TI.com では Internet Explorer をサポートしていません。環境や使いやすさを最適化するために、他のブラウザの使用をお願いいたします。

ビデオ・シリーズ

DC/DC スイッチング レギュレータを使用して低ノイズと低 EMI の特性を実現

エンジニアにとって、スイッチング レギュレータの EMI (電磁干渉) とノイズの低減が、困難な課題になることがあります。古代中国の風水は方角や色などに多くの制約を課しますが、PCB レイアウトで同様に過度の制約が生じ、開発したシステムが (EMI などを規定する) CISPR 規格に合格しない可能性があります。このため、多くの電源設計者は単純に、確立済みの方法としてリニア レギュレータを採用しています。そうすれば、電磁波低減という頭痛の種を回避できるからです。ただし、効率の低下が原因で、リニア レギュレータは熱に関する問題を引き起こす可能性があります。スイッチング レギュレータを使用すると、この問題を回避しやすくなります。両方の利点を活かすにはどうすればよいですか?このトレーニング シリーズと資料は、エンジニアがスイッチャの優れた効率を実現すると同時に、EMI やノイズの課題を克服する方法を説明することを意図しています。EMI やノイズに関する概念的な例と実際的な例の両方、およびベスト プラクティスを紹介するほか、車載システムやノイズに敏感なアナログ回路への電力供給に関する具体的な説明を取り扱います。伝導型 EMI や放射型 EMI の両方を制御する目的で使用されている各種規格に精通しておくと、役に立つことがあります。

Video Player is loading.
Current Time 0:00
Duration 0:00
Loaded: 0%
Stream Type LIVE
Remaining Time 0:00
 
1x
  • Chapters
  • descriptions off, selected
  • subtitles off, selected

      講演者

      [MUSIC PLAYING]

      Hi. I'm Marc Davis-Marsh. I'm part of the Systems Group at Texas Instruments. Today we're going to show you some tips and tools to solve your EMI problems before you even start designing.

      So first, let's talk what is EMI? So EMI is Electrical Magnetic Interference. And unfortunately, it's what your design is putting on to other susceptible systems. Now there's also Electrical Magnetic Compatibility. That's what other systems and the noise that they're putting onto your system.

      So today, we're going to be talking about EMI and not EMC. But it is important to note that EMC, if you solve EMI issues, you'll actually solve a lot of EMC issues as well.

      So EMI has two components. There's conducted noise and radiated noise. Conducted noise, of course, is going to be noise that you conduct onto a wire or other kind of system, onto the susceptible part. And then radiated noise is going to be noise that you radiate onto the other component.

      Now, there are several different types of tests that you can do for EMI. And they're broken up into end equipment. So you'll have a different test for a commercial device than you're going to have for an automobile than you're going to have for something that's in a factory.

      And in America, those tests usually fall under FCC 14 part B. And in Europe, they're going to be European Norms. And you also have CISPR 22 and CISPR 25. CISPR 22 is more for factory and home use, and CISPR 25 is going to be for automotive.

      Today we're going to be primarily focusing on CISPR 25 class 5. It's one of the hardest tests to pass. And we're going to give you some tips and tools to pass it.

      So here we're showing just the coupling mechanisms of the noise. So you have a noise source. In this case, it's your design. It's your buck converter and the associated equipment and electronics around it. You're the noise source.

      Now, there are four different types of coupling mechanisms. You have conducted coupling, which is going to be directly through a wire. You're going to have electrical coupling, which is also known as capacitive coupling. That's pretty much where you're in close proximity to something and you have an electrical field that's causing the coupling.

      You have magnetic coupling from a magnetic field. And then also, in far field, then you start to get radiated coupling. And all of those coupling mechanisms allow the noise from your system to go onto the susceptible system. And that's what we're trying to prevent.

      Now, for a test, what you're going to do is you're going to model the susceptible system as something that you can measure. So in the case of a radiated test, you're going to use an antenna to pick up the noise. And then you're going to measure how much noise was measured. Or if you're doing a conducted test, then you're going to use a LISN, or a Line Impedance Stabilization Network. And you're going to use that in order to measure the amount of noise that you're conducting onto the system.

