1

Practical
transformer
winding
In the good old times it was a matter
of fact that every electronic hobbyist
or technician would wind himself any
power transformers he needed, and
rewind any that burned out.
Unfortunately, nowadays transformer
winding is fast becoming a lost art,
and I have seen many people despair
about where to find some very
specific transformer, or pull their hair
out about the cost of having one
professionally wound to
specifications.
Since I started in electronics, as a 12
year old boy, I have always wound
my own transformers. I started using
the basic, but useful instructions
provided in The Radio Amateur's
Handbook of the time, and later I
came to better understand how
transformers work, which enabled me to optimize a given transformer for the intended application.
Following a request by many readers of my web site, I've added this page, which is complementary to the
previously published Transformers and coils. You should first read (and understand!) that page, before trying
to design any transformer. Then come to this more practically-oriented page, to learn some tricks and hints
about the design process, and about hands-on winding.
This page addresses mainly single-phase power transformers in the power range from about 1 watt to 10,000
watts, operating at line frequencies, but much of what's described here can be applied to a wide range of
other transformers too.

Let's start with the materials. To
make a typical transformer, you need
the iron laminations for the core,
enameled copper wire of several
different diameters for the windings,
a bobbin (or some material to make
one), insulating material to apply
between wire layers, between
windings, around the whole winding
assembly, and on exposed wires, and
2

in most cases it's also a good idea to use an impregnation varnish.
The photo here shows several stacks of iron E-I laminations, two coils of wire (with cardboard protecting the
wire from damage), one roll of thick, stiff Pressspan, another roll of NMN laminate (we will soon see what
that is), two small bundles of spaghetti for wire protection, and a can of transformer varnish. Add to this
some glue, cotton straps, ropes, adhesive tape, terminals, bolts, angle iron, and other small material, and
that's it.
All these materials are sold by companies specializing in transformers and parts for transformers. Enameled
wire is also sold by many other distributors, but is usually cheapest at the places that sell it together with the
other materials. You will have to dig into the phone book or some other directory to find these companies,
since they don't usually have a shiny nice store in the downtown shopping mall!

Transformer iron is an alloy of iron
with silicon and some other minor
components. It's characterized by a
relatively high permeability, very
high saturation flux density, relatively
low hysteresis loss, and relatively
high specific resistance. This latter
factor, along with the practice of
using the material in thin, insulated
sheets, reduces the power losses
produced by eddy currents.
The most common shape of these
sheets is shown at right. It's the
classic "economy E-I" shape. Why it's
called E-I should be pretty obvious
when looking at the photo! But the
explanation for "economy" might be a
bit more elusive: It's because at the
exact proportions shown in the photo, the I's are nothing else than the cutouts to make the windows in the E's,
when two E's are cut facing each other! This allows stamping E's and I's out of a large steel sheet, without
any wasting of material, except for the little round bits cut out of the bolt holes. By the way, small
laminations often don't have such bolt holes, and such cores are held together by clamps instead of bolts, or
even welded.
The lamination in the photo is a large one, as the comparison with my hand shows. It's an E80 (the center leg
is 80mm wide), typically used for transformers in the 3 to 10 kilowatt range!
In any E-I lamination you are likely to encounter, the center leg is twice as wide as each of the other parts.
This is because the entire magnetic flux has to go through the center leg, but then splits up, with one half of
the flux returning through each of the side legs. If you ever come across a lamination that has all three legs of
the same width, then you are looking at a lamination intended for three phase transformers!
Such an economy E-I lamination like shown here has completely fixed proportions, beyond the rule above,
that stem from the need to cut the I out of the winding window of two E's facing each other: If the center leg
is 2 units wide, then the window is 1 x 3 units, the total E is 6 x 4 units, the I is 1 x 6 units, and so on.

3

Not all laminations follow the
"economy" proportions, though. Here
is an example of a lamination that
comes in one piece, instead of being
divided into an E and an I, and that
has the windows proportionally much
larger than the E-I lamination shown
above. Such a lamination is a bit
more expensive to make, because the
steel cut from the windows is wasted,
unless the manufacturer can find
some other use for it. But being able
to accomodate a much large winding
assembly, it has some advantages in
certain cases.
These "non-economy" laminations
were quite usual in Europe, many
years ago, but nowadays copper is so
much more expensive than steel, that
transformers are usually designed to use more steel and less copper. And for that goal, the economy
lamination is very well suited. So you won't very often come across a lamination like this, unless you are
restoring antique equipment.

The laminations should be thin, and
reasonably well insulated from each
other, to reduce eddy currents to an
insignificant value. Typical
thicknesses vary from 0.2 to 0.5mm,
but higher frequency transformers
(audio) use much thinner ones, while
extremely large transformers might
use slightly thicker ones.
The insulation is often applied at the
factory that makes the big rolls of
steel sheet, even before stamping the
E's and I's. Different kinds of
insulation are used: A thin oxide
layer, a thin layer of enamel, or any
of several chemical processes.
Antique transformers sometimes even
used very thin paper!
When I was young, patient and overly eager to do things right, I painted each and every E and I for my
transformers, using diluted transformer varnish, to make a thin, nice layer. The photo shows the steel for a
200 watt transformer, drying. Later, getting old and lazy, I noticed that the layer of rust on old, recycled
laminations is more than enough insulation, and that the very thin and imperfect insulation that comes on
new laminations is enough too, even if it takes only a light scratch with the multimeter's test probe to
4

puncture it and get through to the steel. We don't need perfect insulation between the sheets! We only need
enough resistance to reduce eddy currents to a low level.
Transformer steel is not all born alike. Manufacturers will provide data sheets about their products (often on
their web sites), where you can see what they offer. There are usually many grades, with vastly different loss
characteristics. At a given flux density and frequency, a good material might have ten times less loss than a
cheap material! So it pays to look, investigate, and decide intelligently what to buy. Thinner sheets normally
have lower loss, and the rest of the secret lies in the exact alloy. In any case, you need to know what material
you have, to be able to make a meaningful transformer design!
Some transformer steel is grain-oriented. That means that when rolling the steel sheets, a process is used to
align the crystalline grains in the direction of the rolling. This kind of material has particularly good behavior
when the magnetic flux is aligned with the direction in which the sheet was rolled, but is worse than standard
material in the perpendicular direction. Such grain-oriented material is ideal for toroidal cores, which are
made by coiling up a long strip of steel, but is not a large improvement for E-I laminations, because in these
a significant portion of the material has to work with the flux perpendicular to the rolling direction.

Enamelled copper wire comes in
many different diameters, and with
several different kinds of enamel. The
diameters vary from less than that of
a hair, to about that of a child's finger.
Different standards exist for the wire
diameter. A very common one is
American Wire Gauge, shortened to
AWG, which is used in much of the
world. Britain has its own standard,
and in many countries the wire is
specified simply by its diameter in
millimeters.
Thick wires usually are coated with a
sort of enamel that is very tough, an
excellent insulator, highly heat-
resistant, highly resistant to solvents,
and that clings to copper even better
than dirt does to children! This enamel is usually yellowish clear, so that the wire coated in it looks mostly
copper-colored, but many exceptions exist. To solder the ends of these wires, it's necessary to scrape off the
enamel, using a sharp knife or similar tool. This procedure would be too difficult with a thin, fragile wire, so
that these thin wires are instead covered with an enamel that has most of the same characteristics of the other
one, except the heat resistance: It will melt and turn into solder flux at a temperature a common soldering
iron easily achieves! This allows easily soldering these wires, without previously stripping them. But
transformers using this latter kind of wire enamel cannot survive temperatures as high as those using only the
former kind of wire enamel. The red wire on the right side in this photo has this kind of enamel. But be
careful with colors! The clear wire on the extreme left side also has solderable enamel, while the dark violet
one in the middle is of the non-melting variety!
The thickness of the enamel layer depends on the wire thickness, the manufacturer, and can sometimes be
chosen: Some manufacturers will offer the wire with seeral different thicknesses of enamel. In any case, the
diameter specified by a certain AWG number refers to the copper diameter, so that the complete wire, with
5

enamel, will be slightly thicker than what the AWG standard tells!

Here is a wire table for AWG wire. It
shows the AWG number, the
diameter in millimeters excluding the
enamel, the approximate typical total
diameter including the enamel (but
this varies somewhat), the cross
sectional copper area in square
millimeters, the area of the square of
window space occupied by that wire
in a transformer (including the
enamel, of course), the current
carrying capacity at a typical, average
value of current density, the
resistance in ohms per meter, and
finally how many meters of that wire
come in one kilogram, because
enamelled wire is usually bought by
weight, not length.
This table has wires from AWG #1 to #40, and for the thickest ones I didn't calculate all data. But you should
be aware that there are wires exceeding this range! The thinnest I have ever used was #46. It breaks when
you blow at it! The photo here shows a #39 wire lying on a #7 wire. The hairy thing below is my floor carpet.
Note that even this #39 wire is not much thicker than the hairs of this carpet!
It's interesting to note that every three AWG numbers, the cross sectional area exactly doubles. Any
deviation from this in my table is due to approximation errors.

Modern transformers of small to
moderate size are usually wound on
plastic bobbins. Here you can see
simple ones. Some bobbins have pins
or terminals molded into them, others
have one or two divisions. Some don't
have the slits for terminals, which the
ones shown here do have.
Typically for a given size of E-I
laminations, bobbins will be available
in two or three sizes, accomodating
different numbers of steel sheets. So
you can vary the amount of steel in
your transformer not only by
choosing the lamination size, but also
the height of the lamination stack!

6

Here is a little transformer using a
divided (or split) bobbin. This is very
practical, because it completely
separates the primary from the
secondary winding, making it much
easier to achieve the degree of
insulation required for safety. More
about that later.

If you cannot find a plastic bobbin in the proper size, don't despair! Bobbins can be easily made from
materials such as strong cardboard, or Pressspan, which is nothing else than a particularly strong cardboard.
The bobbin shown here was made from 1.5mm thick Pressspan, which is really too thick for this small
bobbin, but I had nothing better on hand. The pieces are cut to size using a sharp knife (X-acto or the like),
and glued together with cyanoacrylate adhesive (instant bonder). The clever structural design of this super
high tech bobbin holds it together perfectly while the glue sets!
You must make the inner dimensions of the bobbin core a tad larger than the transformer center leg, but
JUST a tad, no more, unless you want to waste valuable winding space! The sides can be made pretty tight to
the size of the laminations, because if they don't fit at the end, they are easily enough cut or filed down, even
7

after the winding has been made. But the length of the bobbin must be smaller than the window length of the
core, by as much as 2 or 3%, plus any tolerances of your manufacture! Because it is critically important that
the E's and I's can touch each other properly, without being kept separated by a bobbin that deformed during
winding, and grew!

Be sure to at least break the corners
as shown here, or even better, round
them off. Otherwise the wire is
guaranteed to tangle at the sharp
corners during winding, and a wire
loop sticking out of the completed
winding can ruin the whole thing!
Note that the junction of the bobbin's
center piece is placed in the middle of
one side, and not in a corner. It's next
to impossible to produce a reasonably
symmetrical and precise bobbin when
placing the junction in a corner.
If the material is thin compared to the
bobbin size, the junction should be
made by overlapping the material. Of
course, the overlapped junction is
always placed on one side that will
end up outside the core window, so
that the added bulk has little
detrimental effect.
To bend this thick material in reasonably clean right angles, my technique is to use a sharp knife to cut out a
90 degree wedge from the inside, along each bend line, leaving only the outer third of the material
intact. After that admittedly cruel treatment, the Pressspan eagerly bends to my will.

