Introduction
Conducted emission is the easiest electromagnetic compatibility test to set up and the easiest to get wrong. It measures the noise a product pushes back into the mains between one hundred and fifty kilohertz and thirty megahertz, and the result depends strongly on the impedance of the supply, so the same unit reads differently in a different socket. The line impedance stabilisation network exists to remove that variable. It isolates mains noise, presents a defined impedance to the equipment under test, and couples the disturbance voltage to the measuring receiver. This article explains its three roles and the meaning of the fifty microhenry inductor, gives the method for separating common mode and differential mode paths, explains the difference between quasi peak and average detection, ties together the full chain from excess level to required filter insertion loss, and closes with the set-up points of a pre-compliance test and four common mistakes.
1. What conducted emission measures: the band, the path and why impedance has to be defined
Conducted emission measures the disturbance voltage that the equipment under test pushes out through its power leads, over the standard band from one hundred and fifty kilohertz to thirty megahertz. Above that band the radiation efficiency of the cable rises, the disturbance travels mainly through space, and it is controlled instead by the radiated emission requirement. That boundary is not a physical discontinuity but a line the standards draw to assess the two phenomena separately.
Within this band the disturbance comes from two quite different sources. The pulsating input current of a switched mode supply is the classic differential mode source, concentrated at the switching frequency and its low order harmonics, while coupling through stray capacitance to the reference ground is the classic common mode source, including the capacitance between heat sink and chassis, between transformer windings, and between the cable and ground. The two have different spectral distributions and different suppression components, so deciding which one dominates is the first step of any fix.
Why does the supply impedance have to be defined? Because a disturbance voltage is a drop produced across an impedance, so a different impedance gives a different reading. Mains impedance varies violently with location, time and load, and the same unit can differ by more than ten decibels between two sockets. The role of the line impedance stabilisation network is to present a defined, frequency independent impedance at the equipment terminals, to keep the noise already on the mains outside, and to deliver the disturbance voltage to the receiver through a known coupling factor.
2. Three roles of the network and the fifty microhenry inductor
The standard sets three functions for the network. First, present a defined impedance at the equipment terminals over the radio frequency range. Second, isolate the test circuit from unwanted signals on the supply mains. Third, couple the disturbance voltage to the measuring receiver. All three are necessary, because without the first the measurement cannot be reproduced, without the second mains noise mixes in, and without the third there is no usable measurement signal.
The fifty microhenry value is a compromise. It presents a high impedance toward the mains and keeps mains noise out, while on the equipment side it works with a capacitor network to form the standard impedance curve. The standard defines several networks by band. Fifty ohms plus fifty microhenry plus five ohms for nine kilohertz to one hundred and fifty kilohertz, fifty ohms plus fifty microhenry for one hundred and fifty kilohertz to thirty megahertz, and a low impedance fifty ohms plus five microhenry plus one ohm for vehicle and high current setups. Which one applies is decided by the product standard, not by the person running the test.
Three usage details are easy to miss. First, the network carries a rated current and a peak current, and the inrush of a capacitor input rectifier can exceed the peak rating, after which the measurement is no longer valid. Second, the measurement port must be terminated in fifty ohms, and an unused output must be terminated as well. Third, a ten decibel attenuator is usually built in or added externally, and its attenuation is part of the voltage division factor, so it must not be dropped from the reading conversion.
3. Common and differential mode: two paths and three ways to tell them apart
Differential mode is where current flows out on one conductor and returns on the other, in the same direction as the power current itself, and it is produced mainly by the pulsating current of the converter. Common mode is where current flows in the same direction on both conductors and returns to the reference ground through stray capacitance, and because it depends on layout, heat sink and distance to ground it is the hardest to predict and the hardest to fix. The two measurement outputs of the network each measure one line against ground, so what the receiver sees is a mixture of both modes.
There are three ways to tell them apart and they cross check each other. The first is by frequency, with differential mode dominating at low frequency and common mode above a few megahertz. The second compares the two lines, because if the disturbance levels on line and neutral are similar then common mode dominates, while a large difference points to differential mode. The rigorous method applies arithmetic to the two measurement outputs, with common mode equal to half the sum and differential mode equal to half the difference, which gives two separated curves directly on the instrument.
The third is to add a component and see. Place a capacitor of zero point one to zero point four seven microfarads between the two lines, and if the low frequency end drops the problem is differential mode. Add capacitors from each line to ground or a common mode choke, and if the high frequency end drops it is common mode. This is the fastest method, but safety and leakage current constraints apply, because a capacitor across the lines has to meet the safety standard and capacitors to ground cannot grow without limit because of leakage current.
