1
\$\begingroup\$

EMC/EMI standards (conducted/radiated emission/susceptibility) are well established for system boxes. However there appears to be no standards, nor methods for deriving a specification for internal modules that are inside the RFI filter protected Faraday cage.

I am looking at frequencies below lambda/100 relative to the module size. I am particulalrly concerned with sensors that are noise floor limited where it is difficult to detect additional injected noise, or noise that is exported to other modules within the enclosure.

Being below lambda/100 the coupling will be essentially near field i.e. capacitive, and occasionally inductive, and the main threat will be the chassis itself because of the system box's RFI filters (where the Y capacitors induce half supply noise on the chassis relative to the '0V' supply return).

Possible modules include pluggable PCB's, Camera and Image Intensifier modules (which are near photon noise limited and will alias and mix any injected noises).

I'm looking for documented/public methods and test procedures that can be used as formal references for a Statement of Work and for sub-Module specifications.

\$\endgroup\$
4
  • \$\begingroup\$ Have you considered that EMC can be a systems level problem, with "spooky action at a distance" taking place due to cavity effects etal? \$\endgroup\$ Jun 30, 2017 at 3:38
  • \$\begingroup\$ @ThreePhaseEel, I've looked at the systems issues, and part of those issues is the lack of a way of doing the flowdown that is "to a standard", and therefore something a supplier would know about in advance. At the moment they simply deny any knowledge or responsibility for the susceptability, and even if a flowdown is attempted they (essentially) deny the validity of the flowdown (in the nicest possible way, obviously). Having a standard allows both parties to rest on a common foundation. It is all scientifically predictable once the stray components are noticed;-) \$\endgroup\$ Jun 30, 2017 at 6:44
  • \$\begingroup\$ By the way a good way to mitigate ground noise is proper placement of voltage regulators\voltage references. \$\endgroup\$
    – Voltage Spike
    Jul 6, 2017 at 22:56
  • \$\begingroup\$ @laptop2d, the question is: how to specify the requirements. If we can't specify what 'proper' means, with a testable requirement, the issues won't diminish. Most of the fixes happen after the fact, and often at the contractors cost because of the change to the sub-contract. \$\endgroup\$ Jul 13, 2017 at 15:17

3 Answers 3

1
\$\begingroup\$

You probably won't find a public spec on near-field EMI as this is a complex geometrical and complex math curl product of emissions * susceptibility threshold * the mutual coupling impedance loss between.

You are looking for near field EMC specs and these are too variant with sub-system design geometry and shielding. System level tests usually done at 10~30 m outside and inside the box you can sniff with near-field loop probes from 10~30 mm for current noise spectrum using various size loops that trade-off sensitivity with bandwidth.

I have designed many complex systems with EMC issues and resolved all of them with instruments such as solenoid-pneumatic robotics with Eddy Current guided at 100k/200kHz as well as designed and performed Corporate System level susceptibility tests for conducted ESD and power line noise, Magnetic current levels from 10Hz to 30MHz , E field interference levels with AM over RF up to microwave with high E levels, ESD impulse conducted to frame and radiated from a plane wave conductor sheet.

These were done on a variety of products such as magnetic disk drives, DS1 transceivers, and thresholds were established by changes in analog window timing margin from injected jitter ( starting from >30% minimum margin) which translates to digital bit error rate (BER) <<1e-12 or very high SNR's.

Initially in the 70's and early 80's I used real-world noise generators like brushed DC motors and ESD discharge generators and then refined the process with dV/dt and dI/dt controlled spectral density from some fundamental swept frequency clock rate into the late 80's.

What I discovered was that the effective shielding of Disk drives was great but the ingress on long cables was poor which became most apparent with AM modulated 1GHz being rectified by ESD protection diodes and other non-linear impedances. Meaning feedthru caps were a necessary part of the design.

But in your case it appears to be surface currents on the photo-tube and interface cable power noise with sensitive circuit impedance or perhaps conductive supply ripple vs frequency which then is limited by decoupling cap ESR.

We also discovered the ground inductance and resistance of earthed coaxial CATV cable greatly affected the ingress from motive power trains within a block of the parallel coax and this affected RF video signal integrity on analog TV.

So my advice is perform margin tests to swept frequency E fields and H fields with linear and loop antenna using modulated carriers to simulate CMOS transition speeds or use square waves in to capacitive loaded loops to define your modules sensitivity on cable interfaces and shielded modules. Then you have measureable criteria to determine allowable limits at a specified short range perhaps using 1cm gap and a 10cm wire or a 10cm Loop at a 5cm gap.

Also recognize that slots or gaps are just as effective as antenna between module shields as linear wires but ultimately your geometry and your sensitivity to ingress ultimately affects your acceptance and failure limits. These ought to be measureable in some analog form such as % margin, SNR , voltage ripple, or spectral density below the wanted signal at some defined power level, area and gap.

