What Does the Metal Can on a Circuit Board Actually Do? Testing a Shield Over a Buck Converter
イチケン / ICHIKEN
Metal cases are sometimes soldered over power-supply or high-frequency sections of printed circuit boards. The usual explanation is that they work in two directions. They protect the components inside from external noise that could cause malfunctions, and they keep electromagnetic noise generated inside from escaping. Ichiken wanted to know how well that works in practice. They built a small switching power supply, covered it with a metal shield, and measured noise near the board, on its unshielded back side, and at a distance. They also checked whether the shield changes conversion efficiency.
The Test Circuit: A Buck Converter With a Deliberately Leaky Inductor
The test circuit is a buck converter built around the Texas Instruments TPS62841 power-supply IC. It has input smoothing capacitors, MOSFET switches, an output inductor, and an output capacitor. All parts were bought through DigiKey, the video's sponsor.
Noise comes in many forms. This experiment focuses on radiated noise, and specifically on how much noise from magnetic flux the shield reduces. To make the effect easy to see, Ichiken chose an inductor whose magnetic path is not fully closed. Flux therefore leaks out of the core whenever current flows. The idea is that a strong, well-defined source makes any reduction from the shield more visible.
The shield is tin-plated iron, 0.3 mm thick, and it is connected to ground.
Reading Near-Field Noise With Loop Probes
The first measurements use loop-type near-field probes, which pick up noise leaking from components very close to them. Ichiken had two sizes, one with a large loop and one with a small loop, and used both. The probe is held in a fixture and moved across the circuit. The signal goes to an oscilloscope in FFT mode.
Ichiken explains how to read the display. The horizontal axis shows noise frequency from 1 MHz to 500 MHz. The vertical axis shows intensity, so taller peaks mean stronger noise. The TPS62841 switches at roughly 1–2 MHz, and the strongest peak appears there. Its harmonics form a row of peaks across the spectrum. Ichiken calls the probe a convenient tool because simply bringing it near the board shows which frequencies are strong and roughly where they come from.
The large-loop probe has a limitation, though. Its loop area is big enough that it picks up flux from the whole circuit, so it cannot show exactly where the noise originates. The small-loop probe sees only the local flux. Scanning the unshielded board with it confirmed Ichiken's expectation: the signal was strongest directly above the inductor. Ichiken notes that "noise" here really means magnetic flux. Because the inductor's magnetic path is open, leaking flux links with the probe's loop.
Adding the Shield: A Large but Incomplete Reduction
The same scan on the shielded board produced a much weaker signal. Side by side, the shielded circuit showed a large reduction from high frequencies down to low frequencies.
The flux did not drop to zero, however. Some components remained in places. Ichiken offers two possible explanations. First, the shield does not sit flush against the PCB, and flux could leak through the small gap. Second, some flux may simply pass through the shield itself.
Ichiken adds an important caveat. The shield reduces how much noise reaches the outside, but it does nothing to the source. The inductor produces just as much flux as before. The shield only limits how much of it escapes.
The Back Side of the Board Is Not Protected
Out of curiosity, Ichiken ran an extra test. The shield covers only the top of the board, so they probed from the underside.
On the unshielded board, the back side showed noise concentrated around the inductor's location, still quite strong. On the shielded board, the back side looked very different from the top: the noise was about the same level as without the shield. Ichiken's explanation is that nothing shields the back side, so the probe picks up the noise there without obstruction. The back of this board is a solid ground plane, yet noise still came through wherever there was no metal shield.
Ichiken draws a general lesson from this. To keep noise from getting out, the circuit apparently needs to be shielded completely. The same reasoning applies to noise coming in. If external noise arrives only from the top, a top shield might be enough. If it arrives only from the bottom, a bottom shield would be the right choice. Shielding both sides seems likely to work quite well.
How Magnetic Flux Behaves Around an Iron Shield
Ichiken then explains, as a conceptual picture, what might happen to the flux near the shield. Without a shield, flux leaves the inductor and spreads outward, which is why the probe detected it clearly. With the shield in place, the flux behaves very differently. Iron is ferromagnetic, so rather than passing straight through the shield, the flux can travel through the steel and loop back around. Ichiken presents this as a reasonable way to think about it, not as something they measured directly.
Not all of the flux follows that path through the iron. Some probably still leaks out, and that may be what the probe detected above the shielded circuit.
The inductor is also not the only source. Any trace carrying current produces a magnetic field around it. The probe picked up some flux directly above the shield and also over parts of the circuit outside the shield. That signal could be flux leaking from the shield, or it could come from current flowing in the circuit's traces. Ichiken says plainly that they do not know where the flux came from or how much came from each source, and that this was not verified in this experiment.
