Zero offset active lowpass filter, part 4

Article By : John Dunn

Circuit suggestions should be rigorously analyzed to avoid adverse consequences.

A little while back, I presented a nearly-zero-offset Chebyshev lowpass filter. Two comments were made that must have seemed valid to their authors, but they were not.

First, we look at the original circuit and its rolloff characteristic:

active lowpass Chebyshev filter Figure 1 Here’s the active lowpass Chebyshev filter from part 3 of this series.

Next, we look at the first comment and the adverse consequence of following the suggestion. The comment read in part, “Why not take the 82nf directly to ground?”

changed lowpass filterFigure 2 This change adversely affects the lowpass filter.

Putting the 82 nF (0.082 µF) directly to ground instead of to the op-amp output utterly undoes the intended frequency response. The revised circuit isn’t even an active filter anymore.

Let’s go back to the original circuit, but with a downward shift of vertical position on the scope. You’ll see why in a moment.

downward shift of vertical position on the scopeFigure 3 Here’s the original circuit with a downward shift of vertical position on the scope.

Here’s another comment that would introduce adverse consequences: “This is the Sallen-Key circuit with an added real-axis (passive) pole (C1/R1). I note that the impedances in that and the next circuit (R2/C2) are in the same order of magnitude. This is bad practice due to interaction between the two circuits. An improved approach in this instance would change R1 to 1.21K and C1 to 100nF in order to minimize these interaction effects.”

unwanted peak Figure 4 This suggestion causes an unwanted peak to arise in the rolloff characteristic.

In this case, the suggested change causes a really big and unwanted peak to arise in the rolloff characteristic.

In both cases, the offered suggestions had not been rigorously analyzed. In my opinion, circuits of this nature are not readily characterized by intuition. One must either do the algebra to derive the exact transfer function or one must use one or another SPICE tool to develop a simulation.

It is unwise to say, if I make “these” values much larger or much smaller than “those” values, there won’t be any interactions to worry about. There very well could be.

John Dunn is an electronics consultant, and a graduate of The Polytechnic Institute of Brooklyn (BSEE) and of New York University (MSEE).

Leave a comment