Electromagnetic Interference (EMI)

 

By R. C. (Bob) Scott, P.E.

Inveng LLC

I-ENG-A of Dallas

 

Electromagnetic Interference (EMI) was once an item of interest and concern for only scientists, electronic/electrical, and telephony engineers.  EMI has now evolved into a major topic of concern, discussion, and news coverage for the worldwide populace.  Think of the plethora of recent Toyota problems and the coverage of such problems by USA Today, CNN, and other primary news providers. Talk about coming out of the darkness and into the light.  Electromagnetic Interference can involve Electro Static Discharge, Radio Frequency Interference, Magnetic Interference, and many types of Power Line perturbations.

 

In the late 1970s and early 1980s, all product personnel (including myself) had to get interested in EMI.  Why?  Primarily because the Federal Communication Commission (FCC) limited the amount of radio frequency energy a product could emit, either in the air (radiated) or on the power line (conducted).  Also, we were designing more and more products used in an uncontrolled environment.  At the time, we correctly believed we would be able to protect our products from those emissions and off we went, oblivious to the ramifications to come.

 

For years we had heard the solid state physicists talk about “hole flow” and other theories that made transistors and integrated circuits work.  We proved those fellows wrong.  It was “little blue smoke.”  If you let the “little blue smoke” out of the case, the part ceased to work.  Worse yet, “little blue smoke” is hard to catch.  Getting it all back in the part’s case is virtually impossible (we never succeeded).

 

As time progressed, through various failures and successes, we learned a great deal about EMI and its effects on computers and related integrated circuits.  For example:  Rather than dying, as initial computers did, a computer’s primary mode of failure now is to hiccup.  I cannot tell you how many times a week I must restart my notebook (several in any case) and it has not died yet.  However, I have thought about expediting the process numerous times.

 

Modern automobiles are equipped with many computers in their controls (ignition, acceleration, cruising, braking, etc.).  Actually, the correct reference for these computers is “embedded controllers”.  But do not kid yourself - they are simply small computers used to obtain sensor information and take the necessary actions to control things.  Of course, it is really important that they do not hiccup - especially in cases where the safety of life or the damage to property is concerned.  If you think about the situation, you certainly would not want a cash register to add $100 to your bill for no apparent reason, just because of a small hiccup.

 

We product engineers have become quite adept at designing circuits and systems that can be immune to some very high levels of EMI, as well as temperature, humidity and other elements of an environment.  There are many techniques we can employ and many tests we can perform to prove a system’s functionality preventing hazards to life and property.

 

Even more interesting (an understatement), consider the case of commercial/military aircraft.  Several of the modern airplanes on which we all fly utilize what is referred to as “fly by wire.”  This term is used when there are a number of embedded controllers communicating with each other acquiring data and controlling the operation of the aircraft.  Hiccups are simply not allowed.  It’s hard to park on a cloud and fix things.  Besides, hiccups occurring during flight can be catastrophic.  Modern aircraft avoid occurrence of these nasty hiccups by heavily employing the principles of redundancy, voting, and other “fail safe” techniques.

 

It is not too difficult to design and build products that operate correctly.  But the engineers building and operating such products must constantly keep their thoughts and actions in the pro-active mindset required to prevent and/or correct interference induced failures.

 

Several years ago, we were contacted by a firm that was experiencing problems with a robotic unit they were producing.  The client’s customer required the robotic unit to perform 120,000 cycles without a mistake.  At that time, their unit could do no more than 10,000 cycles.  We suggested a suite of tests for both hardware and software that could identify their difficulties (there is usually more than one).  These tests encompassed things like the various elements of EMI, temperature and humidity, integrity of communications, software performance and so forth.  The robotic product used several embedded controllers with serial communications lines between the controllers.  Also, each controller interfaced with the sensors and actuators required for correct operation.  Fortunately, the entire robot was enclosed with no public access, so there was no danger to life and property.  Failure, however, could be embarrassing.  And, for an expensive product produced in quantity, failure can have an extremely negative impact on the manufacturer’s bottom line.

 

We began with Electro Static Discharge (ESD) testing.  This testing is the least expensive and easiest to perform.  The unit failed immediately, and at a very low level of ESD.  We discovered that the unit was constructed with various pieces of ungrounded metal.  Construction of the unit did position these pieces of metal out of reach of the user, thereby not presenting a safety hazard.  However, the ungrounded metal acted as an antenna and was radiating noise through the unit.  After grounding the various pieces of metal, the unit was able to pass high levels in ESD (25 KV+ static air discharge at about 0.01 joules).

 

Continuing our testing, we moved on to Radio Frequency Interference (RFI).  During these tests, the unit was subjected to high levels of radio waves from about 10,000 Hertz to multiple Giga Hertz impinging at various angles on the equipment.  Again, the unit failed at fairly low levels.  Various methods of shielding were employed to prevent the introduction of such frequency interference (noise) into the system and its various embedded controllers.

 

During the RFI testing, we also discovered that position sensing for the robotic elements sometimes “lost” position (i.e., the position information was incorrect).  After further investigation, we found that this anomaly was occurring because the client, in their design, had allowed for, but not implemented, the use of quadrature for position sensing.  We were able to change a gate array implementing the quadrature position counting scheme.  By now, the unit was beginning to perform fairly well.

 

Of course, the client by now was beginning to realize the advantage of performing such tests and really wanted to continue the testing.  So, next came testing for power line perturbations.  There are many tests involved in this type of testing which introduces noise into the unit via conduction on the unit’s connection to AC mains.  The unit passed through these tests quite well, including the test for lightning (a pulse of 6,000 volts at 3,000 amps).

 

An examination of the communications between the embedded controllers and the protocol used for that communications link indicated that the scheme did not possess a robust error checking routine.  The scheme was, however, a standard in the client’s industry.  Because all the communications lines were contained within the unit and additional protection from noise introduction was provided by the earlier changes, we were able to prove the unit operated well in spite of the error checking issue.

 

Because construction of the unit incorporated a controlled environment, temperature and humidity testing was performed with no failures, as we anticipated.

 

After all tests were completed, and associated corrective actions implemented, a tremendous improvement was noted in the operation of the units tested.  The last I knew (a year or so ago), the units were at a count of 250,000 cycles and still going without failure, twice the customer’s original requirement.  So, with all the hoopla concerning EMI and its associated problems, we proved, through proper design, testing and corrective actions, those problems can be averted.  It can be done.

 

Electromagnetic Interference