Building a Faraday Cage To Contain EMI, Part 1
EMI (Electromagnetic Interference) is a phenomenon that can plague a design if it is not properly and carefully addressed. One of the most often used but not necessarily thoroughly understood ways to contain EMI is through the use of a Faraday cage. This two-part article will provide an overview of EMI, how it can interfere with other products, and how to address EMI through the successful design and implementation of Faraday cages. Included in this article will be a discussion of what EMI is and where it comes from; grounds, and how they apply to electronic circuits; how to get heat out of a Faraday cage while keeping EMI in; how to get signals in and out of a Faraday cage without letting EMI out; how to get power into a product without letting EMI out, and building a Faraday cage for rack-mounted products that have plug-in cards and backplanes.
Part 1 of this article will cover the various ways that EMI escapes from products and how to build a Faraday cage to contain it. Part 2 will focus on getting power into a product without letting EMI out; building a Faraday cage for rack-mounted products, and whether or not logic ground should be connected to a Faraday cage.
What Is EMI And Where Does It Come From?
As noted above, EMI stands for Electromagnetic Interference. In electronic configurations, EMI escapes from one product and interferes with another. This can happen in two ways:
Electromagnetic energy can be radiated into space because there is an accidental antenna extending from the product.
Electromagnetic energy can be conducted out of the power lines of the product and into the power terminals of another product.
The commercial standard for radiated EMI is measured in the band of frequencies from 30 MHz to 1 GHz or to 5 times the highest clock frequency, whichever is higher. Conducted EMI is measured in the band of frequencies from 150 KHz to 30 MHz.
EMC stands for Electromagnetic Compatibility. This means that a product has been designed such that it is not interfered with by other products as a result of EMI, whether it is radiated or conducted. In other words, noise in the form of electromagnetic radiation coming from another source that can cause a product to malfunction does not affect it. The Faraday cage solves this problem as well.
Stated simply, radiated EMI is an unwanted radio link. If we think about what is needed to create a good radio link (i.e, a transistor), it becomes easier to understand what needs to be done to eliminate an EMI problem. The two necessary elements in a radio transmitter are:
A source of RF energy (transmitter).
A radiating surface (antenna).
Controlling EMI involves eliminating either the source or the antenna.
The rules-of thumb methods passed around the engineering community have tended to focus on removing the source of EMI. These methods evolved in the 1980s when the operating frequencies of products were well below the 30 MHz starting point for measuring EMI. At that time, ASICs occasionally had speeds fast enough to generate noise in the 30 MHz to 1 GHz range. Inserting a ferrite bead in the power lead of such a device prevented it from operating fast enough to cause EMI. Such techniques focused on removing the source of EMI.
Since modern electronics operate well above the 30 MHz starting point for measuring radiated EMI, suppressing it with ferrite beads and other similar methods is not a choice. Therefore, eliminating the accidental antennas becomes the de facto method for eliminating EMI.
The things that make good antennas include:
- Things that stick up above the PCB such as PLCC lead frames and other elements that leave the PCB. This includes unshielded wires going to devices such as mice and monitors.
- Things that don’t make good antennas are things that don’t stick up, such as traces on a PCB.
- Two PCBs joined by a connector, such as a DIMM connector, form a dipole antenna.
- PGAs (Programmable Grid Arrays) and BGAs (Ball Grid Arrays) in sockets make good antennas.
- A two board set will usually behave as a dipole antenna.
- An unshielded wire leaving a Faraday cage will be an antenna.
- Connecting logic ground to a Faraday cage in more than one place often turns the cage into an antenna.
- A component lead frame protruding above a PCB is an antenna.
- Cutting ground planes can turn a PCB into a dipole antenna.
- Connecting a plugin module faceplate to logic ground will turn the faceplate into an antenna.
Faraday Cages and EMI
There are three ways to treat potential antennas (wires). These include:
- Shielding them when they leave the product.
- Placing a low pass filter in series with the antenna where it leaves the product.
- Placing the product in a Faraday cage.
For the purposes of this discussion, we will focus on Faraday cages and their role in suppressing EMI. Faraday cages are metallic enclosures that surround a product that is radiating energy in the EMI band. The Faraday cage reflects this energy back into the product but rarely absorbs it, and it is the ultimate method for containing EMI. A Faraday cage is necessary when a system has multiple PCBs or when there are large components that stick up and can serve as antennas.
