
EMI and transient protection for MIL-compliant systems
Defense systems with 28-V power requirements must meet a number of noise and power related standards, including MIL-STD-461 (Requirements For the Control of Electromagnetic Interference Characteristics of Subsystems and Equipment), MIL-STD-704 (Aircraft Electric Power Characteristics), and MIL-STD-1275 (Characteristics of 28 V DC Electrical Systems in Military Vehicles). Each of the standards encompasses multiple revisions, any one of which may be enforced by the application. Additionally, each standard features subsections that apply as dictated by the end installation. To satisfy the standard, designers need to introduce EMI filters. Let’s take a closer look at how to design an effective filter to ensure compliance.
Of the various power-related standards, MIL-STD-461 is of particular interest. The standard addresses a number of characteristics including conducted emissions, conducted susceptibility, radiated emissions, and radiated susceptibility. Emission refers to the noise a device generates as it impacts the source to which it is connected. Susceptibility is the vulnerability of a system to incoming noise.
The latest substantive revision, MIL-STD-461E, includes a number of substandards for various components such as power leads, antenna terminals, and antenna ports, over defined frequency bands (see table 1). [Ed. note: a subsequent revision, 461F, does exist but it is just a validation and does not change specifications.]


Table 1: Substandards for MIL-STD-461E define performance requirements at a component and system level.
As table 2 shows, not all sections of MIL-STD-461E apply universally; hence, most power conversion suppliers focus on achieving compliance to the subset that affects all installations, and in particular to the conducted, rather than radiated, sections. These standards are CE102, CS101, CS114, and CS116. Frequently, manufacturers will also reference CE101 because the frequency band of interest (30 Hz to 10 kHz) is well below the switching frequencies of most DC-DC converters. Conducted emission and susceptibility requirements, not radiated requirements, are quoted because radiated sections are significantly dependent upon the physical layout, external output circuitry, and the enclosure in which the power supply resides. With a valid filter design and good PCB layout, most components can easily meet conducted requirements.


Table 2: The applicability of MIL-STD-461E substandards varies by platform.
There is not much difference between revision E and the earlier revision D; in fact, of sections CE101, CE102, CS101, CS114, and CS116, only CS101 and CS114 differ as follows:
• CS101 – No change up to 5 kHz; above 5 kHz:
MIL-STD-461D: Required level drops 20 dB/decade to 50 kHz
MIL-STD-461E: Required level drops 20 dB/decade to 150 kHz• CS114 – No change up to 30 MHz; above 30 MHz:
MIL-STD-461D: Required level drops 10 dB/decade to 400 MHz
MIL-STD-461E: Required level drops 10 dB/decade to 200 MHz
Basics of EMI
Now that we have introduced the standard, how do we gain compliance? What follows is a general guide for EMI filter design. We will focus on CE102 for our discussion.
Let’s start with an overview of EMI. We can separate EMI measurements into two parts: conducted and radiated. Conducted measurements are measurements of either voltages or currents flowing in the leads of the device under test, as dictated by the standard. Common-mode conducted noise current is the unidirectional (in phase) component in both the positive and negative inputs to the module. This current circulates from the converter via the power input leads to the DC source and returns to the converter via the output lead connections. This represents a potentially large loop cross-sectional area that, if not effectively controlled, can generate magnetic fields. Common-mode noise is a function of the rate of change of voltage (dv/dt) across the main switch in the converter and the effective input to output capacitance of the converter. Differential-mode conducted noise current is the component of current, at the input power terminal, that is opposite in direction, or phase, with respect to each other.
For our purposes we will concentrate on MIL-STD-461, CE102 that is a voltage measurement into 50 Ω.
E-field radiated emissions are caused by currents conducted by a suitable antenna, such as the power leads of the device under test. If we can greatly reduce the conducted emissions then we will reduce the radiated emissions, as well. The enclosure of the device under test, lead geometry, and other devices running within the device under test will affect the emissions. Radiated emissions due to B-fields are best addressed by shielding with a suitable material and proper layout.
Testing for EMI
To get repeatable results, we need a defined test setup, a known source impedance, and limits to which we can compare results (see figure 1).


