Eric Bogatin's Blog: What I Learned This Month

Next public classes: Essential Principles of Signal Integrity and Advanced Signal Integrity Design, and Multi GigaBit Design, Sept 29- Oct 7 in San Jose, CA.

Next No Myths Allowed Webinar, “Selecting Capacitor Values for the PDN”, presented on Dec  9, 1 pm EDT.

I just learned from Colin Warwick, my buddy at Agilent, of a series of workshops Agilent is presenting on Fast Channel Simulation and Statistical Eye Diagrams.

If you do channel simulation and need to check very low bit error rates or need an accurate measure of the jitter distribution, you may have to simulate millions of bits. If you are also using the eye diagram simulation to explore design options, you may have 10 to 50 different designs to evaluate.  If you’re not doing it smart, you’re simulation is probably still running.

Agilent recently introduced some clever techniques based on the transient response of the channel and statistical analysis to synthesize eye diagrams for LTI (linear time invariant) systems which can show BERs well below 10^-12 in a few minutes of simulation time.

This process could revolutionize the way you explore new design options. To learn more about it, check out the workshop, coming to a city near you.

Next public classes: Essential Principles of Signal Integrity, Advanced Signal Integrity Design, and Multi GigaBit Design, Sept 29- Oct 7 in San Jose, CA.

Next No Myths Allowed Webinar, “Stack-up Design for Differential Pairs”, presented free on Sept  16, 1 pm EDT.

In all the lectures I give on differential pairs and differential pair design, I always say that you should forget the words “differential mode” or “common mode.” Using these terms confuses the use of the words and, I think, significantly confuse our intuition and make it even harder to understand this already confusing topic.

The words we use influence the way we think and the models we carry around inside. Instead of differential mode, we should be using the words differential signals and common signals, or odd mode and even mode.

If you are consistent  using these terms, you will keep clear the concepts of the differential signal component and the common signal component, as properties of the signals, and the odd and even modes as properties of the interconnects.

An EMC engineer buddy of mine told me my quest to change the words we use in the industry is like Don Quixote, tilting at windmills, which Wikipedia defines as “fighting unwinnable or futile battles”.

OK, so I have given up on my unwinnable quest to change the industry, but I still think if you want to calibrate your intuition and really understand what goes on when a differential or common signal propagates down a differential pair and it sees a differential impedance or a common impedance, don’t use the words differential mode or common  mode. It is not needed and will screw up your intuition.

If you want to really understand differential pairs and how to design them, join me for our next No Myths Allowed webinar, Sept 16, 2009. Sign up on www.bethesignal.com

Hope to see you in cyberspace!

Next public classes: Essential Principles of Signal Integrity and Advanced Signal Integrity Design, and Multi GigaBit Design, Sept 29- Oct 7 in San Jose, CA.

Next No Myths Allowed Webinar, “Stack-up Design for Differential Pairs”, presented free on Sept  16, 1 pm EDT.

I was not able to attend the 2009 IEEE EMC Symposium held in Austin last week, but as a member of the EMC Society, I got a copy of the proceedings.  I was delighted to find a number of really exciting papers. Some introduced new ways of modeling or evaluating interconnect structures like vias and the power-ground cavity, while others recounted simple experiments to verify or quantify well know solutions.

Over the next few months, I will try to review some of the more relevant paper to give them a little more visibility.

Two papers in particular showed very simply the problem with connectors in cable assemblies. In the ASID class, I spend a whole module on EMI problems and solutions. While external cables, even shielded coax cables, may act as the radiating antenna, its not the cable that is the source of the noise voltage that drives the common currents that radiate; it’s the connector.

In the paper, “Relationship Between Connector Contact Points and Common-mode Current on a Coaxial Transmission Line,” by Hayahi, Mizuki and Sone, of the Tohoko University in Sendai, Japan, the authors illustrate this principle with a simple construction.

They built a two-section coax cable, with the sections connected in a region where they could vary the number of connecting wires between the shields. Four configurations were constructed, with 1, 2, 3 and 4 connecting wires between the shields.

What I teach in the ASID class is that the more the connector allows the return current to flow symmetrically around the signal current, the more cancellation of external magnetic field lines and the lower the total inductance of the return path. In these four different return path configurations, the more the connection looked like a 360 degree, symmetric path, the lower the common currents around the cable and the lower the emissions.

Though this was not a profound conclusion, nor unexpected, it was a simple, beautiful example of this principle that the connector is often the weak link in radiated emissions from coax cables.

In the paper, “Effectiveness of Shield Termination Techniques Tested with a TEM Cell and Bulk Current Injection,” by Bradley and Hare of NASA Langley Research Center in Hampton, VA, the authors show by direct measurement the radiated emissions from cable assemblies with different shield termination schemes.

When a pigtail or drain wire is used to provide the return path connection, the radiated emissions are 20 dB higher than with a connection that goes 360 degrees around the signal current. In fact, the authors compare four different ways of making the 360 degree termination, using foil, clamping the braid, soldering the braid and an elaborate clamp to the backshell with additional braid overlapping the cable shield.

They find that all the methods that provide 360 degree connection work equally as well.

These two papers did not rock the world, but they demonstrate in simple, clear experiments this principle that when the return currents are not symmetrical around the signal path, the noise across the total inductance of the return path will drive common currents which will radiate.

If you want to learn more about solving EMI problems by understanding the root cause of the problem and the essential principles on which the solutions are based, check out the classes and lectures listed on our web site.