As noted above, the Faraday cage itself can serve as an antenna if the Faraday cage is connected to logic ground at more than one place. This is sometimes erroneously referred to as “chassis ground”. This error is most commonly made if you tie the logic ground to the Faraday cage at the system backplane, and then also tie the logic ground to the faceplates of the plug-in cards.
A clue that this has occurred is when EMI is detected at the “cracks”. We have often heard this described as EMI “leaking out” at the cracks or seams of the box. The Faraday cage is usually composed of parts of the chassis, such as the sides of a card cage. Because of this, the term “Chassis Ground” is often used when discussing EMI containment. These terms are confusing as they can mislead product developers. We use the term Faraday cage only when discussing EMI and represent it with the symbol at the right. It’s true that some parts of the chassis are used to form part of the Faraday cage, but the “chassis” is not the EMI containment vessel.
Using the word “ground” in discussions about EMI causes confusion. With respect to EMI, ground is not a magic place. A good definition of ground was made by Bruce Archambault of IBM to cure engineers from using the word in EMI discussions.
Faraday cages can be built with any common type of metal. Figure 1 is an example of a product with a Faraday cage. When the casework is painted, it’s important to make sure that the paint does not cover the areas where the metal bonds are needed, such as between various parts of the Faraday cage and the edges of the faceplate.
Figure 1. A Typical Card Cage Based Faraday Cage
Getting Heat Out of The Faraday Cage While Keeping EMI In
Once a product has been surrounded by a Faraday cage, it is not only EMI tight, it is also heat tight. Since this heat can cause a product to improperly operate or fail altogether, it’s important to be able to get the heat out of the product.
As an example, in laptop computers the heat is removed through the use of a heat plate that conducts the heat to the outside case of the laptop. In larger products where this is not practical, heat is dissipated with moving air. In products that have modest heat dissipation, the air is moved by convection. For those products that have more heat than convection can handle, fans are used. In all instances, air must be able to enter the product and leave it. This means that there has to be openings in the Faraday cage that are large enough to allow the air to move through while still being small enough to prevent EMI from escaping.
There’s been a lot of speculation as to the size of the hole that can be made without creating an EMI leak. We have not seen any documents that clearly illustrate how to determine the size of the hole that would meet this criterion. Instead, we have employed the methods that we always use—build test structures and then make the measurements. Through our experiments, we have determined that meshes with holes no large than ¼”, 6.35mm, will contain EMI up to least 10 GHz. An array of holes can be punched in the surface of the Faraday cage. Then a screen can be tightly mounted into a hole in the Faraday cage. It’s also possible to have a honey comb such as that shown in Figure 2 that can be mounted in the top and bottom of a card cage above and below the card guides. If screens or honey combs are used, they must be bonded to the Faraday cage all the way around it.
Figure 2. A Honey Comb Screen Allowing Air Out While Keeping EMI In
Containing EMI While Signals Come and Go
A tightly sealed Faraday cage is a sure way to contain EMI. However, the product surrounded by the Faraday Cage is of little value unless the signals can also come in and go out.
One way to solve the foregoing problem is with fiber optics. With this technology, there are no conductive paths into or out of the box on which EMI could travel. As a result, products with fiber optic interfaces, such as large routers, are relatively easy to make comply to EMI specifications. For other products, there has to be another approach.
There are two kinds of signals that enter and leave a product on wires. They are:
- Shielded cables such as coaxial cables.
- Shielded twisted pairs on unshielded wires.
- Signals that run on shielded cables include:
- 10Base2 Ethernet
- Graphics signals on 9 pin DIN connectors
- RF signals to and from antennae
- Infiniband cables
Signals that run on unshielded wires include:
- Mouse connections
- Fan controls
- Ethernet on unshielded twisted pairs (UTP)
- Keyboard connections
- Power cabling
Handling signals traveling on shielded cable is straight forward. This shield is an extension of the Faraday cage so it must be connected with a very low inductance connection to the cage. This is accomplished by connecting the shield of the cable to the shell of the connector on the cable side and connecting the shell of the connector to the Faraday cage on the product side. It’s important to note that you do not want to connect the shell of the connector to logic ground when it is part of the Faraday cage.
There are instances in which it is not possible to make a DC connection between the cable shield and the Faraday cage such as in the 10Base2 version of the Ethernet. If there is no connection between the shield and the Faraday cage, the shield may well function as an unwanted antenna. Figure 3 is an example of a 10Base2 Ethernet exiting the end of a card that plugs into the backplane on the left.