Eddy Currents, Heating, and Efficiency
Ichiken also raises a possible cost of the shield. When changing magnetic flux tries to pass through metal, Lenz's law causes eddy currents to flow on the metal surface in a direction that opposes the flux. This is the same principle an induction cooktop uses. The inductor's flux could therefore heat the shield slightly and waste some power. Because the shield is steel, hysteresis losses might also play a part.
To test this, Ichiken measured power conversion efficiency on the same components and the same PCB, first without the shield and then with it soldered on. The graph plots efficiency against output power. The two curves were almost identical. Ichiken is careful about what this means. Efficiency may have changed slightly, and it would be more accurate to say that their instruments could not resolve a difference, not that no difference exists.
Far-Field Measurement With a Log-Periodic Antenna
The last test looked at noise radiated by the circuit as a whole instead of by individual components. Ichiken used a log-periodic antenna placed about 80 cm from the circuit. They did not have an anechoic chamber, so they moved to a location with as little ambient signal as possible. They were also unsure whether 80 cm really counts as the far field.
With nothing connected, the display showed background noise. Connecting the unshielded circuit clearly changed the amplitude in certain regions of the spectrum. Connecting the shielded circuit also produced visible changes, including around 500 MHz. Overall, though, the readings stayed close to background, with little difference from the unconnected state.
Ichiken suspects the measurement setup. They were using the oscilloscope's spectrum-analyzer function, and it may not have been sensitive enough. Even so, the noise spectrum did differ between the shielded and unshielded circuits, and Ichiken suggests that something else still on the circuit may be responsible. They also explain the difference between the two tests. Close to the circuit, magnetic flux dominates. Farther away, the noise is electromagnetic waves, and this test looks at the whole circuit's radiation, not at individual components. Their conclusion is that the oscilloscope's FFT function made this measurement difficult, and they would like to repeat it with proper equipment and a dedicated spectrum analyzer.
Conclusions and Open Questions
In this setup, the metal shield blocked much of the noise close to the circuit, but not all of it. Leakage appeared through the gap between the board and the shield, from the unshielded back side, and possibly through the shield itself. Even so, Ichiken judged that the shield worked well in the near field. They suggest that it may be more useful to think of it as reducing noise coming into the shielded circuit from outside, and not only as reducing noise going out.
In the far-field test with the log-periodic antenna, Ichiken could not clearly see the shield's effect. They note that this test measured radiation from the entire circuit, not from individual components, and that covering the whole circuit in a metal can would probably give different results. Ichiken describes the whole project as a short, quick experiment, and it left them with the impression that noise suppression is quite difficult.
This video is brought to you courtesy of DigiKey. Hello. I'm looking at the circuitry of the power supply and the high-frequency circuitry. On the printed circuit board, sometimes these metal cases are mounted. Aren't you wondering what this is?
The conclusion is that the metal case protects the electronic components inside against malfunctions caused by external noise and, conversely, serves to keep electromagnetic noise generated within the metal case from going out.
But I'd like to know how effective it really is. I'm sure many of you are wondering. That's why we're using this circuit today. If you put a metal shield over it, like a metal covering, we will examine how effective it is. But first, here are today's sponsors. Go ahead.
Where do you get your electronic components from? This time, for those who are new to DigiKey, we will look at how to select parts at DigiKey together! I will look for the MLCC to be used in this video as well.
Type MLCC in the search field on the top page to search. Your search results were displayed as ceramic capacitors. Click here. Approximately 510,000 items were found in the search. However, there are some omissions in the word search. Selecting Products, Passive Components, Capacitors, and Ceramic Capacitors from the top page is a sure choice. That would bring the total up to about 870,000 items.
In my case, I mostly get stock parts. I often check where it's in stock. I would like to use Murata's MLCC, so select Murata from the manufacturer field. In addition, the voltage can be selected in the range from 10 to 50 V by holding down Shift. And when we did, we found a lot of parts in stock.
Of course, parts can be purchased from a single item. Free shipping when you buy over 6,000 yen. We support qualified invoices. Convenient purchase of electronic components for both individuals and corporations. We hope you will try DigiKey, too!
This time, I made this circuit to verify whether putting a metal shield over it prevents noise from being emitted to the outside. The circuit is a buck converter. Capacitors for input smoothing, MOSFET switches, an inductor at the output, and a capacitor at the output. This time, as a power supply IC, I'm using Texas Instruments' TPS62841.
