Every EMI (electromagnetic interference) problem ultimately starts or ends at an electronic circuit. And since electronic components are the building blocks of circuits, it only makes sense to pay attention to the EMI impact of those individual components.
Probably the most important thing to remember about electronic components is that nothing is ideal. Components change values with frequency, current, voltage, and even physical size. And those changes may be nonlinear, adding a new level of complexity. Like a pilot, you need to know the limits so you stay within the envelope of safe performance.
Two key EMI factors are parasitic inductance and capacitance. As frequencies increase, so do their unwanted effects. In fact, parasitic inductance can convert a capacitor to an inductor, and parasitic capacitance can convert an inductor to a capacitor.
As these factors are not documented on schematics, we are often left looking for the “hidden schematic”. In fact, when diagnosing EMI problems, we often add these parasitic components to the schematic to better understand what might be happening. Simple, but effective.
One helpful hint is to examine circuits at three different frequencies – low, medium, and high. This is like using a microscope at different levels of magnification. For example, in an AC power supply we might focus on the following frequencies:
- 50, 60, or 400 Hz (normal operating frequencies)
- 1 MHz (a good starting point for routine power transients)
- 100 MHz (a good starting point for radiated emissions or immunity)
Of course, if you have specific EMI problems, you can refine the focus as follows:
- Substitute a known failure frequency for radiated/susceptibility test failures
- EFT (electrical fast transient), use 60 MHz (based on a nominal 5 nsec rise time)
- ESD (electrostatic discharge), use 300 MHz (based on a nominal 1 nsec rise time)
Another helpful hint is to consider self-resonant frequencies of your components. Sometimes this data is available from the component manufacturers. If not available, you may need to use engineering judgment to make estimates. We’ll provide some general examples in the article.
Let’s begin with the simple passive components — resistors, capacitors, inductors, and transformers. We’ll also consider the lowly circuit trace. Yes, for EMI purposes we must often consider connections as components too.
Figure 1 shows the frequency response of several passive components. At low frequencies, the components behave as expected. Inductors induct, resistors resist, capacitors capacitate, etc. But as frequencies increase, the components may no longer behave like the text book models we all learned in school. This is often due to parasitic inductance and capacitance within in the components, or is associated with how they are installed.
Figure 1: High frequency behavior of real world components
But is it not just frequency that changes behavior. To further complicate things, high currents and voltages can also affect components. In extreme cases, the behavior may even become nonlinear, such as inductor saturation.
Although the simplest of components, even resistors can exhibit EMI problems. While old-fashioned carbon resistors were pretty well behaved up to several hundred MHz, wire wound or tape wound resistors often show inductive effects at lower frequencies, thereby raising the overall impedance.
At higher frequencies, however, parasitic capacitance (between windings or end-to-end) limits impedance. For example, 3 pF at 100 MHz is about 500 ohms. Put that in parallel with anything over a few kilo-ohms, and you still only have 500 ohms.
High voltage transients, such as ESD, can arc between winding or across the entire component. Though not a well-documented problem, we’ve seen this occur several times. Furthermore, the ESD currents can cause permanent damage to small resistors due to heating effects.
All capacitors resonate! This can be due to both internal and external series inductance. As a result, the capacitor impedance is actually low at resonance, which might be seen as good. But above the resonant frequency, the capacitor actually looks like an inductor, increasing in impedance with frequency. Not good if you are looking for a high frequency short.
The internal inductance is a function of component size. As a result, most electrolytic capacitors are self-resonant in the 1-10 MHz range. This means they may be fine for power frequency filtering or energy storage, but they are useless for decoupling (shorting) at 10 MHz and higher. See Figure 2 for typical safe frequencies for various types of capacitors.
Figure 2: Typical safe frequencies for various types of capacitors
For most higher EMI frequencies, ceramic capacitors are preferred. By themselves, small surface mount ceramic capacitors are typically good up to 1 GHz. However, the inductance in external traces (and even vias) can still limit performance.
For example, assuming 20 nH per inch for a wire or trace, a perfect 1000 pF capacitor will resonate at about 70 MHz with a total lead length of only ¼ inch. At ½ inch, that drops to about 50 MHz. Our constant advice on decoupling capacitors — keep the leads short!
What value of capacitor do I need for EMI decoupling? Enough, but not too much. Since you are trying to provide a high frequency short, something under an ohm is a good goal. No need to go too much lower, as you become limited by the capacitor’s internal resistance.
