Ferrite in EMC applications
(published in Electron #8, 2005)
As usual the motive to this article was to found in my home environment. I like to be (radio-) active on the HF amateur bands and did not very much appreciate a sudden and very high noise level of over 50 microvolt to appear at many frequencies below 10 MHz. Such sudden bursts might be generated locally and since I have a back-up power facility for my radio-shack it was easy to throw the main(s) switch of the house to get a simple picture about any in-house generators. Indeed the burst disappeared and by successive switching-on of power groups the noisy system exposed itself rather quickly.
With the noise source localized, attacking this was the next assignment. As often is the case with EMC problems, packing the noise source in metal was no option since this local noise generator happened to be the electronics in a lap-top docking station of one of the housemates. Other measures had to be taken and if you regard some of the other articles on this web-site it is obvious that a solution with ferrites was more appropriate.
Some of the articles on this web-site are discussing general aspects on Ferrites in HF applications. For this particular issue some of these general aspects had to be inspected more closely to attack this EMC problem.
Electro-magnetic compatibility (EMC) issues arise when (electronic) systems interact in an un-desired way. This always means that at least two systems are involved, of which one is (undesirably) sensitive to the other, which is (undesirably) acting upon the first. A number of measures can be taken to stop or diminish this unwanted interaction, depending on the nature of the active source and/or passive receiver and the channel through which the interaction operates.
Best way is always to tackle the problem at the active source, since this usually is a well defined and spatially restricted emitter. Once 'out-of-the-box' there usually are many channels to broadcast the unwanted message whether through radiation or conduction. Also more than one receiver may be sensitive to picking this up of which you probably only detected the first one with many more to follow (neighboring equipment!).
Even if the disturbance arrives through electromagnetic radiation, there always is a generating source (transmitter) and a radiating transducer (antenna) involved of which the transmitter usually is the spatially smaller 'component'. In many situations however the transmitter signal may only be slightly diminished as the undesired transmission is a byproduct of a different and often powerful, desired action. This is especially true in case of unwanted harmonics of a desired fundamental action. Diminishing these harmonics may only be achieved up to a certain level.
When all measures to the source have been taken and the problem still has not been resolved, the transmission channel has to be attacked by blocking or at least heavily damping. At radio-active sources, cutting the 'antenna' usually is the first and best option again, with screening-off (packing in metal) the entire transmitter plus antenna as a last resort.
The transmission channel may also consist of an unwanted conductor between the source and the receiver. Damping again is the best way to shut-off this channel and this is what we will mainly focus on in this discussion.
A basic model of an EMC 'system' may be found in figure 1. To narrow the field of discussion somewhat, it is conceivable the EMC source to consist of say a switched-mode power supply of which the internal switching action does not allow to extend rise and fall times too much for reasons of switching efficiency. The powerful harmonic content of such a signal will extend over a large range of frequencies, effecting all sorts of 'receivers' of which audio and video equipment are only a few that may not be too sensitive, but our HF receiving equipment usually is.
Since this particular source may not be switched off, next action could be to keep all unwanted energy inside the box, for instance by screening-off this power supply in a metal cage. In a fully enclosed box this might be a successful action, but usually such power supplies are connected to the mains supply on one side and a DC-power outlet at the other, that have to leave the protecting environment of the box. At this position an number of new channels are being created. The wiring may act as an antenna by which the unwanted signals are radiated and reach the HF receiving antenna through the air. The wiring may also act as a conductor to bring the noise signals directly to the receiver when both circuits are connected together through the mains supply or through the (DC) output power. In practice both channels may act simultaneously making the picture somewhat more complicated.
In general it is clear that attacking the unwanted signals at the source is usually the best option and may simultaneously attack the radiation as well as the conduction channel.
For the rest of this article the main focus will be on the damping action in the signal channel and especially the application of ferrite materials to such actions. In this type of application, the signal channel usually is of the conduction type, either from the source to the antenna or directly from the source tot the receiver.
We should be aware that at this type of damping the impedance of the damping device is important together with the impedances at the source and at the receiver. Together these impedances make up for a power divider that determine the amount of signal to reach the receiver. The total amount of damping is more important than the exact phase of the interfering signal, so all type of impedances may fulfill this damping function whether these are inductive, capacitive, resistive or anything in between. So long as total damping is high the amount of interfering signal power may be effectively be diminished.
Ferrite materials will be applied when a high impedance ('AC resistance') should be obtained at a simultaneously low DC resistance. This high impedance is made up from a reactive and loss component and my be written as
ZL = 2 . p . f . n2 . μ0 . (√(μ'2 + μ"2) . A / l (1)
f = frequency
n = number of turns
μ0 = absolute permeability = 4 . p . 10-7
μ' = material permeability (ferrite material quality; numbers by manufacturer)
μ" = material loss (ferrite material quality; numbers by manufacturer)
A = magnetic area (core shape quality; numbers by manufacturer)
l = magnetic length (core shape quality; numbers by manufacturer)
More information on the particular ferrite qualities may be found in the series of articles on Ferrites for HF applications.