Figure 1: The standard test configuration includes one line impedance stabilization network (LISN) per power lead (top). Each LISN terminates into a 50 Ω impedance (bottom).
We realize the known impedance using line impedance stabilization networks (LISN) terminated into 50 Ω impedance (internal to the measurement device). One LISN per power lead is needed.
From the limits specified for 28V systems, we can see that at 500 kHz and above, the limit is 1 mV into 50 Ω (see figure 2). Given the limits, we will need to understand the source of the noise to determine the amount of attenuation required to stay below the limits. It is critical to understand the properties of the noise source in order to design a good filter.


Figure 2: The MIL-STD-461 substandard CE102 specifies limits for radiated emissions. It is beneficial to translate the limits to millivolts in addition to the standard dBµV.
Since in most cases the noise character of a device is unknown, the most effective solution is to have the device in hand prior to the development of a filter. The noise source can then be characterized through experimentation and, once characterized, a model can be generated. A good series of noise voltage measurements are:
• Input to ground — open circuit.
• Input to ground — 100-Ω shunt termination, with DC blocking cap
• Input to ground — 10-Ω shunt termination, with DC blocking cap
• Input to ground — 1-Ω shunt termination, with DC blocking cap
• Measurement of the short circuit common mode current input-output.
Let’s assume the noise voltage measurements are:
• Input to ground — open circuit, 10 V P-P
• Input to ground — 100 Ω, 4 V P-P
• Input to ground — 10 Ω, 580 mV P-P
• Input to ground — 1 Ω, 280 mV P-P
• Short circuit (50 nH) current input-output, 290 mA
The equivalent circuit (model) would be most nearly a 10-V source as found from the open circuit test, with a series resistance of about 35 Ω (10 V from the open circuit test and 0.28 A from the
1 Ω test).
Let’s now investigate adding Y capacitance (from line to ground). This 4700 pF device has an impedance of approximately 13 Ω at 2.3 MHz (an assumed frequency of the ring wave measured in the 1 Ω termination test). We repeat the measurement to observe the amplitude of the waveform. Let’s also assume that the result of this measurement yields 1.3 V.
We now need to check our results. A 10 V noise source with a series impedance of about 35 Ω is the model for the source. The Y capacitor has an impedance of 13 Ω at 2.7 MHz. Solving for the voltage across the capacitor yields 2.7 V. The measured value across the 4700 pF capacitor is 1.3 V.
Although this looks like a huge difference in percentage, we are only off -6.3 dB from the calculations. The good news is the error is in the right direction.
So what do we know?
If we measure the conducted emissions using a LISN, we would see a value of only slightly less than 1.3 V. Our source impedance is still relatively low with respect to 50 Ω, i.e. with 1.3 Vocv and Isc of 0.29 A, Vocv/Isc = 4.5 Ω.
Our target voltage measurement value is 1 mV, so we only need 63 dB of additional attenuation. Is it practical to continue to add shunt capacitance or impedances? The answer is no. Even if we could add as much shunt capacitance as we wanted, the entire impedance, given an Isc current of 290 mA, would require the total shunt impedance of less than 3.4 mΩ. This dictates that a practical filter must be constructed of a cascade of shunt and series devices forming an AC voltage divider (see figure 3).


