r max Vi min le
2n21 0 e Ae f

[49]

During transistor conduction, Fig.19, the magnetizing current changes from + I magM to - I magM. While both
transistors are off, during the interval (l/2f - /f) the transformer primary is open circuit. This forces the
magnetizing current to flow through the output diodes in series. Thus, the load and magnetizing currents
reinforce in one diode and cancel in the other.

22

Part 1

2Iac

IL
Io

Im1
ID1

Im

ID2

/f

1/2ff

1/2f
S2 ON

S1 ON
D1 ON

D2 ON

1/f

Fig. 19 Waveform for calculating the peak choke current of a


push-pull converter.

If, at low output current, one diode ceases to conduct, there is no path for the magnetizing current, which
is then diverted through the conducting diode and the output choke to the output capacitor. This causes
the output voltage to rise. From this, it follows that the minimum load current that can be drawn from the
converter without one diode ceasing to conduct in this way is

I o min = I magM

I ac

The magnetizing current flowing through the output diodes is

I maxM =

max Vi min r
2 L1 f

where
2

I maxM =

2n1 0 e le
Ae

These two expressions together yield Eq. 49. From Fig.19, the peak current through the power choke is

I M = (I o max

I o min + I magM )

Further choke design proceeds with Part 4.

23

[50]

Part 1
3.3 Transformer currents
In a push-pull transformer, only one half of the double winding conducts at a time. The peak current through
each half of the double winding is

I 1M = I M

[51]

The effective primary current in each winding half is

I e1

Io

I e2

Io

o max

[52]

o max

[53]

and, in the secundary,

where

o max = max

Vi min
Vi av min

Transient response time


The transient response time required for a push-pull converter to adjust to a step in the load current
of Io

tr =

Io L
2 ( Vo max + VF + VR ) (tr/ - 1 )

24

[54]

Part 2

Part 2 - Selection of suitable cores for a transformer design


Available ferrite core types cater for a wide range of SMPS application requirements.
If the proper core for a given application is to be selected, a number of factors should be taken into
account. The type of converter circuit used, for example, determines to a large extent the throughput
power capacity of transformer wound on a given core type.
The discussion here assumes that no premagnetization - magnetic bias by a permanent magnet - is applied
to the core.

Part 2

1.1. Core selection for forward and push-pull transformers


The practical throughput power of a transformer is represented as a shaded area, as shown in Fig. 1.

Pth

range of Pth
for good, practical
designs

10

100
f(kHz)

Fig. 1 The powerhandling capability of a core is plotted as a shaded


area extending from 10 kHz to 100 kHz.The vertical boundaries of
this area represent the upper and lower limits of throughput power
capacity achievable by good design, but depending on conductor type
and mains insulation requirements.

The actual throughput power obtainable with a given core depends to a large extent on the following
characteristics:
flux density sweep (Section 2)
the winding configuration (simple or split/sandwiched windings, sensor or demagnetization windings)
conductor type (solid, strip, Litz)
single or multiple output
mains insulation requirements.

Part 2
The upper limit of the shaded area for each core refers to a transformer design with optimized flux-density
sweep, maximum use of the winding window, and Litz wire for minimum a.c. resistance.
The lower limit refers to a design with 8 mm creepage distance for IEC 435 mains insulation, optimized fluxdensity sweep, a (1 + 2) winding configuration (Part 1), and optimized but solid-wire windings. In addition,
the following general conditions were assumed in the calculation of both boundaries :
the hot-spot (peak) temperature of the core is 100oC; the temperature rise is 40 K
the maximum flux-density sweep is limited to 1/1.72 of the maximum permissible flux density for the
core material (0.32 T for 3C94) to cope with transient conditions.
the thermal behaviour of wound cores, but without potting or additional heatsinking was assumed.
core flux densities are calculated assuming minimum cross- section areas.
Where the ambient temperature is lower than 60oC, when feedforward is used to ease the restriction on
maximum flux density, or when heatsinking or potting are employed to improve heat transfer, throughput
power capacity will be increased.
It may happen that, at a given power level, more than one type of the core may be used. The following
criteria may be used to make a choice :
copper foil secondary windings are preferable for low-voltage, high-current supplies; thus, for ease of
winding, a core with a round centre pole should be chosen.
coil formers and mounting hardware are not available as standard for all core types.
production logistics may be improved if one core type is used for both transformer and choke.
Where these considerations do not apply, the choice of the core should be guided by the discussion
of Section 2.

1.2. Core selection for flyback converters and chokes


The magnetic design of flyback transformers and output chokes for forward push-pull converters is
essentially the same: the main design parameter is the energy to be stored. Core selection, therefore, is
made on the basis of energy stored, I2ML. . This leads directly to airgap length and number of turns.

Part 2

2. Operating flux density


When determining operating flux density, a distinction must be made between transformers and chokes. For
chokes, and flyback-converter transformers (which also function as chokes), the most important parameter
is the maximum peak flux density. The flux density sweep follows from the inductance value required.
For push-pull and forward-converter transformers, both a.c. and d.c. components of flux density must be
taken into account from the start of the design process.

2.1. Flux-density levels for forward and push-pull transformers


The operating flux density of a transformer can seldom approach the maximum permissible flux density in
practice since an allowance must be made for transient conditions, such as sudden load increases.
Unless special electronic measures (Part 1) are taken, a transient factor is required to cope with sudden
load changes. This factor is related to the ability of the power supply to accept a range of input voltages. The
input voltage range of mains-fed power supplies may be 215 V to 370 V, or 200 V to 340 V; for telephone
supplies, 40 V to 70V; and for mobile supplies, 9V to 15.5 V. The usual transient factor is 1.72.
For symmetrical excitation of the core (push-pull), the maximum possible flux-density sweep (Fig.2) is,
in principle, twice that for asymmetrical excitation. In practice, however, allowance has to be made for
unbalance when determing the operating flux density.

H
(a)

B
2Bac

2Bac

2Bac

H
(b)

H
(c)

Fig. 2 Flux density excursions and the corresponding flux-density sweeps


for (a) push-pull converter transformer; (b) forward-converter transformer
(with slow-rise capacitor) or ringing-choke flyback converter; and (c)
flyback converter choke.

The maximum operating flux density depends on the protection circuitry. One source of unbalance is
unequal flux linkage between two halve of a centre-tapped winding. For this reason, bifilar windings are to
be preferred. However, this is not possible in mains-fed supplies since the voltage across the winding might
be greater than the maximum voltage between adjacent turns.
The major reason for asymmetry is unequal conduction times or saturation voltages of the power switches
in a push-pull converter. Storage effects can result in different switch-off delays. Core saturation occurring
due to a delay decreases the primary inductance so that the magnetizing current rises steeply. This may
lead to the destruction of the power switches. As a further safeguard, the maximum operating flux density
should be decreased by an additional factor that depends on the efficiency of the protection circuit.
A pratical guide is a 15% allowance for a fully protected converter (unbalance factor = 1.15), but a 100%
allowance unbalanced push-pull converter ( = 2).
In forward converters, remanence should, in theory, also be allowed for. However, to obtain the correct
primary inductance, some airgap is often useful. This airgap, together with the slow-rise capacitor (Part 1),
results in the whole first quadrant of the BH loop being useful in practice, as is shown in Fig. 2.

Part 2
Bac
(mT)

Fig. 3 Optimum peak centre-pole flux density


sweeps Bac opt for a variety of cores for SMPS
applications. Horizontal lines indicate the limits
for various converter types with = 1.72.
The curves are calculated for a transformer
temperature rise of 40 K.

7Z90894

400

saturation
limit

300

200

push-pull

150

1
2

100
90
80
70
60

3
4

forward

50
40
30
10

15

20

30

40

50 60 70 80 90 100
f (kHz)

1. UU15/22/6. EE20/20/5 2. UU20/32/7. EE25/25/7. EE30/30/7


3. UU25/40/13 4. UU30/50/16

Bac
(mT)

7Z90895

400

saturation
limit

300

200

push-pull

150
1
100
90
80
70
60

2 forward
3
4
5
6
7
8

50
40
30
10

15

20

30

40

50 60 70 80 90 100
f (kHz)

1. EC 35 2. EC41, EC52 3. EC70 4. EE42/42/15


5. EE42/42/20. EE42/54/20. EE42/66/20 6. EE55/55/21
7. EE55/55/25. UU64/79/20 8. EE65/66/27

Pth/Pth max
Bac
(mT)

7Z90893

400

1
saturation
limit

300

200

0,9

push-pull

150

0,8
100
90
80
70
60

forward
1
2
3
4

50

0,8

40
30
10

15

Curve 1
Curve 2
Curve 3
Curve 4

20

30

40

50 60 70 80 90 100 150
f (kHz)

0,9

1,1

1,2
B/Bopt

Fig. 4 The effect on throughput powercapacityof a


deviation in Bac from the optimum value indicated in
Fig. 3.

ETD34
ETD39
ETD44
ETD49

Part 2
Where feedforward control is used, the transient factor can be reduced considerably and a higher fluxdensity sweep can often be applied. The actual transient factor is determined by the feedforward percentage
used. Note, however, that application of feed- forward reduces the transient response of the power supply.
TABLE 2 Maximum Values of flux-density sweep for various converter types and control circuitsboundary conditions

flux-density sweep Bac cp (T)


forward
push-pull
maximum flux sweep (100oC)

0.16

0.32

at transient factor

0.32
2

0.32

with unbalance factor

0.32

with x% feedforward

0.32
2(1 + x/100)

0.32
(1 + x/100)

0.32
(1 + x/100)

with unbalance factor and x%


feedforward

Practical flux density and sweep limits are summarized for various converter types in Table 2. The curves of
Fig. 3, show the optimum flux density sweep, where throughput power is maximum, for a range of cores in
the frequency range 10 kHz to 100 kHz. Horizontal lines indicate the maximum allowable sweep for various
converter types. Further lines can be added for other boundary conditions with the aid of Table2. The given
curves are calculated for a transformer temperature rise of 40 K.
The converter operating frequency is set by the required output voltage and current, and by the type
of switch to be used. One this frequency is known, the optimum flux-density sweep can be found for
any core type.
Where the frequency is more or less fixed, and two core types could be used (Section 1), preference should
be given to that core type for which the intersection of the optimum-sweep curve with the set frequency is
closest to the maximum flux density sweep.
Where frequency can be chosen freely, the frequency corresponding to the intersection of the optimumsweep curve with the line for the maximum flux-density sweep represents the optimum use of the core
material.
Operation at the optimum sweeps represented by the curves of Fig. 3 means that core loss and
permissible winding loss are in optimum proportion of core and winding loss, results in a lower throughput
power.When the design is limited by saturation flux density, deviation is inevitable. The effect of deviation
from the optimum flux density is plotted in Fig. 4.
This plot, which applies to any frequency, gives a rough indication for the reduction in throughput power
from the optimum value. Once the optimum flux density sweep has been determined for the core and
converter combination, the number of turns can be calculated as shown in Section 3.

2.2. Flux densities for chokes and flyback transformers


The maximum flux density is taken into account in the design procedures given in Parts 1 and 4. The design
centres around the maximum stored energy I2M L in the choke, which, in turn, is related to the inductance
value from which follows the number of turns.
The flux density sweep in chokes is generally relatively low, and so, in consequence, is the core loss.

Part 2
It may happen, however, especially in ringing choke flyback converters, that the flux-density sweep is
comparable to that in forward converters. Core loss is not then negligible and should be calculated as
shown in Section 4.
Once core type, spacer thickness, and number of turns have been established, the peak flux density sweep
can be calculated:

Bac max =

L Iac
Nprim Amin

where Iac is found during the design process (Part1). In case of a flyback transformer, all quantities refer to
the primary. If Iac is relatively high, core loss will be significant.

3. Number of turns
Once the flux-density sweep is known, the number of turns can be determined.

3.1. Forward and push-pull converter transformers.


The minimum number of turns in the secondary

n2 min =

Vi

r'Amin Bac max f z

z = 2 for forward and half-bridge push-pull converters


z = 4 for full-bridge push-pull converters.
The values of and Vi are obtained from
Vi = maxVi min = minVi max (this and r are found in Part 1).
n2 min will generally not be an integer and must be rounded. If the value of n2 min is small, rounding will
change the flux density sweep significantly. To prevent saturation, rounding should be to the next higher
integer. The effect of rounding can be counteracted by changing the operating frequency.
Where operating frequency is not fixed, iteration will result in a satisfactory design. When the actual fluxdensity, sweep is within 10% of the optimum, throughput power will be within 5% of its maximum (Section
2.1). Use the flux density sweep obtained after rounding to determine core loss (Section 4).

3.2 Flyback converters and chokes


Flyback converters and output chokes:
the number of turns is derived and rounded in the choke design procedure of Part 4.
Flyback transformers:
the rounded number of primary turns is found in the choke design procedure of Part 4. The value of
r' is calculated in Part 1.

Part 2

4. Core loss
Core loss has two main components: hysteresis loss and eddy-current loss. The hysteresis loss in a typical
low frequency power ferrite is for instance:

Ph 8 f1.3 B2.5ac Ve

Both coefficients of f and B are frequency dependent. In the design tools software this type of formula is
used to predict core losses during transformer design.
The eddy-current loss for a ferrite, under the same conditions, is:

Pe 0,8 f2 B2ac Ae Ve /.

Where is a function of frequency and temperature.


The contribution of eddy-current loss in the core is greatly dependent on core size, frequency and flux
density sweep. Below 100 kHz, eddy-current loss may be neglected for small cores. However, for the larger
cores, such as the E65 and EC70, eddy-current loss becomes important at much lower frequencies.
For very large cores, with centre-pole diameters exceeding 35 mm, the expressions for both Ph and Pe
do not hold, even below 100 kHz.
In the above expressions, Bac is the peak effective flux density sweep. It is related to Bac max by

Bac max Acp = Bac Ae

For cores of about constant cross sectional area, such as ETD, and most E and U cores, Bac max and Bac are
similar in value, but they differ significantly for EC cores.

Part 2

5.Thermal resistance
In order to determine the maximum permissible dissipation of a transformer or choke, its thermal behaviour
must be know. This depends on core size, conductor form, and insulation requirements. For standard, more
or less cubical transformers based on cores like E or ETD the constant Cth is about 50. Very flat designs like
Planar E cores have much better thermal properties, the constant is around 25. (Ve is in cm3).

