The evils of current loops.

   This was originally "penned" in reponse to a thread on "Keeping clock noise out of the A-D converters" on the Motorola 68HC11 Applications mailing list.

   A correspondent advised shielding lines to the ADC pins by "wrapping" analog ground traces around them on the PCB but "don't make a circle. i.e., ground still connects to one point."

  Don't make "hum loops", eh?   This is apparently among the worst understood precepts of analog design, so perhaps a little thought is in order.   If this were true, that spurious and evil currents build up in every little circle of conductor, would it not by the same token be important to have the shield on co-axial cable slit all the way along?   Common sense (un-common?) says this is nonsense.   And so it is for "hum loops".

   The story is this:   A "hum loop" consists of a circuit where there are two or more paths between a point sensitive to an EMF and its ground reference.   The point in question is itself a ground return for a signal path, so an induced EMF at this point will be added to the signal and appears as interference at the input of a respective amplifier.

   In this circuit, magnetic fields may now induce an EMF across each of the ground paths according to their length and orientation to the field.   The same magnetic field may also induce an EMF across the signal path.   In addition, DC or AC currents sharing the ground paths will cause voltage drops due to resistance.   This may also be considered a variant of "induction" for the present considerations.

   One ground path may have an induced EMF which is closer in magnitude and phase to that in the signal path, than the other ground path.   This will therefore tend to cancel the interfering EMF.   This is called the "right" ground path, while the other (or indeed, others) is by a simple process of elimination, called the "wrong" path!

   Since (but only if, it need not be so;) the EMF induced across each path is different, and they are connected together, a current will flow.   This will, due to resistance in the paths, cause different potentials to exist across different sections.   It will cause the (DC or AC) potential at the EMF-sensitive point to be some value between that induced in the right and wrong paths.

   If indeed the value of EMF induced in the right path is a good replica of that induced in the accompanying signal path, then this altered EMF due to the existence of the wrong path will cause interference which would not occur if the wrong path was removed.   The interference will nevertheless be less than would exist if the right path were removed instead, because while connected, the right path is shunting interference current from the wrong path.

  The meaning of the "if"s in the above:   In fact, if the signal path takes a course partway between the two ground paths, the "average" return path may well be the best balance to it.   This is most often the case on PCBs where a "loop ground" around the outside of the board with extra crossings where possible is the best shield against externally applied fields (both electric and magnetic).   Of course, a shielding "can" extending this surrounding into the third dimension is even better, grounded at as MANY points as possible.   The analysis is slightly different for fields generated on the board itself (which is to say: a shield of any sort won't shield from interference produced within its own enclosure, will it?).

   But surely you know this already, that it was "common sense" to ground the shield can at more than one point, didn't you?(?)

  The correspondent summed up: "I just have to have faith that a well designed circuit board is going to work!"

  Indeed you do and indeed, it will.   And (you also have to have) the common sense to design it.   Use wide and many earth tracks, placed between the interfering signal and the sensitive sections.

   What's all this mark-up about EMFs?   Well, an EMF refers to an open-circuit voltage.   I'm just stressing that in talking about induction, we refer to loose- coupled circuits and are not primarily dealing with the induced currents, but potentials or voltages.   If a loop connection is broken, there can be no current flow, but an induced voltage may appear across the gap and it is this voltage, generally greater than between any two points on a closed loop, which forms a series- added (noise) signal for amplifiers, ADCs etc.