HOW TO DESIGN A TRANSFOR"ER WITH FRACTIONAL TURNS
Lloyd H. Dixon, Jr.
rractional turns used in high frequency switching power supply transformers
can reduce the number of turns otherwise needed to provide low voltage outputs
and to obtain desired voltage resolution between several outputs. With
fractional turns, half the number of turns or less may be required in all
windings, significantly decreasing transformer size and cost. Unfortunately,
fractional turns have inherently high leakage inductance, ~ing their use
impractical unless special techniques are employed. Several methods of
accomplishing this are described.
The Need for Fractional Turns: The optimum number of turns in a transformer
winding depends upon the maximum allowable flux swing and the operating
frequency according to Faraday's Law (in SI units with dimensions in cm):
"N(optimum) = (VN6t/AeAB).10
where 68 is the flux swing (Tesla), Ae is the centerleg area (cm2), and ~t is
the time (approaching 1/2 the period) that voltage VN is across the winding.
In switching power supplies designed to operate at frequencies below 50 kHz,
the optimum numbers of turns are so large that there is little need to use
fractional turns. At higher frequencies, fractional turns become attractive
for the following two reasons:
1. Optimum transformer design may call for less than one full turn for
the lowest voltage secondary. This is likely to happen at high
frequencies, high power levels, and especially with the 2 to 3 volt
outputs required by the newer logic families.
2. With multIple secondaries, to obtain the desired output voltages
with sufficient accuracy using integral turns may require several
times the optimum number of turns. For example, with a 12 V and 5 V
output, a turns ratio of 2.5:1 or 2.25:1 may be desired. If 1 turn
is optimum for the 5 Volt output, 3 turns will provide too much
voltage for the 12 Volt output, causing excessive losses in a linear
post-regulator. Otherwise, 5 and 2 turns or 9 and 4 turns are
necessary to achieve the desired voltage resolution.
In these examples, the actual number of turns required in all windings may be
2, 3,or 4 times greater than optimum. Slightly larger wire sizes are required
because the larger transformer must operate at lower current density. This
means the winding window area will be 2, 3, or 4 times larger and the core and
transformer volume will be 2.8, 5.2, or 8 times larger, with a corresponding
increase in cost. This can be a powerful incentive to use fractional turns!
lllPlementinq a Fractional Turn: A fractional turn is really a full turn
around a fraction of the total centerleg flux. With an E-E core shape having
two outer legs of equal areas, each outer leg has 1/2 the total flux. A
single turn around either outer leg will have an induced voltage equal to 1/2
the primary volts/turn. Such a turn is therefore equivalent to 1/2 turn. In
R6-1
R6-2
The Leakaqe Inductance Problem: The full secondary turns are tightly coupled
to the primary, although there is a small amount of leakage inductance in
series with the full secondary turns due to stray flux between the windings.
Unfortunately, the fractional turn has very high leakage inductance, and its
induced voltage, r.vin/Np' occurs only under no-load conditions.
When load current is drawn through the fractional turn, its voltage collapses.
In fact, when a fractional turn is added to an otherwise stiff winding, the
short-circuit current will probably be much less than the desired full load
output current. Rather than helping matters, the performance of the winding
is worsened by adding the fractional turn because of its leakage inductance.
As shown in Figure 3, secondary current through the
full turns around the centerleg generates a
magnetic potential, NsIs, which is cancelled by
equal and opposite primary ampere-turns, NpIp.
Magnetizing current, Im' and centerleg flux, +1, do
not change significantly. However, current through
the fractional turn creates a magnetic potential in
leg #3 which easily diverts flux +3 to leg #2.
Because flux +3 is diminished, the voltage induced
in the fractional turn is reduced. So the
fractional turn voltage decreases rapidly with
increasing load. At higher load current levels
(usually well below desired full load), d+3/dt will
reverse and the voltage induced in the fractional
turn becomes negative. When this happens, the
total secondary voltage is less than it would have
been without the fractional turn. Figure 3
The leakage inductance of the fractional turn is:
L = F(I-F) ° p = F(I-F) olAAff ° 10-2 Henrys
t (cm) -length of the outer legs
A = A2+A3 (cm2) -combined areas of all outer legs
F = A3/A -fraction of total outer leg area linked to fractional turn
~ = ~o~r = 4..10-7.~r -absolute permeability of the outer leg material
p = P2+P3 = ~A/t -permeance of all outer legs combined
This inductance of the fractional turn Is equIvalent to the Inductance of a
sIngle turn wound on a core consisting of legs .2 and .3 In series. (Leg .1
has no effect.) The worst case Is when the fractional turn links half the
total outer leg area (effectively 1/2 turn). Whether the fractional turn is
in series with one or more full turns around centerleg .1, or whether It is
the entIre secondary winding, it has the same leakage inductance. However,
when the fractional turn is In series with several full turns, the power taken
from it Is only a small portion of the total transformer power. The adverse
effect of the leakage inductance Is then proportIonately less, but It is more
than enough to badly hurt cross-regulation in a multi-output supply.