8

Most antique transformers, and many
of the larger modern ones, don't use a
real complete bobbin. Instead, they
use only the center former, and no
sides at all! It takes some tricks and
practice to wind a transformer like
this without having the whole thing
come apart many times over during
winding, but for people who have
acquired enough practice, it's faster
than making a real bobbin!
Further down, I will show you a trick
to make this kind of transformer, with
high quality.
Note in this photo how the
Argentinian maker of this transformer
(back in 1931!) used insulating
material of several different
thicknesses for the bobbin center, the
interlayer insulation, the interwinding
insulation, and the terminal support.
You might ask why any insulation
material is required at all, if the wire
is insulated by its enamel layer! Well,
the enamel is very thin, and easily scratched. It might survive as much as a few thousand volts, but it might
also break down with a lot less! It depends on type, condition, thickness, temperature, and other factors. So,
wherever the voltage can exceed a few tens of volts, some additional insulation needs to be used. Specially
between the primary and secondary, safety regulations ask for an insulation good for at least 4000 volts, to
avoid electrocuting somebody when there is a lightning transient on the AC power network.
In antique transformers, the most usual insulating material was paper, impregnated with something like
beeswax, tar or the like. This impregnation had several purposes: Mainly, it would seal the pores of the
paper, making it a really good insulator, while without the impregnation it would only insulate as well as the
same thickness of air! But in addition, it kept moisture out, it helped stick the thin wires in place during
winding, and it improved the thermal conductivity of the completed winding assembly.
It was also quite sticky, dirty, messy and gross.
Modern insulating materials are far superior. Plastic sheets such as Mylar provide excellent dielectric
strength and have no pores, so they require no impregnation to realize their high degree of insulation. Nomex
instead, with its fibrous structure, behaves like paper, but both Nomex and Mylar are much better than paper
at surviving high temperatures! This is a key characteristic of insulating materials: The temperature class. It's
coded with a letter. Paper would have an A or B rating, telling that it is fine for temperatures not much above
that of boiling water. Different plastic insulation materials instead are routinely available in classes as high as
F, G or even H! They can safely run much hotter than paper can.

9

The photo shows an NMN insulating
sheet. This is a sandwich of a Mylar
sheet embedded between two layers of
Nomex. The Nomex will eagerly soak
up and distribute the impregnation
varnish (or the oil, in an oil-inmersed
transformer), while the Mylar will
provide safe insulation even in places
that for any reason stayed dry! I love
this material. It's thermal class G, if I
remember right.
Insulation materials come not only in
many different variants, and
temperature classes, but of course also
in many different thicknesses. You
choose the proper thickness so that it
has enough dielectric strength and mechanical strength, without taking up an undue portion of your valuable
window space!
Despite all modern materials, good old paper and cardboard is still used sometimes. Mostly in its form
known by the German word Pressspan, which means "compressed chips", and is simply a very dense paper
or cardboard.

It's very good practice to soak a completed transformer in some impregnation varnish. It will form fillets
around wires, papers, and anything else. It will improve the insulation, make the transformer highly
moisture-proof, glue everything together so that nothing can rattle, come loose, or chafe through, it will
improve thermal transfer, and so on.
Varnish comes in several thermal classes, just like the insulation material, and also it comes in variants that
dry at high temperature, or at room temperature. My experience is that no varnish ever fully dries at room
temperature, and when you start using the transformer and it warms up, the varnish inside will start drying,
and stink! So, it's necessary to apply heat anyway, regardless of what sort of varnish you use.


Now that you have turned into a person very knowledgeable about transformer materials, let's turn to those
pesky questions such as "how many turns do I have to wind?" or "what wire size?" or "how much power will
I get?"
There are three typical situations:
1. You need to repair/rewind a transformer that burned out.
2. Your want to rewind an existing transformer, to produce the voltages and currents you need, which are
different from the original ones. A variation of this case is when you want a certain voltage, at the highest
current that transformer can provide.
10

3. You have fixed specifications, want to design a transformer to optimally meet them, and you will buy the
core and all other materials.

Let's start with the first case. You
MUST find out why that transformer
failed! A correctly designed, correctly
built and correctly used transformer
is, for all practical ends, eternal. If it
failed, there is a reason. If you know
that the transformer was shorted,
overloaded for a long while, exposed
to intense lightning transients, thrown
into water, gnawed through by rats,
exposed to corrosive substances, or
anything like that, then your best
approach is to unwind it, count the
turns, measure the wire sizes, and
rewind it exactly as it was originally
made.
The photo shows an antique speaker
field coil, mounted in my winding
machine. I unwind coils by pulling off the wire while having the thing spin in the machine, so that the turns
counter in the machine will do the pesky job of keeping count. The problem, as illustrated here, is often that
thin wires won't come off nicely! They are stuck in place, and will break, then entire chunks of wire will
come off all together. This often makes it hard to accurately count the turns.
In such cases you might simply estimate how many turns you didn't count. Or you can collect all the pieces
of wire you removed, weigh them, calculate the amount of wire from there, and calculate the turns number
from it. Or, instead of unwinding the coil, cut it with a knife or better a Dremel tool, remove it in one block,
measure the cross sectional area of the entire winding, then remove a little piece of wire to measure the
diameter, and finally calculate the number of turns from this. Any of these methods will usually be precise
enough for non-demanding applications, and none of it will be precise enough when you need anything
critical.
By the way, do you know how to precisely measure the diameter of a thin wire, when you don't have a
micrometer screw? Simple: You wind 10 or 20 or even more turns tightly on a former (such as a screwdriver
stem), measure the length of the coil with a common ruler, then divide by the number of turns to get the wire
diameter. It's highly accurate. Sometimes it's even better than using a micrometer screw, which can flatten
the wire if you apply too much torque!
But there are cases when you have a burned transformer, and no good reason why it burned. It might have
been a manufacturing defect, a huge transient, an overload that went undetected, but maybe - the Gods of
Electromagnetics forbid - that transformer might have been misdesigned! In that case, painstakingly
rewinding it with the same wire gauges and turn numbers as original, will only produce a transformer that
will fail again. So, if you don't know why a transformer failed, re-do the design, and compare your results
with what the manufacturer did! Many manufacturers are cheapskates, and use substandard transformers, in
the hope that most clients will never use them intensively enough to blow them up! This sends you straight to
the section about the third case, further down this page!
11

In the second case, when you want to rewind an existing transformer for new output values, very often the
transformer will already have a properly wound and healthy primary winding. In such a case, keep it! There
is no point in unwinding and rewinding the primary, if it is fine. The calculation work for such a transformer
is quite simple: Before taking it apart, measure the voltage delivered by the secondary. Disassemble it,
unwind the secondary, counting the turns, and calculate the number of turns per volt from this. Calculate the
new number of turns you need for your desired voltage. Calculate what's the largest wire size that will
comfortably fit in the available space. Get the wire, wind it, and assemble the transformer. The power rating
will be the same as before, and this allows you to calculate the current you can safely draw, at your new
voltage.
And if you also want a new primary, well, do the same as above, but unwind and re-wind all windings,
according to the value of turns per volt you found out!

Do you know what you can do with
all that wire you remove from old
transformers? Well, a super trendy
wig, like the one shown here,
modelled by my sister, is sure to
catch everybody's attention!
Otherwise, there isn't really much use
for such wire. It comes out totally
kinked, stretched, broken, scraped,
with varnish, wax or tar sticking to it.
Don't even dream about ever re-using
it in any other transformer!
OK, now it's time to really start
designing a transformer. Because this
is what you have to do in case 3,
which is essentially creating a
transformer from scratch. I hope you
are still fresh and aware of everything
you learned in Transformers and
coils, because you will need it here to
understand what's going on.
The design process starts with
guessing the size of core you need,
for the power you want. If you are
experienced in electronics, you will be able to make a reasonably close first guess. Otherwise, use the data
given by core manufacturers to get this first guess.
The core chosen will have a certain cross sectional area, and will have a certain window area. The two
multiplied are the "area product", which can be related to the approximate power capability, with a
reasonably simple, but nonlinear curve. This curve also depends on the quality of the core material, and
several other factors.
The next step is calculating how many turns per volt you need on this core. To this end, you have to decide
how much flux density you will put through your core, and then you can apply the equations from
12

Transformers and coils. The optimal flux density might be anything from 0.8 to 1.6 Tesla, and sometimes
even outside this range! General rules of thumb are these:
- Larger transformers use lower flux densities.
- Better core material uses higher flux densities.
- Transformers that are always energized, but rarely used at full power, use lower flux densities.
- Likewise, transformers that work at full power whenever energized, use very high flux density.
- Forced air cooled transformers use higher flux density.
- Oil-immersed transformers use even higher flux density!
- Higher flux density produces better voltage regulation.
- Lower flux density produces less base loss.
- Lower flux density is less likely to produce humming noise, and magnetic stray fields.
- Lower flux density produces lower iron loss, but higher copper loss.

I have seen many text books giving design equations that result in a flux density of 1 Tesla in each and every
transformer you calculate by them, like if that were a sacred rule! If you come across any such book, BURN
IT! It's nonsense! While 1 Tesla indeed tends to produce a workable transformer in most cases, in at least
70% of all situations it's far enough from the optimal value to warrant some effort toward optimization!
Specially in small transformers, and in those using the better core materials.
So, I suggest to start with a value chosen from the rules above, and then calculate the transformer based on
this value, analyzing the losses, heating, voltage drop, efficiency, and so on. Then change the flux density, by
10 or 20%, and re-do all the calculations. You will see what I mean! There is a clear optimum value for flux
density in each particular case, and this value is very often sufficiently removed from 1 Tesla to make you
wonder why some book authors still copy that "magic number" from other, long gone authors! Most likely
they have no idea about the matter they are copying.
That said, sometimes I do wind my transformers for 1 Tesla, because there are cases when this is really a
good value!
The loss calculation isn't very hard: The manufacturers of transformer steel specify the loss of their products,
as a certain amount of watts per unit of volume or weight, under certain conditions of frequency and flux
density. And the better of these data sheets also contain curves, or give equations, to calculate the loss under
different conditions. That's about the iron loss. The total loss of the transformer also includes the copper loss,
which is caused simply by the current flowing through the resistance of the wire. For a given flux density
you get a certain number of turns per volt, this allows you to calculate the total turns in each winding.
Dividing the available window cross section between these wires, insulation, bobbin, and inevitably wasted
space, you get the wire cross sections. From the dimensions of the core, you can then calculate the total wire
length, from this and the cross sections you calculate the resistances of the windings, and from that and the
13

currents you calculate the power loss. It's easy!

