Some simple hints . for identifying and fixing EMI troubles. This article covers the essential aspects of a domain which is seldom addressed in current EMC litterature: «What to do when an equipment – or a whole system – is failing the tests or experiencing Interference (EMI) problems ?». Whatever we are dealing with a prototype at the end of its development phase, failing one or several EMC tests, or an already installed equipment that exhibit on-site problems, we face a situation that must be solved quickly, with an equipment that cannot be deeply modified.
Contrasting with a development phase where many EMC solutions are available on a product that is still flexible, the engineer confronted to a failing equipment has to detect, diagnose and fix a problem that could be unpredictable, with elusive symptoms, troublesome and penalizing for the user. RFI, ESD, Transient surges, Crosstalk are complex threats involving many interactive mechanisms. No human brain can see at a glance all the possibilities and limitations of the available solutions, where options are limited anyway.
Here we will explain how to identify an EMI problem and its coupling paths in order to correct it with fixes that must be quick, using components that are readily available and applicable in the field, if necessary. All this by using instruments and accessories that are portable, rugged and relatively unexpensive, not requiring the sanitized environment of an anechoic shielded room.
This Part I of the article is focusing on EMI Emission problems. A forthcoming article, Part II will cover Susceptibility problems, including those occuring in the field, where we do not have the commodities and elbow room we enjoy in a development lab.
Note: Readers interested by this topic should read – or briefly return to – our articles 1 to 8 (EE magazines #2015.3 /September, to #2017.1 / March), such as they acquire a basic knowledge of EMI/EMC. Unless you are already a seasoned EMC specialist, the present article without such basics would be worthless since it contains no theory. Rather, it gives practical guidelines for most EMI crash situations.
1. Various Aspects of an EMI Problem
An EMI problem – assuming it was unexpected – in fact many times it WAS expectable, coming out of a design that neglected EMC, of some deliberate cost savings or lack of installation precautions – may show-up with different situations:
a) The status of the equipment:
– equipment is a well advanced prototype, or early pre-production item, or
– equipment is already in production and sold to customers, with little possibilities, if at all, for modifications
b) The nature of the problem
– equipment is failing on one or several EMC mandatory tests, Emissions or Immunity.
– the stand-alone equipment did not fail (or not yet) the EMC tests, but creates functional problems when integrated in a system configuration.
– the equipment is malfunctioning on-site, in certain installations only.
c) The occurrence of the problem
– problem is continuous or quasi-continuous (occuring frequently, in a repeatable manner)
– problem occurs rarely, in a random, unpredictable manner
Each one of these A, B, C conditions, and eventually their combination will require a different approach, according to the urgency, cost and possibility of investigations.
Note: we intentionally ruled-out the case of an equipment that is disturbing itself (Internal EMI), since such problem is normally discovered soon enough during development phase. Yet, this case can be analyzed using the routines described in this article.
2. Brief Reminder of Basic EMI/EMC Terms and Units
This short paragraph is for those readers having no access to printed or electronic copies of the EE magazine articles listed above.
Traditionnally, voltage, current and fields are expressed in Volts, Amperes, Volt/m (E-field) or Amp/m (H-field). However, in EMC when dealing with sensitive receivers or with Emissions testing, these standard units are much too large and submultiples are used instead, the most common ones being:
MicroVolt (μV), MicroAmp (μA), MicroVolt/m (μV/m), MicroAmp/m (μA/m).
Example: A good FM receiver tuned on a given station has a typical sensitivity of 0.5 to 1μV on its RF input (antenna socket). Given approximately 0.1V per V/m for its rod antenna factor, the minimum discernable field by this radio set is: 1μV / 0.1 = 10μV/m.
The Decibel is widely used in EMC community for many reasons:
– Specifications levels are most often imposed in dB.
– EMC hardware (filters, shields etc…) performances are given in dB.
– Most measuring instruments are scaled in dB.
