2
\$\begingroup\$

I have a question regarding the configuration of opamps for sensor signal amplification. I have a circuit with two sensors going through identical signal conditioning circuits (amplification and filtering). I am using Option 2 topology where I am using two dual-opamp ICs, each one dedicated to a different sensor. However to optimise cost and power I am thinking of re-arranging as illustrated in Option 1 - so that I can use a different opamp for State 2.

Does anyone have experience using Option 1 in designs? Does this lend itself to internal crosstalk and interference within the IC between the different sensor signals? Does this have a negligible effect or is it best go stick with Option 2?

Sensors are piezoelectric, generating 20 Hz sinusoidal signal.

schematic

simulate this circuit – Schematic created using CircuitLab

\$\endgroup\$
6
  • 3
    \$\begingroup\$ Option 2 would seem to allow more freedom in layout, as you don't have to route both signals to the same chip until the very end. \$\endgroup\$
    – Hearth
    Commented Jun 24 at 13:31
  • \$\begingroup\$ What do "131x" and "22x" mean? Are these the gains of these stages? Do you have op amp part number(s) picked out? \$\endgroup\$
    – Null
    Commented Jun 24 at 13:45
  • \$\begingroup\$ Layout won't matter that much since the sensors are placed right next to each other anyways. \$\endgroup\$
    – enrico
    Commented Jun 24 at 13:46
  • \$\begingroup\$ @Null Yes those are the gains, using TLV2333 at the moment. \$\endgroup\$
    – enrico
    Commented Jun 24 at 13:47
  • 1
    \$\begingroup\$ Option 2 provides better power rail decoupling between the channels. \$\endgroup\$
    – RussellH
    Commented Jun 24 at 19:32

4 Answers 4

3
\$\begingroup\$

Option 2 will have slightly less cross-talk between the two signal channels. For me, that makes it the starting point, especially with a channel gain of almost 70 dB, and there has to be a sound reason for changing to option 1.

\$\endgroup\$
2
\$\begingroup\$

This quite depends on what the sensor signals signify. In particular, whether "signal absent" is an important case to handle. When it is, crosstalk between the processing paths of the two sensors is a problem. When it isn't, "crosstalk" between two stages of the same signal with vastly different processing levels may actually cause more trouble than crosstalk between independent signals at similar level.

Also is the fully amplified noise of the sensor input itself more of a problem than the partially amplified other sensor input? At the gains you are talking about and depending on the meaning and interpretation of the sensor signal, that may happen.

So in my book, there is no one-size-fits-all answer here. Depending on the use case, either architecture may be actually preferable, and there may be some use cases where you want to separate the input stages both from other sensors as well as from high-level signals of any kind.

\$\endgroup\$
1
\$\begingroup\$

I don’t expect serious crosstalk levels, until gross mistakes in routing and bypassing are made. In-chip crosstalk usually is less than 60..80 dB (A haven't found such data for TLV2333, but there's a picture for LTC6241). Total effect highly depends on a lot of details (frequency range, complexity of the circuitry, resistor values used, etc), which are external to the chip. Btw, they apply for both options. The decision also depends on the tolerable level of crosstalk for the application. The ultimate decision is only possible after experimental comparison of the options. Hope, a planning production scale may justify such experiment :)

1

\$\endgroup\$
0
\$\begingroup\$

There is no global "better" here. It's a tradeoff, like with most things in engineering.

Option 1 is good when the first stage has very high gain, and feedback from high-amplitude output of second stage is a concern. Channel separation is worse, output-to-input feedback is better.

Option 2 is good when the first stage has moderate gain. Channel separation is better.

There is option 3 for the first stage:

schematic

simulate this circuit – Schematic created using CircuitLab

This improves voltage noise when the first stage has very high gain. The impedance of the input source must be low enough not to convert the input current noise of U1 into voltage noise. Feedback and combining resistors also must be chosen to have thermal noise >10x lower than the op-amp noise for there to be any benefit.

Expected voltage noise on the output is \$\approx\sqrt{2}/2\$ (70%) of the noise the same circuit would have with just one op-amp.

\$\endgroup\$

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.