What? You are banging your head
against the wall? Come on! Don't
cry! I have prepared an Excel sheet
for you, which takes care of these
calculations! It's really too time
consuming to do all this math step by
step, every time. Click on the
screenshot at right to get the Excel
file, and then play with it as you
continue reading.
In this sheet, you enter your data in
the green area, and watch what
happens in the orange one. The first
three values you have to enter are the
ones that specify the core. First is the
center leg width of the lamination
used, in millimeters. The 100mm
width given in the example is a pretty
large lamination. You will typically use values between 12 and 50mm. The sheet is based on the proportions
of the economy E-I core, so that you don't need to enter any other dimensions of the lamination. If you
happen to be using a lamination that has different proportions than the economy E-I, you should still enter
the width of the center leg, but later you will need to manually compensate for the larger available window,
by increasing wire size beyond that calculated by the sheet, reducing the copper loss, and so on.
The second value you have to enter about the core is the stack height, also expressed in millimeters. This is
simply the height of the stack of E's, well compressed. This sheet does not allow you to enter the stacking
factor, which tells how much of the stack is actually steel. There's always some little space used by
insulation, and even wasted space due to imperfect compression. But the effect of this is small enough to
ignore, as long as you compress the stack well enough!
Good stack heights to use are from the same as the center leg width, to close to twice that. Often you have
the choice to use a certain lamination, stacked as high as the center leg is wide, or use the next smaller
lamination, stacked much higher, with both options producing the same output. It pays to simulate both
options, optimize each, and compare the efficiency, voltage drop, weight, and cost! The differences can be
profound.
And the third value is the loss factor of the steel material. This must be taken from the datasheet provided by
the manufacturer of the lamination. My sheet expects this loss factor to be expressed in watts per kilogram of
material, at a flux density of 1 tesla and a frequency of 50 hertz. Many data sheets include the value in this
exact form, but those published by US companies might instead express the loss factor in an eclectic mixture
of metric, CGS and Imperial units! If that's what you have, you will need to convert the value into its fully
metric equivalent. You might want to modify the spreadsheet to do that.
The value of 2 W/kg @ 1T and 50Hz is pretty representative for modern low cost laminations. A modern
ultra low loss material might be a lot better, while an antique or ultra-cheap material might be significantly
worse.
14

The next three values in that column are pretty obvious: You have to indicate the primary and secondary
voltages, and the frequency of operation. The secondary voltage refers to the open circuit (no load) voltage.
And the frequency will usually be either 50 or 60 Hertz. If you enter a frequency far away from this, it's quite
possible that the loss calculated for the material will be rather imprecise, so use this sheet with caution if you
need to design a transformer for a very different frequency.
In the right hand column, you have four values which are design decisions which you can vary somewhat.
The first is the all-important flux density. Just try varying that value, and watch how things change in the
orange output area! Specially, see what happens with the iron loss. I have already given guidelines about
what flux density to use. Use them and see what happens in your case, when you change it.
The next is the amount of copper cross section you will allow for each ampere of current in the windings.
Reasonable values are about 0.25 mm^2/A for very small transformers, increasing to 0.5 for large ones. 0.35
is typical for medium sized transformers (50 to 300 watts or so). When you adjust this value, the design of
the transformer doesn't really change, but the sheet will calculate a new set of currents, power, voltage drop,
efficiency and loss. With this parameter, you basically are telling the sheet how much you will stress a
particular transformer.
The fill factor expresses how much of the lamination's window will actually be filled with copper. It can
never be very high, because a lot of that area gets inevitably filled out with the bobbin, the wire's lacquer, the
air around the round wire, the insulation between layers, between windings, and some space is always lost
due to sloppy winding, even if you are careful! The value of 0.4 used in this example has proven in practice
to be achievable without much trouble. If you wind very carefully, and minimize the amount of space
devoted to insulation, you should be able to get up to 0.5. But don't push this number too much, or you will
end up with a transformer design that you cannot actually wind! On the contrary, if you have never before
wound a transformer, and will do so by hand, in a ragged, ugly winding, it might be a good idea to design the
transformer with an even lower fill factor, such as 0.3, to make sure you will be able to fit all the turns! Of
course, using a lower fill factor means simply using thinner wire, and this means that at a given amount of
loss and heating, you get less current.
The fill factor can be pushed beyond 0.5 when you wind a transformer with square wire (instead of round), or
with copper tape separated by thin layers of insulation. But square wire is hard to find and a hassle to wind
properly, and tape winding is acceptably easy only for transformers that have rather few turns. This is often
the case with high frequency transformers used in switching power supplies, but not at line frequencies.
The temperature rise defines how many Kelvins (same as degrees Celsius in this case) hotter than the
surrounding air and objects you want your transformer to run. You need to carefully choose this value,
according to the highest ambient temperature (inside the equipment!) at which your transformer will have to
work, also taking into account the highest temperature your wire, insulation material, varnish, glue, etc, can
survive. And what's most difficult, you will also need to estimate the thermal gradient from the innermost
wire turns (the hottest ones) to the transformer's surface! Calculating all this can be quite hard, and I can't
give you simplified equations for it. Maybe you can find them elsewhere. The value of 70 Kelvins which I
used in the example design is relatively high. This is so because this transformer would work in open air, not
inside a housing, where the ambient temperature is never above 25 degrees Celsius; also, I used class G
or higher insulating material, wire and varnish throughout; and finally, this transformer was carefully
impregnated with varnish, giving it a reasonably good thermal conductivity between winding layers.
If your transformer will not be impregnated, or use class A or B insulation material (paper), or run inside a
cabinet that can be hot, then you will need to use a lower value for allowed temperature rise then my 70
Kelvins!
15

Finally, you can enter your local and current prices for enameled copper wire and transformer steel
laminations, to have the sheet calculate the cost of these main materials for your transformer. All the
additional cost, for the bobbin, insulation material, terminals, bolts and so on, is usually small compared to
the copper and steel cost. The most expensive item is usually the wire, by far.
The first two lines of the orange output area of my spreadsheet show some basic results for that transformer:
The cross sectional area of the magnetic core and of the winding window, also the total copper area (after
applying the fill factor), the turns per volt constant that will be valid for all windings on this transformer, and
the average length of one turn, which is calculated as the average between the length of a wire that goes
around the center leg touching it, and one that goes around the entire winding package, touching the outer
legs.
Then you have a line for the primary winding and one for the secondary. Each of these lines tells you the
number of turns, the length of the wire need to wind it, the copper cross sectional area of the wire, and the
nominal current that will flow at full rating. The number of turns isn't rounded off, so you will have to do
that, because you can't wind a fraction of a turn. You can fiddle with your voltage data to get the sheet to
show round numbers of turns. The wire length is based on the length of the average turn, so this will be
correct only if you wind the primary and secondary side-by-side, on a split bobbin. If instead you wind
the secondary on top of the primary, you will need less wire than calculated for the primary, and more than
calculated for the secondary. Anyway, these lengths are not very useful in practice, because wire is bought by
weight, not length. The main situation where they are useful is when you have to wind a transformer with
several thin wires in parallel, because these are much easier to bend than one thick wire. In such a situation,
it's great to know how long the total winding is, so that you can cut the strands and twist them together,
before you start winding.
The remainder of the orange output area is divided into two columns. The left one shows some important
performance data of the transformer: There is the input power, expressed in voltamperes, which is really
more correct than watts. The value calculated by the sheet does not include the magnetizing current;
Calculating it would need additional information about the core. But in medium to larger transformers, at
least, the magnetizing current tends to be small enough to be ignored.
Then comes the percentual power loss of the transformer, at full load. This includes both the iron loss and
copper loss. The output power is of course the input power minus this loss.
Then we can see the voltage drop at full load. The value is calculated only from the resistances of the
windings. Any additional loss caused by imperfect coupling between the windings is not considered here. So,
if you use a poor core or winding technique, that results in bad coupling, you should expect a somewhat
higher voltage drop. Just for user convenience, the sheet also calculates the output voltage under full load,
which is based on the voltage drop calculated above.
Below comes the weight of iron laminations and copper wire used in the transformer. This has several
purposes. One is knowing how heavy the beast will be, of course. The other is knowing how much material
you have to buy! If you use side-by-side winding of the primary and secondary, you need to buy one half the
calculated copper wire of each size (plus some extra, of course, to be on the safe side). If instead you wind
the secondary over the primary, you need a little less than half of that weight of the primary wire, and a little
more than half that weight of the secondary.
And then, the sheet will calculate the total cost for laminations and wire, and also divide this by the power, to
give the cost per watt for your transformer, which is a good figure of merit which you might want to
optimize. Even while this calculation doesn't include the cost for insulating material and other odds and ends,
it's still a good reference.
16

The right side column of this area is about thermal matters. These tell whether your transformer will survive,
so don't take them lightly! The sheet calculates the power loss in the iron, in the copper, and adds them to
get the total power loss. Copper loss is calculated at ambient temperature, though. When the wire heats up,
its resistance increases, and so its loss increases too! For this reason, take the calculated value with a small
grain of salt. The same is true for the voltage drop end efficiency calculations!
You need to be aware of the fact that the iron loss is essentially constant, regardless of the load placed on the
transformer, except for a little effect caused by voltage drop in the windings reducing available magnetizing
voltage, which causes a slight decrease in core loss when the load gets higher ! The loss in the wire instead
increases with the square of the current taken from the transformer, and the value calculated by the sheet is
for the full rated current.
This gives you some big help in optimizing a transformer design. For example, a transformer that will spend
lots of time plugged in, but idling or loafing along at low load, will see very little copper loss, but the iron
loss will be there all the time. So, you should design that transformer with a relatively low flux density,
resulting in low core loss, accepting a higher copper loss instead, by setting a smaller value of copper cross
section per ampere. After all, most of the time the rated design current won't be present, so that the very high
copper loss resulting in the calculation will be present only very rarely, for short times! Transformers used in
radio communication equipment, in audio amplifiers, and many other uses, are best designed in this way.
The opposite case happens with transformers that are energized only briefly, but run at full output power
whenever energized. Examples of such use are microwave ovens and spot welders. In such a transformer,
iron loss always happens at the same time as copper loss, and you can optimize the transformer to get the
lowest total loss, regardless of how it distributes between the iron and the copper. Even more, you might
intentionally place more loss into the core than the windings, based on the fact that the core is less prone to
be damaged by heat, has more thermal mass, and that the short operation time won't allow the peak heat
to distribute through the transformer! And then, such a transformer that operates only for short times can
be designed to have a really huge loss, because it will have time to cool off between uses! These things are
what makes microwave oven transformers that deliver 800 watts be as small as a 200 watt transformer
intended for continuous service at low rate, and run at a flux density of 2 teslas or even more!
The heat produced by a transformer has to be dissipated to the surrounding air. The spreadsheet calculates the
approximate total surface area of the transformer, and finally calculates a required thermal transfer
coefficient, which expresses how much power the transformer needs to dissipate per unit of surface it has,
and per temperature rise allowed. This coefficient tells you how difficult it will be to keep this transformer
cool enough to survive! The violet area below this coefficient includes referential values (not calculated by
the sheet) which you can use to try judging whether your transformer will be OK, when you have it in a tight
area, in a more open area, cooled by a fan, or immersed in oil. I have my doubts about these values, specially
aout the value for the oil-immersed transformer, so please take these with a big grain of salt, and let me know
if you have any further, better, or more reliable data.
Anyway, my transformers designed for a coefficient of around 12 have all survived so far, even while getting
quite hot at full load, so this value can't be too far from the truth.