But why is it so? Simply because an EMI situation is often facing a huge dynamic range: sensitivity in the μV or mV may be confronted with strong fields, or power transients with amplitudes of kV, that is 6 to 9 orders of magnitude. Logarithmic scale is more convenient than linear in such cases. Also, with decibels, thanks to the logarithms, multiplications become additions and divisions become substractions.
By definition, the ratio of two Power is expressed by: KdB = 10 Log (P1/P2)
where P1: power in Watts (or mW) of measured or computed phenomena
where P2: reference power in Watts (or mW)
Power is not commonly used in EMC parlance, where amplitudes are more the rule. However, power is mentioned in RF applications where power amplifiers or Radio transmitters are used.
The ratio of two amplitudes (Voltages, currents, E field or H field) is expressed by:
KdB) = 20 Log10 (A1/A2)
where A1: amplitude of measured or computed phenomena
where A2: reference amplitude
In EMC, the decibel is not just used as a dimensionless term expressing gain or attenuation. We associate the dB to a unit, in order to express an amplitude. This way, voltages in μV can also be expressed in dB above 1 μV which writes dBμV, currents expressed in dBμA and so forth.
1μV = 0dBμV, 100μV = 40dBμV
200μA = 2 x 100 μA = ( 6dB + 40dB) above 1μA, that writes: 46dBμA
A 60dBμV RF noise, once passed through a 26dB filter will appear as 60dBμV – 26dB = 34dBμV. Notice that we have substracted dB (dimensionless ratio) to dBμV, therefore the result is dBμV. Speaking in linear terms, we’d have divided a voltage by a number (the filter attenuation), so the result is a voltage.
When dealing with power, the Watt is often a too large unit, and the practice in radio, telecom and EMC has been to use the milliWatt, that expresses in dBm:
1mW = 0dBm, 10mW= 10dBm, 1000mW (or 1 Watt) = 30dBm
Converting dBm into dBμV is possible if we define the impedance where this power is applied. For instance, into 50Ω (the typical impedance in
the EMC instrumentation):
0 dBm (or 1mW) into 50Ω corresponds to 107dBμV (or 223mV)
Why frequency domain?
Except for transient pulses like Power line spikes, lightning, ESD etc, EMC problems are most often treated in the frequency domain, because:
– Most EMC Specifications levels are shown on frequency scales or curves.
– EMC hardware (filters, shields etc ..) performances are characterized in frequency domain.
– Most measuring instruments and sensors are scaled in frequency.
Thus, many EMI emission problems or measurements end-up in measuring at some discrete frequencies. Also many calculations (field reflections, skin effect, transfer functions, resonances, Crosstalk etc..) are simpler to perform in frequency domain. Even with a single pulse, quick calculations can be carried using a sinewave at equivalent frequency ( i.e. bandwidth) reciprocal to the pulse risetime (Fig.1). Therefore, in many cases where the signals are known by their time waveform, the EMC specialist will translate them in frequency domain, using Fourier conversion.
The Source / Victim Concept
Given the complexity of the intercations in an EMI situation, a clear and simple way for addressing the « who-does-what » is the «source and victim» concept (Fig 2 ). An EMI problem can be viewed as a theater act staging 3 players. These three actors are needed on the stage for the performance to exist. If only one is missing, there will be no playing.
- The source of EMI, which can be a natural phenomenon (lightning or ESD), or man-made devices that generates high frequency intentionnaly (authorized RF transmitters) or as a byproduct of their operation (digital circuits and switch mode power supplies).
- The victim of EMI, which can be any analog or digital circuit whose low-level input can be activated, and eventually damaged by undesired signals.
- The coupling between source and victim, which can be a conducted path, a radiated path or an in-between like cable-to-cable crosstalk.
Common Mode voltage or current, a key definition which is the crux of many EMI manifestations and solutions. The simple circuit on Fig.3 shows a wire pair carrying two sorts of currents:
a) the intentional current flowing towards the load then back to its source is called Differential Mode (or «balanced») current. The amplitude difference between the upper and lower wire opposite currents is null, since it is the same current. A corresponding Differential Voltage is found across the pair, or the load.
b) Currents coming from an outside source, or resulting from a non-perfect balance vs ground are flowing on the two wires of the pair in the same direction, returning by some ground path (ground wire, earth plane etc ..). This is called the Common (or «unbalanced») Mode. A corresponding Common Mode voltage is shown, as being the driving source.
