3
\$\begingroup\$

I’m working on a non-inverting op-amp circuit with:

  • AC input: 1 Vpk at 1 kHz

  • DC offset: 2 V in series with the AC signal

  • Gain: 3 V/V

  • ±12 V power rails

There is significant simulated noise on the rails (thermal noise, resistance 10 MΩ, bandwidth 10 GHz), and I want to reduce it while maintaining the same gain.

Constraints:

  • I cannot use a T filter (I already tested that filter circuit).
  • op amp gain must remain as 3v/v

I tried decoupling capacitors (0.1 µF and 10 µF on each of the +12 V and -12 V rails), but after adding them, the op-amp output became more noisy and the top of the amplified sine wave was flattened.

Circuit details:

  • Non-inverting op-amp with gain = 3 V/V

  • Input signal: 1 Vpk AC + 2 V DC

  • Power supply noise: thermal noise with resistance 10 MΩ and bandwidth 10 GHz

  • Only decoupling capacitors added (no T filter, pi filter, or ferrite beads yet)

Questions:

  • Why is my op-amp output distorted (“flat top and noisy bottom”) after adding decoupling capacitors?

  • What alternative filtering solution(s) could I use to reduce high-frequency power supply noise while keeping the same gain, without using a T filter?

  • Would a pi filter, ferrite bead, or other method work better for this type of noise and how would you calculate the required component values?

Any guidance or suggestions on proper filtering and decoupling for this op-amp would be greatly appreciated.

The op amp circuit with thermal noise The op amp circuit with thermal noise

Oscilloscope depicting op amp output with decoupling capacitors in the circuit Oscilloscope depicting op amp output with decoupling capacitors in the circuit

Oscilloscope depicting op amp output without decoupling capacitors in the circuit Oscilloscope depicting op amp output without decoupling capacitors in the circuit

\$\endgroup\$
3
  • 3
    \$\begingroup\$ S1 and S2 aren't going to do a whole lot. \$\endgroup\$ Commented Nov 10 at 11:58
  • \$\begingroup\$ S1 and S2 should actually provide the option to short-circuit the capacitors out of the circuit, to provide the intended feature (instead of the current setup). I.e., one leg where they already are, and the other at Vcc/Vss. If I get the intention correctly, that is. And they should be simple ON/OFF switches, not commutators. \$\endgroup\$ Commented Nov 11 at 6:18
  • \$\begingroup\$ I'll remove the switches to simplify the circuit and to avoid those issues. Just wondering if anyone knows what filter I could use to filter this noise instead of a T filter? \$\endgroup\$ Commented Nov 11 at 7:11

2 Answers 2

Reset to default
10
\$\begingroup\$

Regarding this schematic:

enter image description here

The decoupling capacitors are connected to the opamp's supplies no matter what the switch positions are. The only difference between the switch positions would be the switch resistance, if modelled. Since there is a difference between the two traces, that's probably the issue, for example the opamp could be powered in one case, and not powered in the other.

I'm not familiar with this particular simulator, but the thermal noise source looks like a voltage source. If it is modelled as an ideal voltage source, it will have zero output impedance. Simulated wires also have zero impedance. The impedance of the switch is unknown. If it is also zero, the noise source will control the supply voltage, and the capacitors won't change anything about that. The effect of decoupling caps can only be modelled if the supply impedance is known. Otherwise, they form a voltage divider between the unknown supply impedance, and the impedance of the caps.

Some opamp models include power supply effects, but not all of them do. Without knowing if this one does, I have no idea what's happening in the circuit.

If the simulated noise source attempts to sim up to 10GHz in transient mode, and the run lasts several milliseconds, the simulation time step will be way too large to model the noise source, so I'd expect convergence issues and unpredictable output.

This would be better handled by an AC simulation when the power supplies being DC+AC.

\$\endgroup\$
4
  • \$\begingroup\$ Thanks for the explanation, I'll stick to using AC sweep for analysis. I was also wondering if you knew what type of filter would be better at filtering noise apart from a T filter? \$\endgroup\$ Commented Nov 11 at 7:09
  • \$\begingroup\$ Depends on the type of noise and the source impedance of the noise. \$\endgroup\$ Commented Nov 11 at 8:37
  • \$\begingroup\$ It's thermal noise with 10M ohms resistance and 10GHz of bandwidth. \$\endgroup\$ Commented Nov 11 at 13:12
  • \$\begingroup\$ Nevermind I worked out how to make a pi filter. \$\endgroup\$ Commented Nov 12 at 11:09
1
\$\begingroup\$

One possibility is that the 2K ohm resistor is too much of a load on the op-amp. I would try increasing the 1K and 2K by a factor of 10, thus they become 10K and 20K. Normally the 741 will drive loads of 5K or more with ease, but in this case the 2K is part of feedback loop, so it is a defacto load to gnd.

\$\endgroup\$
1
  • \$\begingroup\$ Thanks I'll swap out the resistors for those larger values. \$\endgroup\$ Commented Nov 11 at 7:10

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.