I have a band pass filter with a center frequency of 100kHz, a bandwidth of 10kHz, and a gain close to 1. Input signal is 100mV-150mV peak to peak with frequencies varying from 80kHz to 120kHz. I am using an OPA355 op-amp.

Here is my circuit on Proteus software: enter image description here

Here is the expected output,with an input voltage of 130mV: enter image description here

I then use the output of this filter as the input to my comparator. The comparator works perfectly in simulation and when tested alone. I am using LM339:

enter image description here

As soon as I connect the two, the filter output immediately changes due to an added noise component. On close inspection, the amplitude of the filter output decreases and has visible noise spikes.

This messes up VTH, VTL and hysteresis settings of the comparator. Hence the expected output is not attained.

enter image description here

and at 1ms/D:

enter image description here

I noticed that:

  1. If I remove the 2.5V from the potential divider from pin 4 of the comparator, the band-pass filter output goes back to it's original correct state.
  2. if I remove the wire connecting the output of the filter to pin 5 of the comparator, the bandpass filter output also goes back to normal.

I have tried, adding a decoupling capacitor across R1, changing resistors used for the voltage divider, using different LM339 comparators, and also using different channels of the comparator. None of these methods have worked. I am really stuck here, any suggestions on how to reduce/eliminate this noise would be appreciated.

How is noise being generated on the output of the bandpass filter, and not on the comparator output?

Please do let me know if I need to provide further information. I also have pictures of the (crude) setup on my breadboard if those would help identify a possible culprit. Thanks

  • \$\begingroup\$ What do you need R55 for? \$\endgroup\$
    – Linkyyy
    Commented Jan 10, 2019 at 21:33
  • \$\begingroup\$ R55 is a standard part of the multiple feedback topology. It allows tuning of the resonant frequency. \$\endgroup\$
    – rrz0
    Commented Jan 10, 2019 at 21:39
  • \$\begingroup\$ It is supposed to be connected to ground. \$\endgroup\$
    – Linkyyy
    Commented Jan 10, 2019 at 22:01
  • \$\begingroup\$ @Linkyyy Assuming we are talking about \$R_{55}\$'s end tied to a voltage source and also keeping the (+) input to the OPA355 at \$2.5\:\text{V}\$, does it matter? Seems like the main difference would be during start-up and the reference the OP uses is probably slightly better in that sense, though I've no idea what that voltage source actually is... so it could be worse from some other perspectives, depending. \$\endgroup\$
    – jonk
    Commented Jan 10, 2019 at 22:15
  • \$\begingroup\$ @Rrz0 I'm confused a bit with the scope pictures. Am I correctly reading milliseconds per division on the first such image? If so, doesn't it seem odd that it is about 3.3 milliseconds per cycle if you are using 100 kHz? Your bandpass looks right for the frequency. But your scope picture makes less sense to me given the timebase. \$\endgroup\$
    – jonk
    Commented Jan 10, 2019 at 22:21

1 Answer 1


The impedance and noise from the 2.5V potential divider from pin 4 of the comparator is too high, for a 82 Load (R55). Use a better source Z(f) with 0.1 Ohm impedance @100kHz and < x ohms otherwise.

Rev A

Consider scale Filter up 10x to 100x in impedance and decouple 2.5V with a big cap.

enter image description here

There is not need for low R value to go to 2.5V since it is AC coupled. This makes this node a noise source from 2.5V with high gain. ~R8/R55 ?? 50dB gain

For better bandstop rejection I might consider a quad OA with an 8th order filter. Here testing with 25mVp signal. enter image description here

  • \$\begingroup\$ 2.5V going to R55 is not coming from R1 and R2 voltage divider biasing the cmoparator, but from a separate divider. I'm not sure what you mean by "Use a better source Z(f) with 0.1 Ohm impedance @100kHz and < x ohms otherwise.". How can I implement this in my design? \$\endgroup\$
    – rrz0
    Commented Jan 10, 2019 at 21:31
  • \$\begingroup\$ What is Zf(2.5V) ? What do you need ? < 1% of R55 What is C for this Zf? \$\endgroup\$ Commented Jan 10, 2019 at 21:47
  • \$\begingroup\$ Looks like power supply or ground noise to me, like the scope probe ground is the only ground connections between the circuits. also, neither schematic shows any power supply decoupling, which is critical. Also, can you post an integrated schematic that shows all power sources and connections, grounds, and reference connections for both circuits? \$\endgroup\$
    – AnalogKid
    Commented Jan 10, 2019 at 23:23
  • \$\begingroup\$ I agree with >50dB gain thru the ground or supply to R8. So THAT gnd is critical. But if R divider picks up noise, my fix should improve it. \$\endgroup\$ Commented Jan 10, 2019 at 23:38

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.