Fixing EMI across the Printed Circuit Board in Switching-Mode Power Supply (SMPS)

An ounce (or two!) of prevention are always better than cure. So here we look at some of the practical design aspects involved in controlling and testing EMI.

The Role of the Transformer in EMI

Very often an engineer resolves a stubborn EMI problem by just 'playing' with the transformer. The transformer comes into the picture in the following ways:

++ With its windings carrying high-frequency current, it becomes an effective H-field antenna. These fields can impinge upon nearby traces and cables, and enlist their help in getting transported out of the enclosure, via conduction or radiation.

++ Since parts of the windings have a swinging voltage across them, they can also become effective E-field antennas.

++ The parasitic capacitance between the primary and secondary windings transfers noise across the isolation boundary. Since the secondary side ground is usually connected to the chassis, this noise returns via the earth plane, in the form of CM noise. The situation is very similar to the tradeoffs required in heatsink mounting issues. In this case, we wish to couple the primary and secondary very close to each other in order to reduce leakage inductance (especially in flyback transformers), but this also increases their mutual capacitance, and thus the CM noise.

Here are some standard techniques that help prevent the preceding:

++ In a safety-approved transformer, there are three layers of safety-approved polyester ("Mylar") tape between the primary and secondary windings -- for example the popular #1298 from 3M (at www.3m.com). In addition to these layers, a copper 'Faraday shield' may be inserted to "collect" the noise currents arriving at the isolation boundary, and diverting them (usually to the primary ground). See Ill. 1. Note that this shield should be a very thin strip of copper foil, so as to avoid eddy current losses and also keep the leakage inductance down. So, it's typically 2 to 4 mils thick, consisting of one turn wound around the center limb.

A wire is soldered close to its approximate geometric center and goes to the primary ground. Note that the ends of the copper shield shouldn't be galvanically connected together, as that would constitute a shorted turn from the viewpoint of the transformer. Some designs also use another Faraday shield, on the secondary side (after the three layers of insulation). This is connected to the secondary ground.

However, most commercial ITE (information technology equipment) power supplies don't need either of these shields, provided adequate thought has gone into the winding and construction, as we will soon see.

Ill. 1: Screens used for transformers

++ There is usually also a circumferential copper shield (or "flux band") around the entire transformer. See Ill. 1. The ends of this shield can be, and usually are, shorted (soldered) together. It serves primarily as a radiation shield. This is often left floating in low-cost designs. However, it may be connected to the secondary ground if desired. And if so, safety issues will need to be considered, in regard to the requirement of reinforced insulation between primary and secondary, and also the required primary to secondary 'creepages' (distance along the insulating surface) and 'clearances' (shortest distance through air) as valid. When the transformer uses an air gap on its outer limbs, the fringing flux emanating from the gap causes severe eddy current losses in the band. So this band is also usually only 2 to 4 mils thick.

Like the Faraday shield, this too can often be omitted by good winding techniques.

++ To reiterate, from the point of view of EMI, a flyback transformer should be preferably center-gapped, that's , no gap on its outer limbs. The fringing fields from exposed air gaps become strong sources of radiated EMI besides causing significant eddy current losses in the surrounding copper band.

Ill. 2: Low-noise Transformer Winding Technique

++ There is usually an auxiliary winding present on the primary side, which provides a low voltage rail for the controller and related circuitry. One end of this is connected to primary ground. Therefore, it can actually double over as a crude Faraday shield if we a) wind it evenly and spread it out over the available bobbin width, and b) we help it collect and divert noise by ac-coupling its opposite end (i.e. the diode end) to primary ground, through a small 22 to 100 pF ceramic capacitor as shown in Ill. 2.

Ill. 2 also reveals a low-noise construction technique, as applied to a typical flyback transformer. We should compare the right-hand schematic with its equivalent "winding" version on the left. In the following discussion, we note that though transformers with split windings are not being explicitly discussed here, the same principles can be easily extended and applied to them too. Here are some observations on Ill. 2.

++ Since the drain of the mosfet is swinging, it's a good idea to keep the corresponding end of the primary winding buried as deep as possible, that's , it should be the first layer to be wound on the bobbin. The outer layers tend to shield the fields emanating from the layers below. For sure, the drain end of this winding shouldn't be adjacent to the 'safety barrier' (the three layers of polyester tape), because the injected noise current is proportional to the net dV/dt across the two "plates" of the parasitic capacitor (formed by the windings on either side of the interface). Since we really cannot reduce the capacitance much (without adversely impacting the leakage inductance), we should at least try to reduce the net dV/dt across this interface capacitor.