CM voltages and currents are a major cause of EMI problems, since they often originate from invisble, non-intentional sources and follow invisible or non-intentional paths.
Basic EMC Requirements, imposed by law, or Industry/ Military standards
- Electrical/Electronic Equipment/System must operate satisfactory in its intended environment.
- System must be self-Compatible (intra-system EMC).
- System must not interfere with neighbour systems.
- System must have a sufficient immunity to potential neighboring interference.
3. A Few Facts Leading to Troubleshooting Optimization and Time Saving
For both Emission and Susceptibility specifications, Conducted and Radiated aspects are treated separately, since the former are generally the dominant mode below 10-30 MHz region, while the radiated concerns are generally the driving issues above 30 MHz.
Advantages of early EMC testing during the design phase
Statistics from EMC test labs reveal that 50% of the products submitted for final compliance fail the first time, at least on one test. Using the simple workbench tests described here, that statistic can be reduced to only 10 or 15% (H. Ott, Ref 2). Although not as accurate as legitimate measurements at a certified lab, workbench EMC measurements are simple, inexpensive and can be performed early in the development phase of a product in order to get a preview of its EMC performance or weaknesses. They can be run in the designer’s laboratory, with limited, relatively inexpensive equipment. From now-on, the equipment of concern will by designated as EUT (Equipment Under Test).
When planning an EMI problem investigation, one should consider that:
a) EMISSION MEASUREMENTS are faster, easier to do than susceptibility ones,
– You do not try making the equipment fail, you just let it run.
– When limit is exceeded, it is rather easy to trace the culprit source.
– No risk of damaging the EUT or associated equipments by excessive stress.
– Improvements you will apply are generally beneficial to immunity as well.
b) CONDUCTED MEASUREMENTS are always faster, easier to do than radiated ones:
– Less instrumentation.
– Set-up is simpler.
– Less prone to measurements uncertainties / errors.
c) BEFORE STARTING ANY INVESTIGATION
– Make yourself familiair with the EUT features relevant to EMC: main frequencies of the digital signals and switchers, type of I/O interfaces (balanced or not) etc …
– Get a figure of how many dB of improvement are needed, at which frequency (or frequency range)? Having to harden a device by 6dB or 60dB will put you on two different ball parks!
4. Troubleshooting Emission Problems
According to our previous list, this (a) choice is the faster to perform if you have such a chance. Yet, several situations may occur:
4.1. Prototype or a Pre-Qualification item, prerequisites
Here, the EUT is designed, but some aspects are not completely frozzen, thus room exist for minor changes. You are probably not (or no longer) on an EMC test site, and in any case, an EMC test chamber is not the place for cut-and-try investigations. Yet, you will need a location with the following minimum characteristics:
– A quiet RF ambient, not in close proximity ( at least > 3m) to powerful noise sources like fluorescent tubes, air-conditioning compressors, elevators, power converters etc … Ground level or basement rooms, away from the building façade are preferrable to upper floor locations.
– Noise-free power mains, with the EUT, associated peripherals and instruments being fed from a same ac branch that does not supply other noisy equipment. Since this is sometimes difficult to ascertain, a good precaution is to install an Isolation Transformer (IT) plus an EMI filter near the distribution panel. In addition to a good isolation from the rest of the ac distribution noise, the IT is generally needed to avoid triggering the Ground Fault Detector when you will connect the LISN (artificial network) for some tests.