A typical design sequence using this spreadsheet would be to first enter the tentative core size and loss, the
required voltages and frequency, then start with something like 1 tesla and 0.35mm^2/A, leaving the fill
factor at 0.4 and setting the temperature rise according to your transformer's materials and environment. Then
you can observe the power and current it would operate at, and the losses, efficiency, voltage drop, and also
you would get the thermal transfer coefficient which you can compare to the table to gain an idea of whether
17

the transformer will survive. You can then tweak the flux density and current density, trying to get the
characteristics into the range you need, without exceeding the thermal possibilities. If you just can't find a
combination that provides what you need, you will have to try with a larger (or lower loss) core. Then you
might want to explore several different core sizes, optimizing each, and watch the cost, finally settling for the
design that best provides what you need, at the lowest possible cost, while staying in the survivable thermal
range.
This work with the spreadsheet is only the first step, though. When you have arrived at a good design using
the sheet, you need to tweak it to make it buildable with real, available material! For example, you cannot get
wire in any desired diameter. The sheet doesn't know that; It might ask you for a wire measuring 1.2345
square millimeters, or anything else. It's your job to see what wire you can actually buy, or maybe what wire
you happen to have in stock, and adapt the design. The AWG standard is quite finely stepped, so you don't
need to change the design very much to adapt it to standard AWG sizes. But I have heard that in the US
many stores only sell the even numbered AWG sizes of wire! That's odd (pardon the pun), since even down
here in less developed Chile I can easily buy all AWG sizes, even and odd. If you are limited to even sizes
only, you will have to make bigger compromises.
A good approximation technique is to round the wire size to the nearest AWG size, or if the values calculated
fall just in the middle between AWG sizes, you might want to use the next thicker wire for the primary, and
the next smaller wire for the secondary. That way the final losses and the amount of space required will be
almost exactly the same as calculated by the sheet.
If your transformer uses lots of turns of thin wire, you are about ready to start winding at this point. But if it
uses a winding that has few turns of a thick wire, you should check how well (or how poorly) that wire fits in
an integer number of layers, considering the width of the bobbin, and about 5% of lost space due to the wire
not being perfectly straight. The problem is this: If the sheet calculates you need 48 turns of a certain wire
size, and it happens that you can fit only 15 turns per layer, then you will end up with three complete layers,
plus one additional layer in which you have just three turns! So the total height taken up in the window by
those 48 turns will be as much as 4 complete layers, that could have 60 turns! Consequently, the winding
might end up too high, and won't fit the window! Then you cannot assemble the core, and you have to
unwind that winding, throw the wire away, get new (thinner) wire, do it again... you get the idea. Try to
avoid such frustration!
When you are in the situation just described, it would be wise to try the next smaller wire size. It's very likely
that this smaller wire size would accomodate 16 turns per layer, thus allowing you to wind the 48 turns in
three nice, clean, complete layers, and using up a little bit less space than calculated by the sheet. This can in
turn allow you to use the next larger wire size for the other winding, which will almost completely
compensate for the higher loss and voltage drop of the smaller wire you used for the 48 turns!
Putting it in short words, you have to pick the best wire sizes for your transformer so that their cross sections
are close enough to the calculated values, but still allow a good, space-saving distribution on the bobbin and
thus inside the window area of the core.
Sometimes you might end up with a high current transformer requiring a very thick wire. Such a wire is very
stiff! If the bobbin is small, you might not be able to bend that thick wire tightly enough around the corners
of the bobbin. The result would be a huge loss of space, and the completed winding wouldn't fit in the
window, making it unusable. To work around this problem, you can replace one thick wire by a bundle of
seven wires, each of which is one third as thick as the single wire. Such a seven wire bundle twists very
nicely into a round cable, and is more then 20 times as flexible as the single thick wire! It costs only very
slightly more money, and performs great. So, this is the way to go when you have a need for such thick high
current conductors. It's good to know that an AWG number 10 units higher is roughly one third the diameter.
18

So, if you would need a #7 wire and this is too thick to wind comfortably, you can use 7 strands of #17,
lightly twisted together.
Sometimes instead of using one thick wire you will also find it convenient to wind with two or three thinner
wires in parallel, without twisting them. This technique can significantly ease the distribution of a winding in
entire layers.
Another hint: Most transformers use some thin and some thick wire. For transformers that have the windings
on top of each other (instead of side-by-side), I suggest to always first wind the windings that use the thin
wire, then the ones using thicker wires, regardless of which will be primary and secondary. This allows to
wind the stiffer wire on the outside, where the bending radius required is a lot larger and thus less
demanding. Whether the primary is under or over the secondary has no significant effect on performance.
You might have noticed that my spreadsheet only considers transformers that have a single primary and a
single secondary. But many transformers use several secondaries, and some use two or more primaries! In
such cases you will have to do some more work manually. You should use the sheet to calculate the
transformer, simulating only the main secondary, tweaking it for the total power, and then manually reduce
the wire size of that secondary proportionally to the percentage of the total transformer power this
secondary will have to deliver. Then you can add the other windings, calculating their turns number from the
turns per volt calculated by the sheet, and the wire size from the current they have to carry, and the mm^2/A
you selected.
Or if you have two equal secondaries (or two equal primaries!), you can let the sheet calculate a single
secondary (or primary) of twice the voltage. That will produce the correct number of total turns and wire
sizes. You only have to remember to cut the wire after having wound half of the turns, bring it out,
start again and wind the second half!
You might have noticed that I didn't assign additional wire cross section to the primary, to account for the
magnetizing current. The reason is that the magnetizing current is normally much smaller than the main
current, and on top of that, the magnetizing current is 90 degrees out of phase with the main current! The
vectorial sum of the main and magnetizing currents is so little higher than the main current alone, that there
is usually no need to consider the difference.
You might also miss any discussion of core saturation. The problem is that's quite hard to discuss saturation
of silicon steel cores, because they just don't saturate at a well defined level of flux density! Instead, the
saturation is quite gradual: It might start at a level as low as 0.5 tesla, become more noticeable at 1 tesla, then
the curve bends further, but even at 2 tesla there might be a significant amount of permeability left! The
effect of this is that with increasing flux density, the magnetizing current increases more sharply, but it would
be really hard to reach a level where the saturation makes the transformer stop working. So, the most
important consideration about flux density is the sharply increasing core loss. Only when you are using very
high flux density, would it be a good idea to allow some additional cross section for the primary wire, to
accomodate the larger magnetizing current.
Enough calculations. Let's go to winding!


19

Very often, winding a transformer for
a hobbyist will start with unwinding
an old, burned one! The photo shows
part of the guts of a Heathkit tube
tester from the 1950's, which was
given to me in damaged but
restorable condition by my friend
Renato Menare. Its power transformer
worked, but got extremely hot in just
one minute of operation, and started
smelling burned after two minutes.
After restoring the tube tester,
without touching the transformer, I
used it for about a year, switching it
on only for a minute at a time to test
one tube, and then letting it cool off.
But eventually the inevitable
happened: The transformer burned
out, filling my home with smoke.
Such a tube tester transformer is one of the worst nightmares any transformer maker can get: It has two
secondaries, each of which has a huge number of taps to provide all the different filament and plate voltages
needed by any of the thousands of different tubes around in the 1950s! It takes patience to rewind such a
thing. Every few turns you have to install a tap!
The first step is removing the transformer from the circuit, unsoldering all wires and taking notes which wire
goes where. The fact that may wires are the same color doesn't help. Then, the core has to be disassembled.
To do that, you first remove the bolts, then push a sharp knife between the first and second lamination,
prying them apart to break any glue or varnish or rust between them, then grab the lamination with flat pliers
and wiggle it out. Depending on how much the core was compressed during manufacture, this can be quite
hard to do, and one or two laminations might be damaged in the process. That's not the end of the world, the
transformer should later work even with one or two laminations less. Usually, after removing the first few,
the others come out easily. Sometimes each lamination needs a little help with the knife to come loose, while
in other transformers the core falls apart on its own as soon as compression is relieved.

Once the winding assembly has been
freed of the core, it is mounted on a
suitable wooden core in the winding
machine. Then the windings are
carefully unwound, letting the turns
counter of the winding machine do its
job. You should take notes on the
number of turns of each winding,
between each tap and the next, and
anything else you find worthwhile
noting down. Shooting pictures with a
digital camera can also be very
helpful, should you later suddenly get
20

doubts about where each tap connection should be located!
You should also save the wire, so you can measure its diameter and find out what sizes of wire you need to
buy.
As the layers of wire and paper come off, things usually turn darker and ever darker! This is because
transformers work hottest at the inside. This picture shows severely carbonized insulation, both the paper
and on the connection wires. Probably the paper slowly carbonized, becoming slightly conductive, causing
further loss, further heating, until the beast failed for good. This is the problem of paper insulation!


When I was done unwinding this transformer,
I found that even the cardboard winding core
was totally carbonized, crumbly and
definitely unusable. I was left with the heap
of burned, charcoal-like insulation shown
here, kilometers of thin, crumbly, burned
copper wire, burned pieces of hookup wire,
these photos, and a piece of paper where I
had written down wire sizes and turn
numbers.
Don't do such a job in your parent's bedroom,
because as you can see, it's messy! Even the
kitchen might not be the best place to do it.
Neither mothers nor wifes tend to like this sort of transformer autopsy!
When not even the bobbin is usable, you will need to make a new one, or find one that fits. I was lucky this
time, and found one in my junk box (aka treasure chest) that just fit this core, and even had side walls! It's a
great thing that core sizes are standarized, even if several different standards exist...


21

It's probably high time
to introduce my
transformer winding
machine. Here you
can see it in all its
glory, somewhat
obscured by the
messy background
(my workbench).
Amateur radio friend
Enrique Villanueva,
CE5FSB, gave me
this machine when I
was still a schoolboy.
That was many years
ago, Enrique is no
longer in this world,
but I remember him
every time I wind a
transformer! He was a
true gentleman, and
always helpful.
This machine is
simply a motor and switchable two-speed gearbox with central neutral position, that drives a three-
pronged thorn, opposed to which is an adjustable centering screw. It's quite handmade, but does the job well.
The fast speed is about 120 rpm, and the slow one is about 15 rpm.
I added a turns counter to it. It didn't have one when I got it. This turns counter came from a surplus store,
out of a broken gas meter. It advances 4 counts per revolution, so I coupled it to the machine's drive shaft via
a 4:1 speed reduction, getting exactly one count per turn. To make this coupling, I made two gears! One was
cut from the lid of a coffee can, has 16 teeth, and is visible in the picture. The other one was made from a
piece of wire, has 4 wire loops acting as teeth, and is hidden in the photo. It's crude, but was cheap and has
worked well for about 30 years now! I can hardly even try to calculate how many transformers I have wound
wit this machine!


When winding a transformer, you need to
have the wire unwind straight and cleanly
from the spool on which it's sold. If you
simply stand the spool on the ground, the
wire will twist while unwinding, curl up, and
kink. At that point, you can throw it away and
start anew! So, don't. Instead, you need to
fashion some device that allows the spool to
rotate, letting the wire unwind without
kinking. This photo shows one of the
22

methods I often use: A screwdriver, serving as axle, held in a vise.
Sometimes I'm too lazy to set up the vise. In those cases I simply hold that same screwdriver between my
knees while winding the transformer!

Here you can see the junk box
bobbin, made from Pertinax, mounted
on a wooden core in my winding
machine, and with part of the primary
winding already wound.
There are several styles for winding
transformers. The highest quality one
is to painstakingly lay the windings
out in neat, perfect, orderly layers,
with every turn of wire placed
precisely next to its neighbor, in close
contact, and never crossing over.
Each layer is separated from the next
by a sheet of thin insulation material,
cut precisely to the width of the
bobbin, and to such a length that it
will overlap a bit. This overlap is
placed at either of the outside sides of
the bobbin, not on the sides that will
end up in the core's window. Such a
winding is gorgeously beautiful to look at, produces an excellent fill factor, but is very time consuming to
make. I tend to use that technique when winding few turns of thick wire, but not when winding many turns of
thin wire. One time only I wound a 78,000 turn high voltage transformer in that way, because no other
technique would do, but I almost turned crazy, and never repeated that feat.
Modern small split bobbin transformers are normally wound in a very untidy way: The operator at the factory
simply lets the machine run, and fill the bobbin section with wire, letting it build up wherever it wants
to. There are no layers, no insulation between layers, it looks poor, is less reliable, the fill factor is less good
but still acceptable, and it's fast and cheap to do.
The winding shown in this photo, instead, is a hybrid of both techniques. The winding is split up into just a
few layers, two or three or so, but each layer is wound thick, with the wire turns actually building up on each
other. I let the machine run and guide the wire to let it slowly fill the layer from one side to the other (never
going forth and back several times!), letting the windings build up to the desired height, about two
millimeters or six wire diameters in this case. When that ugly layer is complete, I apply a coating of self
adhesive tape as insulation, and then wind the next layer. It's almost as quick to do as a totally wild winding,
but significantly more reliable.
A word about tapes: Transformer parts stores sell self-adhesive mylar tape, usually yellow, in several
thicknesses and widths. This is a good material, and you should use it. But I often don't have it on hand, and
use alternative materials. Vinyl insulating tape can be pressed into service, but isn't great , because it gets
very soft when hot, and the wires can press through it. Much better, even if you don't believe that, is painter's
masking tape! It has just about the optimal elasticity, is thin, cheap, comes in several widths, holds up well
when hot, and looks pretty good on transformers for antique equipment, on which any plastic tape would
23

look out of place! Masking tape is what I used for this transformer.