++ Comparing the diagram on the left with its schematic on the right, we see that the "start" and "finish" ends of any winding have also been indicated. In particular, all the start ends have been shown with dots in the schematic. Note that in a typical production sequence, the coil winding machine always spins the bobbin in the same direction, for every layer and winding placed successively. Therefore, since all the start ends (i.e. dotted ends) are magnetically equivalent - if one dotted end goes high, the other dots also go high at the same moment (as compared to their opposite ends). We can also see that from the point of view of the actual physical proximities involved, every dotted end of a winding automatically falls close to the non-dotted end of the next winding (with the usual fixed winding direction). This means that for the flyback transformer of Ill. 2, the diode end of the secondary winding will necessarily fall adjacent to the safety barrier. And because of that we will have a certain amount of dV/dt still present across the barrier. But note that this dV/dt is much smaller than if the drain end of the primary winding was brought adjacent to the safety barrier (because of the bigger voltage swing on the primary side, due to the large turns ratio). However, the transformer as shown in Ill. 2 now has the advantage that the "quiet end" (ground) of the secondary winding is now the outermost layer. That is by itself a good shield. So we can safely drop the ubiquitous circumferential shield (flux band). Consider the alternative - suppose we had wound the transformer the "wrong" way, that's , by reversing all the start and finish ends shown in Ill. 2. That would have brought the drain end of the primary winding right next to the safety barrier, with the secondary ground end (which is usually connected to the chassis) directly across the isolation boundary. With this winding arrangement, we would have a healthy dose of CM noise injected directly into the chassis/earth - not the best way to achieve compliance for sure!

++ When we go through the same reasoning for a forward converter transformer, we will find that with the described winding sequence, we will automatically have the quiet ends of both primary and secondary windings overlooking each other across the safety barrier (isolation boundary). That is because the relative polarities between the primary and secondary windings in a forward converter are opposite to those of a flyback transformer. So now, very little noise will be injected through the parasitic capacitance. That is good. But the outermost layer isn't "quiet" anymore, and we could have a radiation problem. So in this case, the circumferential shield may become necessary.

++ Another way out of the forward converter "outer surface radiation problem" is to ask production to reverse the direction of the secondary winding (only). So for example, if up to the finish of the primary winding, the machine was spinning clockwise, for the secondary we should specify an anticlockwise direction (with expected resistance coming, not from the transformer, but our production staff!). If we do that, the reasoning given previously for the flyback will now apply unchanged to the forward converter transformer too. So we would now have a "quiet" exterior (without any flux band necessary), though some more common mode noise transferred across the isolation boundary, due to the dV/dt. Note that in general, aiming for a "quiet" exterior seems to be a better option than trying solely to prevent noise injection through the interface capacitance, because the latter can be overcome by various tricks - like having the auxiliary winding double over as a Faraday shield, and so on. But a radiation problem can be hard to manage. We don'te however, that a forward converter transformer has no (or very small) air gap, so it's generally considered "quieter" in terms of radiation to start with (as compared to a flyback).

Tip: We don't need to draw any current at all from this 'Faraday winding' to make it work. So it need not even be required by our circuitry (for an auxiliary rail). In that case, we could just wrap a few turns of thin wire (spread out evenly), with one end of it connected to primary ground, and the other end via a small 22 pF capacitor to primary ground. This technique certainly saves production costs associated with the making and placing of a formal Faraday shield - not to mention the improvement in efficiency due to the reduced leakage inductance (as compared to what a formal Faraday shield may lead to). In that sense, this informal Faraday shield is a very useful idea, certainly worth trying out.

Ill. 3: Cancellation Winding to Reduce CM Noise

++ When the transistor is mounted on the chassis for thermal reasons, there is a technique that's used to actually try and cancel the current injected through the heatsink capacitance. This is done by placing another winding, equivalent to the main winding, and opposite in phase (though it can be of much thinner wire).

See Ill. 3. The idea is that if the noise current is being pushed out from the primary winding, the cancellation winding gets the same current pulled in.

