Fri May 22 15:30:18 CEST 2015


Basic idea: interrupt a loop by placing a low-value resistor in the
loop path (e.g. 5 Ohm) separating internal and external grounds, then
use a differential input/output to reference the external ground.

It seems that the cause is *not* transformer coupling which is what I
originally thought.  From Appending C in [1] the way to explain it is:

  1. power supply currents (operating and decoupling "spikes") are

  2. relative to size of current, the power supply resistance is not
     negligible, creating ground levels that are not the same

  3. generally, the impedance of a shield cable connecting two such
     grounds is large enough to not not pull the grounds to the same
     potential, i.e. a voltage drop is present.  this signal is the
     audible noise.

[1] http://www.ti.com/lit/an/sloa143/sloa143.pdf