Rth = Cth Ve-0.54 K/W

These thermal resistances were measured with the transformer mounted on a printed circuit board with no
local heat sources. They are referred to the hot-spot temperature in the middle of the centre leg of the core
(as deducted from the measured value on the ferrite surface).For hot-spot temperatures up to 100oC, the
maximum permissible winding dissipation may be obtained from :

Pw =

Rth

- Pc

This expression may be used for any ratio of core to winding dissipation.
In a choke with a large d.c. component of current, where core loss is negligible, P w = T / Rth.
The temperature rise should be checked again once the winding design is complete. Winding design is
treated in part 3 for transformers and Part 4 for chokes. If the actual temperature rise is too high, the
estimated throughput capacity was optimistic and a larger core should be used.

Part 2

6. Data necessary for winding design


Before it is possible to move to the design of the windings, the following data should be available.
Forward and push-pull converters
core type(s) (Section 1.1)
coil former dimensions (Ref. 1)
mains insulation requirements (Section 1.)
preliminary turns ratio (Part 1)
rounded number of secondary turns (Section 3.1)
operating frequency
secondary current and waveform
winding dissipation permitted (Section 5)
other windings - auxiliary outputs or sensor windings

Flyback transformers
core types (Part 4)
coil former dimensions
number of primary turns (Part 4)
preliminary turns ratio (Part 1)
operating frequency (Part 3)
primary and secondary currents and wave forms
other winding - sensor or auxiliary outputs
winding dissipation permitted (Section 5)
core loss (Section 4)
mains insulation requirements (Section 1.1.)

Output and flyback chokes


core loss (Section 4)
Proceed to part 4 for remaining design.

10

Part 3

Part 3 - Transformer winding design


1. Introduction
At ultrasonic frequencies, the simple design rules used for low-frequency transformer windings are no longer
valid. Winding resistance and thus winding loss are increased through eddy-current effects, the proximity
effect. At mains frequencies these effects can be ignored. The design methods and aids presented here aim at
finding the optimum winding design: that is, the winding geometry that yields lowest loss.
In drafting these design aids, the number of decisions made for the designer has been restricted to a
minimum. He thus obtains a maximum of freedom of choice. The essential background information will be
given after presenting practical design procedures. A certain amount of background information is required
to use that freedom properly.

2. Definition of the problem: boundary conditions


2.1. Given boundary conditions
The following boundary conditions are set once a core is selected:
frequency
primary and secondary r.m.s. currents
primary and secondary turns numbers
winding breadth
available winding height
permissible winding loss
Coil former and winding dimensions are defined in Fig. 1.
These parameters can only marginally - if at all - be varied in winding design, therefore, they are termed
given boundary conditions.

BCF
bw

c/2

HCF

c/2

H
screen

CL

7Z80234

Fig. 1 Definitions of coil former and winding


dimensions.

Winding breadth bw follows from coil former breadth BCF and the required creepage allowance c as set by
the insulation standard. Available winding height is the height of the winding window HCF in the coil former

Part 3
less the height occupied by the screens with the insulation, and the auxiliary windings (control windings not
taking part in power transfer.) The available height should accommodate the power-transfer windings. It is
not possible to determine beforehand the maximum height of individual windings.

2.2. Chosen boundary conditions


In the winding design process the designer can make his own choice of
winding configuration (Section 4.1.1.)
conductor form (wire, strip, bunched wire, etc.)
At the moment of choice, it is not always clear which choice is best. When a cheap design has an acceptable
loss it may be preferable to a more expensive design with even lower loss. It may therefore be desirable to
work out more than one design by making preliminary choices for these boundary conditions.

2.3.The design problem


The actual winding design is now to determine the most suitable winding geometry.
Solving this problem requires (preferably convenient) means and methods to determine.
for strip winding (strip width equal to winding breath)
- strip thickness
for wire windings
- number of layers
- wire size.
Having chosen the interleaving thickness, all further desired parameters can be calculated. Such means and
methods are presented here. For the prevailing set of (given and chosen) boundary conditions they are
directed to finding the geometry giving minimum loss.

Part 3

3. Design procedures
The design process is organized in the following five phases.
I

Collecting the boundary conditions.

II

Determining the ideal power windings.

III

Their evaluation for winding loss and required height. (According to the result, the design
process then branches to Phase IV of Phase V, or a new start is required.)

IV

Determination of the optimum combination of non-ideal designs that will fit into the
available height.

Design finalization.

Note: here, the term ideal is used to describe windings designed for minimum loss regardless of the height
of coil former required. Non-ideal windings have a height less than that required for ideal windings and the
lowest possible loss that this height permits.

3.1. Phase I :
Collect the boundary conditions
3.1.1. Given boundary conditions
Collect the given boundary conditions as set by circuit requirements and the core selected.
1. Core related :
operating frequency
maximum permissible winding loss Pw max
preliminary primary to secondary turns ratio r.
For forward and push-pull transformers:
number of turns in secondary Nsec
full load r.m.s. secondary current Ie 2 .
For flyback transformers:
number of turns in primary N1
full-load r.m.s. primary current Ie 1
full-load r.m.s. secondary current Ie 2.
2. Circuit related :
in the case of mains isolation, creepage distance c
data on windings other than power windings,
such as demagnetizing and sensor windings
single or multiple secondaries.

Part 3
3.1.2. Chosen boundary conditions
1. Winding configuration:
Choose between the simple configuration (where C = 1) or the split/sandwiched configuration (where
C = 2), see Sects 4.1.1 and 5.1. In the latter case, decide which winding will be split, and which sandwiched.
(It is usual to split the winding with the larger number of turns.) Sketch the complete winding arrangement
(not to scale), including any auxiliary or multiple-output windings, as in Fig. 4, 19, and 20. Draw also the
leakage flux (or NI) diagram and determine for each of the power windings the value of , the ratio of
minimum to maximum leakage flux density over the height of the winding. Normally = 0, but in multiple
output transformers it occurs that 0, see Sect. 4.5.1.
In the case of a flyback transformer, draw the leakage-flux diagram for both the period of primary
conduction and the period of secondary conduction, see Sect. 4.5.3. For push-pull transformers see
Sect. 4.5.2.
2. Conductor form :
Assume single round wire initially. The results thus obtained will guide the final choice of conductor.

3.1.3. Derived boundary conditions.


1. Effective frequency:
If the waveform of the current through the windings is known from previous designs, the effective frequency
fe can be obtained by analysis of the current waveform , as given in Sect. 6. If the waveform is unknown,
assume that fe is the switching frequency in kHz.
2. Effective skin thickness in copper at 100oC is
e = (5,62 / fe) (mm)
(with fe in kHz)
3. Number of turns
Forward and push-pull transformers :
secondary, N2 = Nsec/C
primary, n1 = Nsec r.
Initially, round n1 to the next even integer Nprim.
It is essential that Nprim is even if the winding is to be split, and it avoids a further rounding if there are
two layers in a simple winding. Then,
N1 = Nprim/ C.
Flyback transformers :
A flyback transformer may use a split/sandwiched winding configuration to reduce leakage inductance, but
this will not, however, reduce winding loss. The design procedures are always those for simple windings.
But, of course, the number of turns of the split winding must be rounded to an even number.
N2 = N1 /r
rounded to an integer.

Part 3
Multiple-output windings (see alsoSect. 4.5.1)
Calculate for the output with the lowest voltage, that is, with the smallest number of turns, and
round up to the next integer.
To obtain the number of turns for the other outputs divide the required output voltage by
the volts-per-turn value for the lowest-voltage winding. Round the numbers of turns of all output
to the nearest integer.
Check whether the various output voltages are within the required limits with these numbers
of turns.
If not, increase the number of turns of the lowest-voltage winding by one, and repeat the
calculations, starting with the volts per turn.
When the numbers of turns for the correct output voltage have been established, correct the
number of primary turns using the volts-per-turn value last found.
4. Winding window
Winding breadth bw = BCF - c, see Fig.1.
Available height: the height available for the complete power windings is HCF less
- an allowance for imperfect contact between layers, interleavings and screens;
- the height of screens and their insulation (C times), see Sec. 4.5.3;
- the height of auxiliary windings handling little or no power, such as :
a sensor winding on the primary side of the screen that may also supply a little power for
the control circuitry.
a demagnetizing winding.
Such windings are not designed for minimum loss. To save height, single layers are assumed of winding pitch
t = bw/(N + 1).
Choose a wire size such that tmin t and d e. Then with an interleaving of thickness i suitable for the
winding pitch, the required height is
H = do + i.
The height that remains after all deductions divided by C is Ha, the height available for one primary and
one secondary winding portion.

Part 3
3.2. Phase II:
Determine the ideal power windings
For push-pull multiple output and flyback transformers, first see Sect. 4.5.
Not yet knowing the value of p, multilayer windings are assumed initially If, during the design procedure ,
this assumption proves incorrect, the design procedure switches to that for single-layer windings. Follow the
procedure below for all power windings.

3.2.1. Ideal multilayer windings of solid, round wire


1.Calculate
3

d =

17.1 bw
N fe

2. If = 0, read Cp for the table below. If p is not known, p = 1.5 for a sandwiched winding or
p = 2 in other cases.
p
Cp

1.5
1.06

2
1.03

2.5
1.02

3 to 4.5
1.01

>4.5
1

if p > 3check
for errors

-If 0, calculate

1-
(1 - 3)1/3

Cp =
Then calculate did = d Cp.

3.Select the nearest standard wire size from a wire table (see page 49) and note d, do, tmin and rdc.
4.Number of layers
a

Pid =

N
bw / tmin - 1

Note: this expression is valid only for tmin from Step 3. If Pid 1.5 a strip or foil alternative may be
preferable, see Sect. 3.2.23.
If = 0 and pid 1, the expression for d in Step 1 is not valid. Go to the single-layer winding procedure,
Sect. 3.2.2.
b. Find p by rounding up Pid to the next suitable value: to a multiple of 0.5 for a sandwiched winding,
and to an integer in other cases.
c. Calculate Nl = N / p. If that value is not an integer, consider an adaptation of N to facilitate manufacture.
Start again with the new value of N.

Part 3
d. If = 0, check the value of p used equals that assumed in Step 1. If not, repeat from Step 2 using
the correct value of p.
5. Determine the winding pitch
t = bw /( Nl + 1 ) = pbw / ( Nl + p ).
Nl may not be the same in all layers (Step 3c); this will result in different values of t. Remember that all layers
should occupy the breath bw. do not allow a difference greater than one turn in Nl.
6. Select a suitable interleaving, thickness i, for the winding pitch and calculate the required height
Hid = p ( do + i )
7. Resistance factor FR = 1 + 1/2 (d /did)6 when d/did < 1.25. But when p = 1.5, this expression holds
only for d/did < 1.15.
8. AC resistance per metre length of wire rac = FR . rdc.
9. Winding loss of a complete winding ( of C portions ) Pw = C I2e N lav rac, taking for the average turn
length lav the value for a fully wound coil former in m.
Having found the ideal designs for all power windings using this procedure, proceed to Sect. 3.3 (Phase III).

Part 3
3.2.2. Single and half-layer windings of solid round wire
This procedure applies only when = 0.
Arriving form Step 4a of the previous procedure, take p = 1.
1. Winding pitch t = pbw /( N+p).
2. Select from a wire table the thickest standard wire for which tmin t and note d, do and rdc. Also choose
a suitable interleaving, thickness i.
3. H = p (do + i).
4. = (0.124 fe d3/t),
(fe in kHz, d and t in mm).
Note : If fe differs from the switching frequency, use the lower value of frequency.

FR

7Z89843

4
3
2
1,5

1,2

p=1

p = 0,5

1,1
1,05

1,02
1,01
0,5

10

Fig. 2 FR versus for single and half-layer windings.


Above FR = 4, take FR = p.

5. Read FR from Fig.2. Beyond the chart, FR = p.


6. rac = FRrdc.
7. Pw = C I2e N lav rac, where lav is in m.
In the case of a sandwiched winding, consider a strip or foil winding, Sect. 3.2.3. If a wire winding is preferred,
and if from Step 4 is about 4 or greater, a half-layer winding will probably have about equal rac, about half
the wire size and about one quarter of the copper content; follow the above procedure using p = 0.5.
Having found the ideal designs for all power windings, go to Sect. 3.3 (Phase III).

Part 3
3.2.3. Ideal strip or foil windings
1. If = 0, read CN from the table below.
N
CN

0.5
1.69

1
1.19

1.5
1.06

2
1.03

2.5
1.02

3 to 4.5
1.01

5
1

If 0, calculate
4

CN =

(1 - )3
1 - 3

Then determine fe (kHz).

hid = CN 9.74 /N fe mm, using fe in kHz.

2. Select the nearest available strip or foil thickness h.


3. FR = 1 +(1/3)(h/hid)4 when h/hid < 1.4.
4. rac = FR/(45bwh) /m (bw and h in mm).
5. Pw = CI2eNlavrac.
6. Select a suitable interleaving, thickness i, then Hid = N(h = i).
Note: this procedure is for copper at 100oC, resistivity.1/45 mm2/m. It can be adapted to aluminium
conductor (62% higher resistivity at 100oC). In the expression for hid (Step 1) re-place 9.74 by 15.8; in
Step 4, replace 45 by 28.
Compare Pw from Step 5 with that of the wire version of the winding (Sect. 3.2.1, Step 9 or Sect. 3.2.2,
Step 7) and choose the ideal solution for lowest rac. Having houdn the ideal design for all power windings
go to Sect. 3.3 (Phase III).

Part 3
3.3. Phase III
Evaluate the desigh with respect to winding loss and height.
The success of a design attempt is judged on the basis of winding-loss margin Pm and remaining free height
Hr. The winding-loss margin is calculated from Pw max and the total loss of the power windings as found in
Sect. 3.2.1. (Step 9), Sect. 3.2.2 (Step 7) and Sect. 3.2. (Step 5). The remaining free height is calculated from
Ha as found in Sect. 3.1.3 (Step 4b), and the heights of the power windings found in Sect. 3.2.1 (Step 6),
Sect. 3.2.2 (Step 3) and Sect. 3.2.3 (Step 6).
There are four possible results of the evaluation:
- Pm and Hr both positive : design attempt successful, proceed to Sect. 3.6 (Phase V),
- Pm and Hr both negative : boundary conditions do not permit a successful design. Make a fresh
start with a higher operating frequency or a larger core, for example.
- Pm positive and Hr negative : a frequent occurrence, especially at lower frequencies.
Try non-ideal winding designs requiring less height, Sect. 3.5 (Phase IV).
- Pm negative and Hr positive : reconsider the chosen boundary conditions and start again.
The background of Sect. 5 may be useful; some basic considerations are :
Split/sandwiched designs of offer lower loss than simple ones, specially at higher
frequencies.
Strip or foil windings tend to have lower losses than (single layer) wire windings.
If Hr is large, try replacing a single-wire winding by a multiple-wire or bunched-wire
winding.
Design procedures are given ins Sect. 3.4. A Litz-wire winding may also be possible If Hr is
particularly large. Evaluate the result again.
If adoption of one of these measures results in both Hr and Pm being positive, the design is successful; if not,
try a higher operating frequency or a larger core in a new design.

