[Elecraft] K3 RF Feedback Problem

[Paul Christensen]

> "I know it is good engineering practice often to isolate internal 
> circuitry from chassis ground. Lab quality signal generators, for example, 
> usually tie the shell of their RF output connector to the chassis via a 
> (typically) 100 ohm resistor..."

Moreover, designers of instrumentation-grade circuitry often tie multiple 
returns to one common chassis ground.  I know of one test equipment product 
that utilizes five separate return path references to ground (e.g., digital 
ground, analog ground, power supply ground, control systems ground, etc.), 
but in each case, the return path currents are controlled and reasonably 
isolated for each of the stated purposes.  Nevertheless, each of the five 
returns circuits in this example are tied to exactly one common chassis 
point.  But the current for the respective circuits circulate among their 
own returns and not with others.  They share just one low-impedance common 
point.  By contrast, ground loops exist in other equipment where one common 
return is used with multiple circuit bonding points and various DC, AF, and 
RF current are allowed to share and flow past sensitive devices.

For example, if there's a high current return path in a control circuit 
(e.g., open collector to a solenoid), then the ground returns on a sensitive 
3-stage-input instrumentation op-amp circuit (e.g., a high-gain mic pre-amp) 
should not be sensitive to the current demands of the solenoid.  The 
addition of a 100-ohm resistor in your example may be one such use of 
limiting return current on  circuit paths *within* a piece of equipment.

But what is occurring with the K3's mic returns is a total isolation of a 
common, grounded return path of the front and rear panel mic jacks at RF. 
Yes, from D.C. through audio frequencies the return path is unimpeded as a 
result of L4 and L7.  But in the presence of RF, at some frequency between 
AF and RF, the ground reference looking into the K3 mic connectors is 
completely lost.

> "Perhaps it has something to do with using balanced line audio sources. 
> Minimizing ground loops can be a headache when one has multiple audio 
> devices, powered by different power supplies."

Even in balanced audio systems, the same rules apply.  Had the K3 been 
designed with a truly balanced, 3-stage instrumentation input for its mic 
pre-amp, the inclusion of L4 and L7 on the shielded return paths would have 
the same effect.  The saving grace in an instrumentation-input circuit (or 
in the alternative, an audio transformer input) is the inherently large 
common-mode rejection ratio (CMRR) across a very broad frequency span that 
limits the presence of RF on a twisted-pair audio line, even in the total 
absence of the cable shielding.  For nearly 100 years, the Bell System and 
its progeny have used unshielded twisted-pair balanced audio systems in the 
presence of outrageously-high RF fields with no measurable detriment to 
performance in many instances.

Paul, W9AC