Therefore in effect, the injected current does a quick "U-turn" back to its noise source. Note that this additional cancellation winding should be very closely coupled to the main winding. Often it's wound bifilar with the primary winding (i.e. both wound simultaneously, rather than one on top of the other). However we should be aware that in that case, we will have a high voltage differential between the two windings at points along their length. So if, for example, there are pinholes in the enamel insulation, there is a danger of flashover and resulting failure of the power supply. The solution is to use wires with "double insulation." Note: This technique does nothing to cancel the noise injected through the interface capacitance (i.e. between the primary to secondary windings). But despite that limitation, a 10 dBµ V reduction in conducted EMI is still possible (at various points in the EMI spectrum). So this could certainly be worth trying out, if there is a last-minute problem and a major redesign of the board needs to be avoided.

It therefore may be prudent to plan for this winding in advance, including a PCB placeholder for the additional capacitor.

Note: The above idea can clearly be applied to any off-line topology (and also all high-power dc-dc converters) - whenever the switch needs to be mounted on the chassis/enclosure (and its tab is swinging).

A similar technique may be useful on the secondary side too, if suppose the catch diode is to be mounted on the chassis. However this secondary-side heatsink noise injection is of concern only when the tab of the diode (which is almost invariably the cathode of the diode) happens to be the switching node for that particular topology/configuration. So we can work out that the normal boost and flyback topologies don't have this problem, since the cathode end of their diodes are "quiet." However, the (positive-to-positive) buck and the forward converter do have swinging cathodes (tabs), so we should be careful when chassis-mounting their diodes.

++ Rod inductors are often used in LC post-filtering stages on the output. Because of their open magnetic structure, they have been called "EMI cannons." But they are nevertheless still popular because of their low cost, and also the low "real estate" they need. Some tricks have therefore been developed to control their ill-effects.

First, they should be placed vertically (as they normally are). Further, if two such rods are being used on a given output, we should wind the two rods identically, but reverse the current flow in one of them, as compared to the other (by suitable modification of the PCB). See Ill. 4. So, looking from the top, one rod should be carrying current clockwise and the other anticlockwise. This helps redirect the flux from one, back into the other ("U-turn"). In that way, much less "EMI-spilling" occurs.

Ill. 4: Correct Way to Use Rod Inductors

EMI from Diodes

Here we list some of the things to keep in mind and try out, regarding diodes:

++ Diodes are a potent source of low- to high-frequency noise. Slow diodes (like those in a typical input bridge) can also contribute to such noise.

++ Input bridges that use ultrafast diodes are available, and their vendors claim significant reduction in EMI. But in practice they don't seem to provide much advantage. They also typically have much lower input surge current ratings. In fact in a front-end position, any component always needs to be able to handle a lot of stress (if not abuse), such as the stresses occurring during power-up at high line.

++ To minimize EMI, ultrafast diodes should be selected on the basis of softer reverse recovery characteristics. For medium- to high-power converters, RC snubbers are also often placed across these diodes (at the expense of some efficiency). In low-voltage applications, Schottky diodes are often used. Though these diodes have no reverse recovery time in principle, their body capacitance can be relatively high, and can end up resonating with PCB trace inductances. So an RC snubber is also often helpful for Schottkys. Note that if any diode has fully recovered (i.e. zero current) before the voltage across it starts to swing, there is no reverse recovery current. In that case, diodes really don't have to be 'super-super-fast.'

In fact many engineers have reported much lower EMI by choosing slower diodes for snubbers/clamps. A popular choice for snubber applications is the soft-recovery fast diode BYV26C (or BYM26C for medium power) from Philips.

++ It is advisable to have the mosfet switch roughly two to three times slower than the reverse recovery time of the catch diode - to avoid shoot-through currents - that will produce strong H-fields (in addition to causing dissipation). Therefore it's not uncommon to intentionally degrade the mosfet switching speed by adding a resistor (typically 10 ohm to 100 ohm in off-line applications) in series with the gate - maybe with a diode across the gate resistor so as to leave the turn-off speed unaffected (for efficiency reasons).

++ Small capacitors may often be placed across the mosfet (drain to source). But this can create a lot of dissipation inside the mosfet, since every cycle the capacitor energy is dumped into the mosfet. (P = 1/2 × C × V^2 × fSW).

++ Ultrafast diodes also exhibit high forward-voltage spikes at turn-on. So momentarily, the diode forward voltage may be 5 to 10 V (rather than the expected 1 V or so).

Usually, the snappier the reverse recovery, the worse is the forward spike too.

