Hello!
I designed a PCB trying to read from a sensor. The signal is being modulated to three tones, one at 500Hz, one at 800Hz, and one at 1.1KHz The one at 800 Hz is about 40dB higher than the other two. The information I care about is their individual amplitude. I am using an ADC to sample the data, and then FFT to get the spectrum. I tried to use a notch filter to suppress the peak at 800Hz. I followed my text book to design a 6th order high pass and a 6th order low pass and then summed the outputs to create a notch filter. The simulation worked fine on LTSPICE, I have a notch of 34dB at 800 Hz.
The problem comes when I actually make this into a PCB implementation. The circuit uses two OPA4209 opamps and capacitors and resistors. The output of the circuit looks like simulation except that the 800Hz tone slowly increases its amplitude over time. From my observation, the rate of change is decreasing exponentially.
Does anyone has knowledge of what might be the source of this behavior?

 

hi f11,
Welcome to AAC.
Please post your working LTSpice asc file and any models, we can then run the Sim and compare results.
E

 

Here it is.

 

Here it is.

@fishba11
What kind (e.g. ceramic x7r) of caps are you using?
My naive assumption is that if you are concerned about the absolute value of 800Hz passed, then you should be using a filter that has a flat frequency response in the region of 800Hz (a band-reject filter), not a notch. With a notch having steep walls, even a slight variation in frequency response of the low- & hi-pass filters results in a large change in output.
How much change in amplitude of the 800Hz signal are you seeing?

 

@fishba11
What kind (e.g. ceramic x7r) of caps are you using?
My naive assumption is that if you are concerned about the absolute value of 800Hz passed, then you should be using a filter that has a flat frequency response in the region of 800Hz (a band-reject filter), not a notch. With a notch having steep walls, even a slight variation in frequency response of the low- & hi-pass filters results in a large change in output.
How much change in amplitude of the 800Hz signal are you seeing?

That is a very good point. I agree that it is not very wise to put it in a very steep region. However, I suspect that might not be the root problem, because I see a monotonic increasing in amplitude, from -22.5 dBV to -20.9 dBV in about 10 min. If it is due to frequency shifting, this implys that the frequency only changed to one direction?

 

That is a very good point. I agree that it is not very wise to put it in a very steep region. However, I suspect that might not be the root problem, because I see a monotonic increasing in amplitude, from -22.5 dBV to -20.9 dBV in about 10 min. If it is due to frequency shifting, this implys that the frequency only changed to one direction?

@fishba11
I/We have been assuming that the 800Hz signal itself is stable and its frequency is not varying. If that is not true, then indeed whatever is happening in the filter is irrelevant at this time...and you will certainly never achieve what you seek using any notch filter.
Assuming that the 800Hz signal frequency is not varying, then I was suggesting that a gradual change in the filter characteristic over time might be occurring. That is why I asked about the capacitors. Ceramic capacitors are notoriously unreliable for filters; the capacitance changes with voltage, temperature, age, and mechanical stress and is different for different ceramics. If you are using ceramic capacitors with any dielectric other than C0G/NP0, you should replace them with stable film capacitors. Each kind of film dielectric also has its own characteristics that must be taken into account, primarily temperature coefficients...but all film caps are more stable than ceramic caps (except for C0G/NP0).

Regardless of cause, why do you believe that a monotonic change is surprising? Monotonic drifts in electronics are very common.