I consider using an isolated BNC connector for a precision application, in which the BNC Shell will be either grounded via an internal switch or used as a Guard electrode. The motivation behind this is to avoid the cost and "nicheness" of Triax connectors. The circuit looks like this:
U1 is a buffer amplifier with high impedance. Signal frequency is DC to kHz, source impedance can be from low to >GOhm. Therefore, the two different operation modes of the Shell. If Switch W1 is closed, the BNC would work as a regular shielded, single-ended connector, which is the main purpose. For very high source impedance inputs to be measured, a guarded input is needed. Opening switch W1 allows this.
When using the guard function, of course the ungrounded cable is a fasttrack for common-mode noise into and out-of the chassis bringing all sorts of EMI problems. The answer of Dan Mills to this post suggests that isolated BNC connectors are a bad idea for EMI. I suppose, the same would apply to other measurement devices such as table-top multimeters using unshielded banana connectors for example. So how do such devices tackle EMI with the unshielded cables?
a) One option I thought of was tackling this with a common-mode-choke. But: Won't that commom-mode-choke hinder the guard drive function. And when using the grounded shell, won't that same common-mode-choke couple the chassis noise into the sensitive circuit?
b) Another idea would be to RF-bond the BNC Shell to the chassis anyway, leaving the guard function only functional for low frequencies (sub-kHz). But I wonder how to achieve this? A capacitor is baaad at RF-bonding from what I have read.
I would be happy to hear your thoughts on my two ideas (a) and (b), or about ideas of your own on how to make isolated BNC ports work well EMC-wise.