Therefore, at mosfet turn-off, the diodes become strong E-field sources (voltage spike), whereas at mosfet turn-on, the diodes will generate strong H-fields (current spike). A small RC snubber across the diode will help control the forward voltage spike too.

++ In integrated switchers, access to the gate of the mosfet may not be available. In that case, the turn-on transition can be slowed by inserting a resistor of about 10 ohm to 50 ohm in series with the bootstrap capacitor. The bootstrap capacitor is in effect the voltage source for the internal floating driver stage. At turn-on, it's asked to provide the high current spike required to charge up the gate capacitance of the mosfet. So a resistor placed in series with this bootstrap capacitor limits the gate charging current somewhat, and thereby slows the turn-on.

++ To control EMI, ferrite beads (preferably of lossy nickel-zinc material) are sometimes placed in series with catch diodes (often slipped on to their leads), such as at the output diode of a typical off-line flyback. However, these beads must be very small, as they can have a significant effect on the efficiency of the power supply.

Note: In multi-output off-line flyback converters, we may find larger beads (possibly with more than one turn, and made of the more common manganese-zinc ferrite) in series with the output diodes belonging to some of the auxiliary outputs (i.e. those not being directly regulated). But the purpose of these beads isn't EMI suppression, but to block some of the volt-seconds and thereby improve the "centering" of the outputs.

++ A comment about split/sandwich windings. In general, the primary winding may be broken up into two windings, which are then positioned on either side of the secondary winding - so as to reduce leakage inductance in flybacks, and proximity effect losses in forward converters. This is acceptable for EMI provided the two split winding sections are in series. In general, putting windings in parallel isn't a good idea (especially from the EMI point of view). In high-current power supplies, the secondary winding is also sometimes broken up into two windings (or foils).

The intention is usually to increase the current handling capability. See Ill. 5.

But these split secondaries are also usually placed physically apart, on either side of the primary winding. However, in paralleled windings, the two supposedly "equal" sections are actually always magnetically slightly different - because of their different physical positions inside the transformer. Plus, their DCR is also just a little different (different lengths), creating the possibility of an internal current loop.

The designer may be completely unaware of this, except for the severe tell-tale ringing on the voltage waveform, and a mysteriously bad EMI scan. So if paralleling is really needed, it's better to use the scheme in the right-hand side schematic of Ill. 5. Here the forward-drops of the two diodes help "ballast" the windings, and this also helps "iron out" any inequality between the two halves.

Ill. 5: Correct Way to Parallel Windings; Beads, and an Industry Experience-the dV/dt of Schottky Diodes

In a very high-volume power supply design and manufacturing house, the following situation arose. The output Schottky diode of an off-line 70 W flyback had a small but unacceptably mysterious "ppm" (parts per million) failure rate in production testing. Finally, by careful analysis, this was traced to a very slight wiggle (ringing) somewhere in the middle of the turn-off waveform of the diode. Then, by drawing asymptotes, it was seen that the dV/dt rating of that particular diode was being momentarily exceeded at the point of the wiggle, thus probably causing its failure (no other overstress could be seen). A small ferrite bead was inserted on the leg of the Schottky diode. This smoothed out the wiggle, reduced the EMI dramatically in the bargain, and as a proof of the hypothesis, no Schottky failures were seen thereafter. There wasa1to2% loss in overall efficiency though. So, not all Schottkys are created equal. We must be conscious of the different dV/dt ratings they can have, depending on their vendors. And also of the leakage current through them, which can't only cause loss in efficiency, but in certain cases anomalous behavior - like premature current limiting at turn-on (especially in integrated switchers where the switch drop is being sensed for implementing current limiting).

Note: On the topic of beads, note that beads also sometimes have been put in series with the mosfet. But we shouldn't put any such bead in the source. If we do so, then during crossover transitions, the source pin (with bead) can develop spikes. And since the gate is referenced to the source, not the drain, this can lead to a spurious turn-on, resulting in reliability issues. Therefore, a bead, if necessary, should be placed only on the drain side of the mosfet. True, this extra uncoupled inductance can also cause a small spike in principle, but in practice, that's rarely an issue. For the same reason, if we want to monitor the current in the mosfet by means of a current probe, we should place the loop of wire (to slip the probe tip on to) on the drain side, never on the source.

