Published and Presented, Fondazione Ugo Bordoni, Symposium on Electromagnetic Security for Information Protection, Rome, Italy - November 21-22, 1991

Abstract - With costs and scheduling issues associated with laboratory based electromagnetic design and testing always a problem, there is a continuing need to increase productivity at the basic engineering level without forcing equipment manufacturers into major test equipment investments. In general, most reduced emission design is really much simpler than hardware designers would expect. By shifting the major design burden from the "test house" to "in-house", significant development time and cost savings can be achieved. This paper describes the expected cost, required equipment, and general techniques necessary to quickly enable in-house engineering personnel to perform emission corrective actions prior to utilizing expensive outside test services,

In most instances, the manufacturer's initial baseline equipment intended for low electromagnetic sensitive applications is mechanically well enclosed, contains "protected" wiring assemblies, and uses printed circuit boards that are designed to be functionally correct. Functionally correct means that with the minimum number of layers necessary to allow for easy modifications by the circuit board design house. Upon receiving such equipment, it is the emission engineers job to find and fix any last minute problems prior to sending the final package out for product build and testing.

Unfortunately, most equipment manufacturers don't have their own in-house test laboratory, and thus are usually forced to schedule very expensive test time at laboratories specializing in emission reduction work just to perform engineering scans on their equipment. Many organizations, willing to take substantial risks-, schedule formal testing prior to performing quick look scans. As it often happens, these equipment manufacturers have noise emission problems with their equipment that are not readily discovered by lab testers, resulting in a great deal of unusable data derived at substantial costs. Not only is the test data unusable, but the combined costs of supporting in-house engineers in the field, plus the engineering costs associated with test laboratory troubleshooting results in massive development costs for the final product.

Since equipment manufacturers normally can't support specialized low noise designer talent on a full time basis, a mental acceptance of the costs associated with this type of product design has developed within the commercial industry. Manufacturers just accept their fate, and plan their programs accordingly. However, not all emission design is a "Black Art" as it appears at first, and developing "quiet" and secure equipment is really much simpler and cheaper then it might otherwise seem. This paper focuses on methods that reduce the engineering costs associated with developing protected equipment, while at the same time enhancing the manufacturer's in-house technical base.

There once was a time when electronic equipment manufacturers spent considerable effort and expense having specialized laboratories reduce equipment niose levels to meet commercial (FCC) emission compliance requirements. So much money was involved that many engineers, once they learned how to troubleshoot equipment for EMI problem identification, and once they could use EMI fixes to reduce the identified emissions, put small labs together and went into the FCC certification -business. Competition was increased among test houses and practitioners, and costs to certify equipment dropped drastically. There even grew up a large cadre of consultants that would EMI engineer boxes in-house prior to sending out for certification testing, saving manufacturers from both schedule delays and the cost of lengthy troubleshooting test time at the certification lab.

Equipment manufacturers that don't perform in-house emission reductions prior to sending their equipment to FCC test labs still pay premium costs for troubleshooting, but even these costs are considerably less than the cost associated with extensive laboratory troubleshooting. However, the lessons learned from the FCC test industry can be applied by equipment manufacturers in order to greatly reduce their ultimate costs.

While not all emission related equipment problems would be identified with a comparatively inexpensive and simple in-house test setup, the vast majority of problems can be quickly found and fixed with the approach proposed herein. First consider the cost of a basic set of test equipment as listed in the table below.

In justifying the cost of the equipment described, extensively equipped specialty test labs on the average charge about $1000 per day for test time. It is a simple calculation to discover that savings of just under two months of development testing more then pays for the equipment needed to troubleshoot your own products. In addition, the equipment you purchase may qualify for a tax deduction, and can also be depreciated over time. It should also be noted that this suite of equipment can be rented for less then $10,000 per month. For the company that has only infrequent (less than 2 months/year) need of the equipment, rental may be the preferred (and also deductible) cost option.

Notice that some items commonly thought of as essential are not necessarily required for an emission-troubleshooting lab. For instance, there is really no need for a shielded room (about a $20,000 cost) if no formal testing is to be performed. Unless your test laboratory is very noisy, simple testing can often be performed "around" the areas where ambient noise is a problem (and with a little practice "within" the noisy environment). Problem signals that appear at one frequency will normally show up all over the spectrum, allowing ample opportunity for detection elsewhere. In addition, if ambient noise is a problem, local foil shielding will provide considerable attenuation at minimal cost.