A question many newcomers have is
how to handle the ends and taps of
windings. When the wire is rather
thick and robust, you can simply let
the end stick out of the bobbin! In
that case it's good practice to protect
that wire by a piece of spaghetti (I
don't mean an Italian noodle, but a
piece of plastic or fiber hose that
looks much like it). It should ideally
be only slightly thicker than the wire.
You slip it over the wire, and anchor
it in the winding assembly, with
adhesive tape and the pressure of the
windings. It will protect the wire from
chafing, and provide additional
insulation at places where the wire
might pass very close to other wires
or to the core.
But thin wires should not be treated like that. They are too fragile, and might very easily break later, forcing
you to rip up and rewind the whole transformer! Instead, you should take some pigtails of stranded, plastic-
insulated wire, strip and tin the ends, solder the thin enamelled wire to the end of a pigtail, and embed it in
the winding like is shown in this photo. You need to place enough layers of insulating material both under
and over the connection, to make sure that no sharp tip or edge of the wires may puncture through the
insulation. This adds quite a lot of bulk, so of course these connections must be done on the sides of the
bobbin that will end up outside the core's window!
When you solder these connection, make absolutely sure that the enameled wire stripped properly in the
solder bath. It's very frustrating to end up with a nice new transformer, that has one winding that doesn't
conduct, because of a badly done connection! In this photo, the first one and a half loops of the thin
enamelled wire around the pigtail didn't strip, but the other three loops did, so the connection is secure.
One word about safety: Between the primary and secondary windings, and at any place where there might be
high voltage, you need really good insulation. After connecting this pigtail, which is the end of the primary
winding, I had to apply such safe insulation. When doing it with adhesive tape on a bobbin like this, it's done
by winding several layers of that tape, and winding it not only all the way to the side walls, but even a bit up
on these walls, forming a kind of cradle bedding for the secondary winding. You need to get either a perfect
seal between the tape and the side walls, or having so much tape, and the wire moved far enough toward the
middle of the bobbin, that the creepage distance from primary to secondary, around the insulation, is at least
4 millimeters. This is even a legal safety requirement!

24

After winding the two
secondaries with their lots of
taps, and re-assembling the
core, the finished transformer
looked like this, in all the
glory of its dozens of
connecting wires! In this
Heathkit tube tester, the wires
connect directly to the circuit.
In other cases, these wires
would be neatly dressed and
soldered to terminal strips,
which are either embedded in
the outer layers of
insulation in the winding
assembly, or bolted to the
core.


A winding machine with turns counter is a great tool,
but it's not absolutely necessary. Transformers can be
wound with much simpler tools too. And despite
having my machine, sometimes I have to wind
transformers that are larger than what the machine can
handle!
Such was the case in 2008, when I had to build two
transformers for 10kVA each, starting from scratch.
These are the transformers I chose as an example to put
into the spreadsheet before uploading it to this page! In
the sheet, the primary voltage is 230V and the
secondary is 2000V. In truth, one of the transformers
works in this way, while the other is opposite, the
2000V winding serving as primary. In the following
discussion, I will use "primary" to refer to the high
voltage winding, regardless of how the transformer will
be used later.
I built the simple but effective setup shown in the
photo. A big wooden bobbin was made, with
dimensions such that the complete winding package for
the transformer would precisely fit inside. Note that
this bobbin is just the support for winding! It will
be removed before final assembly of the transformer. For this purpose, it's held together by screws, and the
wooden pieces on the inside are designed so that they can be easily removed from the finished coil assembly.
This big wooden bobbin got fitted with a hand crank, and mounted on a steel tube serving as axle, which was
clamped to the workbench.
25

Each wooden piece of the bobbin was separately wrapped in kitchen wrap (saran wrap, Sichtfolie) before
assembling the bobbin. This assures it can be disassembled after varnishing the winding assembly, without
the wood sticking to the coil assembly!
The spools of wire for these big transformers weigh 30 kg for the secondary, and 25kg for the primary. So I
made a simple but sturdy support structure for them and placed it at the work site.

Instead of a complete bobbin with side walls, I chose to
employ only a basic bobbin, made from a sheet of
1.5mm thick Pressspan (I couldn't obtain a material
strong enough with a better temperature rating). The
Pressspan was cut to size, half-depht cuts were made
with the knife at the bending lines, then it was bent
around the wooden bobbin. The overlapping side was
smeared with epoxy glue, and the whole thing was
compressed with wooden boards and clamps like
shown here, to get it nice and straight.

26

Since the Pressspan might carbonize
and become slightly conductive at
high temperature, I wound two layers
of high temperature NMN laminate
over this core, to insure permanent
safe insulation between the winding
and the core. It's temporarlily held in
place with plain office type adhesive
tape, but this tape is later removed
when winding the wire, to keep it
from carbonizing and possibly
causing trouble.
This transformer has sufficiently few
turns and thick wire to wind it in
orderly layers, separated by insulating
sheets. To keep the whole thing from
falling apart, I bound the layers
together with cotton straps, and to
make sure the wire turns don't get
closer to the edge of the bobbin than
5mm or so, I wound cotton ropes at
the edges, used as spacers. These
ropes will be removed later.
The photo shows the assembly, ready to start winding. A double rope makes one turn on each side of the
bobbin, enough spare rope for the successive turns is coiled up on screws driven into the bobbin, and the
cotton straps are installed and coiled up too. The wire end, protected by a piece of high temperature red fiber
spaghetti, is anchored in a hole drilled into the wooden bobbin. Time to start winding!

The first layer has to be wound very
carefully, pushing each winding into
tight contact with the previous one.
When that first layer is complete, be
sure to count the turns to make sure
you actually got as many turns as you
calculated! Otherwise, you need to
compress the winding a bit more, and
then add the remaining turns.
When not using a turns counter, such
as in this case, a good thing to do is to
wind a full layer, then count the turns,
and write it down. Don't forget!
Otherwise it's just too easy to loose
count of the layers! That sounds
stupid, but it's very real. A good way
to count the turns is to lightly draw a
pointy, non-scratching object (plastic
or wood) over the wire turns, and
27

count the clicks as you feel them. Counting by sight tends to be confusing.
Since the wire isn't ever perfectly straight, it's impossible to completely eliminate air between the turns.
That's why you have to consider a safety factor when calculating how many turns you can fit in each layer!
The next layers instead are easier to wind, because the wire will embed itself slightly in the depressions
between two turns of the previous layer, automatically assuming the correct spacing. Note that this produces
either winding layers which alternatingly have one more or one less turn, or else the windings with the same
number of turns in each are alternately offset sideward by one half wire diameter, respective to each other.
You need to consider this when designing the transformer, when you intend to do a neat, tight winding like
this.

After each layer, the wire is pressed flat, using boards
and clamps, on the sides of the bobbin that will end
up inside the core's window. This further helps to
actually make the winding fit. It doesn't matter at all
if the windings bulge out a bit on the other two sides,
because these will end up outside the core, where
space is plentiful.
If you do this, be careful not to damage the wire's
insulation. It's quite tough, but not infinitely so. The
boards should have a smooth and not too hard
surface. I used a thick paper as cushion between the
chipboard and the wire.
After winding and compressing a layer, a single layer
of insulating material (NMN laminate in this
transformer) is wound, overlapping it a bit on one
side that won't be inside the core window. Then the
four ends of the cotton straps are crossed over, pulled
tight, and reattached to the bobbin, and the cotton
ropes are wound another turn, taking care to avoid
tangling the straps with the ropes! The straps will
remain in the transformer, while the ropes will be removed. So the straps should hold just the wire and
insulation sheets, but not the ropes.

28

If you simply let the wire unwind from the spool and wind it
on your transformer, it will get a bow shape instead of lying
flat, and have so much air below it in the center region of
each bobbin side, that later when compressing the layer you
would get excessive bulging on the other sides. To prevent
this, the wire gets a pre-bend in the opposite sense while
winding, by guiding it with the thumb like shown here. I
used a woolen glove for this, because the wire has to be kept
pretty tight, is quite stiff, and it would wear through my skin,
despite being quite smooth!
This photo was shot during the winding of an intermediate
layer of the primary.
The wire and the cotton ropes are placed between the ends of
the insulation sheet, for the transition from one layer to the
next. Said in a different way, the insulation sheets are simply
wound into the coil of wire and rope.

Here you can see how a center tap is made. The winding was arranged in an even number of layers, so that
the center point ends up at the transition from one layer to the next. At this point, the wire was brought out
and back in through a single hole, and protected with spaghetti. The spacing ropes were also brought out and
back in through two holes, to keep them from interfering with the wire tap. Finally, the cotton straps are
threaded so that they will hold both wire ends in place.
After this, the insulation layer can be wound, and then the ropes and the wire for the next layer.

29

After winding the last layer of the
primary, the spacing rope is anchored
in a hole in the bobbin. The end of the
wire is treated just like the beginning
of it, and brought out through a hole
too. I drilled all these holes where
needed, while winding.
Then several layers of insulating
material are wound, since the voltage
between primary and secondary can
reach 10kV or more during a nearby
lightning strike! These transformers
were built to send power over a 600
meter long outdoor transmission line,
in a mountainous and lightning-prone
area.
Since the secondary winding uses
much thicker wire, the spacing rope
needs to be thicker too. I used one strand of thick rope for the secondary, and two strands of thinner rope for
the primary. As a result, both windings are spaced about the same 5mm from the sides.

The secondary is wound just like the
primary. In this case it was a little
more complicate, because I actually
wound two secondaries, each of them
having a wire half the cross section
calculated in the sheet. On top of that,
each of the secondaries is center-
tapped too.This allows me to do
several clever things with my
transformers, and as a bonus lets me
wind with a thinner, more
manageable wire. This is #7, still stiff
enough!
The photo shows the completed
secondary, before applying the final
layers of insulation. You can see the
spacing ropes, and how the cotton
strap will hold the last turn.

30

The spacing ropes are pulled through holes and
secured, several layers of insulation are wound, the
final layer is secured with a loop of masking tape,
and then the cotton straps are pulled tight one last
time, and knotted together using a special
pulling knot, the same sort truckers like to tie down
their cargo. If you don't know how to tie such a knot,
ask a trucker, a sailor or a boy scout.
At this point, the winding assembly is ready to
remove part of the wooden bobbin, and start the
varnish impregnation process.

Removing one of the side walls of the wooden bobbin reveals not only the structure of the inner part of the
wooden bobbin, designed for strength and easy disassembly, but also you can see the spacing ropes partially
falling out, leaving the insulation protruding nicely between the turns of wire. This produces plenty creepage
distance for safe operation!

31

After removing all those
ropes, the winding assembly
looks like this. Note how the
wires are nicely tucked in
between the insulation layers.
While the cotton straps do a
great job holding the end
turns of each layer in place,
preventing them from falling
out and the whole thing
springing open, this assembly
is still far too weak to be used
like this. Considerable forces
act on wires of big
transformers. They need to be
secured very well.