You can calculate capacitor values for different frequencies, but we use the “rule of one”. If the product of MHz x uF is one, then Xc is 0.16 ohms. Thus, at 100 MHz, a 0.01 uF is suitable. You can even back this off to 0.001 uF and still have 1.6 ohms of reactance. At 1 MHz, you need to scale up to 0.1 to
1 uF to get the same results.
Inductors resonate too, due to the interaction of the inductance and capacitance between turns. This forms a parasitic parallel resonant circuit. Thus, at resonance the impedance increases, but then decreases above the resonant frequency. In effect, the inductor becomes a capacitor above resonance. Once again, not a desirable trait.
As a rule of thumb, we use 50 MHz as a default resonant frequency for small wire wound single layer inductors, common for EMI applications. You can get a better estimate with the following formula:
f = 200/sq rt (L)
f = the self-resonant frequency in MHz, and L = inductance in uH
Using this formula, a 1 uH choke would be self-resonant at about 200 MHz, while a 100 uH choke would be self-resonant at about 20 MHz. These numbers are pretty close to measured values.
At frequencies above about 50 MHz, we prefer ferrites. As permeable materials, they increase the inductance over air core devices. With ferrites you get more inductance with fewer turns, which means better high frequency performance.
Unlike air core inductors, ferrites become quite lossy as the frequencies increase. While this is seen as a negative by RF (radio frequency) designers, we embrace the loss for EMI uses and actually prefer to use them in their lossy range. In effect, they become high frequency dependent resistors.
There are numerous ferrite materials, but the most popular for EMI applications are nickel-zinc ferrites. Common vendor nomenclatures are Fair-Rite type 43/44 or Steward type 28/29. In single-turn configurations (beads or cable ferrites), these materials exhibit a fairly flat resistive loss between 100 MHz and 1 GHz. In fact, they are often specified in terms of ohms @ 100 MHz.
High currents can affect all inductors, wire wound or ferrite. Power frequency inductors can saturate, which means their impedance can drop to about zero. Ferrites exhibit this characteristic too, but are much more forgiving. For example, under heavy currents a 100 ohm ferrite (@100 MHz) doesn’t drop to zero, but may drop to 20-25 ohms. As such, we typically derate EMI ferrites in high current applications by a factor of four.
The main transformer EMI problem is parasitic capacitance between the windings. This is true for both power and signal transformers. Fortunately, Faraday shields between the windings can break up this capacitance. Unfortunately, these shields do not completely eliminate this unwanted coupling, but they do significantly extend the operating frequency range.
Experience suggests that while isolation is very good at power frequencies, unshielded transformers are very leaky by the time the frequency reaches 1 MHz. Since lightning and many power transients have equivalent frequencies in this range, these transients can easily pass through the transformer.
Adding a Faraday shield between the windings, however, extends this range into the tens of MHz, providing significant protection against most power related transients. This protection does not extend to higher frequency events, such as the EFT or ESD, nor to radiated emissions and susceptibility. In those cases, you will need filters and other high frequency protection.
A secondary EMI problem with transformers is voltage breakdown between the primary and secondary windings. This can be due to arcing at high voltages or insulation breakdown over time at lower voltages. We’ve seen the latter a few times, with failures after months of operation. Thus, it is very important not to exceed the manufacture’s voltage isolation specifications, even if device appears to work.
Yes, the lowly traces (and their cousins, interconnecting wiring) are components too. At frequencies above about 10 kHz, the inductive impedance exceeds the resistance, so the traces start acting as small inductors. When the length exceeds about 1/20 wavelength, the traces start to exhibit transmission line and antenna effects. At multiples of ¼ wavelength, resonances can occur, and at many wavelengths, the traces can even act like long antennas with gain!
Are these effects serious? Much of the time these effects can be ignored, but you still need to be aware of these potential problems. Even traces can be part of the “hidden schematic”.
Now that we’ve looked at basic electronic components, let’s look at two special EMI components – transient protectors and optical isolators.
These components are typically used to protect power and signal lines from voltage spikes. Some are faster than others, so the choice will often depend on the type of transients involved. Here are three of the most common types of transient protectors:
These devices are quite rugged, but relatively slow. Once the arc ignites, the voltage across the protector drops to a low voltage, resulting in a near short across the protected line. As a result, very little power is dissipated in the device — rather, the energy is reflected. Gas tubes fall into this category.
A major drawback to these devices is speed. As a result, they are generally not fast enough for the EFT or ESD transient, but are suitable for lightning and other slower power line transients.
MOV (Metal Oxide Varistors)
These devices are moderately rugged, and moderately fast. Since they clip the excess voltages, the excess energy is dissipated in the device. Nevertheless, even relatively small MOVs can dissipate a lot of energy at low cost. As such, they are widely used to protect lower cost consumer electronics.
