PCB design recommendations for high impedance signals

[bateau020]

in my quest to design a SMU, I'm confronted with the PCB design that allows correct measurement and control of low currents, while maintaining an acceptable bandwidth. As a result, signal shielding, parasitic capacitance and leaks become important.

To simplify the circuit I'm planning to design:

Imagine an AC voltage source, having a variable source impedance (up to 100 MOhm), driving a low noise/bias/capacitance precision opamp (OPA140,OPA145,OPA189,... with some compensation circuitry that is not really important to this question. This question is about the PCB, not so much the circuitry).

I'd like to have a high as possible bandwidth (reasonable, but I'm talking about max 100kHz). Now I know that the input capacitance of the opamps will already put a severe limit to what I can achieve, but the PCB design will also influence it, be it in parasitic capacitance or leaks. I'm just trying to see how far I can go with reasonable means.
I don't want to go into wizard territory with teflon and air circuits.
Reasonable also means: no exotic materials, up to 6 layer PCBs. Prefer to avoid buried vias and many grooves in the PCB

What would be some practical guidelines for PCB design in this use case?

This is what I came up with so far:
* surround the high impedance signals with a low impedance guard that is driven from the signal itself (buffered output signal of the mentioned opamp)
* use thin lines for the signals
* put the PCB in a nicely shielded box (of course)

My main interrogations so far:
* what reasonably priced PCB and practically usable material would be better than FR4?
* what reasonably priced box material would be better than a simple copper foil cladding?
* what layer stacking would be best? (signals and the guard on an outer layer vs putting guard also underneath and/or above, would stitching have any advantage at these low frequencies?)
* any other guidelines?


EDIT: no, 100kHz is not what I'm expecting. It's just the upper limit. I'll be happy with anything that I can get at a reasonable price.

[moffy]

100M Ohm and 100kHz BW means a total input capacitance of 15.9fF, that's some pretty impressive bootstrapping. I've seen the BF998: https://www.infineon.com/dgdl/Infineon-BF998-DataSheet-v01_00-EN.pdf?fileId=8ac78c8c7e7124d1017efcdc67410783
used as a bootstrapped high impedance input device for a scope input.

[moffy]

While not directly applicable, this forum post shows an effective high impedance bootstrapping technique: https://www.eevblog.com/forum/projects/help-me-understand-photodiode-amplifier-jfet-bootstrap/msg5070655/#msg5070655

[David Hess]

what reasonably priced PCB and practically usable material would be better than FR4?

If you find something, let me know.  I think Tektronix used polysulfone at one point in place of FR4 for high impedance circuit boards.  Maybe Kapton would be good, although it has poor mechanical properties.  I always wondered if any of the various Rogers substrates are suitable for high impedance circuits, but maybe they are terrible.

Some substrates, including FR4, display circuit board "hook" which will prevent fast settling of high impedance circuits.  Hook is a change in dielectric constant at low frequencies, so it will need to be considered in your application.

what reasonably priced box material would be better than a simple copper foil cladding?

I like using doubled sided copper clad board to make shielded boxes.

what layer stacking would be best? (signals and the guard on an outer layer vs putting guard also underneath and/or above, would stitching have any advantage at these low frequencies?)

Minimizing parasitic capacitance means having one layer and no ground plane.

Guard traces can use vias to guard against leakage through the bulk of the substrate.

[Sensorcat]

what reasonably priced box material would be better than a simple copper foil cladding?

I like using doubled sided copper clad board to make shielded boxes.

Adding a soft magnetic layer, like the tin cans found in many scopes and RF equipment, would not hurt.