And a further closeup showing how the straps hold the wires, and bend the insulation, always maintaining a
safe creepage distance.
In this photo you can also see the overlap of the Pressspan bobbin, complete with some epoxy glue that
seeped out before setting. Since the bobbin was made 2mm shorter than the core's window length, such
seepage and similar imperfections cause no trouble.
32

Without removing the other remaining parts of the wooden bobbin, I soaked the entire winding assembly in
impregnation varnish. I poured the varnish in from the top, until it flowed out profusely from the bottom.
Then I let the assembly soak, let the excess drip off, and let it dry somewhat on the surface.
But this is an oven-drying varnish, so it needs to be heated, or it will never dry. I do have a thermostatically
controlled oven - but not large enough to fit a winding assembly of this size! So I applied enough DC current
to the primary winding to slowly heat up the whole thing to a temperature high enough to dry the varnish.

Every transformer varnish has a
recommended drying temperature. To
get it right, I occassionally removed
the power supply, and measured the
resistance of the primary winding.
Comparing to its resistance when
cold, and considering the thermal
coefficient of the resistance of
copper, it's easy to calculate the
internal temperature! Then I adjusted
the DC current to keep the
temperature just right.
I removed the other wooden bobbin
side, and the spacing ropes on that
side, when the exposed side of the
assembly seemed to be dry. The
newly exposed side was still fresh, so
drying continued with the assembly
supported only by the wooden parts in the center.
When the varnish stopped smelling, signalling that it was dry, I removed all of the remaining wooden parts.
This was quite easy, thanks to the kitchen wrap, which stayed partially adhered to the Pressspan core. In this
photo you can see that. The brownish color is that of the varnish, while the near black sections inside the
assembly are partially carbonized kitchen wrap. This stuff doesn't endure the temperature the varnish needs
to dry! But no harm was done, except to the wrap, which is of course irrelevant.
At this point, the winding assembly has a monolithic feel, with everything firmly glued together by the
varnish. It could actually be used without further work, but the edges of the thin insulation material are quite
fragile. Any object striking them could bend or even break them, compromising the creepage distance and
thus the safety of the transformer! 2000 volts at 10 kilowatts is not something to take chances with.

33

So I filled these areas with silicone caulk.
This material is available, inexpensive, easy
to work with, tixotropic, permanent, an
excellent insulator, and endures very high
temperature. In short, it's a nearly ideal
material for this task! There is only one point
to watch: Be sure to use the neutral curing
type, not the more common, acid curing one!
The last thing you want to do is releasing
highly corrosive acetic acid into your new
transformer!
The color of the silicone is your (or in this
case, my!) choice.
I didn't do a particularly neat caulking job,
but the point was to get the edges sealed and
protected, not to do a work of art.
The winding assembly is now ready!


The next step is assembling the core. I usually do this in the same way for every transformer: I place the
winding assembly on one side, then I start inserting E laminations from alternating sides. Distributed
throughout the core, I occassionally insert two consecutive E's from the same side. When almost all E's have
been inserted, things tend to get pretty tight, and forcing additional E's in between the last one and
the winding assembly, can lead to damage to the latter!

This is where the E pairs inserted from the same side come into play: They form excellent guides for
inserting a third E between each two neighboring E's, even using a wooden block as a hammer to drive them
in, if necessary! This is the best trick I have found to obtain a nice, tightly compressed lamination stack.
After all E's have been inserted, the
34

I's are slid into the voids. After that, two wooden blocks and a big hammer are used to knock all the
laminations into their correct positions, aligning them with each other, and specially, align the screw holes
of all the laminations!

At this point, the transformer is
functional, and you can hook it up
and try it, if you want. But it will hum
like crazy, because all those loose
steel laminations will magnetically
repel each other at twice the line
frequency rate, and vibrate. The core
needs to be tightly compressed to stop
all that noise. The performance will
also improve when the core is
correctly compressed.
In small transformers this is done
with clamps. But big ones, like this,
use bolts. Often you won't find real
bolts long enough and thin enough,
though! In those cases, buy some
threaded stock, which is sold in great
lengths, cut it to size (a Dremel tool
with a heavy duty cutoff disk does this very nicely!), and use it with washers and nuts on each end.

In large transformers, these bolts sometimes have
enough shorting action to cause significant additional
loss and heating! For this reason it's a good idea
to insulate the bolts from the core. You can use
tubing for that purpose, or like I did here, slide in a
sheet of NMN laminate (or plain paper) rolled into a
tube.

35

Installing the bolts just on the core
leaves you with an unevenly
compressed lamination stack, and
what's more important, with nothing
to mount the transformer! For that
reason usually some angle stock is
used . It distributes the force over
much of the core, and provides
convenient surfaces to drill mounting
holes into.
Some transformers use formed caps
instead, or complete steel frames.
Note the insulation sheets under the
washers of the bolts! Depending on
the specific transformer, they might
make any difference, from barely
noticeable to dramatic!
At this point, the transformer is truly
ready for trying. If it still hums, you
can try tightening the bolts even
further, and inserting wooden or plastic wedge pairs between the winding assembly and the core center leg,
to compress the latter. If even this fails, which is often the case, then you have no better option than once
again getting your varnish can, and soak the core in varnish! You can loosen the bolts, let varnish flow into
every space, then tighten the bolts again and warm up the whole transformer by applying a suitable amount
of DC for at least a full day. I have yet to see a transformer that still hummed after that treatment!

Depending on the application,
different things can be done with the
connections. If they are made of
flexible wire, they might be directly
wired into the circuit. In small
transformers, often plastic bobbins
with pins are used, and the windings
are connected to these pins, like
shown here, and then the whole
transformer is soldered to a printed
circuit board. The example shown
here is a current sensing transformer,
which also has a one-turn, high
current winding, which is brought out
separately from the bobbin pins.

36

Often terminal strips are attached to
the winding assembly, and the
windings are connected there. But in
larger transformers, the most common
practice is bolting terminal blocks to
the transformer, and connecting the
wires there. I did this with my large
transformers, leading to the final
product shown in the very first photo
of this web page. That system is more
solid and reliable then most others,
and allows repeated solderless
disconnection and reconnection,
which I need a few times per year to
reconfigure my transformers for
different conditions of usage.

Some people wonder whether they should try winding their own transformers, or if they should instead shell
out big $$$ to have the job done by a professional winding shop. I can only tell you one thing: If you had the
patience to read this long web page from start to end, then you most certainly also have the patience
required for winding your own transformers!!!





Transformers and coils
There are many electronicians, both hobby and professional, who are at war with electromagnetism. Whenever they
need to design a coil or a transformer, an abyss of desperation opens in front of these poor people. The worst thing is
that usually these poor victims are not really at fault, since the authors of electronic textbooks seem to have struck a
plot to explain these things in such a messy way that nobody can really understand them! Or maybe these authors
themselves didn't have a clue about the matter?
Well, internet to the rescue! I will explain the basics, in simple, understandable terms, taking some freedom
in regard to standard physics notation. Here you will find most information required to design
electromagnetic parts for electronic use.

The units
There is one request I have: When you enter this page, you have to leave out all obsolete, absurd units of which most
textbooks and catalogs are full: Most notably, inches, gauss and oersted. Delete these three words from your
vocabulary! They have no place here. They are the principal culprits in confusing people attempting magnetic design
to the point of driving them crazy. Now, that we have gotten rid of them, we can start.
37

The first unit we will use is the weber, written as Wb. This is the official unit for magnetic flux. If you take a
loop of wire, and apply 1V to this wire during 1s, then the magnetic flux inside this loop will have changed
by 1Wb. Note that this is true regardless of size or shape of the loop, and regardless of the matter that's inside
the loop! Officially this definition of the weber is written as:
Wb = V * s
but I prefer a modification of this equation that is more practical, taking into account the number of turns of a
multiturn coil. So, this is one of our basic truths:
(1) Wb = V * s / t
meaning that the change of magnetic flux (weber) is the tension (volt) multiplied by time (second) divided by
the amount of turns. This is one of the most powerful and useful formulas we have.
If we squeeze a certain amount of magnetic flux through a certain area, then we can speak of magnetic flux
density. The unit for this is the tesla, written as T. Its definition is simple and obvious:
(2) T = Wb / m
2

Note that talking about square meters in electronics may look a little removed from practice, as we use parts
that have cross-sections more in the square centimeter range. But please, believe me that accepting such
"impractical" things is very much more practical than using dozens of different conversion constants! Using
the basic units has the large advantage that absolutely no conversion constant are needed.

The basic characteristic of any coil is inductance. It is measured in henry, written as H, and its definition is:
(3) H = V * s / A
or, in words, one henry is the amount of inductance that will allow the current to rise by one ampere when
applying one volt during one second. This equation is also very useful for our purposes. And now we can
start to play! We can join equations 1 and 3 to make the following:
H = Wb * t / A
Such algebraic conversions are always true too, and can give us the means to calculate some value we did not
know how to determine!
But now let's go to practical things.

Designing power line transformers
While almost any electronician knows that the voltage ratio of a transformer depends on the turns ratio, the question
that arises for most novices is: "How many turns-per-volt must I use???"
It's very simple. You have an iron core, and you want to wind it. First, measure the section of iron through
which the magnetic flux will flow. Say, the central leg of the transformer core measures 2cm wide, and the
entire stack of laminations, well compressed, measures 3 cm. This gives you 6cm
2
, or 0.0006m
2
of cross-
sectional area.
38

Now you need to decide how much flux density you will put through your iron. At low frequencies, like
those of power line applications, the limiting factor is saturation of the core. The most humble transformer
iron saturates at 1T, but more typical values are 1.2 or 1.3T, and good grain-aligned material may work up to
1.6 or even 1.7T. If you really don't know what material you have, better stay at 1T, so you are on the safe
side.
For this example, let's assume your iron is good enough to put 1.2T through it.
Applying formula 2, the maximum magnetic flux must be limited to 0.00072Wb. But now, before going
ahead, stop a moment, and think!!! Iron can be magnetized in both senses. So, the total change of magnetic
flux, from maximum negative to maximum positive, can be 0.00144Wb! This said, let's go to formula 1, and
calculate the turns. Let's assume that we are talking about Chile, or any other country where the power lines
work at 220V, 50Hz:
0.00144Wb = 220V * 0.01s / t
which is 220V * 0.01s / 0.00144Wb = 1528 turns (for the 220V primary)
Simple, isn't it? In fact, the above is too simple to be true. There is another factor which I skipped... The
above would be true if the power line was carrying a 220V square wave!! But in truth, it carries a sine wave,
that has 220V RMS, while the average is slightly different. And magnetic flux buildup depends on the
average, not on the RMS! So, we have to introduce a small correction factor, which is derived via calculus
from the sine function. Instead of bothering with the exact math, I suggest you simple take my cookbook
recipe: 11% in our favor! So, we can get away with just 1376 turns!
Where does the 0.01s come from, you may ask? Think again. The change from maximum negative to
maximum positive flux happens during one semicycle. And at 50Hz, a semicycle lasts 0.01s.
We can well reduce all this to a simple, universal formula, valid for calculating turns number for all sine
wave transformers and coils:
(4) t = V
(RMS)
/ 4.44 / m
2
/ Hz / T
The 4.44 factor is not a conversion constant, but a factor composed by 2 * 2 * 1.11. One of the "2" is for the
fact that the total magnetic swing is twice the single-sided one (so we can give the saturation limit as input
data), the other "2" reflects the two semicycles in each full cycle (Hertz refers to full cycles), and the "1.11"
compensates for the sine wave ratio of RMS to average voltage.

