Elektor_Mag_-_January-February_2021

lektor January & February 2021 81

Choice of components
As we already noted at the beginning, there are no degrees of
freedom when dimensioning the components for passive filters.
When computing these reasonably advanced filters the use of
specialised software is essential. At [2], [3] and [4] you will find a
selection of suitable filter programs. The values for the inductors
and capacitors that are calculated this way will not be available in
most cases. With capacitors this problem is quite easily solved:
simply connect two or three capacitors in parallel to get as close
to the desired awkward value as possible. Capacitors are, after all,
small and not that expensive. You do, however, have to take toler-
ances into account.

In a simulation [5] you will have to verify whether the frequency
response of the filter is close enough to the ideal when using the
actual component values. When building a prototype you cannot
fail to measure and verify the value of the capacitors before fitting
them. Passive filters do not really behave any differently to active
filters and the same is true for both types: a filter is only as good
as its capacitors. SMD capacitors of the X7R type have therefore
no place in filter applications. Good capacitors are (unfortunately)
usually large and not very cheap. You can only make a start on the
layout of the circuit board once the circuit is finished and you know
exactly which components you will be using.

Unfortunately the situation with inductors is much more diffi-
cult. You shouldn’t even try connecting them in parallel. Even a
series connection will only work if the inductors are not magnet-
ically coupled. To prevent magnetic coupling you have to fit them
perpendicular to each other and arrange them in a zigzag pattern
on the circuit board. You can admire this in the filter of Figure 23,
although the third inductor at a relative angle of 45° is not really in
a optimal position. In the filter of Figure 24 the small blue induc-
tors are not connected in series and there are also no capacitors
connected in parallel. The manufacturer probably has these compo-
nents specially manufactured so that they fitted perfectly.

A possible way out of the problems with inductors is the use of
adjustable versions. Here you can change the self-inductance within
certain limits by turning the ferrite cap or core. This is, for example,
the case with the Neosid inductors that are used in Figure 16.

There is something else that needs to be taken into account with
inductors: the windings have an ohmic resistance that often cannot
be ignored. We have to measure this or find it in the datasheet and
carry that over into the simulation when dimensioning the filters.
The non-ideal behaviour of the inductors, as a consequence of the
resistance of the windings, can lead to a reduced frequency response
in the pass-band. This has to be either corrected or compensated
for in the circuit that follows.

In general, larger inductance values use coils with ferrite cores that
act as a kind of ‘amplification’ compared to the self-inductance
of the winding on its own. Depending on the size of the core, the
number of turns, and the size of the current that flows, the core
can become magnetically saturated and therefore generate an
enormous amounts of distortion. If more harmonics are measured
at the output of a filter than at the input, the signal level should
be reduced or inductors with larger ferrite cores should be used.

L1

68n

L2

120n

C 1

47p

C 4

47p

R 2

50

R 1

50

L3

120n

C 5

47p

C 2

12p

C 3

12p

200522-020

L4

68n

Figure 20: Double inverse Chebyshev filter of the fourth order.

Figure 19: Common-mode frequency response of the double SE-filter of

Figure 17 (red) together with the combined filter of Figure 18 (green).

Figure 21: Frequency response of a fourth-order double inverse Chebyshev

filter.