Guard rings and Virtual ground

[Tj138waterboy]

I have a general question about whether or not a battery powered precision opamp voltage buffer would need a guard ring around all componets such as the perimeter of a pcb so dut and test equipment have a common bond or if the guard should originate from virtual ground of the split power supply of the batteries or a combination of both. The only info that I can find to help is dealing with mostly rf signals. Im posting where I currently have as a layout for a buffered hamon voltage divider with most layout copied from thedefpom youtube channel https://youtu.be/gUikShvjyis?si=tPgrN6vpv3caRvgm mostly from parts 2, 3, and 4.

[moffy]

It is hard to answer your question without a schematic showing what you are trying to accomplish. Guard rings are typically placed around sensitive nodes to prevent leakage current across the surface of the PCB to the node, I'm not sure if that applies here.

[Tj138waterboy]

In simplest form its just going to be a voltage buffer so that multimeter impedance will not load the divider bridge just a little side project to learn and nice tool to have to start volt nutting. May get another bridge for giggles on the same board for 100:1 but all depends on this iteration and parts cost. May even add a transistor to drive output of opamp if needed.

[moffy]

I can't see how the guard ring would work since the supply, Vcc+ and Vcc- is inside the ring, I also cannot see its purpose as it does not appear to be guarding a sensitive node from leakage current. The greatest error I would anticipate, apart from temperature drift of the resistors is thermally induced emfs, but then a sealed screened box with little internal dissipation should stabilise with time.

[ivo]

It's easy to know, when you figure out what the guard ring is actually doing.

A guard ring can be for two different purposes - either to help with EMI shielding, or to reduce current leakage.

In the former, it would usually present a different voltage (not necessarily) and different source with low impedance that can help absorb some of the EMI energy before it gets to the inside trace. Leaves the inside signal trace unaffected.

The latter, you want the guard trace to be at the same voltage as the protected trace, and it should be connected to whichever path has the lower impedance (but same level). Luckily, for opamps, usually they have both their input terminals by design at the same voltage. If you have an inverting opamp, usually the +in would be connected straight to ground (call it "true zero"), and you'd connect your guard to that. The opamp will create its own "virtual zero" at the -in by opamp action, but this might be connected to high impedance nodes. So you draw a guard, connected to the +in/true zero/low impedance net, around the virtual zero net. Since both true and virtual are practically the same voltage (call it 0V, for example), then there is no differential voltage between the signal path and the guard path, and so no PCB current leakage can flow from one to the other (otherwise possible to happen if the guard path was not surrounding, and there was some other separate voltage rail near it).

[CatalinaWOW]

A particular kind of EMI control which may be easier to determine if you need a guard ring is capacitance shielding.  Any traces on your board will have some tiny capacitance to your inputs.  If the signals on these traces fed through that capacitance will interfere with your circuit you need to isolate them.  Distance is one way, a guard ring another. 

[moffy]

A particular kind of EMI control which may be easier to determine if you need a guard ring is capacitance shielding.  Any traces on your board will have some tiny capacitance to your inputs.  If the signals on these traces fed through that capacitance will interfere with your circuit you need to isolate them.  Distance is one way, a guard ring another.

Personally I would call it shielding and use a metal shield, just a preference.

[Terry Bites]

Given the low impedance of your circuit leakage currents are unlikely to have much effect.
EMI susceptibility/ emission will be very low. Thermal effects and bias currents will likely dominate the over all precision when using "zero offset" opamps.

If your comon mode voltage is zero (good idea) then just ground the guard rings.
To reduce capacitance the guard ring will need to be driven with a voltage equal to (as near as possible) as the opamp input nodes.
If there is no dv/dt, no current flows in the stray capacitance
Note that the inv terminal is driven by the the fb resisitor to create zero common mode voltage and may be used to drive a guard ring.
It can be useful to put a pcb shielding enclosure over the attenuator and input opamp.
Input guarding may be helpful in controling leakage currents at the point of measurement. 

See https://www.analog.com/en/resources/technical-articles/layout-for-precision-op-amps.html

[Tj138waterboy]

I was thinking my layout seems more like a faraday cage and it will be mounted on aluminum enclosure so the plated mounting holes will then make the enclosure a shield as well. It would appear im over thinking needing this around the opamp input and probably should just attach the trace to Gnd/-input.

[Tj138waterboy]

You are right about thermal effects probably should be concerning but from simulation I did It would appear that less than 10nA should be flowing for 10 volt input. I would think that shouldn't cause much of a temp rise to be of concern but I will be using 10-15ppm resistors. Would it then be a better option to make a seperate copper pour around each resistor group with vias to bottom layer for dissipation/equilization?