0
\$\begingroup\$

I have a 4th order bandpass filter which needs to be biased using 1/2 Vcc. I am trying to achieve this using a voltage follower. The bandpass filter was constructed using Multiple feedback topology as shown:

enter image description here

On Proteus I get a 100kHz center frequency and a 4kHz band pass at a gain of 1 AV.

Using a +2.5V simulation goes as expected, and gives the desired response.

enter image description here

enter image description here

I designed two voltage followers using LM324, and tested these in a separate project. I get 2.5V output as expected.

On adding the voltage followers, and connecting just one output of the voltage follower to the second stage of the op-amp, simulation still takes place without errors, but the response is not correct, as I need to bias both stages.

enter image description here

On connecting both voltage followers, I get an error. I cant understand why.

enter image description here

enter image description here

Any suggestions why this is happening would be appreciated.

\$\endgroup\$
8
  • \$\begingroup\$ What is difference between #2 and #3 diagrams? \$\endgroup\$
    – analogsystemsrf
    Jan 17, 2019 at 19:09
  • \$\begingroup\$ @analogsystemsrf, output of LM324 in #2 is only connected to the second stage of the filter. In #3, U1:A is also connected. \$\endgroup\$
    – rrz0
    Jan 17, 2019 at 19:22
  • \$\begingroup\$ There's a lot of problems with this design and no specs. results indicate low Q and -40dB @ 85kHz on my sim ... meanwhile ground noise is 60dB gain above signal . What are your specs?????? Is this for WPT? (hunch) \$\endgroup\$
    – Tony Stewart EE75
    Jan 17, 2019 at 19:28
  • \$\begingroup\$ Updated question with current response. Center freq is @100kHz, as was calculated. Any suggestions on how I can reduce ground noise? \$\endgroup\$
    – rrz0
    Jan 17, 2019 at 19:40
  • \$\begingroup\$ Yes dont use R7/R55 ratios ... What shape is the passband ( Cauer, Bessel, )? or in otherwords, what is the stop band attenuation \$\endgroup\$
    – Tony Stewart EE75
    Jan 17, 2019 at 19:41

2 Answers 2

Reset to default
1
\$\begingroup\$

The lousy old LM324 has slew rate problems above a few kHz. It also has hardly any open-loop gain at 100khz. Your circuit needs the output of the LM324 to have a low impedance but it can't do it at 100kHz. You need a fast opamp to work at 100kHz.

\$\endgroup\$
1
  • \$\begingroup\$ Or just AC couple the signal and connect R55 and R3 to ground. Then U6 and U2 + inputs can be biased with resistor dividers, and U1 A and B aren't required. \$\endgroup\$
    – Bruce Abbott
    Jan 3, 2020 at 5:13
0
\$\begingroup\$

There are better design tools now. TI is releasing an upgraded online filter designer with a tolerance stackup that is now critical with Q > 10 here >25.

  • Don't duplicate a 2nd order filter to make a 4th order filter when you can get better results for passband ripple and stopband skirt attenuation or linear phase or maximally flat group delay or etc.

  • Sallen & Keys filters are preferred for high Q but uses an extra R but at the advantage of much lower R ratios and common C values. More important is that the input is not severely attenuated and ground noise severely amplified even when unity gain overall.

enter image description here Consider your specs wisely. Be critical about tolerances and costs.

  • Unity gain tolerance error can be -1.5 dB even with 0.5% R and 2% C enter image description here
\$\endgroup\$

Your Answer

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

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