Engineer‎ > ‎Electronics‎ > ‎

Electronics - Examples of filtering methods

Created 03-11-2015, Updated; 01-07-2018, 05-07-2018

Electromagnetic compatibility is and always has been wrapped in some mystique. Rules that people don't know or understand but must adhere too. Star Point, Bonding everywhere, coils in earth bonding that electricians and plumbers do. This part of electronics is art and is comparable to mechanical design and considering where to put a rubber coupling (soft magnetic bead or common mode choke or C+RC damping) Or stiff coupling (thick conductor) Or a metal spring (high Q inductor or Polypropylene capacitor)

If you are unsure then put options into your design such as zero ohm link resistors that can be removed to all your PCB mounting points. It will work better if designed in and left out later than left out and added later.
  • An old cheap fix for worn out car shock absorbers was too empty the fluid and replace it with treacle - this fix illustrates an understanding of filtering between road vibration and the car ride. This treacle fix makes the ride quite hard and sporty I am advised. The Advanced Passenger Train was pulled before the design was refined but a tilting train was developed decades later (Virgin Pendolino). Anyway the strategy for dealing with rail vibration in the APT was to have very lightweight boggies so that the wheels could bump up and return down to the track quickly - consequently to keep the boggies light there was very little oil in the gearboxBear these sorts of anecdotes in mind they can be apply to understanding electronics.

Drawn using OrCAD Capture 16.6 - like a London underground schematic diagram/map lines and blocks are not staggered but aligned making them plain to see but also like a map also placing things in realistic geographical locations. Works well for up to desktop size equipment. Not always ideal but a commonly used set of compromises that work well.

This page shows more ideal solutions than this diagram above and how the function blocks could be implemented. Hopefully I have labelled enough to be clear but unfortunately the diagram does look busy.

In general to make a low powered instrument CE compliant is fairly straightforward requiring minimal filtering but always good PCB design. But to do nothing will not only fail and be unreliable and in an unpredictable way. 
  • That is even if you do not bond everywhere you bond 0V to the casing near the i/o use common mode noise methods; differential amplifier with filtering, common mode choke or opto-isolation.
  • If you daisy chain other circuit boards only follow that strategy for one board distance from the and within.
  • If the PCB or module contains switch mode power supplies, motor drive do make sure that it is screen bonded to the casing and use the common mode filtering as shown below. Or at least enclose it and filter the Inputs and Outputs.

Near Fields

It is not possible to measure emissions or susceptibility with sources and detectors close to a printed circuit board under test. But it can be useful to measure things with near field probes and this will give a good indication of hot (radiating) components. I found that A near field magnetic loop and a analogue 30KHz - 30MHz valve voltmeter worked very well for me many years ago. But I only used it when EMC first became part of CE marking and learnt what I need to learn from the exercise.

To restate another point; 
Small metal enclosures can produce very unpredictable results. This is again put down to near fields. I do not know the mechanism of this situation but warn of it.

And in any case engineering is full of analogies to things we may assume a physicist understood. But they work and sound engineering is built on them because they fit well.

To restated some of what I said in the Mechanical design page; Contain high frequencies and susceptibility to near the source and point of susceptibility;
  • Therefore use an embed micro-controller rather than a conventionally bused microprocessor.
  • PCB canning with the lid ultimately soldered all around,
  • PCB with at least an 0V power plane,
  • Filtering on the PCB at the boundaries eg connectors,
  • Bandwidth limit circuit stages with active and passive filtering if necessary.

Circuit below; A & C are optional interconnections B & D are PCB's or a single PCB;

The diagram above will work well with just block D the PCB with all functions and I/O connectors mounted on that PCB. Achieving good results with an insulated or metallised plastic enclosure. It also works fine with a single 0V connection to a metal enclosure near the I/O provided the filtering and screening is contained within the PCB. But many connections to 0V is most desirable but can be bad in the case of a metal enclosure. I will show other ways below;
  • Put the OV or earth, input, output and power reasonably close together and any PCB close to where those enter the enclosure,
  • Place associated parts that are particularly sensitive the closest together,
  • Conversely keep power magnetics away from sensitive electronic circuits - distance first then add screening enclosures (for example PCB canning).
  • Use transmission lines strategy and ensure 0V planes are not fragmented as a PCB design strategy,
If it is necessary to put a cable between the PCB and the I/O connection insert block B or D.

For Automotive applications which are much more stringent than is required for CE marking insert block A. The capacitors add nothing but do increase the margin for compliance.

If the enclosure is metallised there is no good reason for not to make many 0V connections to the metallised enclosure. The metallisation will be resistive and therefore low Q thereby any voltage differential will not cause significant current to flow between 0V connections.
  • Rout wiring away from magnetics.
    • Screened enclosures (such as a PCB can) should also be kept a little distant from magnetics.
    • General rule; Aluminium, copper and gold for high frequency, mild steal or mu metal for low frequency,
  • The power line chokes shown above are not low Q, soft magnetic types (that I have recommended generally). Placing resistor in series or a resistor in parallel with the inductor would lower the Q of the inductor. I have not experiment with adding a resistor though I have placed pads to do that but using an easily available high Q inductor works fine.

