Avoid Misuse of Sallen-Key Filters - John Dunn, Consultant, Ambertec, P.E., P.C.

Using a two-pole Sallen-Key filter on an input signal that has a rise time and/or a fall time that is too fast is a mistake. If the filter's input were a square wave for example, the essentially infinite rising and falling dv/dt of that square wave place demands on the op-amp to deliver step changes of feedback current in zero time which the op-amp simply cannot do.

See this Multisim-SPICE result to understand the issue:

 

What you must do is turn the two-pole filter into a three-pole filter. The added passive RC pair, shown here as R3 and C3, diminishes the dv/dt that gets presented to R1 and thereby relieves the op-amp of the demand to be instantaneously fast.

  

The op-amp remains in control and the output spikes that were there before go away.

 

Comments

[George Storm]

Too true. The second-order filter is an optimum zero-overshoot design. Perhaps it would be even nicer to use the extra pole to make a third-order equivalent?

[Paul Rako]

This is a stopband leakage problem. C1 will pass very high frequencies directly to the output. You can fix it by feeding C1 with a buffer amp from the output. See
edn.com/article/459456-Eliminate_Sallen_Key_stopband_leakage_with_a_voltage_follower.php

[John Dunn]

Stopband leakage arises because C1 does indeed pass high frequencies directly to the output of U1A, but adding a buffer for feeding back to C1 would still require that buffer to be infinitely fast to accomodate a zero-rise-time input from V3. The symptom could be diminished, but not eliminated.

[Ralph Williams]

This reminds of the issue of slew-induced distortion versus bandwidth limiting when oversampled audio DAC's interface to analog circuitry. Presumably one of the advantages of vacuum tubes is that they bandwidth limit before they slew-rate limit.