In this formula a number of parameters are changing with frequency, e.g. f (frequency itself), μ' and μ", the latter may be found in figure 2.
At low frequencies (100 kHz.) this 4A11 (43) type of ferrite material starts-off at a permeability (μ', purple curve) of around 700, to rise to around 900 at 2 MHz. after which it will drop-off quickly to around 60 at 30 MHz. An inductor made on this material with an inductance of e.g. L = 100 μH, would exhibit a reactance of 62,8 Ohm at 100 kHz., of 1621 Ohm at 2 MHz. (L = 129 μH.) and 1615 Ohm at 30 MHz. (L = 7 μH.) This however is only part of the story.
In figure we also find the loss curve (μ", red curve), that is negligible at 0,1 MHz, to rise to maximum at around 5 MHz. and dropping off at higher frequencies. The graph of figure 2 with μ' and μ" usually are provided by the manufacturer. For total impedance as in formula 1, the values of μ' and μ" have to be added as vectors, the factor between the brackets. To determine total impedance this vector summation has been applied in figure 2 as μc , the blue curve.
This blue curve is starting off almost equal to the purple curve, rising to a maximum value a little before the cross-over of the μ' and μ" curves (ferrimagnetic resonance) after which it is dropping off to almost equal to the red curve.
Around this ferrimagnetic resonance frequency, reactance and loss are about equal which means this ferrite material is no longer qualified for resonating circuits because of high loss (Q < 1) and around the same frequency it is also becoming less useful for application in flux-transformers, for the same reason.
For choking applications the material still
is very useful as may be appreciated from the green curve which depicts total
impedance of a one turn inductor at
Since impedance keeps on rising for a long time after the material is almost entirely lossy, this ferrite material is very much qualified for choking applications, i.e. as a component to reduce EMC problems.
A simple calculation
Let's perform some calculations using the general EMC model. For this calculation we take 50 Ohm as the input impedance of the disturbed system and a choke-on-ferrite as the damping actor. In the complete model the output impedance of the EMC source should be taken into account as well. In practice this impedance is in series with the damping factor that usually is very much larger. For this simple calculation the output impedance will therefore be left out of the equation. The picture of figure 1 now evolves into figure 3.
The voltage at the EMC source is Us and the voltage at the EMC receiver input to be Ui, the effective damping of the signal from the source to the receiver input is:
Us / Ui = |50 + ZL| / 50 (damping is the reverse of the voltage division)
The section between vertical lines is to be taken as the absolute value (root of squares of factors) as this depicts a vector summation. If one of the factors is more than three times the other, the last one is becoming insignificant and may be omitted. If we need for example a damping factor of ten times, we find for the damping impedance:
10 = ZL / 50, from which ZL = 500
The choke impedance may be calculated
directly form the green curve in figure 2, that is depicting total impedance
of a one turn choke at a
If we require a damping impedance ZL = 500 at a lowest frequency of 1 MHz., and we are aware this impedance is scaling with the number of turns squared, we calculate the required number of turns from the graph of the one turn inductor according to:
n = √ (500 / 7) = 8,5 (9).
Effective impedance then becomes: ZL = (9 / 8,5) 500 Ohm = 532 Ohm, for a damping factor of 10,6 x., that we like to express in deci-Bells according to 20 log <damping>, so 10,6 x at 1 MHz is equal to 20,5 dB
At a frequency of 10 MHz. we again may be using the values of the one turn coil:
ZL = (40 / 7) x 532 Ohm = 3043 Ohm, for a damping factor of 60,9 x, in dB: 35,7 dB and at 30 MHz.
ZL = (60 / 7) x 532 Ohm = 4560 Ohm, for a damping factor of 91,2 x, in dB: 39,2 dB.
If our receiver from the beginning of this story had been showing an S9 noise signal at the input (50 μV over 50 Ohm) at 10 MHz., and we would have applied the damping choke we just calculated, the new noise reading at the S-meter would have been S3 (0,82 μV over 50 Ohm), which usually makes this noise disappear into the atmospheric back-ground noise.
Now we found the powerful choking properties
of this 4A11 (43) type of ferrite, we might be curious as to other ferrite
types. Therefore I measured the damping factor of different types of ferrite,
all of the
In figure 4 we may notice a number of interesting events.