      Primarily today I'm going to focus on buck converters. Now. All of the DC-DC converters that are switchers are going to have noise issues related to them. And generally, the problem is the same.

      You have a switching action. And the switching action is causing pulses of current. And those pulses of current can either be radiated or conducted onto your susceptible system.

      So let's see where the pulses of the current are coming for a buck converter. So for a buck converter, the general construction, you have an input capacitor. You then have a high-side FET and a low-side FET, which are going to do your switching actions. So that's chopping your DC input voltage and making a pulsed voltage from it.

      Then you have an L and a C. And that forms a filter, which turns that pulsed voltage into a DC voltage. So you're taking a higher input voltage, making a pulsed pattern, and then low-pass filtering that to get a lower output voltage.

      So how does that create a pulsed current, which is causing our noise? What you can see is when the high-side FET turns on, the current is passing through the LC filter and returning to the input cap. Then when the high-side FET turns off and the low-side FET turns on, you can see the current path is now through the low-side FET, through the LC filter, and back to the bottom return, the source of the low-side FET.

      Now, what this causes is a loop, where you get a pulsed current as the current is transferring from the input cap as the source and the low-side FET. So this pulsed current can create noise in a couple different ways.

      So if this input capacitor doesn't bypass all the current, then you're going to get a pulsed current that goes down your line. So that would be conducted noise directly onto your line. And if this loop is fairly large, then if you have a loop area and a current, and you pulse it at a frequency, then you've made an antenna. And so then you're going to get radiated noise from this loop.

      Another way you can get radiated noise is this pulsed current is going down a long input wire. So that pulsed current down a long input wire is also an antenna. And also, what you have here, if you look at the shape of this pulsed current, this node here is creating this chopped pattern. That chopped pattern is fairly high voltage. And that can capacitively couple to your outside system.

      So it's very critical that this loop remain as small as possible and you do a good job of bypassing this loop. When you do a good job of bypassing this loop, then you've solved most of your EMI problems right from the beginning.

      Definitely as you get to stricter and stricter standards, then you're going to want to do things like add differential filters at common mode filters. But definitely, the first part of the problem is getting this loop as small as possible.

      So besides placement of the capacitor right next to the device, what can we do to reduce the loop size? So really, a ground return is your best bet for reducing loop size. Now, it doesn't seem intuitive. But if we have a loop, like shown here, how does adding a ground return under that lower the area of that loop?

      And what happens is-- as you're going around the loop, what happens is your return is now directly underneath the previous return. So now instead of this large rectangular area being your antenna, the area is now the difference between the return path and the initial path. So that area is actually much smaller.

      And that's going to lower your inductance. It's going to lower your voltage spikes. It's the easiest way to get a low loop area. And as we showed before, low loop area is going to give you better conducted, better radiated EMI.

      So now what I want to do is go to ti.com We're going to start a WEBENCH design. And I'm going to show you on the part that we picked, the LM7360X, where you place the input capacitors.

      So here we are on ti.com. What I'm going to do now is I'm going to go to WEBENCH. I'm going to enter minimum input voltage of 4 volts and a max of 32. This is pretty standard for an automotive kind of design. What I'd like to do is show a CISPR 25 class 5 design, since that's a very stringent requirement for EMI. I'm going to say 3.3 volts out and 6 amps.

      WEBENCH gives you a couple of different options. It gets you the option of using a module, an integrated design, or a controller. Basically, think of it as a controller gives you the most flexibility. An integrated design tends to be a little less flexibility, but it gives you a more compact design. And then a module, of course, is going to be something where you have the least control over all of the attributes, but you get a design that's the easiest to do. And definitely, from an EMI point of view, we've shown that the modules, because of their tight construction, actually helps with EMI. Mine

      Today we're going to do an integrated design. I'm going to choose the part. WEBENCH has selected the LM73606. And there is also a Q grade. So I'm going to pick the Q grade of this device. And we're going to look at the datasheet.