Another common problem encountered when purchasing test equipment is the concern often expressed that "none of my people know how to use the equipment for specialized emission testing." While this is true to some extent, remember that engineers troubleshooting equipment to reduce emissions don't need to perform a formal acceptability type test. If there is a major concern about using the equipment, hire a consultant to come in and show them how to use it. Otherwise, remember all that is really required from engineering personnel is the ability to find, identify, and reduce problem signals, similar to what is done in the US for FCC emission suppression work. In addition, people who design the equipment 1) know how it works, 2) know how to use the unit specific test equipment, and 3) are the best suited to find and fix the problem, once they know what to look for.

If testing according to some protected formal techniques is desired, it would be a problem to test in an open lab environment. However, testing in a closed lab with restricted access, except for persons with the required need-to-know, is not too difficult to implement, and may be worth the effort to implement in the long run.

The final concern often expressed by non-emission designers is "I have no idea what to look for". If you know what your problem signal looks like on an oscilloscope, leave a scope probe connected directly to the data line, and use the spectrum analyzer at a wide bandwidth to find this same signal at some point in the frequency domain. Obviously, reducing the spectrum analyzer bandwidth will enable the tester to more clearly identify the signal. The detected data signal is the signal to be reduced, and it can show up in radiated form at almost any frequency.

The easiest and least expensive way to support a redesign and trouble shooting effort is to start by building a finger wound antenna probe. To make a BNC trouble shooting H-field probe, using a barrel connector, solder a wire to the center pin, wrap the wire around your finger about 35 times, then remove your finger and solder the other wire end to the outside surface of the barrel. A large tip on your soldering iron is necessary to heat the barrel enough for solder to stick. Use masking tape to secure the loops from spreading out too far or unraveling. More sensitive probes are available commercially.

The troubleshooting method described here is nearly identical to the method described by Berger', and is based on the assumption that a noise signal that radiates an appreciable emission due to current flow can be readily detected at close range with an inefficient H-field probe. Also, a second assumption is that the radiated emission exists at many harmonics, and can probably be detected with a mid-range (biconical) type antenna at some measurable level by monitoring the spectrum analyzer and pre-amp output on an oscilloscope. The test equipment is shown in Figure 1.

In nearly every case, the radiated condition will become observable when the unit containing the signal source is exposed by opening its cover. Therefore, by reducing this easily detectable signal at its source initially, an engineer can greatly enhance the chances of passing accreditation testing the first time through.

Open the cover on the box to be tested, and first, using an oscilloscope and pre-amp while looking at the schematic diagram and layout drawing, find one of the noisy data lines to be emission controlled. Next, determine the RF spectrum of this signal with a pre-amp connected to the input of the spectrum analyzer if necessary. Connect your finger probe to the spectrum analyzer and move the probe around the board to locate and map where the highest readings are for the signal being analyzed.

At each point where a high H-field level is detected, measure the signal amplitude using the spectrum analyzer and a fixed commercial E-field antenna located at some point near to the unit under test, exactly as would be done in a FCC troubleshooting test lab. The E-field is measured at a fixed location since it is easier to identify and measure than its corresponding H-field using the uncalibrated and less sensitive finger probe. The measured E-field signal should be easily identified since it will look nearly identical to the signal located with the finger probe. Repeat this procedure for each of the other signal lines to be analyzed.

Once the radiated E-field emissions from each of the problem signals has been measured, the next step is to systematically reduce these radiated emission levels using source suppression components mounted directly to the pc card. Many sources' are available describing card level noise reduction techniques and noise testing techniques. If a manufacturer can correct emission problems during the design stage of a program, it is a simple matter to incorporate these fixes within the overall product development program. However, if a laboratory finds problems and recommends fixes after a product has been built and delivered to the test lab, both schedule delays and redesign costs increase dramatically.

If a detected signal is closely associated with a specific integrated circuit, and not just an output pin, the initial suppression technique could include a ferrite pad under the entire integrated circuit (IC). This fix often works with F series logic operating at high data rates. Another approach at the integrated circuit level which does not affect the output waveform is to add a resistor (try 10 ohms) in series with the IC power input (DC bus), and increase the value of the decoupling capacitor (- .1 AF) located between the resistor and the IC. The preferred capacitor for high frequency DC decoupling is a ceramic disk.

Often the detected signal follows a printed circuit board trace, and is highest at the location of the line driver for the trace. Therefore, beginning with an uncovered operating board and the largest emission detected, and using a schematic and pc board layout drawing, observe the signal waveshape on the trace using an oscilloscope. Chances are that the signal waveform will have ringing and overshoot as shown in Figure 2.

To reduce the trace ringing, begin by adding simple wave shaping (loading) components adjacent to their source while monitoring the amplitude of the emission under investigation. Start with a series RC filter of 410 pF and 50 ohms as the loading components. Adding a capacitor only is not a good idea as it has the effect of shunting noise into the ground plane, and will increase the radiated antenna farm effect from common mode noise in the ground plane.

