Magnetic Circuits

Magnets can be air cored or iron cored. The magnetic core plays a vital role in raising the level of
magnetic flux produced by a given current flow in the windings. That is to say, it increases the
inductance.

Problems with ferromagnetic core

Unfortunately, the increase in flux provided by ferromagnetic materials is never a simple linear
factor. At first, as the field strength is raised, the flux increase is a modest one. At higher currents
the increase is more rapid, but at yet higher currents the flux increase slows again. Finally,
saturation is reached and any further current increase generates but a minuscule increase in flux.

These changes in inductance with current are a nuisance at best. In high power transformers,
however, saturation must be avoided at all costs because the ensuing drop in inductance leads to
a sharp increase in magnetizing current, further core overload and a 'runaway' situation that often
ends with smoke and flames Also, ferromagnetic materials all suffer from temperature
dependance. In resonant circuits, for example, you normally prefer that the inductance remains
constant with temperature and not detune the circuit as the environment changes.

In a high-Q tuned circuit you almost certainly want to run the core well below its saturation level
- not through fear of flames but because you cannot afford to loose power to core hysteresis
(which increases rapidly with B).

These three problems are usually alleviated by using a gapped core.

What is a gapped core?

 In an ordinary inductor core the magnetic flux lines are guided all the
way round the windings by way of the ferromagnetic material in the core. However, in a gapped
core a small section of the flux path is replaced by a non-magnetic medium - such as air. The
term 'air gapped core' is still used even if the gap is filled not by air but by nylon or some other
material immune to saturation.

Some core types may be purchased with a gap pre-fabricated, such as the
RM7 shown above; but the photo at right shows a 100 mH EC70 core in which the gap has been
created by inserting sheets of cardboard between the core halves. Adhesive tape can hold the
core together adequately because clamping force requirements are less than for an un gapped
core.

How does the gap help?

Your first thought may be that if the core is overloaded then what you need is a larger one, and
that the last thing you would want to do is remove part of it! Well, magnetics is a curious
business.

Consider an analogy: you have a battery with an EMF of 12 volts and a two ohm load. However,
you don't want six amps to flow round this circuit - that's too much; you only want four amps.
What's the simplest way to achieve that? The answer is a no-brainer: you insert a one ohm
resistor between the battery and the load. You now have a total of three ohms in the circuit and
so all that a 12 volt battery can deliver into it is four amps. Job done.

The problem which an air gap solves in a core is the excessive flux produced by a high level of
current in the windings. Transferring the previous analogy, suppose that your windings produce
an MMF of 12 ampere turns and that your magnetic core (your 'load') has a reluctance of two
ampere-turns per weber. However, a flux of six webers is too much; you can only tolerate four
webers. To solve the problem you just add in a 'resistor' (call it a 'reluctor') of one ampere-turn
per weber. Now you have a total reluctance of three ampere-turns per weber 'in circuit' so that
now four webers is all that the MMF can push round the core. Job done.

You may have guessed that the role of this extra reluctor is taken on by the air gap. Its task is to
increase the reluctance of the core so that less flux flows for any given level of MMF. The virtue
of the air gap is that, because it is free from ferromagnetic material, it does not suffer any change
of reluctance with flux level. Its reluctance depends only on its length, lg, and cross sectional
area, Ae; and both of those parameters you can make very stable.

In summary, the gap reluctance is -

Rg = lg / (μ0 Ae)
Equation AGI

while the ferrite reluctance is -

Rf = le / (μ0 μr Ae)
Equation AGJ
These results combine in series to give the reluctance of the inductor as a whole -

RL = ( le / μr + lg) / ( μ0 Ae)
Equation AGK

From which you see that a large value for μr leaves the air gap as the only significant contributor
to the overall reluctance.

If you use a spacer then lg will be twice the spacer thickness because flux must cross the spacer
twice in a complete circuit of the core.

It's that easy?

Almost. Remember why you use a core in the first place: to raise the inductance of your
windings. When you add an air gap to increase the reluctance of the core then it is almost as if
you have decreased its permeability, and thereby lowered the inductance of a winding on it.
Indeed, when you buy a core with a pre-fabricated gap then the manufacturer may specify what
is called the effective permeability of the core, μe. It may be 20 times less than the 'raw' value for
mu of the ferrite. However, you need only substitute this value of mu for that of the actual
material and then calculate your inductor design (using μe) in the same way you would normally.

An analysis of the core and gap as a two component series circuit gives

μe = μr / (1 + ( μr × lg / le))

Adding the gap therefore spoils the inductance. However the inductance depends also on the
number of turns, so what you normally need to do is wind on more turns. At this point you may
raise the following objection. With more turns the magneto-motive force will increase which will
lead to higher flux once more, and you are hit up again by core saturation. This argument is only
partially correct because inductance increases according to the square of the number of turns
whereas mmf is linearly proportional to N. Eventually, therefore, your additional turns strategy
will always win through - assuming you don't run out of space for the winding.