R6-3
The Solution to the Problem: The solution is simple --maintain the flux in
outer leg portions .2 and .3 in exactly the same ratio regardless of secondary
current; in other
word
word文档格式规范word作业纸小票打印word模板word简历模板免费word简历
s prevent the flux from escaping from leg .3 to leg .2
when the load current increases .
One technique used to keep the flux balanced in the two outer legs of an E-E
core is to put one turn around ~ outer leg as shown in Figure 4. The two
outer core legs have the same area (and permeance) .Each of these turns links
half the centerleg flux and acts like a half turn. If these turns were
connected in series with the correct polarity,
together they would become a full turn. But
connected in parallel (with the same polarity)
they act together like a single half-turn.
Because of the parallel connection, the
voltages induced across each turn must be
identical, forcing equal flux in the two outer
legs. This requires the opposing magnetic
potentials in each outer leg to be the same
ampere-turns which means the secondary current
is shared equally by the two turns .
It
""'-'
Vin
Figure Lf
If the two outer legs had unequal flux, the voltages induced in the two
paralleled turns would differ. This would cause a differential current to
flow between these turns, applying magnetic potentials to each leg in a
direction to eliminate the original flux inequality. Essentially, the cros~
connection between the two turns forces the flux to divide equally between the
two outer legs.
Note that even if the two outer legs have different areas, the flux in each
leg is forced to be half the total flux, SO that the paralleled turns still
act like half turns.
While this technique eliminates the huge leakage inductance of a single half
turn, it is far from ideal because there is much stray flux outside the core
which is linked to the primary but not to the windings around the outer legs .
This results in significant leakage inductance. Normally, to minimize leakage
inductance caused by stray flux, good practice dictates the secondaries should
be wound as intimately as possible with each other and with the primary.
Figure 5 shows a big improvement on the above technique which provides much
better coupling to the primary , minimizing the leakage inductance of the half
turn secondary. Two half
cylinders of copper foil or strip
are placed directly over the
primary winding, separated only by
the minimum insulation required
for primary-secondary isolation.
The half cylinders must not
directly contact each other. They
are paralleled by means of a pair
of tabs off one end of each half
cylinder, cross-connected over the
outside end of the core. output
from the winding may be taken from
across this pair of tabs.
1+1.fJ+
~~
I :
-.-~
+ -
Fioure 5
R6-4
The series inductance of this half turn approach is not quite as good as one
full turn of copper strip because of the inductance of the cross-connected
tabs. Further reduction in series inductance may be obtained by putting
cross-connected tabs at QQ!h ends. The ultimate improvement is to divide the
primary into two portions and interleave the secondary structure between the
two primary portions.
Because the cross-connected half turns in Figure 5 force equal flux division,
the outer legs are "stiffened", so that a half turn added to any other
secondary(s) will also have low leakage inductance.
Using a Separate Flux Balancino Windino: BnY windings that cross-connect the
two outer legs will force the flux to divide equally between the two outer
legs. It is not even necessary for the flux balancing winding to be one of
the output windings. As shown in Figure 6, it may be a completely separate
winding dedicated to the sole purpose of flux equalization between the outer
legs. This enables a single wire half-turn to be added to ~ secondary with
minimal series inductance by forcing the total flux to remain equally divided
between the two outer legs.
This technique is useful when fractional turns
are added to more than one secondary, and
especially with the center-tapped secondaries
used in push-pull converters, where a fractional
turn must be added each side of center-tap.
These situations are difficult to implement by
the method shown in Figure 5.
FigurE' 6
As shown in Figure 6, the flux balancing winding has two coils with equal
numbers of turns cross-connecting the two outer legs at the point where they
join the centerleg. Actually, this winding can be ~ single turn on each outer
leg or many turns. It is better to use multiple turns because finer wire can
then be used. By laying these fine wire turns side by side along the outer
legs, interference with the bobbin is minimized, and eddy current problems are
eliminated.
