0
\$\begingroup\$

enter image description here

(Image is my own creation.)

I notice that an inverting amplifier can be very sensitive to capacitive and inductive noise coupling on a PCB. I was wondering whether the placement of the R1 resistor plays any role in that.

Say I have a certain length of trace from Vin to the negative op-amp input, does it make sense to put R1 as close as possible to the op-amp on the PCB, because the trace between Vin and R1 is less sensitive to noise coupling (assuming Vin is provided by a low impedance voltage source like another op-amp)? Or does it not matter?

\$\endgroup\$
5
  • 1
    \$\begingroup\$ Should minimizing "noise" by using a "ring guard" around - and + inputs ? \$\endgroup\$
    – Antonio51
    Commented Oct 12 at 16:34
  • \$\begingroup\$ @Antonio51 Thanks! I think that's a strategy if Vin is close to the op-amp, right? Because in my case Vin is about 15 cm away on the PCB, and the long trace in question is picking up noise. \$\endgroup\$ Commented Oct 12 at 16:39
  • 1
    \$\begingroup\$ Not only is there a potential noise problem but capacitance at the inverting input can cause instability of the amplifier. I once did a design where I accidentally didn't keep that trace short and the amplifier oscillated. \$\endgroup\$ Commented Oct 12 at 17:13
  • 1
    \$\begingroup\$ Wouldn't the trace between R1 and the amplifier input now be the problem? I recommend shielding techniques and keeping the lumped elements as close as possible to avoid unmodeled phenomena. \$\endgroup\$
    – a360pilot
    Commented Oct 12 at 17:26
  • \$\begingroup\$ @a360pilot Sure, but it's possible for me to keep that trace extremely short, hence my question whether the physical location of R1 would matter. Right now it's close to Vin and there's a lot of noise coupling from other parts of the PCB. \$\endgroup\$ Commented Oct 12 at 18:51

1 Answer 1

2
\$\begingroup\$

The issue you've heard about is that noise couples easily to any node that has a high impedance to ground at the noise frequency. Here, that would be your inverting input, as you say.

There are a couple things you can do. The simplest is to make the high-impedance node less high-impedance; right now, its impedance is something close to 50 kΩ, but you can reduce that by an order of magnitude by using 10 kΩ resistors instead of 100 kΩ. Of course, this increases the loading on your op amp, and reduces the circuit's input impedance, so you can't just go all the way down to using 1 Ω resistors.

Another method, commonly used in high-precision circuits, is to create a guard ring around the high-impedance node. This consists of a trace surrounding the node that's driven with a low impedance to a fixed voltage (ideally driven to the same voltage as the high-impedance node's voltage, which also eliminates leakage currents, but any fixed voltage is better than nothing). This shields the high-Z node from outside interference, because the guard ring will absorb the incoming noise instead of the high-Z node. For this circuit, the guard ring can simply be grounded.

Keep the high-Z node as small as possible. This reduces the coupling to external noise, and makes the guard ring easier to implement. Here, the guard ring would need to be positioned like this:

schematic

simulate this circuit – Schematic created using CircuitLab

It needs to cover the right side of R1, the left side of R2, and the inverting input of the op amp. That means that ideally, R1 and R2 should be as close to the op amp as possible, otherwise you'd need to extend the guard all the way down a cable or something.

It can also help to have the guard ring extend to below (and even above, if possible) the high-Z node as well, on a multi-layer PCB. If this is impractical, or adds too much capacitance to the node, you can also try putting cutouts in power planes below the node to reduce power noise coupling into the node.

\$\endgroup\$
5
  • \$\begingroup\$ Thanks for the thorough reply! I'll give this a try. Just to be sure, by "high-z node as small as possible" you mean the physical trace length between the op-amp and R1, as per my question, right? So I'm not wrong with my suspicion that R1 should be close to the op-amp? \$\endgroup\$ Commented Oct 12 at 19:14
  • 1
    \$\begingroup\$ @one_three_three_seven Yes, both as short a length and as small an area as possible. \$\endgroup\$
    – Hearth
    Commented Oct 12 at 19:44
  • 1
    \$\begingroup\$ As you hint at in the first sentence, the inverting node a low-impedance point, until the open-loop gain starts to roll off or the output saturates. So the concern would be with noise outside the bandwidth of the amplifier. If the OP's noise of concern is lower in frequency than the bandwidth of the op-amp, the coupling may happen through a different point. \$\endgroup\$
    – John D
    Commented Oct 12 at 21:03
  • \$\begingroup\$ @JohnD Thanks a lot, I think it's clear to me now. The noise is well within the bandwidth, so I should be fine with the suggested changes. \$\endgroup\$ Commented Oct 12 at 21:10
  • 1
    \$\begingroup\$ @JohnD Yes, I kind of forgot about that aspect as I was writing the rest of it, so it's a bit unclear. I can't fix it right now, but I'll try to get to it soon (unless someone else edits to clarify it first). \$\endgroup\$
    – Hearth
    Commented Oct 12 at 21:12

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.