Looked at another way, what you are doing is to shift the performance burden away from the
core and onto the copper wire - after all, an air cored inductor can never saturate. Unfortunately,
all those extra turns take up extra space in the winding aperture, and they can't be made any
thinner because they still need to carry the same current. So you may find that you need a larger
inductor anyway. There's no free lunch :-(

Gap energy
Because air has a permeability in the order of 103 times lower than ferrite it so happens that the
gap often accounts for almost all the reluctance seen by the magnetic field, leading to an
interesting property of practical gapped cores: the gap contains nearly all of the field energy.

You may think it strange that energy should prefer to reside in thin air rather than the core
material. If so, consider again the current flow analogy developed to explain the method of
soothing an overloaded core. Energy dissipation (ie power) in a resistor is given by P = R × I2.
The current flow, I, is the same all round the circuit: the battery, the load and the added ballast
resistor all carry it. Clearly, then, the power dissipation in the two ohm resistor is twice what it is
in the one ohm ballast.

Although the reluctor analogues dissipate no power they do store it -

W = ½ R m × Φ2

Equation AGB

To a reasonable approximation, you have the same flux, Φ, 'flowing' round all parts of the
inductor: the windings, the core and the air gap. This means that the greatest energy storage takes
place in the larger 'reluctor'. For all but the shortest of gaps it is the air which has more
reluctance.

The reason this is worth stressing is that you may usually equate the energy in the gap with the
total energy possessed by the inductor -

W L = ½ L × I2

Equation AGC

Also, you should not be surprised to learn, there is, for every core type, a limit to the amount of
energy that may be stored. In the next section you will discover that there is an optimum air gap
length, lO, which will deliver that energy. Table AGT shows lO for a range of core types, under
the following conditions -

 The gap length is an overall one, ie twice your spacer thickness.

 Fringeing and leakage flux is totally ignored. For some cores this gives significant error at lO and
longer.

 Standard coil formers are used.

 Their entire winding aperture carries an average of 2 A mm-2.

  The peak flux density quoted is low enough only to avoid approach to saturation with readily
available ferrite grades over a realistic ambient temperature range. Flux limits which avoid core
heat over-production may need to be set lower.

Table AGT: Energy capacities for various cores

Basic Ae le lO Al Fm Bpk WL
Size 2
type / mm / mm / mm / nH / A-t / T / mJ

ER 9.5/2.5/5 8.47 14.2 0.025 312 5.88 0.3 0.0103

RM 7 44.1 30 0.18 274 43.5 0.3 0.319

RM 10 96.6 44.6 0.35 318 84.0 0.3 1.32

ETD 44 173 103 1.83 115 437 0.3 11.8

EC 70 279 144 3.88 87 927 0.3 38.8

E 80/38/20 392 184 7.69 63 1840 0.3 110

U+U 93 840 354 25.7 38 6130 0.3 772

U+U 141 1350 377 26.1 72 6240 0.3 1260

Gap length
Table AGN: Parameters for an RM7 core

Param Unit
Value
symbol symbol

Ae 44.1E-6 m2

le 30.0E-3 m

Bpk 0.3 T

μr
1500 1
at Bpk

Aw 21.7E-6 m2

As an inductor designer, then, you appear to be caught between a rock and a hard place. Make
the air gap too short and you end up with core saturation. Make the gap too long and you run
out of space for the windings. It's time to calculate a course which avoids these monsters. Your
best approach to this problem is to consider how the inductor energy, W, varies with gap
length.
Consider first what might be dubbed the 'aperture limited case'. This presumes that, although
the total magneto motive force is fixed by the area of the winding aperture, the core flux can
assume whatever value is then determined by the gap. The inductor energy is then

W = ½ Fm2 / RL

Equation AGD

Clinging on to the current flow analogy, this is akin to finding the power dissipation to be P =
V2 / R. Substituting equation AGD into equation AGK -

lg = ½ μ0 Fm2 Ae / WL - le / μr

The term in le may usually be ignored. The energy becomes an inverse function of the gap length
(see figure AGO).

Next, consider the 'flux limited case'. Under this assumption the core flux, Φpk, is set to the peak
value which the ferrite can support; but you don't worry about the MMF necessary to produce it.
The variation of energy with gap length is now different

WL = ½ Φpk2 RL

The current flow analogy finds the power dissipation to be P = I2 × R. Substituting equation
AGH into equation AGK -

lg = 2 μ0 WL / ( Bpk2 Ae) - le / μr

The term in le may again be ignored. This time, core energy is linearly related to the the gap.