– A test ground plane, extending at least up to and beyond the EUT footprint, associated cables and measuring instrumentation. This ground plane will be the artificial RF reference for the entire set (LISNs, Spectrum analyzer). It can be any solid metal sheet, not necessarily copper (aluminium or galvanized steel); thickness is not important. By default, heavy-duty kitchen or barbecue-type alum. foil can do, fold in double layer for tear-off resistance. It will also allow for a well defined height-to-ground distance of the EUT cables, improving the test repeatability. For safety reasons, this plane should be connected to the nearest accessible earth reference (earth bus of the room power panel for inst.)
All the instruments /accessories will be grounded to this test plane using wide, short straps. The EUT is simply grounded via its power cord safety conductor, if any. Unless it is a normal practice for its use (f. inst.: military equipment), do not ground the EUT chassis directly to the test plane.
4.2 Minimum Instrumentation for checking Conducted and Radiated Emissions
Conducted Emission Specs generally cover the 0,15 to 30 MHz frequency range, with some military, vehicle or aeronautic equipment requiring 10kHz up to 100MHz coverage.
Radiated Emission Specs generally cover the 30 to 1000 MHz frequency range, for civilian regulations, extending eventually up to 6 GHz. Military, vehicle or aeronautic equipment require 10kHz up to 18 GHz. Yet, as of today, it is very unusual to find an EUT exceeding radiated emissions limits below a few MHz or above a few GHz, except for rare cases of intererence to sensitive UHF receivers. Given the above coverage the following test gear is a minimum:
– Spectrum Analyzer: the most useful, and expensive piece of equipment. However rugged, portable and easy-to use Sp. Analyzers are available today for less than 3000E( Fig.4). As a minimum you need the following features:
- frequency coverage of at least 0,1MHz to 1.500MHz, or 10kHz to 3GHz if you have to check EUTs for military or airborne applications.
- 10kHz and 100kHz selectable resolution Bwidths, with 1MHz being also recommended. Choose a model with an internal tracking generator option, that will be useful for quick evaluation of some fixes (ferrites, filters etc ..) whose characteristics are not well known or doubtful. Prefer a model with N or SMA style RF input. BNC inputs tend to become undependable and leaky for repeated use above 30MHz.
– EMI current probe: another most useful piece, quite unexpensive. Select a model with well calibrated Transfer Impedance (Zt), preferably flat between 1-100 MHz and characterized up to 300MHz. Eventually one can make his own current probe from a snap-on ferrite ring (Mardiguian, Ref 1), with a shielded core preventing pick-up of external fields.
– Low noise pre-amplifier, with ≥ 20dB gain and noise figure < 4dB. Although not necessary for powerline conducted emissions, it can be useful for detecting very low EMI signatures < 10μV, especially against stringent Radiated Emission limits.
– LISN (Line Impedance Stabilization Network). This artificial network is an important device that simulates a standard, typical impedance of the power mains, for both CM (L1, L2 vs ground) and DM (L1 vs L2) current paths. This prevents that a same EUT, tested in different places or labs could show different results because of different impedances of sites power mains distribution.
– Small proximity Magnetic field probes. Traditional field measurement are made with calibrated EMC antennas which are large (a 30-300MHz wideband biconical antenna is 1,30m long), and sensitive to reflections from surrounding metal objects as well as pre-existing site RF ambient. Our temporary test site has many fortuitous reflective surfaces (shelves, metal desks, chairs, lab benches) that may cause field peaks and nulls: it is impossible to perform a dependable RE test in such environment. Instead, we will measure some parameter that is proportional to the radiated emission, not the radiated emission itself (Ott,Ref 2). Small magnetic loops, much smaller than a wavelength, can be brought very close (< 10cm) to the EUT for «sniffing». Such shielded loops are available off-the-shelf at very affordable prices. As an alternative to a commercial one, a simple homemade probe can be constructed from a 50-ohm coaxial cable (Fig. 9).
– Good quality coaxial cables. Although trivial as a detail, experience tells that a significant amount of time and effort is often spent chasing odd and non-reproducible results caused by low quality or worn-out coaxial cables and especially coaxial connectors; The integrity of the braid and perfect, circumferential contact of connector backshell and mating parts with the receptacle are important for dependable measurements. Brand-new RG58 and BNC set can do a fair job, but coaxial cables of dubious origin with worn-out BNC can ruin a series of test records. For dependable, accurate results, especially with emission tests, prefer double braid coax with N or SMA connectors (male AND female), because they have threaded instead of bayonet fittings. In addition, double-braid coaxial cable exhibit lower losses above 100 MHz.