Power
Another question that usually arises is how much power a transformer of a given size can deliver. Let's analyze this.
The magnetic flux in the core depends on the voltage applied to the transformer's windings, the frequency,
but NOT on the current the transformer is delivering! Oh, well, yes, there is a small dependence caused by
real-world effects: As you draw more current, the resistance of the wires causes a voltage drop, slightly
reducing the effective AC voltage applied to the windings, and thus reducing the magnetic flux by the
proportional amount. But my point is that the transformer core is not involved in limiting the power
delivered. This limit comes from the windings, and has two faces: One is the voltage drop, which is
proportional to load current and will at some point be so much that the voltage is no longer sufficient for
your load, and the other is heating. As the load current increases, the power dissipation in the windings
39

increases to the square, and if you draw enough power from the transformer for enough time, you will burn it
up.
A purist might argue that there is also another factor limiting the power output of a transformer: Imperfect
coupling between the primary and secondary, expressed as leakage inductance (uncoupled series inductance)
in both the primary and secondary windings. In practical power line transformers, though, this is usually not
the limiting factor, and has relatively small effect, except when a transformer is specially designed to limit
the coupling, usually by adding magnetic shunts (ballast transformers, microwave oven transformers,
welding transformers, etc).
All the above makes clear that a transformer's power handling ability depends as much on its magnetic cross
section (because more section requires fewer turns, allowing thicker wire to be used), and on its window size,
that is, the cross-sectional area of the space where the windings go. But there is no linear formula relating the
product of the two areas to power! As the transformer grows larger, the path length for heat evacuation
becomes longer, and so the increase in power handling ability is less than proportional to the area product.
Given all this fuzziness, I will abstain from providing guesstimations, but rather invite you to do the real
calculation: For a given iron core, calculate the amount of turns needed, then consider the space available for
them, calculate the wire size, and from the copper volume resistivity of 1.75 * 10
-8
Ohm meter, calculate the
total winding resistance. Now it may help to know that for small transformers a maximum loss of 10% (5%
in each winding) is usually considered acceptable. This should allow you to calculate the power you can
safely extract from a given transformer, if you are geometrically knowledgeable enough to do these
calculations! It requires nothing more than the math you learned in school, by about the fifth grade.
Hey, I can hear your cry!!! OK, OK, to make things clearer, I will give you an example!!! Let's suppose that
the core mentioned above, with 6cm
2
cross-sectional area, has 10cm
2
available for windings, and that the
average turn length is 20cm. We will distribute the room evenly among the primary and secondary windings.
And we will assume that 40% of the space gets actually used for copper, the rest being insulation, air, and
lost space. This is approximately correct for practical work, and gives us 2cm
2
of copper area for each
winding. At 1376 turns, the primary will have 0.14mm
2
for each turn, and the total wire length will be 275m.
The number given for copper resistivity refers to the resistance across a block of copper measuring 1m on
each side. But our winding has a cross section of only 0.00000014m
2
, and a length of 275m. So, its resistance
will be:
0.00000001.75Ohm m * 275m / 0.00000014m
2
= 34 Ohm.
We will allow 5% loss in each winding. At 220V, 5% is 11V. Now simply apply Ohm's law, and the
maximum primary current comes out as 0.32A, which multiplied by those 220V, gives a gross maximum
input power of 70W for this transformer.
Cool, huh? :-)
Note that the magnetizing current is not being considered here. You may say that even if it is only 10 or 20%
of the maximum current, it should be considered! If you say that, you are quite wrong. The magnetizing
current is 90 degrees out of phase with the reflected load current, and so, even if it is 20% of the load, the
magnitude of the vectorial sum of the two is very close to that of the load current alone. It doesn't pay to care
for the small difference!

40

Switching power supply transformers
The above chapter can be applied almost in full to the design of higher frequency transformers for switching power
supplies. There are just some practical differences which I will mention now.
For one, at frequencies much above a few hundred Hertz, saturation is no longer the limiting factor when
choosing the maximum flux density. The reason is that the losses in the magnetic material become so large
that the flux density has to be kept much below saturation simply in order to keep the losses at an acceptable
level!
You really need the core manufacturer's data, to determine how much flux density is acceptable. To give you
a rough idea, first consider that at these frequencies almost always ferrite material is used. And ferrite
saturates at about 0.3 to 0.4T, so this will be the absolute limit. For typical power-type ferrite material, at
25kHz you will have to keep the flux density below 0.15T, and at 100kHz below 0.05T. But a lot depends on
core size too. A larger core will have to run at lower flux density in order to avoid overheating.
Normally switching power supply transformers operate at square wave, which means that you must eliminate
the 11% "sine factor" from the equations. And then, many of these things don't apply full AC to the
transformer, but instead use only one side of the magnetic excursion! In short, you will have to understand
the circuit you are designing for, and then use a little gray matter to determine which of the "2" factors apply,
if any. For all the rest, the calculation is the same as for power line transformers.
Don't be surprised if you end up with very few turns. In fact, it's quite common to have just 10 or 20 turns for
a 300V primary winding in a large switching power supply!

Broadband RF transformers
Maybe you have seen those ferrite transformers often used at the output of solid state RF power amplifiers. They
look like two ferrite tubes, side-by-side, with two copper tubes inserted in them, forming a 1-turn primary winding.
Through these copper tubes a few turns of insulated wire are looped, forming the secondary. Let me use such a
transformer as a further design example.
Our hypothetical case will be a 100W push-pull amplifier for 1.8 to 30MHz, powered from 13.8V, like there
are millions in daily use by radio amateurs and all kinds of commercial services.
Each transistor can pull its side of the transformer primary quite close to ground, but not fully, because of its
saturation voltage. RF transistors typically saturate at about 1V, so it's reasonable to assume that the
transistors can swing over a 12.8V range, giving a 25.6V peak value over the primary, or about 18V
RMS
.
On the other hand, the secondary is expected to deliver the RF power to a 50 Ohm load, and 100W over 50
Ohm equates to 70.7V. Thus, we need a voltage (and turns) ratio of about 3.9. As we have a one-turn
primary, we can only implement integer ratios, and thus we will choose a 4-turn secondary. The effect of this
is that at 100W the transistors will be running at 17.67V
RMS
, or 25V peak between them. Thus they will
swing over 12.5V from the power supply voltage, giving a headroom of 1.3V for saturation and power
supply voltage drop. So far, so good.
At 1.8MHz, our lowest frequency, a typical ferrite can be safely driven to about 0.012T. We have a nice,
pure sine wave, so let's use equation 4:
1 turn = 17.7V / 4.44 / m
2
/ 1800000Hz / 0.012T
41

or, rearranging, m
2
= 17.7V / 4.44 / 1 turn / 1800000Hz / 0.012T = 0.00018
So, we need a total core cross-section of 1.8cm
2
. A smaller core would overheat when running at full power
for a long time, while a larger core would be more expensive, but bring us an advantage in terms of spectral
purity, as lower flux density means lower distortion too! But for this exercise let's keep the 1.8cm
2
figure.
We still have work to do. We could use long and thin ferrite tubes, or short and fat ones. And we can choose
among several different ferrite types! To narrow our choices, let's see the inductance requirements. The rule
is that the transformer should exhibit a self-inductance that's high enough to be of little effect when placed in
parallel with the load resistance. An old rule of thumb is to have 10 times as much inductive reactance as
load resistance. You can choose if you want to calculate using 4 turns and 50Ohm (secondary side), or 1 turn
and 3.1Ohm (primary side). The result will be the same. I will pick the primary side.
The inductive reactance is calculated as
X
L
(ohm) = 2 * PI * Hz * H
So we end up with H = 31 ohm / 2 / 3.14 / 1800000Hz = 0.0000027
We need 2.7H primary inductance to satisfy the rule of thumb for 10 times as much transformer reactance
as load resistance. Now we must go to the manufacturer's data tables and pick suitable cores. For this
example, I will use the Amidon catalog.
Let's give the very common FT-50-43 a try. This toroid has 0.133cm
2
cross sectional area. Two stacks of 7
each would fill the request for flux density. Its A
L
is 0.52H per turns squared, so 14 cores with 1 turn would
give 7.3H, several times as much as needed. Since broadband amplifiers tend to oscillate at low
frequencies, where the transistors have a very high gain, it may not be a good idea to provide more low-
frequency performance than necessary! Let's try another core type.
The 43 material has a permeability of 850. A core of the same size, but with a permeability of only about 330
would be nice. But Amidon sells no core in this size in any permeability close to that... Hey, you can't get
free rides in every case! :-) The next lower permeability available from Amidon in sizes that are usable for
this project is 125, and that's too low... So we will have to stick to the 43 material. Let's see what we can do.
There is the FT-82-43, made of the same material. It's much thicker, has 0.25cm
2
of cross-sectional area, and
an A
L
value quite close to that of the other core: 0.55H per turns squared. Two stacks of four each of these
would give more than enough cross-sectional area, at 4.4H. This could be a usable solution, also providing
more space for the windings.
On the higher frequencies the flux density will be lower, always staying below the limit of the material at
each given frequency. The ratio of inductive reactance to load resistance will improve as the frequency goes
up, but at the highest frequencies the parasitic capacitances may take over, so it's a good idea to consider this
in the design of the amplifier.

Energy storage in magnetic cores
Do you know how much energy a coil is storing? It's defined by the same old formula that shows up so often in good
old Newtonian physics: "a" is equal to one-half "b" times "c" squared!
(5) J = 1/2 * H * A
2

42

Energy is expressed in Joule (J). H is the inductance in henry, while A refers to the amount of ampere
magnetizing the core. In the case of a transformer, this current must be calculated as the net current after
subtracting any opposing currents, such as primary and secondary ones, giving proper care to the number of
turns. In short, this current is the magnetizing current.

In most typical transformer applications this storage of energy is not really desired, but is an unavoidable
side effect. But there are applications that make good use of this storage! One very important example is the
fly-back topology of switching power supplies. Basically, such power supplies make the transformer store
energy coming from the primary circuit, and then dump this energy into the secondary circuit - often at a
voltage that is unrelated to turns ratio! As primary and secondary currents do not flow at the same time, it is
no longer true that the voltage ratio must be equal to the turns ratio!
Let's suppose that we will make a switching power supply using this principle. We want 13.8V output, while
the input should be 110 or 220Vac. The logical approach in this case is to use a primary rectifier that can be
configured as bridge for 220V, or as voltage doubler for 110V, so we will end up with about 300Vdc in
either case, and the rest of the switching power supply will be the same, regardless of line voltage.
Let's further suppose that we have a ferrite core of 2cm
2
cross-sectional area, 12cm path length, sporting an
initial permeability of 2000, saturating at 0.35T, and we want to run it at 100kHz. For designing this unit, we
need one more piece of information: The A
L
value, which relates turns number to inductance. If we don't get
this value from the core's manufacturer, we can calculate it from physical dimensions and ferrite properties,
or we can wind a test coil and measure it, but it's definitely easier to get it from the catalog! Let's suppose
that for our core, this value is 6H per turns squared. Which means that 1 turn will give 6H, 10 turns will
give 600H, and so on.
The values assumed above are quite typical for practical cases.
In order to reduce voltage strain on the primary switching transistor, we will assign 30% of the time to
transformer charging, and 60% to discharging. This allows to perform the discharge at half the voltage level,
so the switching transistor will only see 450V instead of 600V. Also this reduces peak current on the
secondary rectifier, while increasing requirements for primary current handling and secondary voltage
handling, both of which are not a problem in our supposed situation.
The 10% of time remaining is there to account for time lost during switching, dead time of the controller, etc.
At 100kHz, our charge time will be 3s, and the discharge time will be 6s. A look into the ferrite's data
table lets us know that at 100kHz, maximum single-sided flux density should be limited to 0.1T. Applying
formulas 2 and 1, we quickly end up with the following:
t = 300V * 0.000003s / 0.1T / 0.0002m
2
= 45
So, 45 turns will load up this core to 0.1T in 3s, when applying 300V. Nice and easy. On the secondary side
we need 13.8V, plus about 1V for diode drop, giving about 15V. We can use the same formula again,
inserting the different values for voltage and time:
t = 15V * 0.000006s / 0.1T / 0.00002m
2
= 4.5
Did you like this? The turns ratio is 10:1, while the voltage ratio is 20:1, because the time ratio is 1:2!
Feel free to use either 4 or 5 turns instead of the fractional number. This will simply cause a slight alteration
in the charge-discharge times.
43