Let's start-off by looking at the purple curve for 3E5 ferrite material. According to the manufacturer this material exhibits a very high initial permeability (μi = 12000) and is meant for LF applications. The graph is showing this material to exhibit highest damping around 700 kHz. to go down slowly at higher frequencies. Still at 20 MHz. we notice damping figures that are only half of the peak value, meaning this still could be applied successfully for choking purposes. Also the older 3E1 type (green) may still be around at many places and is a very good material for a wide range of frequencies.
In figure 4 we also find the better material for the lower HF frequencies is 3C11 with 3E25 as a good runner up. Best material for middle and higher HF frequencies is 4A11 (43) type. It may come as a surprise that well known 4C65 (61) type is not so favorable at HF frequencies for choking actions and is becoming interesting starting at 25 MHz. and above.
Since a wire through the hole in a toroide already accounts for a full turn, combining ferrites is effective in combining the qualities of various types of materials to obtain overall properties as desired. I tested this idea by measuring various ferrite combinations, again as one turn inductors. Most interesting combinations are depicted in figure 5.
In figure 5 it may be found the well known 3E1 + 4C65 combination not to be the most optimum around and is even showing a 'valley' around 10 MHz. Even the purely LF combination of a double 3E25 toroide is performing better, almost up to 30 MHz.
Optimal combination appears to be 3C11 at the lower range plus midrange 4A11, with 3E25 plus 4A11 as a runner up. Where 4A11 type is not available a combination of 3C11 plus 3F4 or 3E25 plus 3F4 also will do a good job for the larger part of the HF range between 1 and 30 MHz.
For still higher frequencies, 3E25 type is less than optimal since this is not performing satisfactory anymore around 100 MHz. and up. Nevertheless this material is doing a great job for lower frequencies which is surprising considering this to be a presumably LF type of frequency
Next I constructed a common mode filter for EMC testing purposes.
Such a filter is a practical tool for quickly determining sources of radiation with the mains connection as an aerial or direct conduction path. The filter is of the common mode type with differential currents at the mains frequency to pass undisturbed since these cancel in the core and common mode currents to be blocked by the choking action.
As it happens, 'standard' three-conductor
mains cord easily passes with seven turns through a
In figure 6 it may be seen the 3C11 + 4A11 filter with 7 turns of mains wire to exhibit a nice and constant 25 dB of damping between 0,5 and 30 MHz. When comparing the damping over frequency behavior of this combination with the same combination at the one turn 'choke' of figure 5, the difference shows at 30 MHz. and above where the 7 turns choke is degrading. Background to this phenomenon is parasitic capacitance across the choke.
A quick calculation is showing the importance of this parasitic effect. For a damping of 25 dB (17,8 x re 50 Ohm), the impedance of the choke should be: 17,8 x 50 Ohm = 889 Ohm and a parasitic capacitance of 7,2 pF at 25 MHz. already is equaling this. Of course when the parallel capacitance resonates with the choke inductance , total impedance will go up although this is not a prominent effect since the choke is very lossy at this frequency. After resonance, the capacitor is becoming the dominant parameter which takes total impedance down at 6 dB / octave to end choke performance.
To compare I constructed the same choke, this time on the 3E1 + 4C65 combination, since this is still widely regarded as a winning combination. In figure 6 it is obvious the latter combination is inferior by a factor two (6 dB) at 10 MHz. Above 20 MHz. the same parasitic capacitance is ending the choke action which makes this combination overall a lesser performer.
Next I constructed the common-mode choke with the 3C11 + 4A11 combination, this time using ten turns of twin mains cord, with the ground safety wire as a separate conductor. It is clear overall damping is increased by a factor of 2 (6 dB), but falling-off now starts at 15 MHz. because of the somewhat larger parasitic capacitance and also because the same capacitance has an larger effect at this higher impedance (upper curve).
A well known suggestion is to diminish the parasitic effect by first winding half the number of turns and then crossing over to put the next number of turns in the opposite direction. The idea is that input and output of the inductor are staying further apart, thereby diminishing total parasitic capacitance. As it happens, total performance of this choke was somewhat worse, probably since the crossing-over wire is 'seeing' more turns of increasing voltage before the next windings start at lowering this again. Cross-over windings to diminish parasitic capacitance therefore appear to be not such a good idea.
The final twin-cord choke has been constructed making each next turn at the inside to partly overlap the previous; this will allow a choke with a total of 10 turns with a relatively large distance between first and last turn which are carrying highest potential difference (largest effect on parasitic capacitance). This 10 turns choke is trading 6 dB more damping over the largest part of HF frequencies for 1,6 dB less performance at 30 MHz. when compared to the seven turn filter.
From above tests and measurements we may conclude best common-mode mains damping choke to consist of:
- combining 3C11 and 4A11 type of ferrite
- maximum 10 turns of twin mains cord with the safety wire as a separate conductor
- ensuring a relatively large distance between first and last turn, this not to spoil later-on by allowing input to cross output wires.