4.3 Conducted Emissions (CE) on power cord
The majority of CE specifications are addressing only the HF noise present on the main power cable. We will see later that for RE investigation, substantial time savings and progress can be done by measuring also the noise present on I/O cables. You should prepare in advance a coarse list of the potential HF sources and their basic frequencies. An intelligent test program will anticipate what type of repetitive (or eventually random, non-coherent) noise could be present on EUT cables (Ref 4). This will facilitate the identification of BB versus NB nature of the emissions (see further discussion, Peak/Quasi-Peak). With the help of its designer, list the EUT operating modes that will exercise the maximum of its internal or I/O functions, and retain the ones that are likely to cause the highest activity.
Since you are not in a Faraday cage, check the ambient HF noise that could corrupt your measurements. EUT cables and your own instruments can pick-up emissions from radio stations and other ambients. All the same, if the EUT is associated with ancillary equipments, make sure these items, if supplied by their individual power cables, either do comply withg the very CE spec limit we are looking for, or are equipped with an efficient line filter, even as an add-on. The dressing of all cables at 5cm above the ground plane (see Fig. 6) will also restrain their ability to pick-up RF ambients.
Check for ambient background noise by running several sweeps of the spectr. analyzer, with the LISN, and/or current probe in place and the EUT turned off. The read-out of voltage (dBμV) or current (dBμA) should be at least 6dB below the CE limit. If this condition is not met, a CE evaluation is not feasible. However some tricks are worth trying, if there are only a few frequencies where the background noise is too high:
– Record those frequencies where the background noise exceeds our limit.
– Change the scan width of your analyzer down to 50kHz or 100kHz per div.
Since your receiver bandwidth for a CE below 30MHz is 9 or 10kHz, chances are that when you will turn the EUT «On», its signature will show out of the background spectral lines, or in-between. This procedure is safer than turning the EUT «ON» then «Off» , watching for the differences.
With the EUT «On», sweep the prescribed frequency range using the peak, max-hold function if available, overlaying 5 to 10 sweeps. Record the noise voltage (dBμV) at each LISN port, and retain the worst value (Phase vs Neutral, or L1 vs L2). If the limit is in current, do the same with the current probe successively on L1 then L2 line. Make sure that the non-tested port at the LISN set is fitted with its 50Ω load (many commercial LISNs do this automatically). It is safe to keep at least 10dB attenuation at the Sp. Analyzer input, since dynamic range is seldom a problem with CE measurements.
Where a limit violation appears (ΔdB), return and zoom to these specific frequencies and try catching the culprit element by turning off, or disconnecting temporarily one of the following:
– Each one of the switch-mode regulators that could genrate harmonics corresponding to the limit violation
– Same for the processor card, if it can be unplugged or set on standby with EUT still «On»
– Some loads that are notoriously noisy: motors, discharge lamps etc…
Caution: unless there is a surge limiter on the RF input, NEVER turnoff the EUT while the spectrum analyzer input is connected to the LISN port.
Receiver detection mode: Peak, Quasi-Peak or Average?
Normally, CE limits on power line are defined in a 9kHz Bwidth and average detection. However there is a second, more liberal limit for BroadBand emissions, using the Q-Peak detector. Since it allows for faster sweeps, start with Peak detection:
a) If the Pk detection display is compliant with Average limit, the EUT is compliant.
b) If the Pk measuremen exceeds the Average limit at some frequencies, you get a second chance,
c) Repeat the measurement with the QPk detector at the sole frequencies of concern, because it obliges to slower sweeps). Compare with the QP limit, which is 10dB more permissive:
- if the QP limit is exceeded, EUT is non-compliant
- if the QP limit is met, repeat the mesurement in AVERAGE mode:
- if the Aver. Limit is met with the Aver. Detector, the EUT is compliant otherwise it is not.
