[time-nuts] Query about insulated coax panel connectors

[Charles Steinmetz]

Corby wrote:

> My NBS 106D DMTD originally had all BNC connectors but after finding out
> BNCs are not the best for DMTD units I retrofitted it with insulated SMA
> connectors.
> I used nylon washers on both sides. See PIX.

Note that this can cause problems with RF noise and interference (both 
ingress and egress) if it is not carefully done.

The problem is that RF can get into (or out of) the enclosure on the 
shield of a connected cable, on the signal conductor(s) of a connected 
cable, or (usually) both.  One needs to bypass (short/shunt) the RF 
energy on the cable shield and and on the signal conductor(s) to the 
instrument's RF shield (i.e., the conductive enclosure) *without* 
causing appreciable ground loop currents to flow between the instrument 
and other instruments and devices connected to it  (DC and LF -- usually 
the first 10-30 harmonics of the mains frequency ).

This is a much more involved topic than can readily be covered here, but 
there are some fundamental principles.

Generally speaking, it involves (i) connecting the isolated cable 
shield(s) to the instrument enclosure with low-impedance RF paths (e.g., 
capacitors) *very close* to the connectors; (ii) introducing 
high-impedance RF paths in series with the signal conductors (e.g., 
series resistors, chokes, ferrite beads, and/or common-mode RF chokes); 
and (iii) bypassing the signal conductors with low-impedance RF paths 
(e.g., capacitors) *very close* to the connectors, on the instrument 
side of the high-impedance series RF element(s).

As the frequencies of concern get higher, connecting the isolated cable 
shield(s) to the instrument enclosure effectively with low-impedance RF 
paths gets harder and harder.  I often use two or three capacitors 
equally spaced radially around the connector as a useful approximation. 
  Leads must be kept *short* to minimize the effective antenna loop -- 
2mm or less for high VHF to low-GHz signals (or use SMT parts to build 
out the connector-to-enclosure link).

There is another approach, which I often prefer.  You can bond all 
connector bodies (cable shields) together on the smallest area of the 
enclosure (RF shield) possible, then make sure you don't make *any* 
mistakes in the entire system that create ground loops.  When done 
correctly, this method can often provide superior results.  I'll try to 
post a general overview of that solution in the future, modeled on the 
above -- but there are far more variables and opportunities to go wrong 
on that side of things, so the designer must be correspondingly more 
diligent.

Best regards,

Charles