3.4. Alternative designs: multiple, bunched and Litz-wire windings


3.4.1. Ideal multiple-wire winding
For background and possible loss reduction, see Sect. 5.2.2.
1. Choose the number of parallel strands ns : usually 2 or 3.
2. Following the design procedure for ideal multilayer windings of solid, round wire, Sect. 3.2.1, as if the
winding had nsN turns. If step 4a results in pid 1, follow the single-layer design procedure, Sect. 3.2.2,
again replacing N by nsN.
3. Divide the value of rac by ns to find the a.c. resistance of the ns parallel strands.

10

Part 3
3.4.2. Ideal bunched-wire windings
Bunched wire consists of a few strands of insulated wire twisted together to form a bunch. The strand
diameter is not so mall that eddy-current loss is negligible. For background see Sect. 5.2.3.
The minimum winding pitch of a bunched wire of ns strands is nttmin, and the layer height without
interleaving is nhtmin, where tmin is the minimum winding pitch given in the wire tables (page 27) for a single
strand.Values of nt and nh for crude experimental bunches are given for guidance in the table below.
nS

10

nt
nh

2,45
2,31

2,94
2,69

2,98
2,93

3,11
2,93

3,61
3,16

3,89
3,16

4,34
3,26

3.4.2.1. Multilayer bunched-wire windings


1.Choose ns. A bunch of 7 strands gives the best space (copper) factor.
2.If = 0 read Cp from the table below. If p is not known, take p = 1.5 for a sandwiched winding,
and p = 2 in other cases.
p
Cp

1.5
1.06

2
1.03

2.5
1.02

3 to 4.5
1.01

>4.5
1

if p > 3check
for errors

If 0, calculate

1-
(1 - 3)1/3

Cp =

Then
3

did =

17.1 bw
ns N f e

3. Select the nearest standard wire size from a wire table (page 27) and not d, do, tmin and rdc.
4. Number of layers
a.

Pid =

N
bw / (nt tmin) - 1

Note: the expression holds only for tmin from the previous steps. If Pid 1, the expression in Step 2 does
not apply: go to Sect. 3.4.2.2.
b. Find P by rounding Pid up to the next suitable value. For sandwiched windings, round to multiples
of 0.5 ; in other cases, round to integers.

11

Part 3
c. Calculate.Nl = N / p If this does not yield an integer, consider adapting N to facilitate
manufacture. Start again using the new value of N.
d. If = 0 check, that the value of p found is that assumed in Step 2. If not, repeat from
Step 2, using the correct value of p.
5. Winding pitch t = bw / (Nl + 1) = pbw / (N + p).
Nl may not be the same in all the layers. This will result in different values of t, since all layers should occupy
the full winding breath. Do not permit differences of more than one turn in Nl.
6. Select a suitable interleaving, thickness i, for the winding pitch. The required height H = p(nhtmin + i)
where nh is given in table in Sect. 3.4.2.
7. Resistance factor FR = 1+1/2(d/did)6.
8. Resistance per metre length of bunch rac = FRrdc / ns where rdc is the resistance per metre length
of strand.
9. Pw = ci2enlavrac, where lav in m.

3.4.2.2. Single-layer and half-layer bunched-wire windings


This procedure applies only if = 0
Arriving from Sect. 3.4.2.1, Step 41, take p = 1.
1. Winding pitch t = pbw / (N + p)
2. From a wire table, select the thickest strand for which tmin t/nt and note d, tmin and rdc.
Also choose a suitable interleaving, thickness i, for the winding pitch.
3. Required height H = p(nhtmin + i).
4. = (0.124nsfed3/t).
Note : if fe differs from the switching frequency, use the lower of the two values (in kHz).
5. Read FR from Fig.2. Beyond the range of Fig.2, FR = p
6. The a.c. resistance per metre length of bunch Rac = FRrdc/ns, Where rdc is the d.c. resistance of a
metre length of strand
7.Winding loss Pw = CI2eNlavrac, Where lav is in m.
If in Step 4 is greater than or equal to about 4, a half-layerdesign might be preferable for a sandwiched
winding.

12

Part 3
3.4.3. Litz-wire windings
Litz wire is here taken to be a kind of unched wire in which the strands are so thin (d < e) that
the eddy-current effects can be neglected. It mains drawback is its low space (copper) factor, often only
25% to 30%.
Litz wire is standardized and commercially available. The Standards specify numbers of strands, strand
diameter, overall diameter and d.c. resistance (see also Sect. 5.2.4).
Mechanical stress deforms the funch so that the breath in the layer and, thus the minimum winding pitch
are greater than the overall diameter of the bunch, but the height of the layer is smaller.
1.Knowing the winding current IX, select a bunch of strand diameter d and number of strands ns such that
d2ns IX. The resulting current density will be about 4 A/mm2.
2. Due to deformation, the winding pitch t 1.2do, where do is the overall diameter of the bunch.
3. Provisional number of layers p' = Nt/(bw - t).
4. If p' is within 10% of the next lower integer, the winding may be feasible with care with this lower value.
Otherwise, take for p the next higher integer. Then calculate n's = ns(p/p')2.
5. Select a standard Litz wire with the same strand diameter, and a number of strands as large as possible,
But lower than n's.
6. The required height of the winding will be less than H = p(do + i), where i is the thickness of the
interleaving.
7. The d.c. resistance per metre length is usually given for 20oC, multiply that value by 1.3 to get the
resistance at 100oC. Then calculate Pw = CI2eNlavrdc, using lav in m. Note: If all windings in the transformer
are of Litz wire, the split/sandwiched winding configuration does not lead to lower losses.
If a suitable Litz wire is not readily available, the above procedure can be adapted to use multiple Litz wire
using similar methods to those for multiple wire, Sect. 3.4.1.

13

Part 3
3.5. Phase IV
Determine the optimum combination of non-ideal designs
When ideal designs overflow the winding space, non-ideal designs must be used. First, collect a few
non-ideal versions of, if possible, primary and secondary. The accommodation procedure will then show
which combinations fit into the available height. The one having the lowest loss can then be used as
the final design.

3.5.1. Non-ideal design versions in solid round wire


Non-ideal design versions of wire windings are generated simply by successively reducing the number of
layers to reduce winding height.
The procedure given in this Section is not valid when at the start p = 1: in that case use the procedure of
Sect. 3.2.2. However, it remains valid if step 2 yields values of p 1.
1. Note, for the foil former, bw and lav (is in m); and for the ideal design did (before rounding), N, Ie, p, t, d, H,
rac and Pw. A table such as that shown below will found a convenient design route.
2. Reduce p by one step
of 0.5 for a sandwiched winding
of 2 in other cases.
3. If Nl = N/p is not integer, consider adapting N to obtain an integral number of turns per layer. If possible,
let this adaptation cancel a previous one. Calculate the winding pitch t = pbw(N + p).
4. Select (from a wire table) The thickest standard wire for which tmin t and note d, do and rdc. Select
a suitable interleaving thickness i.
5. Required height H = p(do + i).
6. Resistance factor FR = 1 + 1/2(d/did)6.
If pid > 1, this expression holds for p 1, but this is not the case of pid 1.
7. Resistance per metre length of wire rac = FRrdc.
8. Winding loss Pw = CI2eNlavrac.
For each design version, repeat this procdure from Step 2.

14

Part 3
Having collected several versions of, as far aspossible, all power windings, follow Procedure 3,5.4. Example:
In this table of design versions, the 3-layer version is the ideal design, the others are derived using the
procedure above.

Primary: split, C = 2, bw =13.4 mm,


did = 0.437 mm, Ie = 0.565 A, lav =0.053 m.
P
N
t
d
do
rdc
FR
rac
pw
i
H

3
57
0.670
0.450
0.223
0.430
1.728

2
58
0.447
0.355
0.414
0.225
1.144
0.257
0.505
0.06
0.948

15

1
57
0.231
0.180
0.222
0.873
1.002
0.875
1.688
0.06
0.282

Part 3
3.5.2. Non-ideal design versions of strip or foil windings
Here, the required reduction in height is obtained by using thinner conductor.
1. Note, for the coil former, bw and lav, and for the ideal design N, Ie, hid (before rounding ) h, i, rac, H and
Pw. Set out a table in the form shown below.
2. For h, take the next available size below that used in the last design, and select a suitable interleaving
thickness in i.
3. Resistance factor FR = 1 + (1/3) (h/hid)4.
4. Resistance per metre length of conductor rac = FR/(45bwh) .
5. Winding power loss Pw = CI2eNlavrac .
6. Required height H = N(h + i) .
For each successive version, repeat from Step 2 onward.

The table below shows an example of design versions of a strip conductor.

Secondary: sandwiched, C = 2, N = 4, bw =13 mm,


hid = 0.222 mm, Ie = 8.05 A, lav =0.053 m.
h
i
FR
rac
pw
H

0.2
0.1
0.0104
0.286
1.20

0.15
0.06
1.069
0.0122
0.335
0.84

0.1
0.06
1.014
0.0173
0.475
0.64

0.073
0.04
1.004
0.0235
0.646
0.452

Once several versions of, preferably all power windings have been collected, follow the procedure in
Sect. 3.5.4.

16

Part 3
3.5.3. Alternative winding designs with reduced height
Multiple wire:
To change ns follow the procedure in Sect. 3.4.1. It is difficult to predict the value of ns that is the best
starting point for the following procedure when the loss margin is positive.
- If pid > 1, use the procedure in Sect. 3.5.1, but divide t (Step 3) and rdc (Step 4) by ns.
- For a sandwiched winding, if pid 1, use procedure 3.2.2 to design a half-layer winding, again dividing
t (Step 1) rdc (Step 2) by ns.
Bunched wire:
Use the procedure in Sect. 3.4.2.1 as follows.
- If pid > 1, commence with Step 4b and continue with values of p < pid.
- For a sandwiched winding, if pid 1 and p = 1, use the procedure in SEct. 3.4.2.2 with p = 0.5.
Litz wire:
Use the procedure in Sect. 3.4.3 with lower values of p.
Having collected several versions of, preferably, all power windings, follow the procedure in Sect. 3.5.4.

3.5.4. Accommodation procedure


This procedure seeks combinations of design versions that fit into the available winding height, together
with their total winding loss.
1. Calculate, for all versions of the primary, the maximum permissible secondary height, as shown in the
table below
H2 max = Ha -H1
Note, also primary dissipation Pw1 of all versions.
2. Select, for each primary version, the secondary that most nearly fills H2 max and note its dissipation Pw 2.
3. Calculate the remaining ree height
Hr = H2 max - H2
and total winding loss
Pw = Pw 1 + Pw 2.
4. Select the combination having the lowest value of Pw.
The following table illustrates the accommodation procedure for a transformer with two power windings.
The same principles apply for transformers with more power windings, even though the number of
combinations is greater. Much labour might be saved if design versions with excessive loss per mm of
winding height are not considered.

17

Part 3
Table illustrating the accommodation procedure for Ha = 1.778 mm and Pw max = 1.23 W.

primary versions
(split. single wire)
see Sect. 3.5.1

p1
pw 1 (W)
H1 (mm).

3
0.430
1.728

2
0.505
0.948

1
1.688
0.282

Height available
For secondary

H2max

0.050

0.830

1.496

Secondary versions
(Sandwiched. strip)
see Sect. 3.5.2

H2 (mm)
h (mm)
pw 2 (W)

0.84 0.64
0.15 0.1
0.335 0.475

1.20
0.20
0.286

Remaining height
Total winding loss

Hr (mm)
pw (W)

-0.01 0.19
0.840 0.980

0.296
1.974

Since the maximum permissible winding loss is 1.23 W, the 2-layer primary combined with the secondary
using 0.1 mm strip results in a satisfactory design. There is no need to consider the secondary using 0.15
mm strip, although the excess height of 0.2 mm (C = 2) could, perhaps, be accommodated by careful
manufacture.
Note: The example is for a split/sandwiched winding configuration (C = 2), so that winding heights are
for one portion of each winding.

3.6. Phase V:
Finalizing the design
In order to prepare the transformer design for production, four further steps are necessary.
1. Check that the design route followed was the correct one, and that the correct value of windingconfiguration factor C was used throughout.
2. Make a dimensioned sketch of a cross-section through the windings in the winding window. Calculate the
conductor lengths required for each winding, using the actual average turn lengths. For wire windings, check
that Nl turns of the wire selected will fit into the winding breadth, that is, that bw (Nl + 1) tmin.
3. Prior to prototype evaluation, make a final estimate of the transformer temperature rise using winding
losses recalculated from the actual winding lengths obtained in the previous step.
4. In preparation for manufacture, collect all required information about core and coil former, windings and
interleavings, screens and their interleaving, interwinding insulation, and terminations. Specify winding pitches
for wire windings: windings are generally not close-packed since all layers should be of equal breadth.
This concludes the practical design procedures.
The following sections contain supplementary information only.

18

Part 3

4. Background
The essential difference between a designer and a computer is creative thinking. Thus, a true designer will
find no lasting satisfaction in following design procedures that resemble computer flow charts.
He cannot help asking why the procedures are as they are, and how they can be extended to solve
related problems. This Section seeks to answer these and other questions with a view to avoiding mistakes
due to misinterpretation of the instructions. It is mainly qualitative in nature : mathematics have been
reduced to a minimum, and higher-order effects, although included in the formulas behind the design aids,
are generally not discussed.

4.1. Eddy-current effects in windings


4.1.1.Transformer magnetics
The main flux that couples primary and secondary windings is core bound. Since primary and secondary
ampere-turns oppose, the instantaneous value of the mutual flux is determined by the instantaneous
difference between them. The magnetizing flux is proportional to the induced m.m.f. per turn. In an ideal
transformer (zero core reluctance and zero winding resistance) with a short-circuited secondary, the
induced m.m.f. and, thus, the magnetizing flux are both zero. Flux also crosses the winding space. this
flux is not common to all windings, and not even to all turns in the same winding, and is, therefore,
known as leakage flux.
Figure 3a shows the leakage flux paths in a simple, ideal, short-circuited transformer (with no main flux).
Within the winding space, primary and secondary leakage flux lie in the same direction. Since they mutually
repel, they are in parallel with the interface between primary and secondary: that is, in parallel with the
layers if the windings are on top of each other. Thus, in order not to jeopardize this parallelism, only
complete layers of equal breadth will be considered.
The leakage flux density is maximum at the interface between primary and secondary, Fig.3b. On both sides
of this maximum, the flux density falls roughly linearly to zero over the height of the winding.