Common Mode (CM) or Differential Mode (DM)? This is important to know for selecting the optimal fixes. DM emissions (L1-to-L2) are generally strong below few hundred kHz, since they correspond to the first harmonics of the switch-mode power supplies. DM emissions are also stronger when the power input setting is lowest (for inst 115V instead of 230V), or more generally when the EUT draws the maximum line current; in that respect, DM signature increases if the EUT is most active, and decrease with EUT in stand-by mode.
In contrast, CM signature does not change with the EUT activity, but decreases if the power input setting is lowest. Segregating CM from DM contributors on a power line can easily be done with the current probe (Fig.5). Monitoring your progress with the spectrum analyzer, try reducing the excessive levels by at least (ΔdB) + 6dB margin, doing the following:
– If the violation is by DM noise, try a larger value for DM filtering capacitor, they are generally less bulky than magnetics. Install it between the filter choke and the power input, keeping ultra short leads.
– If the violation is from CM noise, look at the way the CM filtering capacitors (line to chassis) have been mounted. They are often connected with long traces or leads: a 5nF «Y» class capacitor mounted with 30mm total lead length (2 x 15mm) becomes progressively worthless above 12 MHz.
– Check-out the line filter schematic. Is there enough inductance for CM and / or DM reduction? Improve the filter efficiency ( DM or CM, depending on your finding). If room permit in the equipment, try a filter with more efficient DM or CM choke.
– Look at the filter mounting: often, filters are mounted too far inside the equipment instead of right at the power input. Watch for a possible coupling between input and output wires (or traces). Correct if needed. Validate your progress by a formal re-test.
4.4 Radiated Emissions (RE) check by substitute methods
Complying with RE limits is one major EMC challenges during product development and testing. Often, a product has been developed using the best home-grown experience plus simualtion sofware to «make-itwork », free of internal noise problems. Radiated emissions is one of these secondary concerns that are pushed away to the day of tests to see if it passes. Needless to say, generally it does not, unless a serious EMC analysis has been carried along design phase. The methods recommended here are time-savers for identifying and reducing quickly out-of spec radiations, without bringing first the EUT to an anechoic EMC test chamber for a true radiated emission test at 1 or 3m distance with calibrated antennas, turntable etc…
While the majority of CE specifications limits are addressing only power line emissions, below 30 or 50MHz, the rationale of our investigation method is that ANY external cable, by the Common Mode (CM) current it carries, can easily radiate more than the box itself up to 200- 300MHz. This is because the mere geometric dimension of the I/O cables, usually exceeding 1m, makes them effficient antennas, while the internal EUT’s PCB traces represent dipoles or loops with sizes one or two order of magnitude smaller.
Ideally, with intentional signals (differential-mode signals), the current flows down one wire of the cable and returns via a close, adjacent wire, hence the net current should be almost zero and the CM radiation be almost non-existent. Since an actual I/O interface is never ideal, the CM current is the unbalanced current (current not returning on the cable). If this current is not returned on the cable, where does it flows? Via the cable stray capacitance, which means radiation. Thus, measuring the undesired CM current on each I/O cable is one of the most useful things that you can do (Ott, Ref 2).
Above ≈ 300MHz, were wavelength λ becomes less than one meter (i.e. l /4 < 0,25m), the cable resembles an antenna that progressively «shrinks», while the internal EUT wiring and PCB traces eventually override cable radiation (Ref 4).
4.5 Measuring CM Currents on Cables.
The CM current can easily be measured with a calibrated high-frequency clamp-on probe and a spectrum analyzer as shown in figure. 5. The set-up remains the same as for the true CE test on power line, but this time ALL cables will be measured. Current probe must be moved along the 1.50 m cable section that is closer to the EUT box to make sure you do not miss a maximum of current standing wave.
