This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

LM7301: Multiple Feedback Bandpass Filter - low resonant gain

Part Number: LM7301

Good afternoon! I am trying to impliment a bandpass filter using the Multiple feedback topology with fo = 19khz, q=20, and Resonant Gain = 40 dB:

Theoretical circuit simulated with a filter design tool, ac magnitude response above.

I used a DC supply of 1.2V for Vref, I replaced ADA4661-2 (GBW of 4 MHz) with TI LM7301 that has a GBW of 4 MHz also. I used 1% capacitors and .5% tolerant resistors. I simulated the circuit with TI TINA and got close to a 1V pp output with an input voltage of 10mv pp, 19khz input voltage. This TI TINA simulation respresents a 100x gain (same as theoretical results above).

However, when I impliment the exact circuit above on the breadboard and force 10mv pp, 19khz with a function generator on the input, I get only a 200mv pp scope reading on the output. If I sweep the frequency from 1khz to 30khz, I do see the bandpass frequency response I would expect.

My main question is what would cause the resonant frequency gain to be so much lower then the simulated and theoretical results? I would think TI TINA already compensates if the GBW is insufficient, which I don't believe it is. Any ideas?

Regards,

George

  • Hi George,
    Your attachments did not come through. Can you please try and reattach them?

    Regards,
    Hooman
  • Doc1.docx 

    I seem to have trouble with the attachment feature. I have attached a word file with the simulated circuit diagram and results

  • Hi George,

    I think you hit the nail on the head already. At such a high gain (A0=-100 V/V) and relatively high upper cutoff frequency (Fh=19.5 kHz), you would preferably choose an op amp with a unity gain bandwidth of UGBW>100*Fh*|A0|=100*19500*100=195MHz which is far from the specifications of the LM7301. When I designed a MFB BP filter for your requirements in FilterPro, I was even recommended a UGBW of 3.8GHz.
    In addition, I think that your circuit might have an erroneous Q-factor. This can be verified by carrying out an AC-simulation of the circuit in TINA or by hand-calculations. I would suggests that you choose capacitor values of at least 1nF, and then select the other resistors as

    R1=2Q/(w0*C)
    R2=R1/(2*|A0|)
    R3=|A0|*R1/(2*Q^2-|A0|)

    Lastly, if you wish to bias the CMV to accommodate negative input signals and you have a 2.5V source on your board, I would try to hook this up directly to the non-inverting input. Section 16.5.1.2 in Op Amps for Everyone is a good reference that will give you plenty of practical pieces of advice in addition to step-by-step design procedures.

    Hope this helps. Good luck with your design.

    All the best,

    Gustaf

  • Hi Gustaf,


    Thank you for your insight! It is very helpful to get a clear understanding of requirements for such a filter. I do wonder though if the GBW requirement for the FilterPro is exaggerated. I usually get a GBW this is much too high to impliment in practice when I simulate with FilterPro

    Why does the UGBW need to have an additional 100 multiplication factor with Fh and A0? I am perplexed as to why the difference would not be reflected in TINA:

    If I were to use the FilterPro guidelines, I would not be able to make a filter with even a Q of 2 and gain of 0 with this op amp.

  • Hi Gustaf and George,

    George; I'm curious to know if the bench testing with ADA4661-2 yielded any higher output swing than what you saw on the bench with LM7301? To me, that should not happen and the two devices should behave similarly.

    Regards,

    Hooman

  • I wish I could say Hooman. I did not attempt to use the ADA4661-2. I assumed I would achieve similiar results, as they both have similiar specs. Just as you have assumed.

  • Hi Hooman and George,

    It is possible that the TINA model of the device is not complete, which renders you false results (this is a speculation).

    To further elaborate on the reason why you need a high UGBW, it has to do with the gain error in the passband. Most filter transfer-functions are based on an ideal op amp model with infinite open-loop gain, Aol. In reality, Aol has a pole around 1-10Hz which causes the open-loop gain to decrease over frequency. When this is taken into account, an additional second-order polynomial will show up in the denominator of the filter transfer-function which is multiplied by 1/Aol. If this polynomial grows too large in your passband, the gain will be significantly attenuated. Although, in this case, I wouldn't expect such a large attenuation as 1/5 of your desired output.

    Regards,

    Gustaf
  • Hi Gustaf,

    I believe you are right. I will test the same circuit on bench with an op amp with 200 MHZ bandwidth, and I will also test a circuit with two stages and post the results.

    Thanks for your support,
    George
  • I'm looking forward to seeing your results, George.

    Regards,

    Gustaf