Tbl 1: Good component placement

Input Cap Output Cap Catch Diode Inductor

Buck Critical Not Critical Critical Not Critical

Buck-Boost Critical Critical Critical Not Critical

Boost Not Critical Critical Critical Not Critical

Basic Layout Guidelines

For each topology, we need to carefully figure out which PCB trace segments are "critical" in terms of layout and EMI. "Critical" traces are "high-frequency current sections" - in which current is forced to either start flowing or stop flowing (suddenly) at the instant of turn-on or turn-off. So at each transition we get a very high dI/dt across such traces. Further, from the rule-of-thumb "20 nH per inch of trace," we get a voltage spike according to V = L dI/dt. These spikes can't only cause a lot of EMI, but can also infiltrate into the control sections of the IC, causing anomalous behavior (and possibly switch destruction).

To minimize the fields, and simultaneously reduce the associated inductance, the area enclosed by such high-frequency current loops must be minimized. Therefore, analyzing the topologies from this viewpoint, we get the results summarized in Tbl. 1. All traces leading to components marked "Critical" must be kept very short (and not too thin!).

The corresponding high-frequency loop areas will then be minimized automatically.

We indicated above, that layout concerns and EMI concerns generally overlap. In other words, what is good in terms of layout (ensuring proper performance) is also good for EMI.

There is, however, one possible exception to this trend - in particular we need to be careful about inadvertently making traces that have a swinging voltage on them too wide, as they can become good E-field antennas. The prime example is the trace at the switching node of any topology. We may be wanting to increase its copper area, for the purpose of lowering parasitic inductance and /or helping dissipate heat from the mosfet or catch diode, but we must do this judiciously.

The ground plane is a very effective method of bringing down the overall level of the EMI emissions. On a multilayer board, if the very next layer to the side containing the power components (and their associated traces) is a ground plane, the EMI can drop by about 10 to 20 dB. This is clearly more cost-effective than opting initially for a "cheap" one- or two-sided board, and then having to use bulky filters later. However the integrity of a ground plane should be maintained, as far as possible. For this, we should remember that return currents tend to travel by the shortest straight line path at low frequencies. But at high frequencies (or the higher harmonics of the waveform), the return currents tend to image themselves directly under their respective forward traces (on the opposite layer). Therefore, currents, given a chance, automatically try to reduce the area they enclose - as this lowers the self-inductance, and thereby offers the current the lowest impedance route possible (at low frequencies, trace impedances are resistive, but at high frequencies they are inductive). So in particular, if we make ill-considered cuts in the ground plane (possibly with the intention of "conveniently" routing some other trace), the return currents of the power converter stage (which really need this ground plane) will get diverted along the sides of any intervening cuts. And in doing so, will form effective slot antennas on the PCB.

Last-ditch Troubleshooting

It is helpful to separate the CM and DM components to be able to study them and debug a bad EMI scan. But a standard LISN reading only provides a certain weighted sum of the total conducted noise (CM and DM). Therefore, unless special accessories are available (including a modified LISN), we can only guess which part of the EMI scan is mainly DM and which is CM. So we may never know the root cause of the noise either. In Ill. 6 we have shown two current probes, wired up in such a way that they are actually performing "simultaneous equation" math on the L and N wires - to separate the CM and DM components (also see ill 9-1 in Section 9). Note that by doing these two measurements at the same time (using two probes rather than one), we have also retained valuable information about the relative phase relationship, existing between the CM and DM components.

Note: The bandwidth and current capability of the current probes used for noise measurements are important. Popular choices for such probes are from Pearson Electronics and Fischer Custom Communications at www.pearsonelectronics.com. For very high currents (up to thousands of amperes if necessary), a possible choice are current probes based on the "Rogowski principle." This type of probe is available from several manufacturers, for example Power Electronic Measurements Ltd. at www.pemuk.com. These probes are not the usual current transformer type.

The output from a Rogowski probe depends not on the instantaneous current enclosed, but on the rate of change of current. So, instead of just placing several turns around the wire to be sensed, as in a typical current transformer, the Rogowski probe effectively takes an air-cored solenoid and then bends that in a circle around the sensed wire (like a doughnut). Such probes are also considered virtually noninvasive.

The usual lab active current probes (which also measure dc, and therefore include a Hall sensor) are usually just not suited for these high-bandwidth noise measurements.

