Elektor_Mag_-_January-February_2021

80 January & February 2021 http://www.elektormagazine.com

low-pass filter that is connected directly to the DAC output and

even the anti-aliasing filter at the ADC input are also differential.

The circuit in Figure 17 consists simply of two identical 10 MHz

Butterworth low-pass filters of the fifth order in a conventional

single-ended design. The signal paths have no shared compo-

nents (except ground) so component tolerances can cause differ-

ent behaviours in the two ‘channels’.

The second variant, in Figure 18, looks more like a real differential

filter, but here each of the two capacitors from Figure 17 that were

connected to ground are now combined into one. They therefore

now affect both signal paths in equal amounts. In this case differ-

ences can only occur because of the inductors.

For differential or push-pull signals, both these circuits exhibit

the same filter operation. But for common-mode signals the filter

behaviour is different. Figure 19 shows that, in the combined filter

of Figure 18, only the inductors influence the frequency response

and that, therefore, there is only a weak filter effect for common-

mode signals. Although the common-mode signals from DACs are

reduced by the differential amplifier that follows, this is not optimal

because the common-mode rejection of RF amplifiers reduces as

the frequency increases. Every solution always has its advantages

and disadvantages. To realise filter stages that are as identical as

possible we can, for example, use filter modules from the same

production batch so that the filters are as identical as possible.

Double-inverse fourth-order Chebyshev filter

I noticed this filter when reverse-engineering a filter that I had

bought. It consists of two fourth-order filters joined together

(Figure 20) and has a nice characteristic that can be easily realised

with components from the standard E-series. It also requires few

‘awkward’ values, making building it much simpler. If you would

like to change the corner frequency of the filter, then you can

change the values within the standard E-series. This advantage

is however at the cost of a small dip in the pass-band. Figure 21

shows the frequency response of this filter, which has an impres-

sive minimum attenuation of 70 dB in the stop-band. Both notches

have the same frequency and reinforce each other. In Figure 22

we have enlarged the response in the pass-band where we can see

the characteristic dip at 75 MHz.

L1

198

C 2

367p

C 3

367p R 2

50

R 1

50

C 1

272p

200522-014

L2

118

L1

129

C 1

197p

C 2

636p

R 2

50

R 1

50

L2

129

C 3

197p

L1'

129

C1'

197p

C2'

636p

R2'

50

R1'

50

L2'

129

C3'

197p

200522-017

L1

129

C 1

98p5

C 2

318p

R 2

50

R 1

50

L2

129

C 3

98p5

L1'

129

R2'

50

R1'

50

L2'

(^129) 200522-018
Figure 14: Schematic of a passive all-pass filter of the second order.
Figure 17: Combination of two ‘normal’ Butterworth low-pass filters to filter a
differential signal.
Figure 18: This variant combines the capacitors Cx and Cx’ into one
capacitor with half the value of capacitance that, in Figure 17, were each
connected to ground individually.
Figure 16: DIY video filter with eight Neosid inductors.
Figure 15: Characteristic of the group delay of the second-order, passive
all-pass filter in Figure 14.