Based on simple antennas formulas (Ref.4) Table 3 shows the maximum CM current that can be tolerated, from 30 to 400MHz, on I/O cables for civilian residential (Class B), industrial (class A) and Military (RE 102) compliance. The following assumptions are made for these Pass / Fail criteria:
– For FCC/CISPR
- frequencies > 30 MHz.
- cable length greater than 1.50 m.
- cable height : h ≥ 0.75m A 5 dB margin has been accounted for ground reflection.
– For Mil 461F-RE102, additional factors are coming into play:
- Cables are laid at 5 cm above the ground plane.
- The limit relaxes progressively above 100MHz (Mil 461E, F).
- Actual field measurements are made at 1m distance.
Notice that, up to 400 MHz and regarding cable radiation only, the criteria for the most severe Mil.Std 461 limit are quite close to what would be required for FCC-15 or CISPR22 class B. If on each I/O cables, we satisfy the Icm table limits on every spectral line, we know that, at least, the contribution of the cables will keep us below the spec limit. If we fail this CM current test, we will surely fail the Radiated Emission test.
A word of caution: you are not in a RF-clean environment and your cables may pick-up from external sources such as local FM and TV broadcast stations. All measurements must, therefore, be validated to assure that you are measuring what you think you are measuring. A simple validation test in this instance is to turn the product Off and see if the reading goes away. If it stays, it is due to external pickup. For instance, signals in the 88 to 108 MHz (FM) frequency range should be suspect, and double-checked by turning the EUT «Off».
Numerical example: Using the current probe of Fig.5, the following maximum voltage values have been recorded on the Spectr. Analyzer (50Ω input). Once translated into cable current, do we meet the criteria for class B radiated emissions?
WARNING: the Icm limits are peak reading, as displayed by the Sp. Analyzer. Make sure that they have been measured with a 100 or 120kHz BWidth, with a video BWidth set at 0,3 or 1MHz. For instance, if some of the above-limit lines are harmonics of a 30 or 50kHz swicher, there will be 3 or 2 harmonics adding-up in a 100-120kHz bandwidth, compared to what has been seen in the conducted emission (CE).
This technique works on shielded cables, too. It can be a good way to pin-point possible weaknesses of your cable shield, including its terminations. A question often arises: If our current probe sees a current on the cable shield is it really the shield current that we see ? (typical answer will be «yes»…). NO! It is the CM current that escapes the shield and returns by the outer, invisible path, that is radiation. This gives a measure of the quality of the shield, whatever it is.
Interpreting and reducing CM emissions from I/O Cables.
Once the contribution of the I/O cables has been measured, if they violate our Table 3 criteria – as they often do on a prototype or pre-production item, we will work at bringing them below the corresponding limit. While trying improvements, measure one cable at a time with the current probe. If the product is still amenable to limited changes in the I/O ports areas of the PCB, and /or Power input filter area, use surface-mount CM ferrite, surface-mount signal filters or filtered connector sockets, depending on the function of the faulty cable.
– If the faulty cable (or one of them) is the power cable, do not take for granted that the filter is not guilty just because you passed the formal CE test. Power line filters are often optimized up to 30-50MHz, because this is as far as the CE spec goes, and their attenuation could very well drop beyond this frequency range. Try to install an additional bifilar-wound ferrite, or small ceramic CM capacitors, line-to-chassis, close to the power entry port.
– Check that there is a good, metal-to-metal bonding of the PCB Zero Volt (signal Gnd) to chassis, close to every I/O port.
– Inspect the internal EUT wiring, checking for possible crosstallk between I/O wires and internal wire/traces that do not come out.
– Replace unshielded cables /pairs by shielded ones, or if already shielded, check that the shields are making a perfect (360°) contact with their housing/receptacle, straight to the EUT chassis.
After each fix or set of fixes, repeat the Spectrum scans with the current probe to check if – and by how many dB – you have reduced the CM current, until NOT ONE spectral line exceeds our CM current limit. When you have been through all cables one by one, run an overall check of the CM currents: some may have increased on a previously fixed cable (the classic «balloon effect»). At this point you can feel confident that the cables will no longer cause a violation of the radiated emission test at a qualified EMC facility. However you are left with a last possibility: chances are, especially with fast logic running at clock speeds >30 MHz, that the EUT box itself radiates above the permitted limit.