Well, as figure AGO shows, there can be just one gap length which yields both Φpk together with
the maximum MMF -

RL = Fm / Φpk = (le / μr + lg) / ( μ0 Ae)

Then substitute equation TMS to obtain the notably concise result -

lg = μ0 Fm / Bpk - le / μr

The term in le may usually be ignored. So, broadly speaking, when you select a specific core then
you have set a limit on the amount of energy you are able to store, and on the gap length that
delivers it. Manufacturers sometimes specify this information in the form of Hanna curves,
which show how inductor energy varies with gap length.

Consider a concrete example for the case of the RM7 core, for which typical parameters are
given in table AGN. This gives the energy characteristics shown in figure AGO. A value of
2×106 Am-2 is taken for JW, which is a realistic practical value. That leads to a value for Fm of

43.5 ampere-turns.
Only when the gap length is about 0.18 millimetres will the flux produced by a fully wound
aperture drive exactly 0.3 teslas round the core. One manufacturer supplies RM7s ready gapped
at 0.24, 0.41 and 0.73 millimetres. This spread suggests that the above result is at least of the
right order.

Estimates of reluctance based on a linear relation provide serious overestimates in cases where the
spacer thickness is greater than about 10% of the minor core limb cross section. This is due to fringeing
flux.

The narrow outer limbs of RM cores are especialy prone to fringeing, so a ready built centre limb
gap will conform better than a gap fabricated with a spacer. Reliable estimates of the reluctance
of large air gaps is possible with the assistance of magnetics modeling software. Alternatively,
you should just increase the gap until the inductance factor, as measured, drops to the 'theoretical'
linear model value.

Pros vs cons
The advantages of an air gap can be summarised

 Higher values of mmf can be tolerated before saturation takes place

 Reduced core losses (higher 'Q' factor).
 The flux is less sensitive to changes in current and temperature.
 Cores can be obtained in adjustable versions.

The disadvantages of the air gap are -

  More turns are required to obtain a given inductance.
 Increased losses in the windings (lower 'Q' factor).
 Increased leakage inductance.
 Increased radiated field.
 Increased susceptibility to external fields.

The most serious of these is usually leakage. Values as high as 20% are seen in practice
(particularly if a spacer has been used for the gap) and this can have a disastrous effect upon the
efficiency of switching supplies. Those using the 'flyback' principle with gapped transformers are
especially vulnerable to leakage effects.

As the gap size increases, flux will start to 'spray' into the area of the windings in proximity to
the gap. This will generate high levels of eddy current losses in the wire. To avoid this problem
you can wind on a few turns of insulating tape around the middle of the coil former to exclude
wire from the region.

Example design
Let's suppose you need to design a relatively large inductor: 470μH at 40 amps. Straight away,
equation AGC gives the stored energy -

WL = ½ 470×10-6 × 402 = 376 millijoules

Equation AGS

That's a whopper. Well, table AGT shows that a U93 core pair is the smallest which can store
that much energy. In fact, since the U93 can store 772 mJ with the optimum gap (about 18 mm)
there should be a range of air gaps that might permit the basic specification to be met. The lower
limit for the gap will be decided by the flux limited condition. Putting the datasheet values into
equation AGU -

lg = 2 × 4 π × 10-7 × 0.376 / (0.32 × 840×10-6) - (0.354 / 1500) = 12.3 mm

Equation AGY

The term in le is of little significance. The upper limit for the gap will be decided by the aperture
limited condition. Substituting into equation AGQ -

lg = ½ 4 π × 10-7 61302 840×10-6 / 0.376 = 25 mm

Equation AGX

So, which gap length in the range 12.3 to 25 mm do you choose? The answer depends on what
other constraints you have. For example, are you more worried by core losses than copper
losses? If this inductor is intended for use as an output filter then the flux swing will be small,
and a short gap will be best. OTOH, for use as a flywheel inductor in the main switching circuit
then the flux swing will be large. Even at frequencies as low as 25 kHz a swing of +/- 0.3 tesla
will likely generate excessive core losses. This is reducible through a larger gap to cut down Bpk.

Any of these gaps will be more than 10% of the limb cross section (2.8 mm), so you will need to
experiment a bit. Assume that core loss is your enemy, leading you to favour a gap of 25 mm.
That will mean a spacer thickness of 12 or 13 mm (half an inch will be fine). Substituting this
into equation AGK -

RL = (0.354 / 1500 + 0.025) / ( 4 π × 10-7 × 840×10-6) = 23.9 MA Wb-1

Equation AGL

and this yields an inductance factor factor of 42 nH, so for 470 μH you need 106 turns.
Unfortunately, such a huge gap will lead to high leakage flux, and the simple analysis above will
not be too accurate. Still, it's a start.