7Z80235

0N2I2/b

secondary

H1

primary

0N1I1/b

C
L
(b)

(a)

Fig. 3 (a) Leakage flux paths in an idealized,short-circuited (to


main flux) transformer. (b) Flux-density distribution.

Consider a flux path crossing the winding window at a distance x from the wall of the coil former. Fig. 3a.
Assuming that the current density over the height of the primary is constant,

19

Part 3

B ds = NI
0

x
H1

where B is the flux density, s is the distance along the flux path, and NIx / H1 is the number of ampere-turns
enclosed. The field strength is negligible in the (high-permeability) core, and is assumed constant over the
winding breadth, so that the leakage flux density

B = 0 N I x
b
H1
where b is the flux path length outside core material.
Figure 3b shows how B varies over the height of the wound area. There is a good case for arguing that,
if bw is smaller than the width of the winding window (perhaps due to creepage allowance), b should be
taken equal to bw (Ref. 2, page 355). Note that leakage flux is not due to imperfect core material, but is
an intrinsic property of a winding.
The leakage flux through the windings gives rise to eddy currents in the conductors. In a transformer
of normal construction, the leakage flux lies parallel to the layers of the winding and roughly normal
to the turns.
Note that the leakage flux density varies from layer to layer and the leakage flux density is proportional
to the sum of the ampere-turns in that layer and the ampere-turns in the layer between that layer and the
nearest point of zero flux density. The average flux density and, thus, the eddy-current losses, can be reduced
by suitably mixing primary and secondary windings. One result of this process is the split/sandwiched
configuration of fig. 4b.

7Z80236

primary
(a)
secondary
B
primary/2

(b)

secondary

primary/2
B

Fig. 4 The effect of splitting a transformer primary into two


portions either side of the secondary is to halve the peak
leakage fulx and, consequently, the eddy currents.

Figure 4a again shows the leakage flux distribution of a basic transformer winding arrangement.
In Fig. 4b, the primary winding is split into halves, with the secondary sandwiched between them.
This halves the peak flux density.
To make this construction possible, the split winding must have an even number of turns, and an even total
number of layers. The sandwiched winding, although physically one unit, can also be regarded as consisting

20

Part 3
of two portions; its leakage flux distribution is shown in more detail so that it can be seen that it may
have an odd number of layers. Then, each portion contains a half layer ; a layer having half the height
of a normal layer. Similarly, a portion of a sandwiched winding may contain a half turn if the half layer
has an odd number of turns.
Due to the symmetry of the split/sandwiched configuration, only one portion of each winding need be
considered in calculating the eddy-current effects.
The splitting and sandwiching process could, of course, be repeated to further reduce eddy-current
loss. However, this quickly becomes unpractical: each interface between primary and secondary portions
requires extra insulation, usually including screens to reduce radio-frequency interference. Their presence
reduces the space (copper) factor attainable in the winding window, eventually to such a degree that an
improvement is either lost or not worth the extra complication.
A distinction can thus be made between
simple windings as in fig. 4a.
split windings, such as the primary in fig 4b
sandwiched windings, such as the secondary in fig.4b.
The design process of Sect. 3 deals exclusively with winding portions:
a complete simple winding
one half of a split winding
one half of a sandwiched winding.
In the split/sandwiched winding configuration, the number of turns in a winding portion is half that in the
complete primary or secondary.

21

Part 3
4.1.2. Penetration of an electromagnetic wave into a conductor
Eddy currents are induced in a conductor exposed to an elctro-magnetic wave. They oppose the penetration of the wave and, in resistive conductors, transform electromagnetic energy into heat.
Let plane x = 0 be the surface of a conductor of infinite depth. The electric and magnetic fields of a plane
electro-magnetic wave propagating in the non-conducting semi-space above the conductor, and incident
perpendicularly upon it, are tangential to the surface and mutually perpendicular. If the positive x axis is into
the conductor, the penetrating wave can be described by

{ - x + j ( t - x + )}

Constant A and (part of the incident energy is reflected from the surface) are not of interest here.
The amplitude of the wave decays within the conductor as e-x/, where depends on the properties of
the conductor and the frequency of the wave.
At a depth x = , the amplitude of the wave has decreased to 1/e of its value at the surface and the phase
is delayed by one radian. At a depth equal to a small multiple of , there is almost no field in the bulk of the
conductor and, consequently, no induced current.
At x = 2 where the phase delay is just 2 radian, the field strengths decay to the negligible value
e-2 = 0,0019. This is at a depth one wavelength of the penetrating wave.

4.1.3. Skin effect


A straight, isolated, round wire carrying alternating current generates a concentric, circular magnetic field,
both in the wire itself and in the surrounding space.
Isolated in this context means that there are neither other conductors nor magnetic fields in the vicinity of
the wire. The low frequency field distribution under these circumstance is shown in Fig. 5.

Ix
B= 0 2
2r

x
7Z80237

Fig.5 Low-frequency distribution of flux density B about a round


current-carrying conductor. Here, x is the distance from the centre of
the conductor, and r is the conductor radius.

22

Part 3
The field is tangential to the surface of the wire. This field induces eddy currents that oppose its
penetration, enhacing the current flow near the surface and reducing it near the centre of the wire as
shown in Fig.6.

eddy
currents
magnetic
flux

I
7Z80238

Fig. 6 The magnetic field generated by an a.c. current in a wire


induces currents in the wire that oppose its own penetration.

As was the case with the penetrating electromagnetic wave, there is a tendency for current to flow only
near to the surface of the wire: the skin effect. Figure 7 shows the current distribution carrying the same
alternating current at various frequencies. As frequency increases, reduces and d/ increases.
This current redistribution results in the a.c. resistance of the wire being greater than its d.c. resistance.
The voltage is constant over any cross-section of the wire normal to its axis since there can be no
radial current flow. In general, the voltage across a length of the wire is the sum of resistive and induced
voltage drops. With a relatively thick wire (d/ >> 1), the voltage drop near the centre is mainly induced
by the eddy currents, whereas, near the surface, where the current density is high, the voltage drop is
mainly resistive.
If the wire is replaced by a tube of the same material and diameter, such that the tube has the same d.c.
resistance as the wire for a.c., its wall thickness will be equal to , provided that the curvature of the
surface is negligible (d >> ). For this reason, is known as the (equivalent) skin thickness.
The term penetration dept is often used for , but some authorities prefer this term for the wavelength
(2) of the penetrating wave discussed above. The (uniform) current density in the equivalent tube is equal
to that at the surface of the wire it replaces. The ratio of a.c. resistance to d.c. resistance due to skin effect
can be deduced from the relative cross-sectional areas of wire and tube :
FR d , ( d >> 1, in practice FR 1 ( d + 1 ) if d 5 )
4

The a.c. resistance per meter length of wire is proportional to d-1, although the d.c resistance is
proportional to d-2.

23

Part 3

7Z89855

4
d/ =
20
3
10
7
5

2
4
3
1

0
d

Fig. 7 Current distribution in a wire carrying constant a.c. at


various frequencies.

Skin thickness depends on conductor material and frequency.

0 r f

The current redistribution establishes an equilibrium between inductive and resistive voltage drops. Increasing the frequency does not alter the flux density at the surface because the current remains constant, but
does cause an increase in the induced voltage which results in a smaller skin thickness and, thus, a higher
current density. This, in turn results in a higher resistive voltage drop which opposes the concentration of
current at the surface. A new equilibrium is achieved at a skin thickness proportional to 1/ f.
Similar reasoning can be established for the effect of r and .
The skin thickness itself is independent of the current carried by the wire. Because the wire is immersed in
its own field, there is a fixed relationship between average current density and magnetic flux density.

24

Part 3
4.1.4. Proximity effect
A different class of eddy currents is found in wire exposed to an external alternating magnetic field
normal to the axis of the wire. In that case, eddy-current flow in opposite sides of the wire is in opposite
directions, Fig. 8.
external field

field due to
eddy currents

eddy
currents

7Z80239

Fig. 8 Eddy currents induced in a conductor exposed to an


alternating magnetic field normal to its length. A rectangular
conductor is shown for clarity.

Eddy-current flow is confined to a skin below the conductor surfaces that are tangential to the external
field. No net current flow is assumed in the wire. To avoid the complications of a curved surface, a
rectangular conductor is shown in Fig. 8.
The situation resembles that of a turn in a winding exposed to leakage flux of the type described in
Sect. 4.1.1. The essential difference is that the turn carries an alternating current, and that the leakage
flux density is proportional to this current. The relationship between current and leakage flux is not
the same all over the winding, but depends on the position of the layer which contains the turn under
consideration.
Because the leakage flux to which the turn is exposed originates from other turns in proximity to it, the
eddy-current phenomenon is known as the proximity effect in this case.

(a)

(b)
7Z80240

Fig. 9 Four turns from a layer of a winding: the concentric flux


due to the winding current (a) cancels between turns so that the
resultant flux lines are parallel to the layer (b).

25

Part 3
Figure 9a represents a section through a few turns in a layer with the concentric flux paths associated
with an isolated wire. Between the turns, however, individual flux lines, which are roughly perpendicular to
the layer, oppose and so tend to cancel. The result is the flux pattern of Fig.9b. Such flux patterns from a
number of layers result in the flux distribution of Figs. 3 and 4.
The difference between the two types of eddy-current effect is now evident.
Skin effect: tendency for the current to flow near the conductor surface, no current reversal and thus
no increase in current flow.
Proximity effect: there are two skin regions below the surfaces tangential to the magnetic field: besides
the tendency for the main current to flow in the skin region where the magnetic field is highest, the skin
regions carry opposite eddy currents that increase the effective current flow.
The leakage flux density varies from layer to layer, but is constant over the breadth of each layer (Figs 3
and 4). The proximity effect is greatest in the layers at both sides of the interface between primary and
secondary, where the leakage flux density is greatest. The simplest case is that of a strip or foil winding,
where each layer has only one turn. The number of layers (and turns) is thus known at the start of the
design process. Figure 10, which is based on normalised conductor height, shows the relationship between
proximity effect and leakage flux density.
The proximity effect in a wire winding depends not only on wire diameter, but also on the layer space
(copper) factor. The main geometrical parameters of a turn in a wire winding are wire diameter d and
winding pitch t. Of course, t do . A useful simplification is to regard a round wire as a square wire of equal
cross section and, thus, equal current density, Fig. 11. Then, h = d (/4) 0.886d. Extending this model
allows a layer to be regarded as a strip of thickness h and layer space (copper) factor Fl = h / t carrying a
current Nl times greater than that in the wire. The leakage flux is the same as that of the wire-wound layer,
but the current density is 1/Fl times greater
1stlayer

2ndlayer

3rdlayer

4thlayer

5thlayer

=0,59

(a)

1,01

1,09

1stlayer

1,25

1,49

1,82

2ndlayer

3rdlayer

1stlayer

2ndlayer

3rdlayer

4thlayer

5thlayer

=8

50

10

25

300

3000

3
200

(c)

5thlayer

20

(b)

4thlayer

=2

1,33

1,9

8,4

21,4

40,9

100

5
4
3
2
1
0
-1
-2
-3

5
4
3
2
1
0
-1
-2
-3

7Z89852

2000

66,9

7Z89853

40

105

202

331

27,9

1000
0 137
5
4
3
2
1
0
-1
-2
-3

7Z89856

Fig. 10 The proximity effect in a winding of 5 layers. Effective conductor heights are = 0.59 (left), = 2 (centre) and
= 8 (right). Sections through the ends of the layers of the winding are shown, hatched.The layers are of strip conductor,
but could also be Nl turns of rectangular wire.The current density distribution is plotted in (a).The full line is the
amplitude, the broken line the real part, and the dotted line the imaginary part.The losses are plotted in (b).The loss
distribution (solid line) is proportional to the square of the current density amplitude.The broken line shows the average
loss in each layer, and the dotted line the average loss in the complete winding. Since the scale calibration is multiples of
the d.c. loss, the dotted line shows the value of FR. (c) is the leakage flux diagram, obtained by integration of the current
density. Scale divisions are layer current; line types are as in (a).

26

Part 3
The equivalent conductor height = h/ for strip or foil windings, and = (h/)Fl = (d/)(Fl4/)
1.128 (d/)Fl for windings of solid round wire
Comparison of h/ for strip with (h/) Fl for wire indicates that the skin dept in wire seems 1/Fl
times greater than in strip. This can be understood by remembering that is proportional to and
that increasing or increasing the current density has the same effect on the resistive voltage drop on
which the equilibrium depends.

do

(a)
h

h
(b)

7Z80241

Fig. 11 (a) Principal geometrical parameters describing a turn in


a given layer. (b) Replacing a round wire with a square one of
equal cross sectional area simplifies the discussion since now a
layer resembles a turn of strip conductor.

4.2. A.C. resistance


4.2.1. Resistance factor FR
The increase in conductor resistance due to proximity effects can be expressed in terms of resistance factor
FR: the ratio of a.c. resistance to d.c. resistance.
The chart of Fig. 12, based on Dowells work, Ref.1, gives FR as a function of equivalent conductor height
with the number of layers p as a parameter. Inspection of these curves shows that there is a region
where FR is proportional to . In multi-layer windings, this region commences when is greater than about
3, but for single layer windings, when 1. This is the region of skin conduction, where no current
flows in the bulk of the conductor. The situation is similar to that for skin effect when d/ > 1, where
FR is proportional to d.
The curve for p = 1 (eddy-current loss negligible when 1) also represents the behaviour of the first
layer of a multilayer winding. The curve for p = 0.5 shows similar behaviour, but at double the values of
. In all layers of a multilayer winding except the first, eddy-current loss (proportional to FR - 1) increases
as 4 up to 2, and in proportion to above 3. This most clearly demonstrated by the curve
for p = 1.5. Eddy-current losses in the higher layers evidently dominate the a.c. losses in the first layer,
unless < 1. This explains the difference that will become apparent later between single and half-layer
windings, and multi-layer windings.
It should be noted that eddy-current loss remains proportional to 4 in the low bend of the curves, below
FR 2 The bend is due to d.c. loss not being negligible compared to eddy-current loss in that region (see
also Fig. 2, mathematically a plot of FR 1, scale calibrated for FR).