There are many kinds of noise. This time, we'll look at radiated noise. In particular, we'll see how much noise due to magnetic flux is reduced. So that the results are easy to understand, we have devised a component selection for inductors. I made sure to use this type of inductor in which the magnetic path is not completely closed. Since the magnetic path is not closed, magnetic flux often leaks out of the core when current flows. This way, I expect it will be easy to understand how much noise is reduced when the metal shield is placed.
The metal shield to be placed is tin-plated iron. Thickness is 0.3 mm. This time, the shield potential is connected to ground. By the way, the components used in the circuit we made this time were all purchased through DigiKey.
First, using a near-field probe like the one shown here, we will look at the leaking noise in the immediate vicinity of the component. This time we decided to use the loop type. I have near-field probes with a large loop and with a small loop. I'll use both this time.
First, for unshielded circuits, use a large-loop near-field probe to look at the noise. It's fixed so I can manipulate the probe in this way. Try to move the probe over the circuit. Then you see the noise component like this. Just by bringing it close, you can see where in the circuit the noise is coming from and how strong it is. Useful item.
If you look at the oscilloscope, you can see that there is noise. How to see this screen: now displaying oscilloscope FFT mode results. We are looking at the frequency components in the signal. The horizontal axis is the signal frequency, in this case the noise frequency. We are looking at 1 MHz to 500 MHz. The vertical axis is noise intensity. The more this spectrum extends upward, the stronger the intensity of the noise.
The switching frequency of the Texas Instruments power supply IC we are using in this case is about 1 to 2 MHz. The spectrum there stands strongest. And its nth harmonics appear like this. You can see where the noise component is strong and where components of each frequency are coming from. I think you know exactly what I mean. It's quite convenient, isn't it?
But the probe I'm using now has too large a loop area. Where in the circuit is the noise coming from? We don't know for sure. If anything, you're picking up the magnetic flux from the entire circuit. Therefore, we will try to use a near-field probe with a small loop area. By doing so, we can see the local magnetic flux.
If you actually look at the local magnetic flux on the one without metal shielding, it looks like this. I'm moving over the circuitry a little bit at a time. Where the noise is strongest comes out plainly. Looking at this one, it was what I expected. Noise is highest just above the inductor. You say noise, but this is magnetic flux. This is because the magnetic path of the inductor is not completely closed, and the leaking flux links with the near-field probe.
Now let's try the same thing on the next circuit with the metal shield attached. I'm trying to get this close. Can you see? The signal strength has been greatly reduced. The actual operation over the circuit is shown here. Signal strength is quite weak. Here is the circuit without a metal shield, and here is the circuit with the metal shield. From high frequency to low frequency, it's down quite a bit.
As you can see on the oscilloscope screen, the magnetic flux has not been completely reduced to zero. Some components remain in places. The printed circuit board and metal shield are not completely attached. There is a small gap between the two. Magnetic flux leaks from such places, and it may also simply be passing through the shield.
Also, there is one thing to note here. Noise going out to the outside—in this case, the magnetic flux—was reduced. The total amount of noise coming from the source has not been reduced. The metal shield just reduces the amount that goes out.
I'm curious, so I'll do some additional experimentation. I just placed the shield on the top side of the board only. There is no shielding on the back side of the board. Therefore, I'll also see how much noise I'm getting when the probe is applied from the rear side.
It looks like this. We are looking at a circuit without a metal shield. Noise like this. The area around the inductor is still quite strong. At its strongest point, this is about right. Next, we'll see the back side of the circuit with a metal shield. The results are quite different from those on the top side of the board. I'm getting noise as usual. That's about the same level.
As for why this happened, the back side is not shielded. The near-field probe is picking up that noise well. The back side is a solid ground plane. For areas that are not metal-shielded in this manner, the noise is coming through normally. This result shows that, in order to keep noise out, I think it needs to be completely shielded.
On the other hand, if there is a shield, noise from the outside will not be able to get inside. For example, if some noise is coming from the top side of the board, it looks like you could put a shield on the upper side. On the other hand, if the noise is coming in only from the back side, it looks good if you put a shield on the back side. The best thing to do is to put shields on both sides. It sounds like something that would work pretty well.
Earlier, we used this near-field probe to explore the magnetic flux above the shield in this way. We will discuss how the magnetic flux behaves around and in this shield. First, if there is no shield, the magnetic flux from the inductor leaks out like this. If you approach the inductor with a near-field probe, you can sense the magnetic flux properly.