You can get fibre optic E-field meters normally used inside Faraday Cages for measuring such sub-system egress or applied external ingress but remember that Helmholz Resonators from standing waves inside a shielded enclosure will create nulls on an E-Field detector but limiting these anomalies, you can compare applied noise levels with a simple short whip antenna or wire for calibrating your susceptibility tests.

I recall 30 yrs ago we used more than 10V/m and more than 20A/m in AC to RF susceptibility tests, but don't hold me to that value... It's been a while.

When we didn't have specs, we took responsibility for assuming the worst and designed then verified the results.

( The only time I recall failing was not being able to live test a telemetry SCADA system which I designed in the 70's with a 1 mile RS485 cable and launch control's AM RF radio caused bit errors when used causing unexpected results, long after I left the company for a 25% pay raise.)

Regarding Line filter noise injected into the chassis gnd. Proper isolation of this sub 10kHz noise can be treated with suitable RF isolation circuits to your internal shields such that they become 0V RF gnds in a star configuration or controlling the discharge currents impedances so they exit the box rather than radiate inside. External cable ingress is often your worst nightmare from ESD and power line floating switchers with coupling capacitance in the switcher transformer. But keep in mind Hipot tests with grounded secondary is not done by an OEM supply and this is more stressful on primary insulation for the same reasons ( SMPS XFMR Coupling C)

This probe consists of a small coil which is shielded against electrical fields and isolated on the outside for safety reasons. Two examples are shown below 1D and 3D EMI sniffer probes. The first probe has a linear coil, causing it to be most sensitive across the length of the probe. The second probe has three coils, which are perpendicular to each other and electrically connected in series. Its sensitivity to H-fields is independent of the field orientation. It depends on the particular case to determine which probe is best, but in general, the 3D probe is used to “scan” the application for locations with high field intensity, and the 1D probe is used to get a more precise picture of the nature of the problem. enter image description here enter image description here

If you are really pushed for time and desperate for a spec, use the Automotive sub-system specs for EMI and susceptibility in your location or use USA MIL-standards which with suppliers for references were the earliest source standards for EMC IMHO :)

  • MIL-E-6051D Electromagnetic Compatibility Requirements, System (paragraph 3.3)
  • MIL-STD-461B Electromagnetic Interference Characteristics, Requirements for Equipment(applicable sections)
  • DI-R-7061 Electromagnetic Interference Control Plan (current 1AW MIL-STD-461B)

    • The EMC plan is the starting point for the EMC control program. The EMC plan purpose, scope, etc., is described in section 1. The plan ties the EMC program together. The EMC plan is prepared IAW the requirements of MIL-E-6051D, paragraph 3.3 and DI-R-7061. MIL-STD-461B is also referenced in the Military Standards list of the SOW.
\$\endgroup\$
6
  • \$\begingroup\$ I'd agree that it looks like I'll not find a public spec on near field EMI. I'd sort of disagree about the complexity - partly that we have zero public specs, rather than a few basic starter tests, and partly that most of the explanations always use simple component theory. I do agree that detail complexity can be large once one has many PCBs in the same rack, and the component localisation is flexible during design (e.g. the tfr position in your picture). If all the cards are under one's own design control then a Control Plan can help, though talking to each other helps more ;-) \$\endgroup\$ Jul 13, 2017 at 15:47
  • \$\begingroup\$ I'll have a look for the Automotive sub-system specs you mention - is there a magic number to pursue? The EMC control plan approach is often used by colleagues as a method of ducking (many of) the issue(s) and delaying the inevitable where the requirement flowdown for internal modules is at best 'poor', and ultimately there is a 'seek and conquer' phase after the design is complete and hardware has been built and is being tested. Random earth bonds and de-couplers are added, along with tape foil wrap, until things get better. Eliminating just one of the issues (by design flowdown) would help. \$\endgroup\$ Jul 13, 2017 at 15:55
  • \$\begingroup\$ Magnetic coupling is unfortunately a very complex field (Pun intended) with 3D toroidal lines of forces, so differential designs and raising common mode impedance with chokes is common practice. Also orthogonal orientation between unintended emissions and sensor and Faraday shields. This has always been true , in my broad experience from 10~100kA current sensors with coax to CATV 300MHz LO coils with dual channels to SMPS choke orientation to TTL crosstalk between backplane wires. 1) radiation field vectors, 2) impedance of source, crosstalk coupling and load for immunity 3) shunt shields \$\endgroup\$ Jul 13, 2017 at 17:58
  • \$\begingroup\$ Afterthought usually find that slots and strips radiate more than expected, master Clock looks like a patch antenna, and shields leak with poor grounds require additional tape, with conductive glue and many other bandaids. Reverse Engineer a Laptop for EMC avoidance to help looking forward. \$\endgroup\$ Jul 13, 2017 at 18:04
  • \$\begingroup\$ I've marked this as the 'right' answer for the "You probably won't find a public spec on near-field EMI", which matches my searches elsewhere across a number of large companies. I'd agree that slots and strips are effective radiating antennae, though in the context of my question I'm keeping the relevant frequencies low, so in a sense the slots belong to the HF end... One has to start somewhere. Even getting one spec I can flow down would be an improvement after 30 years of trying! \$\endgroup\$ Jul 28, 2017 at 15:17
0
\$\begingroup\$

The standard foil on PCBs provides useful attenuation above 4MHz (125nS edges) for magnetic fields.. For standard MCU edge rates (1nanosecond, or 500MHz), the foil in planes should provide 80 or 90 dB attenuation, depending on orientation; this is 10 Skin Depths or 10 nepers hence 90dB.

