Why am I measuring such high THD+N on my active high-pass filter?

Question

I have designed a 2nd-order active Sallen-Key high-pass filter for use in an audio circuit. An LTspice schematic is shown below, where VDD is 5 V.

I have also breadboarded the circuit using low quality components (X5R caps and 1/4 W, 5% resistors). Using an Audio Precision audio analyzer, I drove the breadboarded circuit with a 1 Vrms differential signal swept from 20 Hz to 20 kHz and measured the following response:

The sweep looks good in my opinion and the -6 dB cutoff point is at 3.16 kHz (very close to expected 3.18 kHz). I measured the THD+N via the Audio Precision using a 1 Vrms signal and a 200 kΩ measurement load. Below about 12 kHz, the THD+N is pretty bad (0.5% - 0.7%). Above 12 kHz, the THD+N is very close to the datasheet of the TL084 opamp being used (THD+N ~= 0.0006%).

I suspect that the higher value tolerance of the resistors and low-quality dielectric material of the capacitors may be to blame here, but does anything else in the circuit diagram / design seem like a likely cause for undesirably high THD+N?

Is it even necessary for me to worry about the THD+N of this circuit? I've only ever measured THD+N on full-bandwidth amplifiers before, not filters.

Answer 1

The number one suspects are the ceramic capacitors. Here is a SE link that shows the severe voltage dependence that ceramics have. The X5Rs are some of the worst, especially in the small SMD sizes.

For any analog signals audio or otherwise, use NPO (Thanks Spehro Pefhany) ceramic or film capacitors instead. Ceramics are fine for bypass and decoupling. Just be sure to increase the value to compensate for voltage dependence. Manufactures don't all specify so you may have to measure voltage dependence.

The common VCC/2 bias will cross-couple at low frequencies. May not cause THD but interference may increase. Better to buffer the voltage divider (VD) with a unity gain amp or have separate VD for each stage.

@Audioguru makes a good point. You need to keep the signal from VCC and GND for good THD, even with rail-to-rail amps. Where you work with single supply, there are lots of modern choices better than the TLO8x.

Whether breadboarding or production: Always put a decoupling capacitor from VDD to GND as close to the chip as possible. A quad amp needs a bigger cap than duals or singles. Always tie off unused amplifiers to prevent coupling through the supply rails.