AES 2006 – Henry Ott, Saturday

The big event for me Saturday was seeing an all-morning talk by famous author Henry Ott.  I quoted Henry in my Sept 28 feature story and he sure delivered a great presentation.  This was an abbreviated presentation he usually charges hundreds of dollars for.  You could see the whole AES program for that or just pay 35 bucks to see Henry.  This was part of the effort of Bob Moses to bring hard-core engineering classes to the AES.  Expect more of this next year when AES is in New York.

Henry’s talk was about how to lay out mixed signal boards for low EMC and best signal integrity as well as making sure the digital circuits do no interfere with the 18 bit analog circuits.  He made the stunning point that semiconductor manufactures that tell you to cut up the ground planes and connect them under the A to D chip are wrong. Like me, he recommends one continuous plane.  First you partition the board right.  But, and this is huge, you have to have routing discipline to make sure that everything works right.  Henry talked from basic principles and avers that those principles will remain immutable forever.  He did say we are allowed to disagree on his specific implementation but that is another issue.  He mentioned an article he did in the 2001 Printed Circuit Design magazine.  Henry maintains that we have to learn how high frequency currents flow to lay out a good board.  Digital signals always work the way you expect them to but analog signals rarely do.  He says to:

• Return Currents locally and compactly.Those return currents are just as important as the signal side.
• Only use one reference plane.
• It is always better to have a single contiguous plane, unless it is a medical product that demands isolation.
• Remember that two planes hooked by a wire is a dipole antenna.
• Smaller flux copings mean lower inductance.
• Always use small loops.
• HF into a plane will always return.

Like a professor I had at GMI, Henry does not like the word ground unless you are really talking about earth ground, which is rare.  Where is the ground plane in a satellite?  Henry prefers Return Plane or better yet, Reference Plane.  Another good observation Henry made came from James Clerc Maxwell.  Fields produce currents, not the other way around.  By controlling fields you can insure the currents do not get created in the wrong place.

Henry went on to show some previously published graphs that showed how a microstrip line has current in the ground plane for a width of 20 times the distance from the trace to the plane (20H).  This accounts for 97% of the
current.  If you use stripline that has a plane on both sides as opposed to microstrip then 20H will account for 99.999997% of the current in the plane. Lesson—use stripline whenever you can.  Henry produced some new math done by a collaborator that shows the current for asymmetrical planes such as in a four-layer board with the planes on the outside and two signal layers on the inside. Very cool stuff since the now equations converge to the older ones in the microstrip and stripline cases.

Henry did say you could split the plane if you feel you must. If you do and you have to cross the split then:

• Build a bridge—a slab of copper at least 20H wider than all the traces that cross.
• Transformer couple
• Optic isolate
• Magneto-restrictive isolation
• Balanced differential (LVDS) signals.

There was way too much good stuff to cover here—buy his book for a hundred bucks and study it well.  After his talk went over to Chevy’s for a great lunch.  Bill Whitlock of Jensen Transformers was there as well as Jim Brown, a consultant with Audio Systems Group.  At lunch Henry explained the problems of an LVDS signal pair crossing a split.  Fascinating stuff I will try to write up later.