Note: When viewing pulse transition times below 100 nanoseconds, or emission noise frequencies above a few megahertz, it's advisable to keep the cable length small. Thereafter, we must terminate the cable at the oscilloscope, or measuring instrument, with a 50 ohm resistor. However, most modern oscilloscopes incorporate a selectable 50 ohm input impedance. Correct termination of cables prevents standing wave effects.

Note that, with this 50 ohm termination, the measured voltage is approximately half of what it really is because we essentially have a voltage divider formed by the cable and the terminating resistor. Oscilloscopes will usually automatically correct for this, if they 'know' that there is a 50-ohm termination present. Also note that fast-rising pulses can produce spurious ringing, due to high-frequency current crowding on the surface of the cable shield. This can be suppressed by threading the measurement cable through one or more ferrite beads (or toroids). E.g., Pearson reports that they obtained good results by placing three turns through four ferrite cores of about one inch inside diameter, two inches outside diameter, and ½-inch thickness.

Ill. 6: Using Two Current Probes to Separate DM and CM Noise Components

Ill. 7: Analyzing Magnetic and Electric Field Sources; Inside a Power Supply

Note: A quick diagnostic test for understanding a particular high-frequency conducted EMI problem (measured by a LISN at the inputs of the power supply) is to twist the output cables of the power supply tightly together (along with their respective return wires). This induces field cancellation (also called flux containment), thus reducing the radiation from the output cables (if present).

This procedure was actually implemented in full production on a particular high volume commercial design. A few tie-wraps were used to hold the bunch of wires tightly together in the twisted position. This happened to be a last-ditch effort to avoid costly last-minute redesign, just before full production. This "twist-and-tie-wrap" technique is admittedly not very practicable or desirable, in production. But it's cheap.

Note that a ferrite sleeve slipped over the entire output cable bunch was also working equally well, if not better. But it was disqualified simply because it was far more expensive than three tie-wraps! However, it's interesting to note here, that even though a ferrite sleeve may look like a radiation shield (and smell like one too! ) and even produce almost the same results as twisting the cables produces, it actually works by reducing the common mode noise currents themselves, not merely by "shielding" the EMI arising due to them.

Twisting, on the other hand, simply tries to cancel the fields of adjacent wires (with their returns). Looking back, in this particular case, the root cause was that there was obviously a significant amount of common mode noise already present on the output, which was causing the output cables to radiate. The radiation was thereafter being picked up by the input cables, leading to the failed conducted EMI test.

In Ill. 7, we show a practical technique to separate that part of the conducted emissions spectrum that's due to radiation from within the converter. We see how to learn to identify these as E-fields or H-fields. For this, we need to cut the PCB traces just before the input bridge, and then route the ac power from a canned filter outside the enclosure.

The ends of the existing filter are kept either open (to receive E-fields) or connected together through a small loop (for seeing the H-fields). The other end of this EMI filter is then routed as usual to the LISN and spectrum analyzer. We can thus see this "extraneous" (radiation based) noise. It will give us an indication if a heatsink, for example, is causing severe E-fields, or if a certain magnetic component is causing severe H-fields. We can also wave a small plate of thick copper (connected to earth) in suspected areas to see which component may be the actual source of the fields. However, for analyzing the source of magnetic near-fields, a slab of ferrite (from a typical EMI suppression kit) works far better than a copper plate (this can be waved around similarly - but there is no need to earth it).

Caution: AC power is NOT to be applied through the LISN. This will cause a serious hazard. Also the plate/slab must be well-wrapped in tape to prevent accidental contact with components.

Are We Going to Fail the Radiation Test?

Most of the smaller companies cannot afford a pre-compliance setup for radiated emission tests. However a few of them have a fairly good idea beforehand whether they are going to be successful in that test or not - just by looking hard at the conducted EMI scan. What they do is to look carefully at the spectrum in the third region of CISPR 22. This is the flat region from 5 to 30 MHz. They can even scan higher to higher frequencies, if possible.

They realize that even though they may have achieved compliance with the conducted limits in this third region, it's not good enough! So, they look at the overall shape of the plot in this region. If they find that it's gradually rising toward the 30 MHz end, they are quite confident that they have a radiation problem. However, if the plot starts drooping, or remains generally flat as 30 MHz approaches, they are likely to submit the prototype immediately to a lab for the formal radiated limit compliance certification. In other words, we can actually "see" the energy level in the 5-30 MHz region. If there is an undue amount of conducted noise energy in this region, radiation can't be too far off either!