Printed Circuit Board to board interconnection;

The diagram above shows an inter-board connection strategy. The filter is important for both compliance and good function. I found by experimentation that the Pie network formed between the PCBs provided the lowest noise strategy for connection to a very sensitive analogue circuity work for low frequencies and up to 1MHz and beyond. The reason why boards would be segregated other than physical necessity are to; Segregate comparatively noisy digital from very small signal analogue circuit by distance as well as electrical coupling.
Where interconnection is between components within a metal enclosure this strategy of bonding everything will almost certainly will not work and one of the PCB's must float but be well screened and filtered within itself. Such a strategy though frowned on works extremely well consistently (I have found).
  • To provide a good 0V for both power and noise use many conductors the width of a ribbon cable.
    • Power be fed through a choke on one of two interlinked PCB's. This pie formation with the choke one side connected to the interlink ribbon cable between capacitors - I have found by measurement of system performance works very well and better than other configurations when measuring system functional performance.
    • All other signals can be fed through a resistor facing out of the board and capacitors forming a similar pie like network.
    • Use a common mode choke for better speeds and to carry power.
    • The choke filtering is important with digital microprocessor and switch mode power supplies with analogue circuits - in those cases the choke power supply filtering is important.

Strategy for metal enclosures;

The diagram below shows a strategy for a metal case enclosure. Block B would ideally be the input/output connection and block D could be the same PCB but I have shown an interconnection C. There is no benefit in having the interconnection C and there is an advantage if the interconnection A were avoided. 
  • Where PCB's are all mounted on metal in addition place a common choke between each interconnecting cable.
    • The common mode choke must be a soft magnetic and must not saturate due current cause by any voltage differential across the metal work.
To say once again the capacitors shown in block B are not necessary but they help with meeting more stringent automotive industry standards if you are working to those standards the standards for CE compliance.
  • SEPIC switchmode power supply takes a little work to figure out a good track layout. In EMC terms this sort of converter also has the disadvantage that the power is switched on/off at chopping speed as opposed to switching high to lower current.
Each PCB must be connected 0V to the metal casing but it may be necessary to only fit one point. As I have said above in such a case that bond point should be near the connections to the PCB. The PCB must be a self contained as far filtering and screening in that case. 
  • L7 and L8 are not necessarily spilt other than because of;
    • The availability of common mode chokes,
    • They could be split though so by current rating, segregation of small signals and large signal for better noise performance, 
    • Or split because they are mounted on different parts of the units casing.
Some more General guide lines;
  • For high frequency * power product tracks keep them short fat and straight.
    • Therefore use traditional rules of thumb for lightening; Shortest path to earth (is a straight fat copper conductor) And balls diffuse electrostatic fields (reduce the risk of lightning strike) and spikes case electrostatic stress causing sparks and corona. Not that you need to design as a lightening conductor - which by the way a lightening conductor is not meant to conduct lightening but to conduct and reduce the electrostatic ionising field around the build so that the path to earth is somewhere else. 
  • Snubber damping and loop filtering applies to everything.
    • C+ R+C network is good such as 100n + (10n + 100R) - arbitrary values but depends on the situation.
    • Place them near the cause - the switching transistor and wound component termination to the power supply.
    • Where DC isolation is required eg PCB 0V plane to each mounting to metal case.
  • Same same damping circuit is applicable to control loops including software control.

Internal interconnection;

This uses the same filtering strategy as the metallised or insulted enclosure above but with a common mode choke are added.
The last diagram above shows a PCB interconnection strategy for a metal enclosure. There is no problem that I have experienced with using star point similar to the first two diagrams above. Both strategies work with small signals. But this "correct way" does allow you to mount PCBs on to metal surfaces for cooling without concern for noise coupling electrostatically into sensitive circuit. the risk with star point are comparable with using floating electronics within a metal enclosure - My general experience below;

The power line chokes allow one PCB to have a much cleaner supplies than the other PCB which maybe source supply. This is important were the object of segregation also includes the need to put distance between noise sources and small analogue signal circuits. But if you do not include such chokes one PCB becomes a flexible extension of the other and will have largely the same power supply noise level and that may be entirely suitable
  • Floating electronics - so that isolation to earth can be achieved conveniently, But;
    • My experience is that it does not work with very high impedance circuits mains 50Hz is electrostatically coupled in to the circuit.
    • This strategy may be okay with lower impedances and where sensitive circuit is very small and on a ground plane. But providing opto or transformer isolation is the only safe way.
  • In Low Voltage Directive systems an approved type capacitor to the casing is permitted. 
    • I would expect but I have not tried it but to not just adding a capacitor but adding an C + R+C network between 0V and case earth would be better. As I have said though Single point RF bond would be quite effective as is specified in such industrial standards but it is not the best way. Such capacitive coupling of cause leaves high impedance sensitive circuits to low frequency electrostatic coupling. Conversely such high frequency coupling to 0V will minimise the likelihood of coupling low frequency components modulated by high frequencies and this is significant (I have conclude from other observations). 
I have describe 0V strategy but many approaches amount to bonding everywhere to 0V, Bonding one point to 0V case and running 0V wires to every PCB or block, or daisy chaining one to another to another. You should recognise some of each in what I described But not truly a star point strategy.

Star point is very important in protective earthing - using crimped wire, nut bolt and anti-shake washers just for the safety earthing and the bolt sole used for earthing but not construction. The Safety earth path made to every metal component and wired back to one common safety earth incoming point should be done. Then for screening applying many bonds to every metal component.


RF design principles can give useful clues to how and why things work out well. A 50 ohm track on a double sided PCB may require an inconveniently wide track width but if the tack is very short this may not compromise the design significantly. Here is a useful link that I was given;

My blog for discussion - Trade-off for safety and EMC in mains voltage power supplies; Electronics - high-frequency metal-vapor-arc-lamp power supply