Archive for September 2011

Implementation of EMC Fixes

September 25, 2011

This post will discuss the last of my four tenets of EMC troubleshooting (see my 8/28/11 blog), Fix Implementation. When dealing with frequencies in the tens or hundreds of megacycle range, you are dealing with rf frequencies and cannot get sloppy in the way you implement EMC fixes. For example, if you believe that adding a capacitor at a certain point in your system will fix a problem, tack soldering the capacitor in with 2″ or 3″ leads, may work at audio frequencies, however, it will be completely useless at rf frequencies. The inductive reactance (impedance) of two half-inch long leads on a capacitor, at 200 MHz, will be about 20 Ω. Many times capacitor lead lengths must be 1/8″ or less in order to be effective.

Therefore, at the frequencies involved with most EMC problems, a fix must be implemented almost perfectly in order to be effective. This is often very difficult to do on an existing product, having many constraints. A poorly implemented fix can have no effect at all, thereby, leading one to draw the wrong conclusion and leading you astray.

I learned this principle many years ago while troubleshooting a 600 MHz emission problem. I was convinced that the emission was coming directly from one of the IC chips on the board. The IC was about one inch square. I fashioned a shield out of copper tape to cover the offending IC,  and grounded it to the PCB ground plane (with very short wires) at the four corners. The EMC measurements were repeated with absolutely no improvement, which lead me to believe that I was wrong about the IC being the source of the emission.

After 4 or 5 hours of trying all kind of other fixes, non of which worked, I (out of frustration) revisited my original idea that the IC was the source of the emission. This time a made a similar shield to cover the IC, but  grounded it at eight points, the four corners plus half-way along each side. Wala! The emission dropped 6 dB–problem solved!

Sometimes a fix cannot be implemented perfectly enough on an existing product. For example, if in the above case, there were no accessible ground points on the PCB to connect the shield to. In these cases you may just have to redesign the PCB, based on faith and knowledge that it is the correct thing to do, in order to implement the fix correctly. More often than not, it works.

This ends this series discussing the four basic tenants of EMC troubleshooting. Just understanding these four principles should make you a better EMC troubleshooter.

Advertisements

The “Kill it Dead” Strategy

September 18, 2011

In my 8/28/11 blog I listed the four basic principles of EMC troubleshooting, the third of which I will discuss in this post. This principle gets its name from a short one-page article written by Scott Roleson in 1992.* I often refer to this method as the “Giant Step versus Baby Step” approach to EMC troubleshooting. The analogy being, if I want to walk across a room, it will take me longer to do it if I take baby steps, than if I take giant steps.

When troubleshooting an EMC problem, especially on an existing product, the client always wants to make the minimal, least costly, change possible (taking baby steps). My approach is to do whatever is required to make the product compliant first, without regard to the cost or complexity (taking a giant step), then go back and refine the fix to be less costly and/or complex. In the meantime the product is compliant! Many times, potential EMC solutions are not tried because they are considered, at first glance, to be too complex or costly.

For example, let’s consider the case of radiation coming from a metallic enclose which has six different apertures. The “baby step” approach to troubleshooting this product is to cover one of the apertures with copper tape and then repeat the emission test to see if that solved the problem. Nope–that did not work, so let’s put copper tape over another aperture and repeat the emission test again, etc, etc, etc.

The “Kill it Dead” approach, on-the-other-hand, is to put copper tape over all the apertures and redo the emission test, in most cases the problem is now solved in one step. Then you can go back and see if the tape can be removed from some of the apertures without the problem reappearing. One should also see how this approach meshes with the Predominate Effect principal discussed in my  9/11/11 blog.

Another example might be, a four-layer PCB which has many traces that are routed over splits in the power and ground planes. A quick fix would be to add two more solid ground plane layers adjacent to the trace layers, a fix that almost always works but is costly since it makes the four-layer board a six-layer board. This also quickly proves the point that the traces crossing the splits in the planes are the problem. One can then, in many cases, cost reduce the fix by completely relaying-out the four-layer board to avoid the traces crossing the splits in the planes, and accomplish the same objective, without adding the two additional layers. This is a less costly but more time-consuming fix.

I think that you will find by taking the “Kill it Dead” approach, the time required to troubleshoot an EMC problem will be significantly reduced.