Such a common-mode test filter will ensure around 30 dB of damping over the entire HF frequency range. Picture 1 is showing this EMC test tool.
To get an idea about the performance of a dedicated mains-filter, I measured an metal encapsulated filter of comparable volume and quality ( 2A / 250 V). This filter, by the brand name of Corcom, was laying around in the junk-box for some time, so probably not the latest and best around. This type of filter usually consists of pi-type of configuration (C's to 'ground', choke in series) for each mains conductor. Figure 7 is showing the performance of this filter with the EMC-testing tool as a reference.
From figure 7 it may be appreciated a dedicated mains filter to far outperform a simple EMC choke, although the latter may serve its purpose quite well as a quick diagnostic tool and even final solution for various EMC problems (loudspeaker wiring).
Again the effect of the inevitable parasitic capacitance is obviously degrading filter performance starting at 5 MHz. and beyond. At the extreme frequency of 50 MHz. performance of the Corcom filter is approaching the EMC tool as 'external factor's' (parasitic capacitance) are determining filter behavior.
For this type of dedicated mains filter to reach maximum performance, better make sure the connection between the encapsulation and the cabinet is as short as possible and also the mains wiring to reach the filter inside the cabinet, to avoid parasitic coupling. Mains filters therefore often are constructed as an encapsulated box with the encapsulation directly connected to the cabinet and the connection to the mains as socket on the outside of the cabinet.
Damping and transmission-line transformers
At this stage it is instructive to make a link to transmission-line transformers, see also the dedicated chapters.
An important factor to the operation of this type of transformers is to isolate the input from the output at every transmission line. To this extend, everything as discussed around choking performance is directly applicable to these transformers, including very wide band performance using not-so-very wide band materials, as long as input to output damping remains high. This is the ultimate reason why transmission-line transformers may operate over a wider bandwidth than flux transformers, that rely on the core magnetic field to transfer energy. To the latter, core loss is a direct disadvantage, where this as at least insignificant and usually advantageous to the performance of transmission-line transformers.
Although chokes on ferrite core may be applied in a wide variety of systems, a few limiting factor still apply. These limiting factors are discussed more extensively in the articles on Ferrites in HF applications. A few dedicated lines on maximum power loading however may also be useful in this chapter.
Power loading should always be checked at
chokes in transmitter output stages and all other positions with large HF
voltages and currents. Since chokes are being applied for a wide frequency
range especially beyond ferrimagnetic resonance, high choke impedance is
maintained in particular because of the high loss factor into which power may
be dissipated. Maximum choke voltage for
maximum allowable internal core dissipation of 4 Watt in a
UL(dissipation) = √(4 . total impedance)
This simple formula holds for all ferrite materials at HF frequencies, that are applied beyond their ferrimagnetic resonance frequency, which means all materials with an initial permeability of 700 and higher and above 1 MHz. Below this resonance frequency, materials may be loaded to a higher extend and the complete loading formula should be applied as derived in Ferrites in HF applications. These higher loading factors explicitly apply to materials for higher HF frequencies e.g. 4C65 (61) type of material with permeability of 150 and below.
As an example for
'standard' choke ferrite we may find damping is 31 dB at 3 MHz. (factor 35,5)
relative to 50 Ohm, for the ten-turn choke on a
UL(dissipation) = √(4 . 1800) = 84 volt.
This voltage would make this choke perfectly fit to be switched in parallel to the output of a 100 W. HF transceiver, that will generate 70 Volt across 50 Ohm in a continuous mode.
Note 1: This maximum dissipation voltage is calculated in parallel to the choke as a common-mode voltage. In a mains choking situation, the mains voltages appears between the conductors, so in a differential mode. This voltage is only limited by the isolation strength of the (twin, mains) wires and does not influence choke limits.
Note 2: The common-mode filter as in our EMC tool is consisting of two different material toroides, each with its unique frequency dependent qualities. These have been selected to each be optimal at a particular frequency range, with that particular ferrite material taking most of the action. In between, where both materials are acting together, the total load 'burden' is shared between both materials so total load may be higher than calculated (twice).
Many ferrite manufacturers tend to drift away from color coding toroides, and so material distinction has to be guaranteed by the reseller. Still many color coded materials are around in appliances and junk-boxes so a table listing the coding scheme as handled by Ferroxcube may still be useful.
Since color coding is on its way out, it is always useful to determine material type and quality locally. At "Ferrites in HF applications" methods have been described to this extend, that are simple and may easily be performed at home.
Thanks for asking. The the wide band noise problem that started-off this article indeed could be solved after applying the choke techniques to the mains connection lead.
Bob J. van Donselaar