27

Part 3
Figure 12 is most useful for determing the a.c. resistance of a winding of known geometry. However, it is not
a convenient basis for optimizing winding geometry itself.

7Z89845

FR 1000

p=
10
500

200

100

4
3

50

2,5
2
20
1,5
10
1
5

0,5

1
0,1

0,2

0,5

10

Fig. 12 Resistance factor FRversusequivalen conductor height ,


with number of layers p as a parameter, after Dowell (Ref.1).

4.2.2. A.C. resistance per metre conductor length rac


The A.C. resistance per metre conductor length is a useful quantity for comparing various versions of a
given winding. Winding loss is proportional to winding Rac = Nl av rac. In a given design problem, the number
of turns and coil former dimensions are known. Conductor length Nl av is only slightly influenced by the
number of layers or the conductor size. Thus, achievement of the design goal - the winding geometry for
minimum loss - can be reduced to the achievement of minimum rac.

28

Part 3
4.2.3. Windings of strip of foil conductor
The geometry of windings of full-width strip or foil conductors is completely estblished once strip thickness
h is known. Strip breadth is bw, and the number of layers is equal to the number of turns in the
winding portion. These are all set by the boundary conditions, so that the only parameter remaining to
the determined is thickness. Since rac is proportional to FR/, its value can be derived from Dowells
chart, Fig. 12.

rac = rdc FR = FR = . FR (since = h/)


b w

bwh

thus

rac bw = FR

Therefore, it is sufficient not plot FR/ against as in Fig. 13. There, the straight, dash-dot line labelled
FR = 1 represents rdc. Since rdc is proportional to 1 its slope is -1.
Each curve in Fig. 13 has a minim, marked by a dot. Each minimum occurs at the normalized, ideal strip
thickness hid/ and the lowest possible value of racbw/ from which the value of rac id can be calculated.
The term 'ideal' is used in preference to 'optimum' since the former implies desirability but not necessarily
feasibility. A thickness other than hid may have to be used due either to conductor or space availability
limitations.
rAC

bW

7Z89844

20

N = 20 15 10 8

5
4

10

3,5

2,5

2
1,5

0,5

0,5

FR= 4/3
FR= 1
0,2
0,1

0,2

0,5

10
h/

Fig. 13 Plot of FR/ versus which virtually solves the design


problem for strip windings. Given the number of turns N, the
minumum in the appropriate curves gives the ideal (lowest loss)
strip thickness as a multiple of the skin thickness.The resistance
per metre length of strip can also be determined.

29

Part 3
As, for a given , strip thickness increases beyond hid eddy-current loss increases more quickly than d.c. loss
decreases. Note that hid is independent of winding current because leakage flux density is proportional to
current density. All minima in Fig. 13 lie close to the straight line FR = 4/3, although some deviation can be
seen at low values of N. In the practical design procedure, in Sect. 3.2.3, hid is found assuming this straight
line, but a correction factor CN has been added for the deviation that occurs when N < 5.
since, usually h hid, the approximation
FR = 1 + (1/3) (h/hid)4
is included in the design procedure to give the a.c. resistance.
Foil or strip conductor is only recommended for sandwiched windings, not for split or even simple windings.
There is no guarantee that current distribution is uniform over the breath of a strip. In fact, there is a
tendency for current density to be higher near the edges of a strip when << (bwh), which is invariably
the case (Ref. 3). Non-uniform current density cannot occur when the leakage flux is truly parallel to
the layers, so advantage should be taken of the repulsive forces between the leakage flux of primary and
secondary (Sec. 4.1.1). Sandwiching a strip winding between two portions of a wire winding, whose current
density must be uniform over the layer breadth, ensures that there are strong repulsive forces on both sides
of the strip winding, where leakage flux density is maximum.

4.2.4. Wire windings : winding breath-to-turns ratio T.


The a.c. resistance per metre length of wire rac can be derived from a plot of FR/ 2 against , Fig. 14.
For rectangular wire of breadth b (in the layer) and height h (normal to the layer)

= h

b
t

2
2
and FR = rac 2 t = rac t
h
2 rdc h b

For round wire of diameter d

= h

b =
t

3/4

( )

d =
t

d
t

3/2

so that, for agiven ratio t/d

F R = r 2 t = r 2 t
ac
ac
d
h
2

is proportional to rac.
Here, rdc -2 so that the line FR = 1 has a slope of - 2. Whereas Fig. 13 gave the solution of the design of
a strip winding, where the number of layers and conductor breadth were known, the plot of Fig. 14 does not
contain the solution for wire windings. The number of layers is still not known.
The winding breadth is introduced in the form of the winding breadth-to-turns ration T = bw/N.
Physically, T is the winding pitch in a single-layer winding containing all N turns. Note that at the interface
between primary and secondary, the leakage flux density B = 0 I/T, Sect. 4.1.1 and also that T is known
at the start of the design process.

30

Part 3
Since t = pT, for a rectangular wire

= h

b
pT

and for round wire

FR/2

3/4

( 4 )

d
pT

7Z89846

100

50

20
p=
10

10

5
6
2
4
1

3
2,5

0,5

2
1,5

0,2
FR= 1
0,1
0,1

0,2

0,5

1
0,5
5

10

Fig. 14 Plot of FR/2 versus , a first step towards the solution of


the design problem for wire windings .

31

Part 3
4.2.4.1. Close-packed windings of round wire
For close-packed windings, rac2/ can be plotted against T/, Fig. 15; this gives better access to the design
problem. In close-packed wire windings, t = do, so the layer space (copper) factor depends on the ratio d/do
for the wire. This ratio is not constant : standard-wire tables show that d/do is smallest for fine wire, but the
rate of change is rather low. If. Fig. 14 were replotted to give rac2/ against, t, the calibration for the axes
would , of course, change, but the shape of the curves would hardly be affected because the rate of change
of d/t is so low. In order to replace t by T, the values for each of the curves must be divided by p, since T =
t/p. This shifts every curve horizontally a distance -log p.
The introduction of T is an important step forward: knowing T/, it is possible to read the lowest possible
value of rac2/, the optimum number of layers, and, finally, from do = t = pT, the optimum wire size
can be determined.
In order to plot Fig. 15, values of d/do and its variation with wire size were estimated. So, although the
plot of Fig. 15 is adequate for illustrating the theory, its accuracy is not sufficient for practical design
purposes.
rAC2

7Z89847

4 3
87

p=
0,5

0,5

0,2
simple, split or sandwiched
sandwiched only

0,1
0,05
0,1

0,2

0,5

10

20
T/

Fig. 15 Plot of rac2/ versus T/ with p as a parameter for


close-packed wire windings.

32

Part 3
4.2.4.2. Spaced windings of round wire
Windings are also possible with a pitch greater than the overall wire diameter: spaced windings. This
introduces an additional degree of freedom that can be exploited to achieve a further reduction in a.c.
resistance. This is illustrated by the plof of Fig. 16.
The solid curves in Fig. 16 are for close-packed windings (t = do), as in Fig. 15. The dash-dot curves are for
windings of two layers, in which t > do. These show that there is a range of T/ where spacing results in
lower values of rac. However, spacing should not be excessive : better performance is always possible if a
smaller number of layers can be used.The straight dashed lines in Fig. 16 are in further addition : they show
that spacing displaces the curves so that similar points lie on a line of slope -2/3. They also show that the
curves for p < 1 are so steep that spacing brings no improvement.
In effect, spacing allows designs intermediate between close-packed versions.

rAC2

7Z89848

p=3 p=2

p=1

0,5

0,2

0,1
0,2

0,5

10
T/

Fig 16 Plot of Rac2/ versus T/ extended for spaced windings

4.2.5.The basic solution of the design problem


Sufficient information is now available to make a plot that effectively solves the design problem. This is
given as Fig. 17. Figure 17a differs from Fig. 14 in that spaced windings are included where they give lower
loss. Furthermore, those parts of the curves for close-packed windings that are not useful for design
purposes have been suppressed.
The full lines apply to optimum designs with an integral number of layers. The broken lines apply to
optimum designs with a half layer (except in the plot of winging height, these often coincide with the
full-line curves). The dotted lines are for non-optimum, close-packed windings.
There it can be seen that, if T/ is known, the optimum design is completely determined.

33

Part 3
Further inspection reveals that :
- for T/ greater than about 2, the optimum design is a single-layer winding but, in sandwiched designs, when
T/ is greater than about 6, the optimum is a half-layer winding
- for T/ less than about 2, the optimum design has more than one layer.
Both the full and the broken lines for p > 1 show that the resistance minima occur generally for spaced
windings (Sect. 4.2.4), so that it is important to include in the manufacturing instructions a specific statement
that windings should not be close packed. The curves for p 1 have a slope of about -1, indicating that
FR 1, so that they belong to the region of sin conduction (Sect. 4.2.1.). Moreover, they are, in principle,
close-packed windings (Sect. 4.2.4.1). It is clear that multilayer (p > 1) windings and those with p 1 will
require different design methods.
rAC2

7Z89849

7
8

4 3 2 1,5 1

p=0,5

1
0,5

0,2
0,1
0,05
0,1

0,2

0,5

10

20
T/

7Z89850

d/ 10
5

1
8
7 5 4 3
0,5 6
0,1
0,2

He/

2 1,5

0,5

p-0,5
1

10

20
T/

7Z89851

10
8 7
6 5
4,5 4
5
3,5

1
3
2,5

2
1,5

p=0,5

2
1
0,1

0,2

0,5

10

20
T/

Fig. 17 Basic design charts for wire windings: (a)Rac2/ , (b) (a)
d/, and (c) He / versus T/ with p as a parameter. He is the
required height excluding interteaving. Knowing T/ , the ideal
winding geometry and the resistance per meter length of wire
can be determined.

34

Part 3
4.2.5.1. Multilayer windings (p > 1)
The full lines in the chart for rac2/ and d/. Fig. 17, consists of several sections, each for a particular value
of p. Together for p > 1, they form nearly straight lines.
The ideal values of d an rac are, thus, generally independent of p. There is only a slight deviation for the
lower values of p: it can be seen that the lines for p = 1.5 do not coincide with those p = 2.
The resistance factor for optimum or ideal design FR id 1.5, which means that the eddy current loss is half
the frequency-independent resistive loss, also called d.c. loss.
The slope of the straight liens indicates that rac id (T /)-2/3 and did (T / )1/3.
Since rdc d-2, FR = rac/rdc must be constant.
If the actual wire diameter d deviates from the ideal diameter did, the estimated resistance factor becomes

FR = 1 +

( ddid )6

This approximation, derived form the first two terms of the series expansion of Dowells expression, Ref. 3
is sufficiently accurate for values of FR 2. In an ideal winding, where d = did, FR < 1.5.
The expression is not valid in the region of skin conduction. To find the ideal wire diameter for windings
having a large number of layers, the expression d/ = 1,45 (T / )1/3can be used. In a more practical
form this expression reads
3

d =

17.1 bw
N fe

The ideal wire diameter for a practical winding, did, is then obtained using the correction factor Cp.
This factor corrects for the effect of the small steps between the line sections for practical numbers of
layers. Using these expressions, ideal designs can be obtained by simple calculation.
Once knowing did the winding geometry can be determined. After rounding did to the nearest standard
wire size, do is known. Following the design rules in the practical sections, the number of layers and the
required height are easily found. Knowing rdc of the selected wire, rac can be estimated after calculation
of FR.

4.2.5.2. Single-layer and half-layer windings


This is the region where the procedures of the previous section lead to a number of layers p 1.
This definition is given because the criterion T/ > 2 is only an indication. First is necessary to decide
between strip and wire windings. Strip of foil windings are often preferable, see Section 5.2.1. For simple
and split designs, it is known that p = 1. This is also the case in sandwiched designs if T/ is below about 6.
Since windings should be close packed, geometry is fixed. the end of the curve for p = 1 at T/ = 6 is not
a theoretical limit, the curve can be extended. But d may be so great that a different form of conductor will
probably be chosen (Section 5.2). For sandwiched windings, a half-layer winding may be preferable, provided
T/ is greater than about 6. That saves winding height without a loss penalty. In case of doubt (T/ 6), a
choice can be made after having worked out both alternatives.

35

Part 3
The remaining problem is the estimation of rac. To solve that, calculate

3/4

( 4 )

d
pT

or, alternatively,

0.124 fe d3
t

and read FR from the plot of FR versus , Fig.2.


Figure 2, gives the FR plot in a more suitable form than Fig. 12. Do not use the expression for FR given in the
previous section; it is not valid in the region of skin conduction.

4.3. Winding height


4.3.1. Ideal designs
In previous sections, methods were developed for designing minimum-loss windings of both strip and wire
conductors. In developing these methods, possible height restrictions have so far been ignored.
By doing this, each winding could be treated separately, the design of a primary was not influenced by the
secondary, and vice versa. Such designs, affected by a height restriction, are called ideal designs.
Similarly, terms such as ideal windings, ideal number of ayers, ideal conductor size, etc., are also used.
It is, of course, necessary to check whether the situation is ideal : that the height of the winding window
does not impose a height restriction on the windings.
The height of a winding (portion) is calculated from:
for strip windings, H = N (h + i)
for wire windings, H = p (do + i)
For the available winding height (height available for all power-transferring windings) see Section 2.1.

4.3.2. Non-ideal design versions


In the strip winding the only way to reduce winding height is to use thinner strip.Various non-ideal design
versions are found by repeatedly stepping to a smaller available strip thickness. The a.c. resistance is found
from the expression for FR in Sect. 4.2.3. In wire windings a smaller winding height is obtained by reducing
p. This leads to reduced pitch (t = pT) and thus, also, to reduced wire size (do t), so that d < did. To
reduce the difference between d and did as far as possible, a non-ideal wire winding should be close-packed
using the thickest possible wire for the number of layers and winding breadth. The a.c. resistance can be
estimated by means of the expression for FR in Section 4.2.5.1. Several non-ideal design versions are found
by repeatedly stepping to a lower number of layers. Thus, non-ideal as well as ideal design versions can
be found. Although these non-ideal design versions have smaller winding height than the ideal design,
they also have higher loss.
IMPORTANT: There is no justification for using conductor sizes in excess of the ideal size (apart from
rounding a calculated ideal size to the available size). Reduction of current density is a false motive. The
extra conductor material has only and adverse effect.