      So I'm going to open up-- I'm going to open up the datasheet for the LM73605/6. And we're going to look at the pin placement and see how that works with that input bypass capacitor placement. So from what we were talking about, the input voltage, the ground need to be well bypassed.

      So on this part, we have PVIN, P ground. That's where we want to get our capacitor as close as possible. The rest of the loop was created by a high-side FET, which connects VIN to switch, and a low-side FET, which connects the switch to ground. So right here is our loop there we're trying to bypass with that capacitor.

      So that's the layout that you want to make sure is very tight, very close. And on the EVM, those capacitors are well-placed so that you can get a nice tight layout.

      So now that we've looked at the datasheet and we've seen where we want to place the capacitors in the layout, let's go back to WEBENCH. And we're going to see what it recommends for EMI.

      OK. We're back to WEBENCH. I'm going to open up a design. Now, the design that WEBENCH did, it started off with a design that's at a frequency of 500 kilohertz. Now, for automotive EMI, they like to have the switching frequency well below the AM band, closer to 400 kilohertz, or above the AM band, so you'd be closer to 2.1 megahertz.

      So what I'm going to do is I'm going to redo the design at 400 kilohertz.

      So I go back to the schematic and I can see that it looks just like our buck converter design from before. We have input capacitors.

      Here's the schematic. We can see it looks just like our schematic from before. We have input capacitors. Those go to VIN. VIN connects through the high-side FET to switch. Then it goes through the inductor and the output capacitor to create our LC filter.

      So it's just like the buck converter layout there we're looking in our simplified schematic. And now what WEBENCH can do for us, if we click on the Input Filter tab-- so I've clicked on the Input Filter tab. I want the design of the input filter for a CISPR 25 class 5.

      But that's not the only kind of design we can do with this. If we look at the noise standard, I could also go to a CISPR 22 design. And you can see that right here the limit lines have moved up. In other words, it's easier to pass a design a CISPR 22 class A than a CISPR 25 class 5 design. So for a CISPR 22 design, you might not even need to add an input filter in order to meet conducted tests.

      If we go back to CISPR 25 class 5, which is the most stringent, what we're going to do is add a filter to our design. So now what you can see here is that it's added an inductor, Lf. It's added a capacitor, Cf. Those are the LC filter components that it added to do filtering of our conducted noise.

      And it also added this bulk electrolytic. Why did it add the electrolytic? So what happens with an LC filter is if the resistances of the inductor and the capacitor are fairly low, it has a high resonance, or a high Q. The high Q can cause problems with stability of a buck converter.

      So generally what we do is you'll add an electrolytic or other capacitor, which has a high capacitance but also a high resistance. And that tends to damp the filter so that we don't get a resonance issue between the input filter and the buck converter.

      So now what we're going to be doing is we have two designs, one with an input filter and one without. We're going to take both of those into our screen room and we're going to do conducted measurements on them. And we're going to do a test with and without the input filter so that you can see the effect.

      We're going to be using the LM73605EVM. If you remember from the datasheet, this part has the input bypass capacitors very close to the device. Another nice feature of this EVM, on the back we have spots for a differential input filter. There's a spot for electrolytic capacitor, an inductor, and for the capacitors of the LC filter.

      So let's go into the screen room and test both our boards. All right. So here we are in TI's screen room. This is what we use for prescreening before we send it out to an outside test house when we're doing CISPR 25 or other EMI testing.

      If you want to look at our setup here, we have a large copper plate for ground. We have two Line Impedance Stabilization Networks, or LISNs. From the LISNs, then we've got our 200 millimeters of power line that's coming to our board. So we have an edge connector where we're going to plug in the boards. And then on the edge connector, I've attached a load resistor.