But with the shield in place, the behavior of the magnetic flux in this area has changed significantly. The shield material used in this project is iron. Iron is ferromagnetic. Therefore, the magnetic flux out of this inductor doesn't just jump through the shield. It goes through the iron like this, going around in a circle. Let's look at it with that image in mind. That sounds like a good way to consider it.
But not all the magnetic flux goes through this iron. There is still some flux that seems to leak out in part. I was picking up magnetic flux even with a near-field probe in a shielded circuit. Possibly it could have picked up this component.
Also, there are other areas besides the inductor where magnetic flux is emitted. For example, if any part of the circuit has current flowing through it, naturally, a magnetic flux forms around it. With the near-field probe in this case, I was able to detect some magnetic flux right above the shield. It was picking up magnetic flux even on parts of the circuit outside the shield. This could have been picking up the magnetic flux leaking from the shield, or magnetic flux due to the current flowing through the circuit traces. I just don't know where and how much magnetic flux is being picked up. That has not been verified at this time.
Also, when changing magnetic flux tries to penetrate the metal, by Lenz's law, eddy currents flow on the metal surface in the direction to cancel the magnetic flux. This is the same principle of operation as an induction cooktop. Therefore, this metal shield is also affected by the magnetic flux emitted from the inductor. Slight heat generation could lead to power loss. Also, the shield is made of steel. Hysteresis may also be in play.
Eddy currents were mentioned. We also verified how much the power conversion efficiency changes with and without shields. I'm experimenting with the setup here. Using the same components, we compare the unshielded circuit with the one with the shield soldered on. In other words, with the same components on the same printed circuit board, we measure efficiency with or without shields.
Here are the results of the experiment. The horizontal axis is output power. The vertical axis is power conversion efficiency. As a result, conversion efficiency remained almost the same with and without shields. It is possible that it may have changed slightly. Perhaps it would be better to say that I failed to read the difference with the instrumentation used in this study.
So far we've seen a lot of things with the near-field probe. We'll also see how much the noise emitted by the entire circuit changes with or without shielding. Using the log-periodic antenna over here, we look at how much noise is being produced by the circuit as a whole. We were not able to prepare an anechoic chamber this time. I've moved to a place with as little signal as possible. I don't know if this can be called the far field. The distance from the circuit to the antenna is about 80 cm.
Now let's experiment. This is a situation where nothing is connected to the circuit. This is what we call background noise. First we'll connect the unshielded circuit. Has it changed? Did you all see? Things have changed around here. The amplitude level has changed a lot.
Now nothing is connected again. Next, we will connect the shielded circuit. Then it looks like this. Connected. You could see it pretty well. It's been a lot of fun around here. The frequency is just about 500 MHz. Like this, you see? If you connect it again, we have been measuring in this way. It's almost like background noise in this case.
Maybe it's the measuring device. This is not a spectrum analyzer; I'm using the oscilloscope's spectrum analyzer function. Maybe it isn't sensitive enough. Almost no change from background noise. But with or without the shield, the spectral component of the noise was different. I still have something on the circuit. Maybe that's what makes the difference.
I'm not using a near-field probe this time. We are looking at the noise radiation of this circuit as a whole. Of course, in close proximity, the magnetic flux was dominant. When you move away, it comes down to electromagnetic waves. I could see the noise in this circuit as a whole and how much it would change.
I conclude that it was a little difficult to measure with the FFT function of the oscilloscope. I'd like to try using more proper equipment and a dedicated spectrum analyzer.
This time, I've been testing how effective the metal shield on the circuit is. In this experimental setup, I was able to shield noise at close distance from the circuit with a metal shield. It just wasn't completely shielded. I could see components escaping through gaps between the printed circuit board and shield, from the back side of the printed circuit board, and through the shield itself. But as for the near field, overall, I thought the shield worked well.
In this case, for the near field, it may have had the effect of reducing the noise going out. If pushed, I'd say it may be better to look at it as having the effect of reducing noise entering the shield of this circuit from the outside.
For the far field, using a log-periodic antenna, I couldn't get a good look at the shielding effect. I'm not talking about component-by-component noise on this one. I was watching the noise radiating from the entire circuit. I'm afraid that's how it turned out. For example, if you cover this entire circuit with a metal can, I'm sure it will turn out differently. This time, as an experiment, we did it in a very short time. I thought that noise suppression is quite difficult.
If you found this video helpful, please give us a high rating and subscribe to our channel! If you have any concerns, please contact us. Please try to leave a comment as well! Thank you for watching the video to the end.
Article published · Updated