36

Part 3
4.3.3. Accommodation
Once a few design version (the ideal design and one or more non-ideal versions) of the primary and
secondary have been collected, selection of the most suitable combination is made by an accommodation
procedure.The accommodation procedure is given in detail in Sect. 3.5.4.
It comprises the following basic steps:
- list the design versions obtained, the height they require and their losses.
- make combinations of primary and secondary designs having a total height not exceeding the
available height, and note the winding loss of each.
- finally, select the lowest-loss combination.
Demagnetizing and control windings are usually designed for minimum height rather than minimum loss,
since they do not take part in power transfer. Their height is taken into account in determing the height
for the power windings.

4.4. Design evaluation


A design attempt is successful if the transformer does not run hot. Final evaluation is, of course,
experimental. Unnecessary experimentation can however, be avoided if the estimated losses are first
compared with the estimated permissible loss. The permissible loss is estimated from the permissible
temperature rise and the thermal resistance given in the data of the core selected. The thermal resistance
in the published data can be reduced by potting the transformers. If this is to be done, it might be possible
to eliminate the creepage distance, so increasing winding breadth. Core loss was estimated in the process
of core selection and determination of the number of primary and secondary turns. The loss in each
winding can be estimated as
Pw = I2nlavrac
Current densities much higher than are usual in low-frequency transformers may well be acceptable.
As has been mentioned, conductors in excess of the ideal size make matters even worse.
If a design attempt fails, it is advisable to first review the chosen boundary conditions. This would
not invalidate the results of the core selection procedure. An improvement may be obtained by using
spit/sandwiched rather than simple windings. These tend to lower rac, thicker conductors and increased
height. A different conductor form might also be considered.
A new start, including the core selection procedure, is required if the given boundary conditions have to
altered. It may be possible to use the same core size if switching frequency is increased.
This leads to a fewer turns, of thicker conductors and, thus lower winding loss.
To avoid fruitless effort, the practical design process of Sect. 3 is so organized that an initial design
evaluation is made as soon as suitable ideal designs and their losses have been determined. Winding height
can be reduced with the penalty of increased loss (Sect. 4.3.2.) and vice versa (by using bunched wire, for
example, Sect. 5.2), but this will not help if the ideal design has neither loss nor height reserve.

37

Part 3
4.5. Some notes
4.5.1. Multiple-output transformers
Figure 18 shows a possible winding arrangement and leakage flux density distribution for a multiple output
transformer. In both primary and secondary 2, the flux density varies from zero to maximum, so these
can be designed using the present methods. That is not possible for secondary 1 because its leakage-flux
distribution does not agree with the assumptions. The ratio for leakage-flux density to number of turns is
much higher than has been assumed so far.

N
0 sec2 Isec2
bw

N
0 sec1 Isec1
bw

secondary 2
secondary 1
sreens
primary
b
N
0 prim Iprim
bw
7Z80242

Fig. 18 An example of a multiple-output transformer with


corresponding leakage-flux diagram.

The ideal wire diameter for an infinite number of layers d in the winding exposed to the higher leakage
flux calculated as normal, but the layer correction factor becomes

1-
(1 - 3)1/3

Cp =

where is the ration of minimum to maximum leakage field over the height of the winding. In secondary
1 of Fig. 18, = (Nsec 2Isec 2) / (NprimIprim). The factor Cp is always less than unity and decreases with
increasing . The higher leakage field is compensated for by using thinner wire. As is usual ( = 0) for wire
windings, FRid 1.5; the formula for FR still applies. Similarly, for strip conductor,
4

CN =

(1 - )3
1 - 3

As normal for strip windings FRid 4/3. To obtain lowest total winding loss, the layout of
the secondary windings ( 2 or more) in the winding window must be considered carefully.
- Secondaries of the same conductor type, either wire or strip: exposing the winding carrying the
smallest current, that is, the one using the thinnest conductor, to the highest leakage field will
result in lowest loss. Remember that the eddy current loss increases in proportion to 4 ,
Sect. 4.2.1.
- Secondaries of different conductor type (wire or strip): the optimum arrangement is not easy to
predict. Work out the possible layouts and select for lowest total loss.

38

Part 3
4.5.2. Push-pull transformers
In push-pull transformers, Fig. 19, not all windings carry current at the same time. The operation of
push-pull converter is explained in part 1 of this series of publications. Here it is sufficient to split up one
converter operating period into
- interval a : N1A and N2A are conducting, other windings carry no current.
During this interval the current waveforms in N1A and N2A are the same.
- interval b: N1B and N2B are conducting.
- interval c, occurring twice per period: N1A and N1B are nonconducting, N2A and N2B carry
equal and opposite flywheel currents about half as great as the secondary currents during intervals
a and b (magnetizing current neglected). The closer intervals a and b approach a half period (at
full load) the shorter this interval will be. Whether or not the flywheel current pules have an
influence on winding, design depends on the winding arrangement.

N1A

N2B

N1B

N2A
7Z80245

Fig. 19 Push-pull transformer windings divided to define


conduction periods.

Bilifar windings are often used to achieve close coupling between winding halves. Figure 20 shows a crosssection through the windings and the leakage flux diagrams during the three intervals. If wire windings are
used the flywheel currents during interval cancel (higher order effects neglected). That is not quite so if
bifilar strip windings are used (two full-breadth strips and interleaving wound simultaneously). The peak
flux density is then about 1/(2N2) times that during interval a or b. Since the eddy current loss in interval
c may be negligible if the half-secondaries have at least a few turns. It might not be possible to ignore the
contribution of the flywheel currents to that d.c. loss unless interval c is very short.

screen

N2A+N2B
bifilar
N1A+N1B
bifilar

B
interval a

interval b interval c
interval c
(wire windings) (strip windings)
(N2A=N2B=3)
7Z80246

Fig. 20 Cross-section through the windings of a bifilar-wound


push-pull transformer together with the leakage-flux diagrams
for the conduction intervals described in the text.

During intervals a and b, eddy currents are generated in both the conducting and the non-conducting
windings. Both windings of a pair are exposed to the same leakage flux and have about equal eddy-current
loss. Only one of these has d.c. loss at a time. For minim loss (ideal designs), the ratio of eddy-current

39

Part 3
loss to d.c. loss remains one half for wire windings and one third for strip windings. That requires thinner
conductors than in non-bifilar windings have equal numbers of turns, and winding pitch or strip breath.
For multilayer wire windings, the reduction is 2-1/6 = 0.89 (one step in the wire table), and for strip
windings, 2-1/4 = 0.84
The reader can adapt the design procedures of Section 3.2 to cope with bifilar windings.
Multilayer wire windings, procedure 3.2.1:
- in step 2 multiply d by 0.89
- in step 4 use 2tmin in the expression for Pid
- in step 9 calculate the loss in a winding pair.
Single-layer wire windings, procedure 3.2.2.:
- in step 2, use 2tmin rather than tmin
- in step 6, calculate rac = (2FR - 1) rdc
- in step 7, estimate the loss of a winding pair.
Strip or foil windings, procedure 3.2.3:
- in step 1, multiply CN by 0.84
- in step 5, estimate the loss of a winding pair.
Bifilar windings are not possible if the wire insulation cannot safely withstand at least twice the peak voltage
across a winding half. The transformer of Fig. 21 differs from the previous one in that the primary winding
halves do not share the same space. The leakage flux diagrams show an asymmetry. During the interval a
winding N1B, although not conducting is subjected to a high leakage flux and has about three times the eddy
current loss of the conducting winding N1A, because the average of the r.m.s. flux density squared in N1A is
one third of that in N1B. In interval b, however, N1A has no eddy current loss. The primary winding halves
thus have different a.c. resistance due to a difference in eddy-current loss by a factor of four. Moreover, the
leakage inductance is greater in N1A. The energy stored in the leakage field can be expressed as LI2or as
VB2/O, where L is the leakage inductance and V the filed volume (here proportional to the area of the
flux diagram times the average turn length).

screen

N2A=N2B
bifilar
N1B
N1A

B
interval a

B
interval b
7Z80246

Fig. 21 Cross-section through the windings of a push-pull


transformer with a bufilar secondary, together with the leakage flux
diagrams for two conduction intervals. Note the asymmetry.

By arranging the windings as in Fig. 22, only a minor asymmetry due to different turn lengths remains and
eddy current loss in the non-conduction primary is eliminated. In this arrangement the coupling between
the primary winding halves will be less close and the stray capacitance across the secondaries, now situated
between two screens, will be increased.

40

Part 3

N1B

screen

N2A=N2B
bifilar
N1A

B
interval a

B
interval b
7Z80248

Fig. 22 Rearrangement of the windings of fig.21 showing the


reduced asymmetry.

In Fig. 23, the secondaries are also divided: non-conducting windings are not exposed t leakage flux.
Now eddy current loss during interval c occurs in the secondaries. This indicates the effect of different
current waveforms in primaries and secondaries, which means hat their effective frequencies are different
(see Section 6).
For split/sandwiched configurations, similar considerations apply as indicated above for simple configurations. Here, also, leakage flux diagrams will be a great help in the analysis. The discussion in this Section
clearly illustrates the general risk of overlooking important parameters in striving for an optimum in a
particular respect.

N1B
screens

N2B
N2A
N1A
B

interval a

interval b

interval c
7Z80249

Fig. 23 Cross-section through the windings of a push-pull


transformer showing that, with divided primary and secondary
windings, the non-conducting windings are not exposed to
leakage flux.

41

Part 3
4.5.3. Flyback transformers
In a flyback transformer, current conduction in primary and secondary occurs alternately. When the primary
conducts, there is no current in the secondary, and vice versa. Two different leakage-flux density diagrams
are, thus, required, Fig. 24.

outer winding
inner winding

B
7Z80243

Fig. 24 In a flyback converter transformer, primary and secondary


conduct alternately.Thus, two leakage flux diagrams are required
for loss analysis: one for the inner winding conducting (left) and
one for the outer winding conducting (right).

The flux diagrams shown do not pretend to be accurate. In drawing the full-line portions tight coupling
to the core and a leakage flux parallel to the layers were assumed. However, since the repulsion between
primary and secondary flux does not exist, the latter assumption is an idealization. The broken lines
are speculative. It is therefore recommended that the outer winding has the thinner wire. The winding
arrangement is recommended for the reasons given in Sect. 4.5.2. The leakage flux generated by the current
in the inner winding also induces eddy-currents in the outer winding.
As a practical guide, it is recommended that the outer winding be wound with a wire one size smaller
than that calculated using the procedures given in this article. The estimated winding loss will then probably
prove slightly optimistic.
Although the present design methods are not as accurate for flyback transformers as they are for forward
and push-pull types, their use is helpful in preventing excessive eddy-current loss.
Strip or foil windings: since the leakage flux might not be truly parallel to the layers we should not be
surprised if the losses are higher than calculated. The current density near the edges of the conductor
might be considerably higher than that near the middle of the breadth. That effect is nearly independent
of the conductor thickness. The use of multiple wire, bunched or Litz wire might be a better solution
for low-voltage windings.

4.5.4. Screens
Eddy current losses will also occur in the screens between primary and secondary. Screens are always
situated in positions of maximum flux density Screens should not be thicker than necessary; there should
be the minimum of overlap of the ends. copper is not the most suitable material : a (non-magnetic)
conductor of higher resistivity is preferable. Eddy-current loss is directly proportional to conductivity, and
proportional to the thickness cubed. Thus, screens, should be as thin as possible (UL-1244 specifies a
minimum of 0.15 mm).
Although its low resistivity makes copper the natural choice for the winding conductor, it is not the first
choice for screens. Phosphor bronze CuSn8 has a conductivity 10.9% of that of copper at 20 oC and 13.8%
at 100 oC. The use of such material reduces losses in proportion.
Additional loss can be caused by the capacitance of the overlap at the ends of the screen, which causes the
screen to act as a shorted turn. This overlap should be as small as possible.

42

Part 3

5. Chosen boundary conditions


Once the given boundary conditions have been established from the choice of core and numbers of primary
and secondary turns, the designer can choose winding configuration and conductor type.

5.1. Winding configuration


The designer can choose between
- simple windings
- split / sandwiched windings
The lower losses of split/sandwiched windings arising from the halved peak leakage-flux density will be
evaluated further, as will also the required winding height. In the split/sandwiched configuration, a winding
portion has half the number of turns of the complete primary or secondary winding. Hence the value
of T = bw/N is doubled.

5.1.1. Strip or foil windings


The use of strip or foil windings is only recommended for sandwiched windings (Sect. 4.2.3). The following
discussion is given for completeness, because the route of a design process cannot always be predicted.
Analysis of the design formula in Sect. 3.2.3 shows that for moderate values of N (Where CN 1), halving
N increasing hid by a factor of about 2. Since the total number of turns in the complete winding remains
the same, sandwiching increases the height of the ideal winding by 41%. The influence of the interleaving
is neglected. Besides the increase in required height, the extra winding interface with its insulation and
screening reduces the available height. In the extreme case, where N changes from 1 to 0,5, hid doubles
because CN also increases by a factor of 2.
In Section 4.2.3 FRid was shown to be about constant for all values of N. So the ideal winding resistance
tends to decreae as 1/2, a reduction of 29%. For N = 0.5, rac id is half that for N = 1
A smaller improvement can be obtained where the sandwiched version has to be non-ideal.
Even if the simple version has to be non-ideal it may occur that its sandwiched counterpart has a lower
a.c. resistance.

5.1.2 Wire windings


If, for a simple wire winding having at least 2 layers, the ideal design is replaced by an, also ideal, split or
sandwiched winding rac will be improved by about 37%. The penalty is a considerable increase in required
height. The actual increase in height is not readily predictable, since the number of layers may change ( the
effect of doubling T/ is shown in Fig. 17) In many cases the split or sandwiched version will be non-ideal,
so that the improvement in resistance is smaller than 37%. The improvement of a (non-ideal) split or
sandwiched winding replacing a non-ideal simple winding also varies from case to case.
The replacement of an ideal simple, single-layer winding by an ideal, split or sandwiched winding must be
considered separately. Doubling T/ (Fig. 17) results in halving rac id. If the result is two single-layer portions,
wire diameter will be doubled, and required height quadrupled. But, if the result is two half-layer portions,
wire diameter and winding height remain the same.
This will not often occur in practice, because strip or foil windings are generally preferable to single-layer
windings (Sect. 4.2).

43

Part 3
5.2. Conductor form
The designer can choose between solid round wire, strip or foil conductor, multiple wire, bunched wire
and Litz wire.