In my next blog I will discuss the last of the four tenets of EMC troubleshooting, Fix Implementation.

* Roleson, Scott, ‘The “Kill it Dead” Strategy,’ EMC Test and Design, September/October, 1992.

Predominate Effect

September 11, 2011

In my 8/28/11 post, The Four Basic Tenets of EMC Troubleshooting, I discussed the four basic principles, of EMC troubleshooting, the second of which was the Predominate Effect. In this blog I will discuss the significance of this principle.

An emission at any specific frequency is often caused by more than one source, and/or radiation mechanism– one of which is predominate and overshadows the others. If a fix is applied (such as a ferrite added to a cable, or reducing the size of an enclosure aperture) that decreases the magnitude of one of the non-dominant emission sources, and/or mechanisms, and the emission is remeasured–no improvement is seen, since the reduced emission is overshadowed by the magnitude of the predominate emission. This can lead to drawing a wrong conclusion, that is, that the fix had no positive effect. In order to notice an improvement, the predominate noise source must be fixed first. Therefore, in order to see an improvement, multiple noise sources must be fixed in a specific order–a difficult and time-consuming process.

Many times when troubleshooting I have applied a fix that reduced the emission, but not sufficient to bring the product into compliance with the EMC regulations. Then when I propose a second fix the client says: “No that does not work, I already tried that fix.” But the client tried it before I applied my fix, that reduced the emission, therefore, it probably was not the predominate effect at that time. My response is often “humor me and try it again,” and in many cases it is now effective.

There are basically two approaches to overcoming this perplexing Predominate Effect problem:

One is to always revisit old fixes that did not work, after an effective fix has been applied.

The second, and the one that I usually prefer, is to leave all seemingly ineffective fixes in place until the product is compliant, then remove them one at a time. Using this approach the removed fix, if it has an effect at all, will always be the predominate one. Any fix that does not increase the emission when removed, can now be eliminated.

Just understanding the concept of the Predominate Effect will make your troubleshooting less confusing and easier.

In a future blog I will discuss the “Kill it Dead” Strategy of EMC troubleshooting.

Divide and Conquer

September 4, 2011

In my 8/28/11 blog I discussed the four basic tenets of EMC troubleshooting, the first of which was Divide and Conquer. The objective of this approach is to eliminate some of the components, parts of the system, etc. from being associated with the problem, and then concentrate our troubleshooting efforts on the remaining components, parts of the system, etc. In this blog I will discuss this approach applied to the case of a radiated emission problem, and the case of a conducted emission problem.

Radiated Emission Problem. In the case of a radiated emission problem it is advantages to determine if the emission is coming from the cables, or from the enclosure/PCB. Often the simplest way to do this is to remove the cables and see if the product passes without cables. If it does, the emission is coming from the cables. If not, fix the enclosure/PCB first, then worry about the cables. If the product does not pass without cables, it will never pass with cables.

Of course, when removing the cables, you cannot remove the power cable. Therefore, wrap the power cable around a big ferrite core to suppress the radiation from this cable.

Another approach, not requiring the removal of the cables, would be to measure the common-mode currents on the cables using a common-mode current clamp such as the Fischer Custom Communications Model F-33-1, or if the emissions are above 250 MHz, the F-61 current probe. See Measuring Common-Mode Currents on Cables.

Conducted Emission Problem. The first thing to do when facing a conducted emission problem, on the power line, is to determine the frequency of the failure(s). If it is at the low end of the conducted emission spectrum, the source is most likely the switching power supply or an ineffective  power line filter. If the failure is at the high end of the conducted emission spectrum, the source is most likely the digital logic.

If the emission is from the switching power supply, it is desirable to next determine if the emission is from a common-mode or a differential-mode noise current – since different components in the power supply, and the power-line filter, affect differential-mode emissions and common-mode emissions. Performing the conducted emission test as specified in the various EMC regulations, will not determine this.

We can, however, determine this by adding a differential-mode rejection network between the LISN and the measuring instrument. This will allow us to determine the magnitude of just the common-mode noise, see Distinguishing Between CM and DM Conducted Emissions.

If you make a habit of dividing up your EMC problems as discussed in this blog, your EMC troubleshooting will be easier and will go much faster.

In my next blog I will discuss the Predominate Effect principle.