      So if we take our first board without the EMI filter, place it in the edge connector, and now we're ready to run our test. So here we are on the outside of the screen room. Here's the rest of the setup. We have a power supply that we're using to supply the 13 and 1/2 volts that's going to our device under test. That's getting put into the LISNs directly. And then we have a Rhode & Schwarz spectrum analyzer that we're using to measure the noise as given to us by the LISN.

      So what I've done is I've pulled the data from the Rhode & Schwarz and put it onto the monitor here for us to look at. We have on the top a peak conducted noise measurement. And on the bottom we have an average conducted measurement.

      Now, the peak and average have different limit lines. And so what you'll see is that the peak limit line here is higher than the average limit line here. And the reason for that is that peak measurement is a worst case noise seen measurement, so the limit line is a little higher. And an average measurement is an average of several different measurements and the measurement tends to be lower than the peak measurement. So there's two different limit lines depending on which type of sensor we're using in the spectrum analyzer.

      Now, you have limit lines at low frequency and we have limit lines here in midband range. This is several hundred kilohertz to 2 megahertz. And what you can see is here is our switching fundamental.

      So we set the EVM at 400 kilohertz on WEBENCH. And you can see here's 1 megahertz. And we walk it down and here's our 400 kilohertz noise from the design.

      And then we could see that the second harmonic is up here around 800 kilohertz. And we can see that that harmonic is going through the limit line. So for this EVM without an input filter, we are not passing the CISPR 25 class 5 test, which we didn't expect to, since this is a very stringent test. If you look at the limit line, it's below 40 dB microvolts, so really any noise in the hundreds of microvolts will make you fail this test.

      So what we're going to do is now I'm going to go and hook up the EVM that has the input filter and we can compare the results.

      So this board has an input filter. And now we're ready to run.

      So we've placed the second EVM into our setup. That EVM has the differential input filter installed. So I've taken that data off of the Rhode & Schwarz and transferred it back to the computer. Let's take a look.

      You can see that we still have our 400 kilohertz fundamental switching frequency, but the peak has been lowered significantly. Also, our second harmonic, which was failing our peak and our average test, is now well below the limit line. So basically, our input filter has taken all of our noise spectrum that we were seeing in this frequency range and lowered it well into acceptable limits.

      Now, in this circumstance, we allowed to WEBENCH to do the input filter design. But if you want to do the input filter design yourself, let me show you how to do that.

      So you can see on this slide, what we have is basically you're seeing the noise spectrum that we saw for the first EVM with no input filter. So you can see here is your fundamental. And they're well above your average and your peak limits.

      Now, how do you estimate how much filter attenuation to add? As we discussed before, the buck converter makes pulses of current, so you get a pulsed current at that input capacitor. That input current has this general shape, where it pulses up and then you get a triangular ramp upwards and then very sharply down.

      So to design your filter, you would basically take your Fourier transform. And then you're going to take that Fourier transform and you convert it into logarithmic form, so you have 20 log and microvolts. So then that converts your Fourier transform into dB microvolts, which is what the limit lines are in.

      Then you're going to see the difference between that Fourier transformed result and your max limit line. And from there, you're going to do your filter design. You're going to see how much attenuation you would need in dB in order to meet your limit. So you're going to do a standard LC filter design. And that's going to get you under the limit line.

      So after that, what you're going to do is you're going to add an electrolytic capacitor. And that's to damp the LC resonance. So generally, you take your input filter capacitor-- that's the one right next to your part. You're going to multiply the value of that by 4, so if you had 10 microfarads, you're now going to say that my electrolytic needs to be at least 40 microfarads. And the ESR of that electrolytic needs to be equal to the characteristic impedance of your LC filter, so square root of L over C. And that's how you calculate the filter if you don't have WEBENCH doing it for you.

      In summary, what we've shown you are a bunch of tips and tricks to help you with your EMI design for passing the first time. Remember, keep your input capacitor close. Use a ground return. And use WEBENCH to do your first-pass differential filter design. For more information, always go to ti.com/dcdcemi.

      View series

      DC/DC スイッチング レギュレータを使用して低ノイズと低 EMI の特性を実現