Now, how much power can we extract from this switching power supply? No, don't run for the things I wrote
above about line power transformers! Here we have two limits: One is the limit of the transformer proper, as
of heat dissipation, but then there is a functional limit too, which is more important: Our switching power
supply works by energy storage, and for each cycle, only a very specific amount of energy is stored, strictly
limiting the power that can be converted!
According to the A
L
value assumed above, our 45-turn primary winding will have an inductance of about
12mH. Using the definition of inductance, we can calculate the peak current that will flow at the end of the
charging cycle:
A = 300V * 0.000003s / 0.012 H = 0.075
Only 75mA! Doesn't look like much... Let's calculate the stored energy for each cycle:
J = 1/2 * 0.012H * [0.075A]
2
= 0.000034
You can also calculate this energy from another approach: As the current will ramp up linearly from zero to
75mA, its average will be 37.5mA. At 300V, and 3s, that is:
J = 300V * 0.0375A * 0.000003s = 0.000034
Isn't it nice when things agree...? :-)
Considering that at 100kHz we have 100000 of these tiny specks of energy per second, and that Joule is
simply Watt times second, we end up with a sad power level of only 3.4W for our glamorous switching
power supply! Seems like a mighty bad use for a core of that size, doesn't it? That core is rated at "250W
typical" by the manufacturer!!!
So, we must see how to increase the amount of energy stored in the core. If we increase the inductance, then
the current will drop, and the current carries squared weight! No good idea. It's better to reduce the
inductance, so the current increases. Given that the stored energy depends linearly on the inductance, and on
the square of the current, it's obvious that as we reduce the inductance, the stored energy increases in
proportion.
How do we do that? We cannot simply reduce the number of turns! This would bring us into the claws of
equation 1, and increase the flux density to a level that is far higher than what the ferrite can tolerate! Do you
realize the problem? We need to reduce the inductance, while keeping the number of turns in order to
preserve flux density!
There is a very simple tool for doing this: Air! Simply force the magnetic flux to jump through an air gap, by
separating the two core halves by a small distance. The effect of this air gap is reducing the effective
permeability of the core, thus reducing its A
L
value, without affecting other parameters. Let's see what
happens if we add an air gap of just 1 mm total, which is done by separating the core halves by 0.5mm:
The magnetic flux will now travel through 120mm of ferrite having a permeability of 2000, and 1mm of air
having a permeability of unity. 2000mm of ferrite would have as much magnetic reluctance as our single mm
of air! So the air gap has added 16.667 times as much reluctance as the ferrite core originally had. Plus the
original reluctance, we have now 17.667 times as much reluctance as before.Which means that our core now
has an effective permeability of only 120, instead of the former 2000! Which in turn means that the new A
L
value is about 0.36H per turns squared, and that our 45-turn primary winding now has only 0.00072H
inductance. This in turn means that it will charge up to 1.25A, and store 0.00056J per cycle, resulting in an
44

output power of 56W for our switching power supply, which looks very much better than the meager 3.4W
obtained without an air gap! All this is while maintaining exactly the same magnetic flux in the core!
Did you ever think that a 1mm thick layer of air can be so tremendously important???
Your next question should be if there is a limit to air-gapping. Sure, there are two limits! One is simply that
as you increase energy storage and power transferred, you are also increasing the losses in the windings. At
some point you will reach the thermal limit of the copper loss, just as you did with the humble, simple line-
frequency transformer. This amount of air-gapping is often chosen as optimum trade-off by designers. But
then there is another problem: As the effective permeability drops, so does the coupling between the
windings. The transformer will start producing a strong stray field, and will start exhibiting large uncoupled
inductances which can lead to the destruction of power transistors and diodes, and in most cases will require
additional snubbering. So, the designer must in some cases settle for less air-gapping than what the thermal
limits of the windings would suggest. In any case, these coupling problems can be minimized by proper
construction of the transformer: The primary and secondary windings can be interspersed, bifiliar winding is
sometimes possible, and it's often a good idea to add a thick copper foil around the entire, completed
transformer, core and all, forming a shorted turn. Such a shorted turn will force the total external flux to zero,
making sure that the flux through the coil assembly will be equal to that outside it (in the side legs of the
core), thus improving the coupling.
In many cases it's better to use a core material of lower permeability, like powdered iron. Our transformer
above would behave almost the same if we built it on a core made of 120-permeability material, with no air
gap, but it would exhibit better coupling and produce less stray. On the other hand, the large advantage of the
air-gap technique is that the designer can decide exactly how much effective permeability he wants, without
having to order a new core!
Lastly in this air gapping business, you should be aware that air isn't always air! Very often cores are
"airgapped" by inserting sheets of paper, plastic, or other non-magnetic, insulating materials between the
core halves. The effect is the same.

DC chokes
One of the worst things I have seen in electronics textbooks is giving different formulas for AC and DC behavior of
coils. This is complete, utter nonsense!!! There is no fundamental difference between AC and DC. At any given instant
of an AC waveform, you have pure DC, and in a "DC" application, you have an AC component too, at least during
power-up and power-down. So, we can, and should, use the same design approach for DC chokes.
Let's see how this applies to practice. One common task is designing a choke that has to exhibit a certain
inductance, while being able of conducting a certain amount of current before saturating. Note that for DC
applications the limit of flux density is always saturation related. Do you remember what I wrote further up?
At high frequencies, the limit is given by core loss, and at low frequencies, by saturation. And DC is simply a
very, very low frequency... :-)
Let's suppose that we need to make a choke of 100H, that can take at least 10A of current before saturating.
Let's suppose that we have a toroidal core for it, made from powdered iron, that has a cross section of 1cm
2
and a path length of 10cm. The permeability is 75, and saturation begins at 0.5T. The A
L
value of this core is
0.08H per turns squared.
From the A
L
value alone, we can easily calculate that 35 turns would be required. Now, how can we calculate
the flux density? After all, there is no voltage applied across the winding!
45

Think again! There MUST have been a voltage applied, in order to get the current going! If we had applied
1V, it would have taken 1ms to ramp the current up to 10A in this 100H coil. So, now we have a voltage
and a time to insert in our universal formula number 1! The flux would be
Wb = 1V * 0.001s / 35t = 0.0000286
resulting in a flux density of 0.286T in our 0.0001m
2
core! Bingo! This choke could actually conduct almost
twice the required current before saturating.
A cost-conscious designer would run the same exercise with the next smaller core available, which probably
would be just large enough to allow making the required 100H, 10A choke.

Picking cores
There are countless shapes and sizes of magnetic cores available, and each of them comes in many different
materials! It's a good idea to know, at least in principle, what's available.
Materials:
The most ancient material is transformer iron, known also as silicon steel. It usually comes in thin sheets,
which need to be insulated from each other in order to avoid eddy losses. Only in pure DC applications is it
acceptable to use solid steel or uninsulated sheets.
Transformer iron can usually be counted upon as to be able to take at least 1T before saturating, while 1.2T is
OK for most, 1.5 for some, and 1.7T is possible with the best grades. The permeability of this material is
usually around 2000 to 5000. Those iron alloys that allow higher flux density have the lower permeabilities.
The losses are so high that for frequencies above 100 Hz or so they become the limiting factor, rather than
saturation.
Iron is also used as dust, mixed with resin and molded into magnetic cores. The permeability depends on the
iron contents in the mix. Given that even a small amount of resin has much more reluctance than a large
amount of iron, the permeability is usually quite low: Values from 2 to 100 are typical. For the higher
permeabilities the iron grain size and shape becomes very important, since very dense grain packing is
needed to achieve them.
Saturation effects start earlier than for massive iron, because the flux tends to be forced out of the iron
particles. 0.5T may be a typical value, but in any case the saturation is very "soft", there is no well-defined
point at which the core would clearly enter saturation. The losses are low enough to make the lower
permeability versions of this material usable well into the RF range.
These dusted cores also are made with other metal alloys, like Permalloy, in some cases offering attractive
features. And transformer laminations are also available in other materials than silicon steel, making them
usable throughout the audio range and even much higher, if the laminations are thin enough, but such
laminations aren't easy to obtain.
Ferrites are the most versatile materials available for magnetic cores. While they saturate at lower values,
typically 0.3T, they exists in a tremendous range of permeabilities: It's not hard to find ferrites with
permeabilities as low as 20, or as high as 15000! The inexperienced user cannot tell a ferrite core's magnetic
characteristics simply from looking at it. Even if two ferrite cores look exactly alike, one can be 1000 times
different from the other! So, make sure you KNOW what material you have, before starting your
calculations!
46

In any case, the most common ferrites fall into two categories: Power ferrites, used for switching power
supplies, fly-backs, TV yokes, etc, that have a permeability of around 2000 and low loss at frequencies of
about 20 to 100kHz; and RF-type ferrite, with permeabilities of around 100 to 1000, and a loss characteristic
that makes them useful at least through 30MHz. But there are many ferrite types that can work at much
higher frequencies and have lower permeability. The permeability values above 2000 are reserved to rather
special-purpose cores intended for wideband transformers, transducers, or noise absorption.
As for shapes, I will mention just a few:
Toroids: They are simple, cheap, easy to design for, have low dispersion, good self-shielding, but cannot be
airgapped by the user, and winding 10000 turns on a toroid is not something I would like to do.
E-cores: Very convenient for most applications, but the sharp edges cause more dispersion.
U-cores: Slightly cheaper and slightly less efficient (due to longer path length) than E-cores.
Pot cores: Join the convenience of E-cores with the good self shielding of toroids (in fact, it's even better!),
but they are more expensive. Some are adjustable.
RM cores: Sort of a space-saving cross breed between a pot core and an E core.
Solenoids (bars): Usable for chokes. They have really large airgaps! :-) For that very reason, they are not
suited to transformer use. The coupling would be too poor.
Bobbins: These are basically solenoids with end plates, so their airgaps are slightly smaller than those of
solenoids. They make good chokes and are convenient to wind.
E-I laminations: It's pretty much the only shape you can buy transformer iron in!
I suggest you order or download a few catalogs from magnetic material manufacturers and distributors, so
you can learn about the other 992 shapes available... May I suggest Amidon Associates, Ferroxcube,
Ferrinox (Thomson Composants), SiFerrit (Siemens), TDK, Philips, to name just a few. I have worked
mostly with Amidon, Ferrinox, and junk-box cores. The best performances seem to come from certain
Japanese ferrites.

This little treatise of electromagnetism cannot be considered complete, of course, but I prefer to keep it
centered on those aspects that are most important to the practical designer/hobbyist. I have completely
skipped those aspects of electromagnetism which in my view are less relevant for practical work. I have also
skipped a lot of practical hints, which are useful, but would have turned this article too much into a
cookbook. If you have any question, don't hesitate to ask. My e-mail address is on the front page. If enough
questions show up, I will add a "questions&answers" section to this page!





47