Figure 3: Multistage filter for MIL-STD-461 compliance.
For a good design we need to understand the impedance of every part and the potential interaction. It is good practice to keep the Q factor of the inductors and the equivalent series resistance (ESR) of the capacitors low, which provides good attenuation without creating a resonance, or peaking, as it is sometimes called. Layout of the filter is very important to avoid inadvertent parasitic coupling. For the example filter above, parasitic capacitance from input to output could easily be 1 pF, which is about 60 kΩ at 2.7 MHz. If there were no shunt impedance looking back into the filter, this would produce over 1 mV at the LISN, putting us above the limit on its own.
The filter impedance (looking into the input) as well as additional Y capacitance near C1 — either real or parasitic – help mitigate the effects of this parasitic coupling. It is important to note that inductive coupling will have the same effect. Good layout practice is imperative to prevent input-to-output and stage-to-stage coupling.
Having a filter precede a power component has the added benefit of providing attenuation to transient fluctuations in the source voltage. Short duration, high dv/dt events have little energy associated with them and the inductance and capacitance present in a filter is sometimes enough to integrate this energy by reducing the peak and expanding the time as it appears at the output of the filter.
Unfortunately, the power supply to the application , as defined by the standard, can frequently exceed the capacity of the input filter to mitigate these power excursions; and so additional circuitry may be needed to transform them in such a way as to not affect the power device. Transient immunity
MIL-STD-704 and MIL-STD-1275 refer to aircraft and ground-based systems that describe the anticipated power quality of those systems, and the levels a device must meet or exceed in order to perform satisfactorily in the anticipated application (see tables 3-6). Other standards may be required but are not covered in this paper.
As with MIL-STD-461, earlier revisions of these standards may be required, depending upon the installation. Be certain you know which one is being imposed, because the limits can vary greatly.


Table 8: Revision summary for MIL-STD-704 shows variations in 28 Vdc steady-state.


Table 9: Revision summary for MIL-STD-704 shows variations in 28 Vdc searches.


Table 10: Revision summary for MIL-STD-704 shows variations in 28 Vdc spikes.
As the tables show, 704F is readily met if the power device has a normal input range of 16 to 50 Vdc. In this case, no special precautions or circuitry are needed and an off-the-shelf device with a suitable input range will provide direct compliance to the standard.
If the application requires compliance with 704A, the off-the-shelf device may require additional protection, such as an input shunt transorb to clamp the spike to a reasonable level, followed by an active clamp circuit using FETs to reduce the voltage to the output to the maximum level the DC device can tolerate (see figure 4).


Figure 4: An input shunt transorb combined with an active clamp circuit can limit voltage to provide MIL-STD-704A compliance.
In figure4, Q1 is the main clamping element and must be sized appropriately to handle the power dissipation needed during the 80 V (for 50 ms) abnormal requirement specified in table 9. Obviously, if the downstream device can handle a higher voltage, less power must be dissipated Q1. D6 through D8 are 33-V, 600-W devices.
MIL-STD-1275 features a number of revisions, as well (see tables 11-13). MIL-STD-1275D presents a more severe requirement than MIL-STD-704 in that the surge amplitude and duration specification is 100 Vdc for 50 ms. As can be seen from the table specifying the variations in revisions, with the exception of the 600-V spike protection required by 704A, 1275D is the more stringent standard. Therefore, if MIL-STD-1275D is met, 704F is met and because the transorb handles the 600V spike, 704A is also met.


Table 11: Revision summary for MIL-STD-1275 shows variations in 28 Vdc steady-state.


Table 12: Revision summary for MIL-STD-1275 shows variations in 28 Vdc surges.


Table 13: Revision summary for MIL-STD-1275 shows variations in 28 Vdc spikes.
The circuit in figure 4 could be built using discrete components, and an EMI filter could be designed using the methodology outlined earlier, but doing so requires iterations of build, test, evaluate, and modify, dragging out the design phase of a project. To save time and ensure compliance, a ready-made filter module might provide a simpler alternative
Learning to design filters to meet MIL-compliance limits for EMI and transient protection can seem like a daunting task but the steps outlined in this document will allow you to design a filter from scratch. The process can be tedious and time-consuming, however. Depending on the application and the time horizon, you may find it more efficient to choose a filter module designed to complement the power component.
About the author
Jeffrey Ham is principal product line engineer in the Vicor Applications Engineering organization (Andover, MA). He has more than 20 years of professional engineering experience, with particular emphasis on power supplies and microwave systems. He received his B.S.E.E. degree from Northeastern University. He can be contacted at jham@vicr.com.
