0
\$\begingroup\$

I have a sine wave (uV level) that is in varying frequency (1kHz to 8kHz). I like to design a band pass filter that attenuates the other frequencies while amplifying my target frequencies and this can be achieved with design tools from TI, Analog etc. However, all these tools provide several filter types such as Butterworth, Bessel, Linear phase etc.

My plan is to digitize the amplified and bandpass filtered signal at 50kHz and do signal processing on it to determine how the frequency of the system is changing over time. I am predominantly a software guy and I like to make sure I choose the right filter type.

I think preserving the shape and amplitude of the signal is my priority across frequencies I am interested as opposed to having a linear phase difference etc.

Does this make sense considering I am predominantly interested in what are the frequencies at which point of time?

Which type of filter I should choose for my application?

\$\endgroup\$

2 Answers 2

2
\$\begingroup\$

A filter from 1 to 8kHz is not really a bandpass filter. Designing a bandpass so wide will give you a very bad filter - you should probably consider a low-pass at 8 kHz and a high-pass at 1 kHz. uV levels are a pain to filter - it will take very special amplifiers to do that.

You might want to consider passive filtering (LC networks) at the input, then carefully amplifying, and doing the grunt work later, in software.

Again, at uV levels you also want to spend some time thinking about how to shield the circuits from ambient noise.

\$\endgroup\$
3
  • \$\begingroup\$ Thanks for the answer. I believe I will use an active filter with gain to amplify and go from there. \$\endgroup\$
    – Ktc
    Commented Jun 3, 2015 at 5:22
  • \$\begingroup\$ Don't skimp on the OPAMP - many run-of-the-mill amps generate more than uV of noise! \$\endgroup\$
    – jcoppens
    Commented Jun 3, 2015 at 5:34
  • \$\begingroup\$ Yes. I will use a low noise voltage feedback opamp in the circuit. Probably something with nV range. I just need to search. \$\endgroup\$
    – Ktc
    Commented Jun 3, 2015 at 6:59
2
\$\begingroup\$

I think preserving the shape and amplitude of the signal is my priority across frequencies I am interested as opposed to having a linear phase difference etc.

If the input signal is a sinusoid, any linear filter will preserve the shape of the waveform. If you use an active filter you may need to be concerned about making sure the filter remains linear across the range of input amplitudes you will see. But this is true regardless of the filter type.

If you want to preserve the amplitude of your signal as its frequency varies across your passband, you want to use a Butterworth filter. The Butterworth filter has a maximally flat passband response.

Other filter types allow more ripple in the amplitude in the passband, but in return they might have better attenuation in the stopband, or a steeper roll-off between the passband and stopband (for the same filter order).

As another answer mentions, because your passband spans 3 octaves, you may want to consider designing your filter as cascaded high-pass and low-pass filters, rather than as a band-pass filter. You should still simulate your design (including component parasitics) with both filters together to ensure that the two filters don't interact in a way that degrades the performance.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.