5.2.1. Strip or foil versus solid round wire


Strip or foil is only recommended for sandwhiched windings. (Sect. 4.2.3). There will seldom be more than
10 turns in a winding portion. Usually, the total winding may have up to 20 turns with as many interleaving
in the (radial) heat-flow path. Depending on current density, internal overheating may occur if the number
of turns and interleaving is unduly high.
Strip or foil windings can be considered if T/ 2, as an alternative for single layer of half-layer wire
windings, for example. If the winding has only a few turns, they are a must. In practical transformers , bw /
may be assumed to be > 20. With N 10, then T/ 2. The basic design chart for wire windings, Fig. 17,
shows that that is the region for single-layer or half-layer designs, for which rac id.2/ 1.1(T/)-1.
Choosing strip rather than wire reduces the winding resistance by a factor of about 0.9 / N if ideal designs
can be used. If N > 1, the expression for the ideal strip thickness can be written as hid / 31/4N1/2 . Then
rac id = FR id / (bwhid) ( FR id /1.32) N 1/2(bw )1. Since FR id 1.33, dividing this result by that for wire
windings obtained above yields the factor 0.9/N.
With a non-ideal strip design, the improvement of an ideal solid-wire version is, of course, smaller, but the
strip version remains preferable as long as the thickness of all turns in the portion together exceeds .
For a non-ideal strip version which FR 1 (that is, if h < 0.5 hid), rac rdc = /(bwh).
To make this smaller than the rac id of the wire version, Nh > /1.1 or, as a rule of thumb Nh > .

5.2.2. Multiple wire


In a multiple-wire winding, each turn consists of ns strands lying side-by-side in a layer, Fig. 25.
Multiple-wire turns have the height of a single strand, but an area ns times that of strand, and a layer space
(copper) factor the same as that in a layer wound conventionally with wire of the size of each strand.

7Z80244

Fig. 25 A multiple-wire winding of 4 turns, each of 3strands.The


strands of one turn are indicated by black dots.

The design procedure for solid round wire windings can easily be adapted for multiple-wire.
In the basic design chart of Fig. 17, after dividing T/ by ns, read did /, pid and He id / as usual. The value of
rac id 2/ for the single strand must be divided by ns to find the resistance of the multiple wire.
In effect, a single-wire winding is designed with a layer breadth bw /ns or alternatively, the winding geometry
is determined as if there were nsN turns. T/ and rac are thus 1/ns times those for a normal wire winding.
Compared with normal, single-strand, multilayer windings (more than one layer per portion, T/< 2), rac id
and did decrease about ns-1/3 and He in-crease about ns-1/3.

44

Part 3
With 2 parallel strands, Fig.26, the reduction in rac id and did is about 20%: the height increase is roughly 25%.
With 3 parallel strands, the reduction in rac id and did is about 30%: the height increase is roughly 45%.
If the increased height does not permit the use of the ideal design of a multiple-wire version, the singlestrand winding will often have a lower resistance. More than 3 parallel strands are not often used due to
winding difficulty, or inadequate height.

rac id

7Z80250

5
ns=
1
2
3

2
1
0,5

0,2

1,

1T

,-1

0,1
0,05
0,1

0,2

0,5

10

20
T/

Fig. 26 Design chart for multiple-wire windings showing the


improvement in rac id obtainable for ns = 1, 2 and 3 strands. Note
that the improvement quickly vanishes between T/ = 2 and
T/ = 2ns . But if T/ exceeds 2ns, winding height and copper
content are reduced by the use of thonner wire.

In the region of T/ > 2 it is usual to try to use strip or foil conductor. If that is not possible, multiple-wire
may be considered. Strip conductor, although promising lower resistance, cannot always be used: because of
uneven current distribution over the layer breadth, for example or due to constructional problems.
If T/ > 2ns lower resistance cannot be achieved, Fig. 26, but a multiple-wire winding design might be justified
by reduced winding height, the use of thinner, more flexible wire, and saving in copper.
The curve for p = 1 Fig. 17 ends at T/ = 6 only for practical reasons. There is no theoretical limit: the
resistance remains inversely proportional to T/. If p = 1 and T/ > 2ns, or if p = 0.5 and T/ > 6ns, the
overall wire diameter and, thus, the winding height, will be found to reduce in inverse proportion to ns.
In order to understand how less copper can have equal resistance, it is necessary to realize that this is in the
region of pure skin conduction ( >> 1). The breadth of a turn (either a single wire or ns strands) and, thus
the skin area remains the same. Stranding reduces the amount of useless conductor in the winding.

45

Part 3
5.2.3. Bunched wire
Bunched wire comprises three or more enamelled wires, twisted around each of other to form a bunch.
Here a distinction is made between bunched wire and Litz wire, which latter is discussed in the next Section.
The difference is in the strand diameter: in bunched wire it is of the order of magnitude of , whereas
in Litz wire the strands are much finer. In the literature, the term bunched wire is sometimes used for
what is here called Litz wire.
In both bunched and Litz wire the strands undulate within the height of the layer. This ensures equal
current sharing between the strands, because, on average, the position of each strand is in the middle
of the layer height. Bunched-wire windings are designed in much the same way as single wire windings
have Nns. turns, but adaptations of the procedures are required for the determination of winding pitch
and layer height.
Practical design guidance is given in Sect. 3.42.

5.2.4. Litz wire


Litz wire is a form of bunched wire in which the strand diameter is small compared to . Eddy-current
effects are thus virtually eliminated, and the a.c. resistance is equal to the d.c. resistance. Thus Litz wire
windings require no special design procedures: the design methods for low-frequency windings apply.
Litz wire is standardized and commercially available.
For instance, IEC publication 317 -11 (Ref.4) covers Litz wire of 3 to 400 strands of 0.025 mm to 0.071 mm
diameter. Standards and manufacturers catalogues should be consulted for engineering data.
Litz wire is generally only useful where ample winding space is available. The main drawback of Litz wire
is its low space (copper) factor. Typically, only one quarter to one third of the winding space will be
conductor.
Design guidance is given in Sect. 3.4.3.

46

Part 3

6. Non-sinusoidal current, effective frequency


So far, sinusoidal current waveform of frequency equal to the switching frequency has been assumed.
When considering proximity-effect loss compared to d.c (i.e. frequency-independent) loss, it was found that
the total loss is minimum if the ratio of eddy-current loss to d.c loss is 1:3 in strip or foil windings and 1:2 in
multilayer wire windings. In single-layer windings there was no such optimum ratio. The current waveforms
associated with SMPS transformers are generally far from sinusoidal.
A non-sinusoidal repetitive waveform can be represented by a d.c. component, a component at the
fundamental frequency (the switching frequency) and several components at harmonic frequencies. The d.c
component will not, of course, cause eddy-current loss: its current density is uniform over the conductor
cross-section. The fundamental component has eddy-current loss as well as d.c. loss associated with it. This
is also the case for the harmonic components, but the ratio of eddy-current to d.c loss will be higher
than for the fundamental. Neglecting the non-sinusoidal character of the current wave-form introduces an
inaccuracy. The sum of all eddy-current losses should be determined together with the total d.c loss to
ensure that they have the desired optimum ratio. One problem is that, when designing the windings, it is
not possible to make a reliable estimate of the current waveform. Sometimes, it is possible to make use
of experence gained in previous designs.
Many designs have, in fact, been made as if the current were sinusoidal and the results have usually been
satisfactory. The errors due to calculation only at the switching frequency oppose. The component of loss
occurring at d.c is over-estimated, whereas the components due to harmonics are underestimated.
The inaccuracy can be avoided by using an effective frequency in winding design.
The effective frequency is, of course, notional. It should be used only in connection with eddy-current
effects. In so far as eddy-current loss is proportional to 4 (that is to f2), the effective frequency can be
found from an analysis of the current waveform:

fe = 1
2

i
I

where I is the r.m.s. winding current; i is the r.m.s. value of dI/dt, the first derivative of the current
waveform. This expression does not apply in the region of skin conduction, where the eddy-current loss
is proportional to or f.
It is apparent that i contains no contribution from the d.c. component and a more pronounced contribution
from the harmonics, since d (sin nt) /dt = n cos nt. The rise and fall items of the current pulses
have a considerable effect on the value of i. Moreover, some (if not all) of the higher harmonics lie in
the region of the skin conduction.
A simple method of determining fe has not yet been found: fe can be expected to be lower than the
switching frequency f, but higher than f Ie ac/Ie Where Ie ac = (I2e- I20) is the r.m.s. value of the a.c.
component and I0 is the d.c. component. The use of this effective frequency, despite its not being valid in the
region of skin conduction, will yield the optimum winding geometry.
Single-layer windings are chose if pid 1 since fe is valid for determing pid the choice between multilay
er and single-layer windings is clear. Both single and half-layer windings are close-packed, inc principle, so
their layer geometry is independent of frequency.
Since the effective frequency for the skin-conduction region is not known, winding loss cannot be accurately
estimated, and, consequently, a reliable choice between a single and a half-layer winding cannot always be
made. In most practical cases, however, these problems do not occur because, in any case, strip windings
are preferable in this region.
For use in choke design an expression for fe has been derived and simplified, but due to the large variation

47

Part 3
in conditions it is impossible to give a typical example for transformer design.
For the waveform of Fig. 27, provided that the rise and fall times are between 15% and 85% of the
repetition period,

1.3 f

fe

[ 1 + 3 ( Io / Iac )2 ]
and the effective current

Ie = ( I2o + I2ac / 3 )

For a sinusoidal rather than a triangular waveform super-imposed on a d.c. current,

fe =

[ 1 + 2 ( Io / Iac )2 ]
and

Ie = ( I2o + I2ac / 2 )

I
Iac
Iac I
M
I0
1/f

7Z89485

Fig. 27 Current waveform in a smoothing choke idealized for the


calculation of effective frequency.

References
1.
Dowell, P.L. 1966, Effects of eddy currents in transformer windings. Proc. IEE, 113 No. 8 (August)
2.
Snelling, E.C. 1969. Soft Ferrites, properties and applications. Iliffe Books Limited, London
3.
Casimir, H.B.G. and Ubbink, J. 1967. Skin effect -I.Philips Technical Review 28 No.6.
4.
IEC Publication 317-11 Bunched enamelled copper wire with silk covering.

48

Part 3
Appendix: Exaple of tables of standart wires sizes
Enamelled dround coper winding wire, IEC-grade 2

nom.resistance
nominal diameter
d (mm)
0.040
0.045
0.050
0.056
0.063
0.071
0.080
0.090
0.100
0.112
0.125
0.140
0.160
0.180
0.200
0.224
0.250
0.280
0.315
0.355
0.400
0.450
0.500
0.560
0.630
0.710
0.800
0.900
1.000
1.120
1.250
1.400
1.600
1.800
2.000
2.240
2.500

max. overall

nominal

minimum

diameter
d o (mm)

cross-sect.
area (mm2)

winding pitch
t min (mm)

at 100oC
r dc (/m)

0.054
0.061
0.068
0.076
0.085

0.00126
0.00159
0.00196
0.00246
0.00312

0.059
0.066
0.073
0.082
0.091

17.68
13.97
11.32
9.022
7.129

0.095
0.105
0.117
0.129
0.143

0.00396
0.00503
0.00636
0.00785
0.00985

0.102
0.112
0.125
0.137
0.152

5.613
4.421
3.493
2.829
2.256

0.159
0.176
0.199
0.222
0.245

0.0123
0.0154
0.0201
0.0254
0.0314

0.169
0.187
0.210
0.234
0.257

1.811
1.444
1.1052
0.8733
0.7074

0.272
0.301
0.334
0.371
0.414

0.0394
0.0491
0.0616
0.0779
0.0990

0.284
0.315
0.348
0.387
0.431

0.5639
0.4527
0.3609
0.2852
0.2245

0.462
0.516
0.569
0.632
0.706

0.126
0.159
0.196
0.246
0.312

0.481
0.538
0.593
0.659
0.736

0.1768
0.1397
0.11318
0.09022
0.07129

0.790
0.885
0.990
1.093
1.217

0.396
0.503
0.636
0.785
0.985

0.823
0.922
1.032
1.139
1.268

0.05613
0.04421
0.03493
0.02829
0.02256

1.351
1.506
1.711
1.916
2.120

1.227
1.539
2.011
2.545
3.142

1.408
1.569
1.783
1.996
2.209

0.01811
0.01444
0.011052
0.008733
0.007074

2.366
2.631

3.941
4.909

2.465
2.742

0.005639
0.004527

Remark: Values of tmin are based on recommendations for mass production of one particular manufacturer.
Other manufacturers may use different values.

49

Part 3
Medium enamelled copper wires - AWG (B. & S.)
(diameters based on BS 1844: 1952 - mediun covering)

AWG
(B. & S.)

nominal copper
diameter (d)
inch
mm

max. overall
diameter
(d o )
mm

nominal
cross-sect.
area
mm2

nominal
resistance at
at 100oC (r dc )
/m

minimum
winding pitch
(t min)
mm

44
43
42
41
40

0.00198
0.00222
0.00249
0.00280
0.00314

0.0503
0.0564
0.0633
0.0711
0.0798

0.06604
0.07366
0.08128
0.09144
0.1041

0.00199
0.00250
0.00314
0.00397
0.00500

11.190
8.899
7.073
5.594
4.448

0.071
0.079
0.087
0.098
0.111

39
38
37
36
35

0.00353
0.00397
0.00445
0.00500
0.0056

0.0897
0.1008
0.1130
0.1270
0.1422

0.1143
0.1295
0.1448
0.1626
0.1778

0.00631
0.00799
0.01003
0.0127
0.0159

3.519
2.783
2.215
1.754
1.398

0.122
0.138
0.154
0.172
0.188

34
33
32
31
30

0.0063
0.0071
0.0080
0.0089
0.0100

0.1600
0.1803
0.2032
0.2261
0.2540

0.1981
0.2235
0.2489
0.2743
0.3048

0.0201
0.0255
0.0324
0.0401
0.0507

1.105
0.8700
0.6853
0.5527
0.4386

0.209
0.236
0.261
0.287
0.319

29
28
27
26
25

0.0113
0.0126
0.0142
0.0159
0.0179

0.2870
0.3200
0.3607
0.4039
0.4547

0.3404
0.3759
0.4191
0.4699
0.5232

0.0647
0.0804
0.1022
0.128
0.162

0.3435
0.2762
0.2175
0.1735
0.1369

0.356
0.393
0.438
0.491
0.547

24
23
22
21
20

0.0201
0.0226
0.0253
0.0285
0.0320

0.5105
0.5740
0.6426
0.7239
0.8128

0.5817
0.6502
0.7214
0.8052
0.8966

0.205
0.259
0.324
0.412
0.519

0.10860
0.08586
0.06852
0.05399
0.04283

0.608
0.679
0.754
0.841
0.937

19
18
17
16
15

0.0359
0.0403
0.0453
0.0508
0.0571

0.9119
1.024
1.151
1.290
1.450

1.003
1.118
1.247
1.389
1.557

0.653
0.823
1.040
1.308
1.652

0.03403
0.02700
0.02137
0.01699
0.01345

1.048
1.168
1.303
1.452
1.627

14
13
12
11
10

0.0641
0.0720
0.0808
0.0907
0.1019

1.628
1.829
2.052
2.304
2.588

1.737
1.943
2.172
2.431
2.720

2.082
2.627
3.308
4.168
5.261

0.010670
0.008460
0.006717
0.005331
0.004224

1.815
2.030
2.270
2.540
2.842

Remark: Values of tmin are based on recommendations for mass production of one particular manufacturer. Other manufacturers may use different values.

