6-pole Butterworth bp filter simulation. Reality or just an artifact???

[markusw]

Hi all,

some more Spice Qs.

I simulated a 6-pole Butterworth BP filter (Fc=30 Hz, Q=1) in LTSpice. The calculations were done with this tool:
http://www.analog.com/Analog_Root/static/techSupport/designTools/interactiveTools/filter/filter.html

Below you can see the schem and the predicted waves. Once a 30Hz sine is fed, once a 150 Hz.

As you can see the 30Hz gets stronger with every stage (green: at input; blue, red and pink after 1st, 2nd and 3rd stage, respectively).

The 150 Hz as supposed gets weaker with every stage (pink at input; then blue, red and green). Unfortunately,
LTSpice also predicts that the 150Hz signal at the beginning is superimposed with a 30Hz wave that is even stronger than
the 150Hz. Since I want to feed the signal into a comparator the superimposed 30Hz would screw up the square wave.

Now my questions: is this superimposed 30 Hz wave just an artifact or does it reflect reality?? And in case it's just an
artifact why does Spice produce it?

Thanks for your answers in advance :) 

Regards,

Markus


The 30 Hz sine...


and the 150 Hz sine.

[MR COFFEE]

Hi Markus,

The 150 hz signal is probably just ringing the filter at the beginning. 6 pole bandpass filters will tend to output their center frequency when hit with any impulse - even a plain old spike.

Is this the same project? Did you mean to design a Bandpass?

We had talked about a lowpass at a higher frequency, and a low-order lowpass to reduce the harmonic content.

Or something new?

[Sir H C]

Looks like  you might have a little ringing in your filter.  I would say it is a result of initial conditions acting like an impulse on the input.

[markusw]

Hi guys,

thanks for your replies! So the simulations are at least similar to reality?

Is this the same project? Did you mean to design a Bandpass?

We had talked about a lowpass at a higher frequency, and a low-order lowpass to reduce the harmonic content.

Yup, it is still the same project. I just tried some simulations with lp and bp filters. Interestingly, I get a similar result from the lp sims. Initially I started with the lp filters and subsequently tried with the bp to see if the results are different.
Basically I'm trying to figure out (by sims and breadboarding) what's the best way to extract the fundamental from my bass guitar signal. Unfortunately, 31 Hz for the low B is damn low. So just a few cycles that are out of tune at the beginning of the note already cause severe tracking problems. Therefore, I started tinkering also with higher order filters. When I look at the bass signal at my software scope there are still quite some 2nd order harmonics present even after a 12dB/octave lp.

Therefore, I alternatively want to check out tracking filters for fundamental extraction. R.G. some time ago mentioned two different approches that might work in this context: either to split the breadboard in seveal bands and pick out the "fundamental" band with some logic stages driven from the envelopes of the various filter bands. The other way was to use a MXR-like filter approch (e.g. with a switched-cap-chip like the LTC1054).
By these tracking-filter approaches I suppose it should be possible to get a rather pure sine wave that subsequently could be multiplied (e.g. with balanced modulator chips) to get the octaves. Unfortunately, I don't have a idea which approach would be suited better for bass freqs nor if these tracking-filters would be fast enough for a 31 Hz signal.



Regards,

Markus