Once the waveform appears clean from loading, again measure the radiated level of the data line with your fixed antenna and spectrum analyzer. The detected emission level should be greatly reduced. However, if the emissions are higher, then you have increased the problem

associated with antenna farm effects from the ground return, and an absorption rather then loading component is required. In this case, a series ferrite bead, or a series ferrite filter (shown in Figure 3) should be added, with a size and value proportional to the frequency being detected. Also, since the addition of ferrites can sometimes cause harmonics to be generated, it is important to investigate this possibility with the spectrum analyzer before finalizing your design.

If the circuit cannot tolerate the ferrite bead, the next approach is to add a small series resistor in the data line itself. In this case, insure the voltage drop across the resistor is not enough to affect the logic operation. The value of the resistor can be determined by the logic family, i.e. the source current provided.

As is often encountered, a particular emission seems uniformly distributed over the entire card, and only detected at higher or lower values depending on probe location. In this case, the date related emission is likely being modulated on a noise source, such as the system clock, and is widely distributed throughout the circuitry by the power system or ground plane (or trace). To reduce a widely distributed signal, either the signal must be localized and controlled, or the carrier must be reduced and/or localized.

One of the easiest methods of localizing an emission was mentioned previously by using powerline decoupling. Normally a resistor or ferrite on the IC's power input line, plus a larger decoupling capacitor effectively isolates the IC source. In addition, locating the signal source (such as the clock circuit) near the card interface connector during board layout, and then placing a ground plane under the IC with the conductive connection to the rest of the ground plane at only one point, as shown in Figure 4, is also effective.

Controlling a carrier, especially a wideband noise source like the system clock, involves a combination of all the techniques describe above. In addition, the carrier ground returns may need to isolated on Printed Circuit Board be controlled by isolated Ground Plane branching, as shown in Figure 5, to insure the noisy ground does not contaminate all other grounds on the pc board.

Once all identified problem emissions have been reduced as much as possible using the techniques previously described, choose the largest emission detected, close up your box, and see if the signal is still detectable. If it is, chances are that your box leaks or a ground loop exists within your box, and its inherent shielding is having no effect.

With the box closed, again use your finger probe and move around the box to find out if a leak is present. Leaks can usually be fixed with finger stock, sanding to insure good metal to metal contact, or gasketing. In some cases, a mechanical redesign is your only alternative.

If an obvious leak is not detected, and the signal is higher in some places and low but detectable in others, a more serious system related problem exists. In this case, take a look at how the pc board is grounded, and reduce the ground to either a single point ground, or as in the case of a pc, the processor board might need multiple grounds, with only the problem signal grounds controlled to a single point.

Check to make sure the power supply is isolated, and also well grounded to the primary central point ground. This is normally the case near the processor ground for metal enclosed electronic units. If the signal is still detectable, go back to the signal source and further reduce it. If further reduction is difficult or impossible, the only alternative may be to add a ground plane to the outer layers of the pc card in a sandwich approach. This is usually the final alternative when all else fails, but it also usually works to fix the problem.

So far we have only discussed radiated problems. Attacking an emission at its source using decoupling at the IC's power input pin also works to reduce conducted emissions. In addition, choosing a less efficient (and usually less noisy) power supply, such as a shunt or series-shunt regulator, and mechanically partitioning your chassis to provide separation between the power supply and the rest of the electronics, including associated cabling, will greatly reduce your chances of encountering a conducted problem.

Once you have identified a radiated emission, the same technique can be used to identify the noise emission on your powerlines, except that you need a Powerline Impedance Stabilization Network (PLISN) to match the spectrum analyzer input to the powerline.

If the noise source can be detected on your powerline, in most cases it coupled around your filtering or partitioning and contaminated the power supply primary. In this situation, if you can't go back and isolate the signal at its source further, and maximum power supply isolation has been mechanically implemented, either add or increase powerline filtering first at the power supply secondary, and finally at the power supply primary. If the signal is still detectable, call in a consultant.

This paper has attempted to provide suggestions to equipment manufacturers on how to greatly reduce their costs commonly associated with an emission suppressed equipment design effort. Drawing a correlation to what has taken place relative to in-house FCC noise reduction programs, in-house specific emission suppression isn't a whole lot different, only more directed. For an equipment manufacturer not to take advantage of his in-house potential for cost savings is a mistake that can be easily corrected.

[1] Berger, H. Stephen, Using an oscilloscope and Sniffer Probe to Solve EXI Problems, Evaluation Engineering Magazine, February, 1987.

[2] Compliance Engineering Magazine and Application Notes, Boxborough, MA; EMC Technology Magazine, Gainesville, VA; ITEM Magazine, West Conshohocken, PA.