50

Part 4

Power-inductor design

A power smoothing choke which is to carry a significant direct current component or to have a well-defined
inductance is usually wound on a core whose magnetic circuit includes an air gap. The reluctance (magnetic
resistance) of the air gap reduces the effective permeability of the core: either to increase the ampere-turns
at which saturation occurs, or to reduce the effect of variations in the permeability of the core material
on the inductance of the choke.
The traditional rout to the design of a choke with a gapped core involves the use of Hanna curves, or
some derivative of them. (Ref. 1, 2, 3). This design route has a number of disadvantages and limitations.
Initial core selection is uncertain and designs may have to be made using a number of cores before the
optimal solution is found. The design procedure involves considerable calculation and iteration, and the
effects of changes in core operating conditions and mechanical tolerances, especially on the airgap, are
not readily predicted.
To simplify the design of power chokes we have devised a method based on computer-generated charts. The
first step in the design is the selection of a suitable core: this selection usually proves to be final.

Design method
Starting with the peak current IM that the choke is required to pass without saturating the core, and
the minimum inductance required Lmin, the designer obtains directly all the information necessary for
the construction of the choke. Core size, airgap, number of turns, and winding geometry are derived by
straightforward procedures. Of especial interest to those engineers to whom the subject is a black art: the
magnetic properties of the core do not enter into to process at all.

Part 4
Core operating conditions
The selection and design charts are constructed for cores in the power ferrite materials 3C90 or 3C94
operating at a temperature of 100oC. Operation at lower temperatures leads neither to core saturation
nor to inductances lower than Lmin. The design peak flux density BM is 0.32 T, however, the charts can
be used for a lower value by designing for a peak current 0.32I/M/BM. Symbols used in this text are
listed and defined in the Table, symbols for currents are illustrated by Fig. 1. Note that in the equations
the unit of frequency is kHz, not Hz, and the unit of length mm. Some constants in the equations are
based on these units.
I
Iac
Iac

IM
I0

Fig. 1 Symbols for choke


current used in the text.
(See also the Table)

1/f
7Z89485A

Definition of symbols used


Symbol

unit

definition

AL

induction factor L/N 2

bw

mm

winding (layer) breadth

BM

peak flux density

mm

nominal wire diameter

do

mm

overall wire diameter

kHz

frequency

fe

kHz

effective frequency (see text)

FR

a.c. resistance factor R ac / R dc

mm

thickness of foil conductor

mm

winding height

Ha

mm

available winding height

mm

thickness of interleaving

Ie

r.m.s. current at full load

IO

d.c. component of current at full load

I ac

a.c. component of current at full load

IM

peak value of current at full load

inductance

number of turns in a winding

number of layers

PW

winding loss

R ac

a.c. resistance

R dc

d.c. resistance

mm

spacer thickness

Note: subscript id means 'ideal' value.

Part 4

I2L
(J)

E71/33/32
E65/32/27

101

E55/28/25
E80/38/20
E46/23/30
E55/28/21
E56/24/19
E42/20
E47&50
E42/21/15
E36/21/15
E41/17/12
E30&31&32&34

102

E30/15/7

E25/13/7
E19/8/9
E25/6
E20/10/6
E20/10/5
E19/8/5
E13/6/6

103

E16/8/5
E13/7/4

104
101

10
air-gap (mm)

Fig. 2 Selection chart for E cores

Selection procedure
The selection graphs included at the end of this part are used to select a suitable core size for the
intended application.
An example of such a graph is given above. The selection charts are used as follows:
Knowing the value of peak choke current IM and the minimum inductance required Lmin, calculate
the value of I2L.
Choose the shape of core based on application considerations. Draw on the appropriate selection
chart a horizontal line at the level of the calculated I2L.
A core whose curve intersects this horizontal line can be used for the application. The air-gap length
corresponding the intersection is, however, only an indication of the final value.

Effect of core size


Where, as is usual, more than one core could be used, the final choice may be governed by the consideration that operation near the right-hand end of the curves carries the risk of overheating. Moreover,
selection of a larger core will generally result in a more conservative efficient design than one based on
a core that is only marginally large enough.

Air-gap length Part


and 4number of turns
Once an air-gap length is chosen the effective core permeability e and the inductance factor AL can be
calculated with help of the program module " Inductance factor calculation".
The number of turns for the choke design can now be derived as follows.
To avoid saturation the maximum number of turns allowed is given by:
I2L

Nmax =

[1]

I2AL

The minimum number of turns required to achieve Lmin is:

Nmin =

Lmin

[2]

AL

Now select an integral number of turns N between Nmin and Nmax.


Establish the winding geometry using the winding-design procedure in the next setion.

Winding design
The loss due to eddy-currents in a winding carrying a.c. increase rapidly with conductor size (ad d4 for wire),
but resistive losses in a conductor decrease with increasing size (as d-2 for wire). It follows, therefore, that
there must be a frequency-dependent ideal conductor size at which losses are minimum. (This is discussed
fully in Ref. 5). This sets the upper limit to conductor size; there is no reason to increase losses by using a
thicker conductor. The use of a thinner conductor is sometimes tolerable (low current density) or necessary
(inadequate space).
The procures that follow allow the ideal number of layers and wire size, or the thickness of strip, to be
determined for chokes with an operating-current waveform similar to that shown in Fig.1. They also indicate
the course of action in the event of the available winding window being insufficient to accommodate the
ideal winding.
Copper conductors are assumed here and the operating temperature is taken to be 100oC, so that conductor
resistivity is 1/45 mm2/m (30% higher than that at 20oC). Symbols used are defined in the table and Fig. 1.

Effective frequency and effective current


To allow for the effect of waveform on eddy-current losses in the choke windings, it is necessary to convert
actual frequencies and currents to effective values. For sinusoidal currents, the effective frequency fe is equal
to the actual frequency f. For small amounts of waveform distortion, and small d.c. components fe can still
be taken as equal to f. For the waveform of Fig.1, and provided that the rise and fall times are between 15%
and 85% of the repetition period,
fe

1.3 f

[3]

[ 1 + 3 ( Io / Iac )2 ]

Part 4
In design for class I applications fe may be only a few kilohertz. Eddy-current effects are then negligible so
that windings can be designed as if they are to carry d.c. only. Remember to use the correct value for d.c.
resistivity. For the waveform of Fig.1, the effective current Ie is given by

[4]

Ie = ( I2o + I2ac / 3 )

For sinusoidal currents with a significant d.c. component, however,

fe =

[5]

[ 1 + 2 ( Io / Iac )2 ]

and

[6]

Ie = ( I2o + I2ac / 2 )

where Iac is the amplitude of the a.c. component.

Multi-layer wire windings of solid, round wire


In the following design procedure, it is assumed that all layers have the same breadth. However, where
the number of turns in the winding cannot be divided into the ideal number of layers, a difference of
one turn per layer is permissible.
1. The ideal wire diameter is

did = 2.6

( bNfwe )

1/3

[7]

2. Select the nearest standard wire size (for d and do) from a wire table such as that for IEC grade
1 winding wires.
3. The ideal number of layers is now

pid =

N
bw/do - 1

[8]

Note: this expression is valid only for do from step 2.


- If pid 1.5 and the current density in wire did is excessive, make a new design using a larger core.
5

Part 4
- If Pid 1, the expression for did in step 1 is not valid: go to the single-layer winding procedure.
Find p by rounding pid to the next highest integer. This rounding increases the spacing between turns.
4.

The required winding height is H = p(do + i)

5. If H exceeds the available height Ha, or if current density is low:


- reduce p by one layer
- select the thickest wire for which do pbw/(N + p) repeat from step 4, if p = 1.
FR = 1 + 1/2(did)6
Check : if FR > 2 an error has occurred.
Note : FR = 1.5 for did when d < 0.7 did, FR 1

6.

7.

Pw = I2e Rac = I2e FRRdc.

Single-layer windings of solid, round wire


The design procedure is to be used only when pid calculated in step 3 of the previous section comes
out as equal to or less than unity.
1.

Select the thickest wire for which do bw/(N + 1).

2.

FR = 0.33 d fe 1/2N(N + 1), only if pid 1 in step 3 of the last section.FR has no upper limit here.

3.

Pw = I2e Rac = I2e FRRdc.*

Bunched (Litz) wire windings


Eddy-current effects in bunched-conductor (or Litz-wire) windings are negligible and, thus, no special design
procedure is required. Con-ductors of this type are not, however, necessarily the complete solution: their
packing factor, and, consequently, winding thermal conductivity are both low. They might be an attractive
alternative where the ideal solid conductor winding fills less than half the available height. The resistance of
bunched conductors, like that of solid conductors, is 30% higher at 100oC than at 20oC.

Part 4

Foil or strip windings


Chokes for high-current, low-voltage SMPS often use windings of strip or foil conductor. The width bw of
the strip is equal to the available winding width.

hid = 3.1
Nfe

[9]

hmin = 0.8 hid


N

hmax =

[10]

Ha - i
N

[11]

Choose a value for i that suits a strip thickness about Ha/N. if hmax < hmin, try a wire winding
4. Select a strip thickness h such that hmin h < hmax. Aim for h = hid.

FR = 1 + 1/3

( hhid )4

[12]

Check that FR < 1,8; if not, reduce h when h = hid, FR = 1.33; when h <0,6 hid,FR 1.
6.

Pw = I2e Rac = I22e FRRdc.

Note : d.c. resistance of copper strip, is 1/(45bwh)/m at 100oC.

Part 4

Design example
A choke of 30mH minimum inductance is required for a peak current of 1A at 100 kHz with a waveform
as shown in Fig.1 IacIo = 0.1.

Core selection
The value of I2MLmin = 3 x 10-2 J. A Horizontal line of this value drawn on the E core selection chart
intersects the curve or the E55/28/21 core at an air-gap length of about 1.5 mm.

Number of turns and spacer thickness


Thus, from Eq.(1),

Nmax =

0.036
371.10-9

= 311.5 turns

[13]

= 284.4 turns

[14]

The minimum number of turns is, from Eq.(2),

Nmin =

0.03
371.10-9

Since Nmax is, at it should be, greater than Nmin, the design is successful so far and a number of turns can
be selected between these limits.

Winding design
The effective operating frequency for the core is given by Eq.(3),

fe

1.3 100

= 7.5 kHZ

[15]

[ 1 + 3 ( 10 ) ]
At this effective frequency, eddy-current effects can be neglected, and the winding can be designed to fit the
space available, allowing for the coil former wall thicknesses.

Part 4
CBW319

I2L
(J)

E71/33/32
E65/32/27

101

E55/28/25
E80/38/20
E46/23/30
E55/28/21
E56/24/19
E42/20
E47&50
E42/21/15
E36/21/15
E41/17/12
E30&31&32&34

102

E30/15/7

E25/13/7
E19/8/9
E25/6
E20/10/6
E20/10/5
E19/8/5
E13/6/6

103

E16/8/5
E13/7/4

104
101

10
air-gap (mm)

Part 4

CBW320

I2L
(J)

E64/10/50

101
E58/11/38
E43/10/28

E38/8/25
E32/6/20

102

E22/6/16

E18/4/10

103

E14/3.5/5

104
101

10
air-gap (mm)

10

Part 4
CBW321

I2L
(J)

101

EC70

EC52

102

EC41

EC35

103

104
101

10
air-gap (mm)

11

Part 4
CBW322

101

I2L
(J)

102
EFD30

EFD25

EFD20

103
EFD15
EFD12

EFD10

104

105
101

10
air-gap (mm)

12

Part 4
CBW323

101

I2L
(J)

102
EP20

EP17

103
EP13

EP7 & EP10

104

105
101

10
air-gap (mm)

13

Part 4
1

I2L
(J)

101
ER48 & 54 & 54S
ER42
ER42A
ER40
ER35
ER28 & 28L

102

ER14.5

ER11

103
ER9.5

104
101

10
air-gap (mm)

14

Part 4
1

I2L
(J)

101
ETD59

ETD54
ETD49
ETD44

ETD39

102
ETD34
ETD29

103

104
101

15

10
air-gap (mm)

Part 4
CBW326

101

P66/56

I2L
(J)

P42/29
P36/22

P30/19

102
P26/16

P22/13

P18/11

103

P14/8

P11/7

P9/5

104
P7/4

105
101

10
air-gap (mm)

16

Part 4
CBW327

101

I2L
(J)

P26/16/I

102
P22/13/I

P18/11/I

P14/8/I

103
P11/7/I

104

105
101

10
air-gap (mm)

17

Part 4
CBW328

101

I2L
(J)

PT30/19

102
PT26/16

PT23/11

PT18/11

103
PT14/8

104

105
101

10
air-gap (mm)

18

Part 4
CBW330

I2L
(J)

101

PQ35/35

PQ32/20 & 32/30

102

PQ26/20 & 26/25

PQ20/16 & 20/20

103

104
101

10
air-gap (mm)

19

Part 4
CBW331

101

I2L
(J)

RM10

102
RM8

RM6S&R

RM5

103

RM4

104

105
101

10
air-gap (mm)

20

Part 4
CBW332

101

I2L
(J)
RM14/I

RM12/I

102
RM10/I

RM8/I

RM7/I

RM6S/I

103
RM5/I

RM4/I

104

105
101

10
air-gap (mm)

21

Part 4
CBW333

101

I2L
(J)
RM14/ILP

RM12/ILP

102

RM10/ILP

RM8/ILP

RM7/ILP

RM6S/ILP

103
RM5/ILP

RM4/ILP

104

105
101

10
air-gap (mm)

22

Part 4
CBW334

I2L
(J)

U93/30
U100/25
U93/76/16

101

U67/27/14
U30/25/16

U25/20/13
U33/22/9

102

U20/16/7
U25/16/6
U15/11/6

103
U10/8/3

104
101

10
spacer thickness (mm)

23