This is where a miniature near-field probe could be used. Without trying to convert the readout into E-field, crude measurements made with such H-field probe allow making A/B comparisons of the Sp. Analyzer scans, when improvements are attempted. But this implies that we have at least one actual RE measurement reference for the EUT box-only radiation on a qualified test site, as we will see next.
4.6 Measuring Radiated Emissions from EUT box alone (I/O cables excluded)
Once the contribution of the I/O cables has been checked and reduced, an actual RE test in an EMC test lab can be made, with fair chances of success, except for the risk of radiation emanating from the EUT box itself. If the EUT with its EMI -hardened I/O cables still fail the test, we will re-test without the I/O cables (see chart Fig. 7)
Near Field Measurements on PCBs.
What we can do first is to identify the strong magnetic fields close to the printed circuit board using a small magnetic field loop probe and the spectrum analyzer. Scan the probe over the printed circuit board looking for “hot spots” (locations of strong magnetic fields). When some are found, check the PCB in that vicinity for violations of good EMC design practices. One frequently found is an interrupted signal current return path caused by a split or slot in the ground/power plane (Ref.2). If a quick, temporary change can be made to the board, retest to confirm that the H-field has decreased in amplitude. In some cases you may find that it is an IC module that is causing most of the emission. In this case, consider using a
small board-level shield over the component(s), or a small ferrite «tile» on top (Mardiguian, Ref 4).
The magnetic field probe can be held in the 3 axis (X,Y,Z) when performing the tests, such as to capture the maximum field strength. Because such small loops are insensitive to remote external fields, you can assume that what you get is coming from the nearby PCB. This can be validated by moving the probe 2 or 3 times further away from the board: the reading should drop abruptly by 4 to 9 times (in near-field, there is a very strong H field dependency with distance. Because of this, make sure to keep a constant distance from the center of the loop to the PCB zone you are suspecting. A simple way is to stick an insulating spacer used as a distance gage.
Near Field Measurements around the EUT box
If the EUT is housed in a – supposedly – shielded enclosure, with little hope that the PCB could be modified, try to identify electromagnetic field leakages through the apertures, using the H-field probe (Ref 3). Place the probe close to the enclosure with the plane of the loop parallel with the shield. Keeping a constant distance, move the probe along the seams, apertures or connectors areas and search for a strong magnetic field. After making changes to the enclosure (e.g. reduce the aperture size or length of the seam, add more screws or spring contacts, temporarily cover the leakage with large conductive adhesive copper tape etc.), retest to confirm that the magnetic field has decreased in amplitude.
For both PCB and Box leakages treatments, make sure that your H-field probe read-out after the fix has been reduced by at least (ΔdB + 6dB), if Δ was the amount of your limit violation.
Note: In theory more could be done if the H-field probe is properly calibrated. Unfortunately, calibration curves provided by the manufacturers are a nightmare to the user, given in exotic units like «dBm per microTesla», requiring some legwork for a quick translation in dBuV per dBuA/m ! Then with complex calculations or a proper software, an H-field reading at Xcm distance can be transposed into an equivalent E-field at 1 or 3m. This includes near field -to- far field conversion factors which are depending on distance, wavelength and nature of the radiated elements (loops or dipole), a computation that we do not recommend to other than seasonned, full-time EMC experts.
Michel Mardiguian EMC Consultant, France firstname.lastname@example.org
- M. Mardiguian, «EMI Troubleshooting Techniques» (Mc Graw Hill, 2000)
- H. Ott «Workbench EMC measurements» (www.hottconsultants.com)
- K. Wyatt «Troubleshoot Radiated Emissions» Interference Technology ITEM (2010)
- M. Mardiguian Controlling Radiated Emissions, 3rd edition (Springer-NY, 2014)