EMC Failure – What Now?

Steps to get a product compliant with EMC requirements.

If you are going to try and fix the EMC Failure yourself, you might be lucky and the one of the items in the “Easy Items” list might fix your immediate issues. Most likely the lab already tried them. The heading more “Difficult Items” are not really quick fixes, but could resolve things fairly quick. If you get to the “More Painful Items”, you are probably better off hiring one of the many EMC consultants. Only 10% of products pass on the first trip to the lab without modifications, another 50% pass with minor changes on the first trip. If you would like to get your products into the 10% that passes on the first visit to the lab, I have a presentation that can help you achieve that. See details on the Services page. If you want to know more about the failure rates see my home page.

Easy Items:

Use only high quality external cables such as TrippLite, Cables to Go, Dell or HP, or your own if you have good control of the vendors.
For ethernet cables uses Cat 6 (shielded) or Cat 8 cables (Cat 8 as it has multiple shields)
If you go to labs on a regular bases, replace all your test cables / Accessories at least once a year. They do wear and can have shields fail.
Internal wiring harnesses should be twisted and routed away from internal RF sources, such as switching inductors, microprocessors, coils, transformers, etc. Internal cables need a signal return for twisting to work. I have dropped levels >20 dB, buy changing harness.
Proper placement and values of bypass capacitors. One value used for all capacitors in not a good thing. Larger values are not effective at high RF Frequencies. The wrong value at the wrong place can cause a capacitor to go into resonance and significantly increase emission levels.
Make sure painted covers have screws with lock washers to penetrate the paint and provide a good connection between metal parts.
Make sure any internal flex circuits going between internal components have good signal returns. If they are high speed they likely need one side to be a signal return to create a transmission line.
Place ferrite disks on top of high speed ICs. The disks are available with adhesive and are easy to add.
If it’s a single chip micro, a number of the single chip micros allow control of drive power on their external address lines, etc. The drive is controlled by software. Turn the power down as much as your system can accept and run reliably. Slow clocks down if you can. You can soften edges with capacitors.

More Difficult Items:

Change to Spread Spectrum clock oscillators if possible. Some are programmable as to what frequency they run on. So it is possible to have one part used in different products running at different frequencies. They get programed on the test fixtures during assembly test. That can reduce costs due to increased volume of a part.
Make sure you do not have harmonically linked oscillators. For example don’t have multiple 10 MHz oscillators. Or a 10 MHz oscillator and a 20 MHz Oscillator. 100 KHz difference in frequency is sufficient in preventing them from syncing and adding to the RF Emission levels.
Increase the copper thickness of your internal planes. I like 2 oz planes and some motherboard companies (Gigabyte and MSI) now advertise they use 2 oz planes on some of their motherboards. One one product we made the change to improve performance of motors, but found that not only did it help the motors, but the RF Emissions dropped over 6 dB.

More Painful Items:

Look for what I call Unintentional Antennas. What causes them you might ask? Engineers when designing PWBs tend to think in logic levels such as 0’s and 1’s. The catch is those logic levels are voltage levels. To transition from one voltage level to the other voltage level, requires current to flow. Current always has to return to the source of the current. Lack of controlling the path of the return current, is what causes the antenna effect and radiated signals. Control the path of the return currents and there will be no antennas. Traces are either transmission lines or antennas. See http://www.artscipub.com/simpleton/simp.quickie.asp to see the importance of the return path. If you run a EMC probe down a trace and notice that the amount of RF you are seeing on a trace is changing as the probe moves down the trace, that is because the trace is an antenna and not a transmission line.
- To start with, most people refer to planes as VCC, VEE, Ground, etc. Quit doing that, they are current return planes from a EMC point of view. What most people call Ground Planes should be called Zero Volt Plane. Envision your planes as a massive resistor matrix see: Circuit Board Plane Resistor Example. Where you place power sources and power devices impacts voltage losses across the network.
- The return current is a RF current and RF Travels by skin effect. That means a copper plane has two signal return paths, top surface and bottom surface. That makes tracing the path of the current more interesting. The return current wants to follow the signal’s source trace. It prefers to be right below it.
- One of the things that has a bad impact on the quality of the return current path are what EMC Consultants call moats, which are breaks in any of the planes. In all the boards I have looked at that have moats, only once was a moat used correctly. So if you have moats, that is likely why your product failed EMC tests.
- With VIAs they tend to have a clearance to prevent shorts. If you have a high speed trace going between two VIAs, the clearance of the VIAs can affect the return currents path.
Verify good routing of clock lines, high speed lines must be the shortest and best routed traces. Lower address lines are more important than higher address lines, then data lines. Make sure that the impedance of the trace matches what the semiconductor calls for.
Look for T’s in high speed lines. They easily sneak into the layouts.
Make sure USB and Ethernet traces are differential pairs through their entire route (no T’s). I have yet to review a board where first route was truly a differential pair.
Low speed lines going off the PCB should have RF filtering (ferrites)
Get rid of RS232 lines unless they are demanded by your customers. Computers no longer have RS232 ports as standard ports.

Discussion:

Once you get to the more painful ones you should consider hiring one of the many EMC Consultants. Especially if time to market is important. Why me over the others? I can review your EMC Reports, drawings, PWB layout files and photographs of your product remotely and tell you what the probable cause of the issue is. I can typically review your product in less than 8 hours and provide a list of changes need in most important to least important order. I can do that without traveling to your location, or your shipping a product to me or even running tests. That will save you travel and other expenses. You will still have to make changes that are likely time consuming.

Most of the products I was involved with from day 1, passed with more than reasonable margin to Class B levels on the first trip to the EMC lab. That was accomplished through drawing review before parts were made. Some of the products you can find in the FCC database as the had RF Modules in them.