0
\$\begingroup\$

I built this op-amp circuit that takes in 4 inputs and outputs 3 sums by using op-amp adder and difference circuits to achieve a gain of 2x. In a simplified circuit, I boiled it down to the bare op-amps being used for the circuit in this image in an effort to show what I'm trying to achieve.

When I ran this circuit in LTspice, sending small inputs results in expected outputs. For instance, I sent in a 100-mV peak-to-peak sine wave (500 Hz) in at input A (while setting other inputs to ground), and I get a 200-mV peak-to-peak sine wave at the outputs. Furthermore, the output signals don't seem to affect the previous op-amps in the chain. (i.e., when putting a signal into A, the resulting output from OA6 does not affect OA5).

However, when I tested my circuit in real-life on a PCB, the results are strange, specifically when it comes to inputting a signal into C and D. When I input a 100-mV sine wave signal into A and B (again, 500 Hz), I get the expected 2x sine wave outputs. However, when I put in the same signal into C, I get a 3x gain instead of a 2x gain at output 2 from OA7. Furthermore, when analyzing the circuit when inputting the signal into D, somehow OA4 outputs a tiny 50-mV sine wave even though A and B are grounded. I think that somehow, part of my output signal is being fed back to the op-amp unintentionally which causes excess gain and strange signal outputs. What can I do to isolate each op-amp circuit so that they don't affect one another?

schematic

simulate this circuit – Schematic created using CircuitLab

Full Schematic

EDIT: I posted my rough, simplified schematic in circuitlab and showcased my entire schematic from my CAD software. To be clear, my op-amps have a potentiometer in parallel with the feedback resistor, but on the board, I only have one. I just made room for the potentiometer for troubleshooting purposes here; in case the gain was an issue, I could remove the fixed resistor, put in the pot, and adjust the gain to my liking. Regarding the offsets, I'm using a potentiometer to control them, but from what I saw, they don't seem to be a big issue. The voltage at the offsets should be very close to ground in the single-millivolt range or less.

EDIT 2: So, upon observation, I think that the issue stems with the difference amplifiers. Somehow, when I inject my test signal into D, OA4 outputs a small sine wave, even though A and B are shorted to ground. Perhaps this is what is causing all my problems, as all the problems seem to stem from the difference amplifiers. Why does this phenomenon occur and what can I do to prevent this 'feedback'?

EDIT 3: As requested, below is a picture of my 4-layer board. While it doesn't show a ground plane, rest assured that there is a ground plane present. It's just not filled in order to show the traces. I tried my best to add some silkscreen so that you can see the traces and follow along as best as possible. The left side has the inputs while the outputs are on the right side. Each input goes through the appropriate low-pass filter circuit as well as a diode clamp. From there, you can see how the traces are routed to each op-amp as per their scheme. The silkscreen makes each component identifiable, but it may be hard to read given the trace overlaps.

Initial Layout

\$\endgroup\$
15
  • \$\begingroup\$ search for what an op amp in buffer configuration looks like. it's a circuit that provides an alternate current source for a copy of a voltage. \$\endgroup\$
    – Abel
    Sep 12, 2021 at 21:13
  • \$\begingroup\$ Is there any signal on the power supply? Your offset circuitry is one way for crosstalk between channels. \$\endgroup\$ Sep 13, 2021 at 0:04
  • \$\begingroup\$ @Transistor Thanks. Apologies, I was in a rush, so I thought a picture would be fine. I redid it so I added in my schematic in circuit along with a full picture of my schematic. Rest assured that this is indeed my design. Thinking about it now, the issues stem from the LM7372, which are all configured in difference-amplifier configurations....the inverting sum op-amps seem to behave well. \$\endgroup\$ Sep 13, 2021 at 0:47
  • \$\begingroup\$ @KevinWhite If you mean an input signal, no. All it is is just a voltage from my DC power supply. Yes, the datasheets say +15 and -15, but my power supply only allows for +10/-10. Nevertheless, when testing this at +15 and -15, the issues still remain. \$\endgroup\$ Sep 13, 2021 at 0:53
  • \$\begingroup\$ @BestQualityVacuum - The current taken by the amplifiers can cause the power supply to vary in voltage. Any variation could then find its way into the other amplifiers. I don't think it's the problem though. Have you changed the schematic in the question - all the amplifier numbers have changed? There is no OA4. \$\endgroup\$ Sep 13, 2021 at 1:43

