[TowerTalk] isolation and SWR

[Jim Brown]

On 9/10/2012 8:03 AM, Pete Smith N4ZR wrote:
> Is there even an approximate relationship between these various SWR 
> numbers and the degree of isolation afforded by the two switchboxes? 

Something would have to be seriously broken for that to be the case.  A 
more likely cause would be the stray reactances in the signal path.  
I've recently measured several with a vector network analyzer, and 
seeing a lot of strays. When I open them up,  I'm finding that the 
manufacturers are failing to carry the return path on the circuit board, 
instead using the chassis as a return. Ther is, for example, no direct 
path between the coax connectors and the circuit board, so return 
current must flow from the connector via the chassis to wherever the 
circuit board contacts the chassis.

One box uses a two-layer board with a "ground" (return) plane on one 
side, which, when done right, causes return current to flow on the 
ground side directly under the trace. This greatly reduces crosstalk and 
stray inductance. But they break that trace at multiple points under the 
signal wiring, which completely defeats the ground trace by forcing 
return current to find a path around the break  The result is a lot of 
stray inductance added to the signal path, both increasing SWR and also 
increases crosstalk (that is, reduces isolation). They sort of get away 
with it on the lower HF bands, but these boxes look increasingly nasty 
on the higher HF bands and some are unusable on 6M.

  In addition to the stray inductance, most multi-way switch matrices 
will have a fair amount of stray capacitance simply by virtue of having 
wiring from multiple relays connected to the output bus. Thoughtful 
designs will use layouts that minimize the strays on the highest 
frequency output ports.

> And probably an even dumber question.  If you have two devices in line 
> and each has (say) 30 dB isolation between a line in use and an unused 
> one, both of which pass through them both, is the resulting isolation 
> 27 Db?  Some other number? 

Probably not easily predicted.

73, Jim K9YC