For slow edges, such as 1MHz, the copper in planes provides little attenuation of magnetic fields.

Suppose you have 1 ampere at 1MHz (easy to get, in a switching supply). The planes provide little shielding. Suppose you have a sensitive circuit 4" away, with 4" by 4" ground structure or circuit path. What happens?

Vinduce = 2e-7 * Area/Distance * dI/dT

Vinduce = 2e-7 * 0.1meter * 0.1meter/0.1meter * 10^+6 amp/second

Vinduce = 2e-7 * 0.1 * 1,000,000 = 2e-2 = 0.02 volts

\$\endgroup\$
1
  • \$\begingroup\$ I am able to do some personal analysis of potentail stray capacitance coupling and possible chassis noise levels. Many of the issues are relatively low frequency issues (in modern terms), but are coupled directly via the chassis structure (which is isolated from system '0V') into and onto the sub-system module. In my case these are haoused inside a vacumn (metal) casing, so the metal work provides an interference route (as it does inside most systems that are less than one wavelength long). It's locating a public spec/method that can act as a foundation that is the core issue. \$\endgroup\$ Jun 30, 2017 at 6:53
0
\$\begingroup\$

EMC/EMI standards (conducted/radiated emission/susceptibility) are well established for system boxes. However there appears to be no standards, nor methods for deriving a specification for internal modules that are inside the RFI filter protected Faraday cage.

Why would there be? Standards\specs are not needed for modules inside of a product because anything could be in there. These are specs that the designer defines to keep noise out of their system. A good designer knows what SNR is needed in a design and how to attenuate noise from outside sources. Unless this is a standard bus between modules, no one is going to define values or testing for you.

I'm looking for documented/public methods and test procedures that can be used as formal references for a Statement of Work and for sub-Module specifications.

Learn how to write requirements, if its RFI then set limits on what the electric or magnetic field strength can be and a method for test. If its conducted emissions from module to module, something like voltage ripple could be a good metric on shared bus lines.

You'll also have to come up with a method to test for each of your requirements, if equipment does not exist that can detect the levels you desire, that would be something you'd need to design.

\$\endgroup\$
5
  • \$\begingroup\$ You asked 'why'. If one is spending significant sums on a new or updated piece of kit (sub-module), it should be worth specifying the requirement for internal interference. At the moment it appears to be a problem no one wants to touch and would like to pretend doesn't really exist. The lack of reference methods hinders the acceptance of flowed down requirements of the sort "we've never done that test before, so can't commit to that requirement, we can measure it if you want but at a cost". When I talk to fellow designers they quietly agree that it's an issue. \$\endgroup\$ Jul 13, 2017 at 15:29
  • \$\begingroup\$ So what your looking for is an external requirement for internal design? \$\endgroup\$
    – Voltage Spike
    Jul 13, 2017 at 15:31
  • \$\begingroup\$ Yes, This is the specification of a piece of specialist equipment (in my case a few global suppliers) that resides inside our own system box. The system box has the regular EMC requirements (Mil-Stan 461 etc.), with the typical configurations of regular electronic blocks (PSU / Processing / Actuators / etc). The specifications for the specialist units should have suitable near field EMI requirements, but no one can find any, and the suppliers are 'cautious' about accepting any. \$\endgroup\$ Jul 13, 2017 at 15:39
  • \$\begingroup\$ It is unlikely that you will find any because of the restriction that it places on designs. An external requirement regulates aspects of a design that will affect other people (like radiated emissions or conducted emissions) it is unlikely that you will find external requirements that regulate aspects of an internal design. Those requirements need to be set internally, if the contractors won't agree then that is a political problem \$\endgroup\$
    – Voltage Spike
    Jul 13, 2017 at 15:48
  • \$\begingroup\$ We may be at cross purposes here. We (my company's box) generates the internal environment levels, however the test method aught to be via an external standard, to gain supplier acceptance (because they should have seen it before). At the moment the 'black art' card tends to be played, so no real attempt at a solution is made. Yet most black arts are just because folk keep their heads down and fail to look around, such as not seeing the stray components, nor doing simple calcs to get a handle on the magnitude of potential problems, often putting too much trust in their computer models... \$\endgroup\$ Jul 13, 2017 at 16:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.