Two drawbacks to these devices are speed and fatigue. The response speeds are comparable to arc devices, making them generally unsuitable for ESD and EFT, but well suited for power transients. They wear out over time due to cumulative effects and often fail open, leaving you unprotected for future transients.
Based on Zener diode technology, these devices are the fastest, and some can protect against sub-nanosecond transients. As such, they are suitable for all transients – power, EFT, and ESD. The also typically fail short and thus blow fuses or circuit breakers.
The drawbacks are size and cost, but we usually recommend these devices for expensive electronic systems where the cost of failure is high.
Since transient protectors are nonlinear devices, they can contribute to RF susceptibility problems, particularly when they are the first component seen on power or signal lines. In a sense, they act like a crystal radio and rectify the RF energy. As such, high frequency filters may need to be installed ahead of the transient protectors to prevent rectification at the device.
These devices are widely used in rugged environments (such as industrial controls) to provide isolation and transient protection for I/O circuits. While quite effective, nothing is perfect. Wish we had a dollar for every time somebody said “Don’t worry about that interface– it has optos.”
Two EMI problems are leakage across the capacitance and voltage breakdowns. The former can be a problem in the hundreds of MHz, creating sneak paths for radiated emissions and susceptibility. The latter can be a problem with ESD or other high voltage transients that exceed the device breakdown ratings. Thus, additional filtering or transient protection may be needed.
Now that we’ve covered the most common passive components, let’s take a quick look at EMI problems in active devices.
Digital ICs (Integrated Circuits)
Key EMI drivers are speed and size. In simple terms, the faster and smaller the devices, the more likely the EMI problems – both emissions and immunity.
For emissions, both edge rates and clock rates are critical. As both increase, so do the higher frequency EMI problems. Both have been the industry trend for many years and will likely continue in the future.
For immunity, only the edge rates matter, as these represent bandwidth. Simply stated, as the edge rates increase, the window of susceptibility increases. This makes digital circuits more vulnerable to spikes and transients such as ESD or EFT events.
Many modern digital ICs are much faster than they need to be, so slowing things down can yield EMI benefits. Don’t use faster clocks than necessary, and filter critical nodes such as resets or control lines. And pay attention to power decoupling — often the back door for emissions due to the high speed current pulses due to changing loads.
Once again, key drivers are speed and size. Also, many analog circuits operate with small signal levels, and resulting smaller noise margins than digital signals. For example, a sensor input may be upset by a millivolt or less, while digital signals can usually tolerate hundreds of millivolts before malfunctions occur.
For emissions, most analog ICs are pretty well behaved since they operate at low frequencies. In recent years, however, we’ve seen an increase in parasitic oscillations, often occurring in the VHF/UHF frequencies. These free running oscillations are often large enough to exceed radiated emissions regulatory limits. Good high frequency decoupling can prevent this problem.
For immunity, analog circuits are particularly susceptible to high levels of RF energy. The failure mode is rectification, resulting in any modulation now appearing in the normal frequency range of the analog devices. Once the rectification occurs, there is no way to undo it. One must prevent it from occurring in the first place. Good high frequency decoupling and filtering of inputs and outputs can prevent this problem.
Incidentally we usually consider RF and power ICs (such a voltage regulators) special cases of analog ICs, subject to the same EMI issues. We regularly recommend 1000 pF capacitors at regulator inputs and outputs. We also pay attention to very low-level inputs, such as radio receivers. GPS receivers, with their extremely low-level inputs are particularly vulnerable.
When dealing with EMI issues in components, expect the unexpected. Look for the hidden schematic, and be aware of your performance limitations (frequency, voltage, current).
Finally, please remember this article is intended as an introductory overview to EMI problems in components. Entire books have been written on many of these components. Although we’ve been brief, we hope we have increased your awareness, stimulated your thinking, and perhaps even demystified some EMI problems you have may encountered.
|Daryl Gerke, PE and Bill Kimmel, PE
are partners in Kimmel Gerke Associates, Ltd., an engineering consulting and training firm that specialized in EMI/EMC design and troubleshooting issues. Both are degreed engineers (BSEE), iNARTE Certified EMC Engineers, and registered Professional Engineers (PE).
Daryl and Bill have prevented or solved hundreds of EMI problems in a wide range of industries – computers, military, medical, industrial controls, automotive, avionics, railroad, telecomm, facilities, and more. They have also trained over 10,000 engineers through their public and in-house training classes. They just celebrated 25 years in full-time practice as EMI/EMC consulting engineers. For more EMI information, visit their web site at www.emiguru.com.