1 Answer 1

3
\$\begingroup\$

There is a path through R28 and R32 to give crosscoupling between Diff1 and Diff2 that is giving the small signal in OA4. The two opamps share the same offset adjustment.

Any signal coming in the D input goes through R31, the R32 to the offset adjustment. Because that acts like a 2.5k resistor to ground there is about a 4:1 attenuation to this point. (expect about 25mV for 100mV into D).

From there the unwanted signal goes through R28 and R27 experiencing another 3:1 attenuation bringing the signal to about 8mV into the non-inverting input of U4A.

U4A has a gain of about 2 or 3 depending upon the setting of RV4. This should result in an unwanted signal of about 20mV at the output of U4A. This is less than the 50mV you say you have measured, it is not obvious why it would be that much.

You can test it by disabling the DIFF_OFFSET1 signal and putting it to the ground.

The output of the offset adjustment pot does not have zero impedance - since it is a 10k pot it will act like a 2.5k resistor to ground when set to the center position. I'm surprised you think the offset adjustment is necessary. These days you can get amplifiers with just microvolts of inherent offset.

In addition to causing interaction between stages, the resistance of the offset pot will dramatically affect the balance of the differential stages. It adds about 2.5kohms in series with R28 and R32 that need to be accurately matched with R26 and R30. The same issue applies to the other differential stages.

Also, adjusting the gain of a differential stage by only varying one side (RV4,5,6,7) will compromise the accuracy of the differencing.

\$\endgroup\$
6
  • \$\begingroup\$ Thanks for your answer. As I was designing an adder circuit like this, I thought it prudent to add in pots for offset adjustment. I'll remove the pot and go back to the 2k resistors and see how it performs. If it works without the offset pots, this may help (especially as how the 40k-resistors show a noticeable effect on bandwidth, I remember now how feedback cap and resistor form a RC filter cutting bandwidth that way). I'm a bit lost on how they share the path. R28 and R32 don't cross any signal traces, though R38 and R42 seem to. Could you please enlighten me? \$\endgroup\$ Sep 13, 2021 at 19:43
  • \$\begingroup\$ @BestQualityVacuum - I've added a detailed explanation to the answer. \$\endgroup\$ Sep 13, 2021 at 22:36
  • \$\begingroup\$ Thank you for that. It was very helpful. Perhaps that's why I didn't see it with 40k resistors. 2.5k impedance isn't much compared to the 40k I was running. Conversely, if I follow correctly, I could alternatively increase the impedance by using a larger pot value, like 1Meg, to get an impedance of 250k, and that would work against my 2k resistors am I wrong? \$\endgroup\$ Sep 14, 2021 at 1:31
  • \$\begingroup\$ @BestQualityVacuum - no, that would make it worse. I would remove the pot and ground the wiper. That would disable the offset adjustment but stop the interaction. \$\endgroup\$ Sep 14, 2021 at 3:48
  • \$\begingroup\$ It worked. Thanks. I don't see the unwanted signal anymore. Shorting it to ground also seemed to fix my other problems. Perhaps my problems thus far stemmed from the DIFF_OFFSET1 feeding the signal back to where it shouldn't have been. Nevertheless, some issues still remain with the gain and perhaps other minor things, but they seem to be easily fixable by replacing the components and not shorting anything to ground. (Though this does make me wonder whether or not I should also short DIFF_OFFSET2 to ground as well) \$\endgroup\$ Sep 14, 2021 at 14:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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