The ampere-turns of the flux balancing winding will be 1/2 the unbalanced
amperes of the secondary half-turns. For example, assume two secondaries, 12
V, 3 A and 24 V, 2 A, each having a half turn in series with several other
full turns. If the 3 A and 2 A half turns link the same outer leg, the worst
case ampere-turns in the flux balancing winding will be (3+2)/2 = 2.5 A. With
five turns on each outer leg, the current in each turn is 2.5/5 = 0.5 A. On
the other hand, if the 2 A and 3 A half turns link to opposite legs, the worst
case is with the 3 A secondary at full load and the 2 A secondary at no load.
The maximum flux balancing ampere-turns will be half of 3, or 1.5 A-t,
resulting in only 1.5/5 = 0.3 A in the five turn flux balance windings.
For this method of flux balancing to be the most effective in reducing leakage
inductance, the flux balancing winding must have good coupling to the
secondary half turn(s):
Wind the flux balancing coils on the outer legs as close as possible
to where the flux divides -close to the centerleg. If they are
located further out on the outer legs, coupling to the secondary
half turns on the centerleg 1s reduced.
R6-5
2. With a secondary half turn In serIes with several full turns wound
helIcallyalong the centerleg, make sure this half turn Is at the
end of the centerleg adjacent to the flux balancIng winding.
3. When the half turn is foil or strip along the length of the
centerleg , put a flux balance winding at QQ!h ends of the centerleg .
4. When a secondary handles most of the total transformer power and has
only 112 turn or 1112 turns total, the method of Figure 5 works the
best.
Diverse Fractional Turn Values: It is certainly possible to obtain fractional
turn values other than 1/2. Referring back to Figure 1B, a cross core with
four outer legs can provide 1/4, 1/2, or 3/4 turns. A slightly different
technique is required to keep the flux divided equally among all four legs--
a single flux balancing turn is put around each of the four leg;-and these
four turns are paralleled. Because of the parallel connection, the voltages
induced across each turn must be equal, which forces equal rates of flux
change in each leg. Otherwise, current would flow in the flux balancing turns
which would bring the flux changes back into equality.
In reference (1), the author cleverly provides the flux balancing winding by
means of a double sided printed circuit board at one end of the centerleg
where it interferes minimally with the transformer windings. Although this is
a very simple and low cost method, the flux balancing turns are not close to
the centerleg and the coupling to the secondary half turns is not as good as
it might be. Also, cross cores are generally not optimally designed for high
frequency power applications where a long narrow winding window is desirable
to minimize leakage inductance and eddy current losses.
Obtaining Any rractional Value with an E-E Core: It was stated earlier that
the flux balancing winding would force equal flux in the outer legs even if
they have unequal areas. Conversely, it is easy to obtain any desired induced
voltage in the fractional turn by forcing unequal flux division between the
two outer legs, even though their areas are equal. This makes it possible to
take advantage of the better performance and cost available with modern E-E
cores.
Unequal flux division between two outer legs of equal area is obtained by
using unequal turns in the flux balancing windings. Suppose there are twice
as many turns on leg .3 as on leg .2. The induced voltage across both
windings must be equal because they are in parallel. This means the
volts/turn and d+3/dt of leg .3 must be 1/2 of leg .2. Therefore 1/3 of the
total flux goes to leg .3 with twice the turns, while 2/3 goes to leg 12.
Any secondary fractional turn linked to leg #3 will have only 1/3 of the
primary volts/turn induced, while a fractional turn linked to leg #2 will have
2/3 of the primary volts/turn. Similarly, a 1:3 turns ratio in the flux
balancing winding will result in a 1/4 : 3/4 flux division and corresponding
fraction of the primary volts/turn induced in a fractional turn. Depending
upon which leg is linked by the fractional turn, 1/4 turn or 3/4 turn is
obtained. It is possible to obtain 1/2 turn in this configuration by putting
one additional turn around the 1/4 turn leg!
R6-6
When the flux division is made unequal between two outer legs of equal area,
obviously one outer leg has greater flux density (and flux swing) than the
other, and probably greater flux density than the centerleg, as well. This
could theoretically force a reduction in the operating flux level and reduce
the core utilization to avoid saturating the high flux density leg. However,
fractional turns will normally be used above 50-100 kHz, where the flux
density swing is limited by core losses, not saturation. The only adverse
result is that one outer leg will have greater core loss, the other leg less,
for a net small increase in core loSS.