4. Limit lines and detectors: quasi peak and average are not the same thing
The limit tables are given by equipment class and by band. Taking multimedia equipment as the example, class A carries a quasi peak limit of seventy nine decibels above one microvolt and an average limit of sixty six over zero point one five to zero point five megahertz, while class B over the same band falls logarithmically from sixty six to fifty six in quasi peak and from fifty six to forty six in average. Over zero point five to five megahertz class B is fifty six and forty six, and over five to thirty megahertz it is sixty and fifty. These values are the pass criterion, not a design target.
The difference between quasi peak and average detection lies in the weighting. Quasi peak weights by repetition rate, giving more weight to disturbance that repeats often, which matches the subjective impression a listener has of broadcast interference, while average simply takes the mean of the envelope. The same trace can read more than ten decibels apart on the two detectors, so it has to be established which one is exceeded. In practice most failures touch both, and the fix is usually targeted at the quasi peak value.
The uses of peak, quasi peak and average also have to be separated. Peak sweeps fastest and suits fast fault finding in pre-compliance, quasi peak is the basis for the formal verdict but sweeps slowly, and average is more sensitive to continuous disturbance. A standard pre-compliance routine is to sweep once with peak to find the suspect frequencies and then remeasure those with quasi peak, which balances efficiency against reliability. Concluding from a peak result as if it were the verdict is one of the most serious common mistakes.
5. Sizing a filter from the excess: a chain that can be quantified
The core of a fix is converting the excess into a required insertion loss. The method is direct. Measure the disturbance spectrum of the equipment with no filter through the network, read the difference between the measurement and the limit at each failing frequency, and that difference is the insertion loss needed at that frequency. If a switched mode supply shows quasi peak and average levels of eighty four decibels above one microvolt at its three hundred kilohertz switching frequency against a limit of fifty, then at least thirty four decibels of insertion loss has to be added at that frequency.
The next step is allocating that insertion loss between the common and differential mode paths. If the separated measurement shows common mode dominating, the burden falls mainly on the common mode choke and the capacitors to ground, with only a small line to line capacitor and a small differential mode inductor needed. The allocation ratio is not guessed, it is set by the excess shown on each of the two separated curves. That is also why separation is not optional. Without it the only option is blindly growing the filter, and both cost and volume run out of control.
The real performance of a filter also depends on source and load impedance, which is where insertion loss figures distort most easily. The published insertion loss is usually measured fifty ohms to fifty ohms, while in the real circuit neither the supply side nor the equipment side is anywhere near fifty ohms, so the actual attenuation can differ substantially from the nominal figure. A design therefore needs more than six decibels of margin and has to be remeasured in the complete system rather than trusting the curve in the filter datasheet.
6. Pre-compliance set-up points and four common mistakes
The goal of a pre-compliance test is to find in advance where the formal test would fail, so the set-up should reproduce the standard conditions as closely as possible. The relative position between the equipment and the network, the size of the reference ground plane and the distance between the equipment and that plane are all fixed by the annex of the product standard, and those distances must be taken from the standard being reported against rather than from a generic checklist. The earth terminal of the network itself should be bonded to the reference plane with a wide short conductor, which matters most at high frequency.
Three further points on the measurement settings. First, optimise the vertical resolution without overdriving, so if the disturbance peak to peak level is eighty five millivolts the vertical scale should be set to a full scale of one hundred millivolts to give the highest sensitivity. Second, convert the time domain data to the frequency domain with a fast Fourier transform and use decibels above one microvolt on the vertical axis where possible. Third, the worst disturbance usually occurs at maximum output load, so the test has to cover the worst case load rather than no load only.
Four common mistakes deserve their own list. The first is treating the peak reading as the verdict. The second is ignoring the current rating of the network, because once inrush exceeds it the measurement is completely invalid without anyone noticing. The third is making the line to line and line to ground capacitors too large, which pushes leakage current over its limit and creates a new safety problem, the classic case of fixing one issue and raising another. The fourth is concluding from a single run, because disturbance changes with load, temperature and ageing, so after a fix every operating condition has to be remeasured and the filter parameters written into the bill of materials before the design is frozen.
Figure Notes
Figure 1｜Differential mode versus common mode disturbance pathsFigure 2｜Typical conducted emission limits for a class B mains portFigure 3｜Conducted emission spectrum against the limit line before and after a fixFigure 4｜The fix flow from excess level to filter parametersFigure 5｜Five operating disciplines for conducted emission testing