Figure 3. 10Based 2 Ethernet Cable Connection to Faraday Cage
Because the circuit is located at the end of a PCB that is plugged into the backplane, there will be both AC and DC voltage gradients between the backplane “ground” and the faceplate of the PCB that is part of the Faraday cage. In most cases, the ground planes in the backplane will form one side of the Faraday cage while the faceplates of the plug-in cards will form the other side. The AC noise on the circuit will couple from primary to secondary of the output transformer through the parasitic capacitance that exists between the two. As a result, this noise will be impressed on the shield and the center conductors of the shielded cable. If this shield is connected to the Faraday cage, this noise will only travel on the inside of the shield so there will be no EMI.
The problem with the foregoing circuit is the Ethernet requirement that there be no DC connection to the Faraday cage. This leaves only an AC connection in the form of a capacitor connection as a choice. A further requirement is that the capacitor be able to withstand a voltage of 1700 VDC. There are no capacitors that have both the breakdown voltage required as well as the ability to make a low impedance connection between the shield and the Faraday cage over the radiated EMI frequency band. As a result, the emissions shown In Figure 4 can occur.
Figure 4. Emissions From 10Based 2 Cable without Plane Capacitor in Figure 5
In order for the shield to do its job, a method of connecting it to the Faraday cage is needed that meets the electrical conditions of the breakdown voltage of 1700 V and AC impedance from 30 MHz to 1 GHz. A parallel plate capacitor made from the planes in the PCB can do this job. Figure 5 shows how this type of capacitor is constructed.
Figure 5. Parallel Plate Capacitors Formed From PCB Layers
Figure 5 is the right-end of a daughter card PCB that extends to the left and plugs into a backplane connector. The last inch of the area on all four layers has been separated from the rest of the PCB by plane cuts. The area on the top and bottom layers has been flooded with copper and connected to the faceplate using the faceplate mounting screws. This area serves as one plate of a plane capacitor connected to the Faraday cage. The plane area in the two inside layers is split into two segments to create a capacitor plate for each of the two axial connectors. The shield of each coaxial connector is connected to these internal plates forming the second plate of a very low inductance capacitor of approximately 370pf that is connected to the Faraday cage. The minimum insulation thickness is 8 mils for a breakdown voltage in excess of 8000 V. An AC connection has been made between the shields and the Faraday cage that meets both electrical requirements. Figure 6 depicts the emissions after the AC connection has been formed.
Figure 6. 10Base2 Emissions After Adding Plane Capacitor to PCB
The emissions have been dramatically reduced by ensuring that the cable shield has a low impedance connection to the Faraday cage. The reason the parallel plate capacitor was effective and the discrete capacitor was not is due to the very low inductance of the plate capacitor. This may have lead to the thinking that some EMI gurus espouse that this science is “black magic.” Discrete capacitors worked in the past when things were slower and now they don’t because of their parasitic inductance.
Capacitors built from the layers of a PCB can also be used to build low pass filters that function over a very broad range of frequencies. This technique works for control lines that exit the Faraday cage to fan trays or that go to keyboard and mouse peripherals. All that is needed is to attach a large patch of copper in a signal layer to the signal before it exits the box. Figure 7 is an example of this for two fan control lines exiting a Faraday cage surrounding a Terabit router
Figure 7. Plane Capacitor Built Into a Signal Layer of a Backplane
The two rectangular patches of copper in Figure 7 form parallel plate capacitors with the ground planes of the backplane. In this instance the ground planes in the backplane form one side of the Faraday cage. They are attached to the traces carrying fan control signals as they exit the Faraday cage creating a low pass filter that prevents high frequencies from exiting this path. A similar technique can be used on other lines that exit a product without shielding.
A more common version of an Ethernet connection is with UTPs (unshielded twisted pairs). When this kind of circuit is located at the end of a plug-in card such as that shown in Figure 3, an EMI problem is sure to result. This problem can be solved by using a transformer with a center-tapped secondary such as that shown in Figure 8. The plane capacitor is tied between the center tap of the secondary and the Faraday cage shunting the noise to it.
Figure 8. Plane Capacitor Connecting Center Tap of a UTP to Faraday Cage
Next, Part 2: Getting Power In Without Letting EMI Out and Beyond.
Discover how to define the new challenges and how to successfully engineer the Power Delivery System (PDS) so it meets the needs of new technologies. Have more questions? Call an expert at Altium.
Ritchey, Lee W. and Zasio, John J., “Right The First Time, A Practical Handbook on High-SpeedPCB and System Design, Volume2.”