Experimental Results: The data given in Table I was taken with a 20 turn
primary winding over the centerleg of an EC41 ferrite core. Secondaries were
placed directly over the primary (not interleaved) with 5 mil insulation in
between. Leakage inductance was measured on the primary side with the various
secondary configurations shorted because it is difficult to obtaining accurate
measurements on the 1/2 turn low impedance secondary side. Equivalent
secondary leakage inductance with the primary shorted was calculated from the
measured primary values.
TABLE I
Heasured Calculated
Description Primary Secondary
(1) Primaryonly (20 turns) --no secondary
(2) I Full turn copper strip secondary
1480.JJH
1.6 JJH InH (Ideal)
(3) 1 Half turn strip -no flux hal. wdg.
(4) Same with flux bal. opposite tab end
(5) Same with flux hal. wdg. at tab end
944 IJH
144 "
38 "
885 nH
91 "
24 "
(6) 2 Par. half turns, outer leg (Figure 4)
(7) 2 Par. half turns over primary (Figure 5)
42 uH
8 "
26 nH
SnH
Measured
Secondary
185 nH
1580 "
307 "
207 "
(8) 5 turn wire sec. spread across centerleg
(9) 51/2 turn secondary -no flux bale wdg.
(10) Same with flux bal. opposIte end
(11) Same with flux bale same end
2.9 uH
17.5 "
4.2 "
2.8 "
Line (1) of Table I shows the open circuit primary inductance of 20 turns on
the EC41 core. Line (2) demonstrates the lowest leakage inductance that can
be obtained without interleaving the secondary between two primary half-
sections. Dividing the 1.6 uH measured primary value by (20/.5)2 gives I nH
lowest possible leakage inductance for 1/2 turn -the ideal goal. (3) shows
how bad a single half turn strip is without a flux balancing winding. Adding
flux balancing opposite the tab end of the half turn (4) provides much
improvement, but at the tab end (5) coupling between flux balance winding and
the half turn is much better. Still, 24 nH is a long way from the I nH goal.
It is just not possible for the flux balance winding at the one end of the
centerleg to couple more effectively to the half turn along its entire length.
Line (6) shows the technique of figure 4, with one turn of strip around each
outer leg. The large amount of stray flux between these turns and the primary
cause high leakage inductance. Best is (7), with the two half-cylindrical
strips directly over the primary. Most of the 5 nH is in the cross-connected
R6-7
181 nH
1320 "
317 "
211 "
tabs at one end, and this could be further reduced by putting.tabs at both
ends. This is the best approach when most of the transformer power is in the
half-turn (or I 1/2 turn) winding.
Lines (8) -(11) show the results of adding a half turn to several full
secondary turns, using wire instead of strip. The secondary impedance levels
are high enough to take measurements from the secondary side as well. (8)
shows that the 5 full turn secondary does not couple as tightly to the primary
as the shorted strIp in (2). This is because the 5 turns were spread across
the centerleg with large spaces between turns (but this is much better than
bunching the 5 turns in the center of the primary). Several parallel wIres
should have been used to fill the centerleg. The 185 nH is almost twIce what
it should be compared to (2) .Note that with the additional half turn placed
at the same end of the centerleg as the flux balance winding (11) the coupling
is good and the additIonal leakage inductance of the half turn Is only 22 nH.
'l11is compares to the half turn secondary alone in (5), but in (11) it is small
by comparison to the leakage inductance of the 5 full secondary turns .
References:
G. Perica, "Elimination of Leakage Effects Related to the Use of Windings
with ~ractions of TUrns," Proceedings of Power Electronics Specialists
Conference (PESC), 1984, pp. 268-278
I.
R6-8
IMPORTANT NOTICE
Texas Instruments Incorporated and its subsidiaries (TI) reserve the right to make corrections, modifications,
enhancements, improvements, and other changes to its products and services at any time and to discontinue
any product or service without notice. Customers should obtain the latest relevant information before placing
orders and should verify that such information is current and complete. All products are sold subject to TI’s terms
and conditions of sale supplied at the time of order acknowledgment.
TI warrants performance of its hardware products to the specifications applicable at the time of sale in
accordance with TI’s standard warranty. Testing and other quality control techniques are used to the extent TI
deems necessary to support this warranty. Except where manda