Hi, everyone. Thank you for joining Teledyne LeCroy today for our webinar on real-world high-voltage probing comparisons, highlighting basic specification and performance differences.

This is part two in our three-part Oscilloscope, Probing in Power Electronics Webinar series.

Just a little bit about Teledyne LeCroy.

We were founded in 1964 by Alabama native Walter LeCroy. Our corporate headquarters is located in Chestnut Ridge, New York, where many of our oscilloscopes are designed and manufactured.

We have sales and service offices all across the United States as well as the rest of the world, and we offer a wide range of test tools for electrical, physical, logical, and protocol layer testing.

Without any further ado, I will turn things over to Ken Johnson to begin today's presentation.

Okay. Thank you. We'll just start with a little bit about myself. This is who I am.

I've been around for a while. I graduated from RPI, played a little ice hockey—not at RPI—and I know a little bit about probing and power electronics. So, I'm looking forward to sharing info with you today. And this is our other presenter, Jon.

And, Jon, a few words from you.

Hi, everybody. I'm on the product management team here at LeCroy, focused on power electronics and the power application space, and I'm excited to present this contact alongside Ken here.

Okay. Thank you.

Alright. So we've got a lot to cover. We started this webinar years ago. It was one part.

We rushed through it, then we went to two. We thought we had plenty of time. We didn't. We went to three parts, and I think we still felt a little rushed in part one.

So we've got a lot of content to cover. We're going to dive in. This is how it splits out between parts two and three.

So we're going to cover a lot of electrical measurements and really bear down on looking at gate drive signals for the most part today and just helping you understand some of the things we talked about in part one and how they literally ? these different probe specs and characteristics ? impact what you see on your signal. Just because your probe says it's true, doesn't mean it's true. You have to understand the signature of the probe and know a little bit about it. So we're going to cover a bunch of examples to make that more real for you, and then we're going to start to talk about higher voltage switching measurements at the end of today, and we'll compare some electrical probe topologies. And then part three is all going to be about a lot of optical and some specialized, 60 V common-mode probes.

And if we have a little time, we'll do a little power rail stuff in that webinar. So without further ado, let's jump in. So as I just said, if you know something about the probe's performance and its characteristics, and you can understand how the specs impact the signal, then you can connect that to the signature of the probe and know what's real and what might be caused by the probe in your circuit, not necessarily in your circuit, which is very important.

Okay. So the first thing we're going to do is look at upper-side and lower-side gate drive signal measurements. Upper-side gates are really very challenging signals; they're floating at the DC bus voltage and they require high-voltage isolation.

There's a lot of switching and interference that can be passed through from one side, the upper side to the lower side or vice versa. And you really need to have a probe with good common-mode rejection ratio and not a lot of probe loading. So, some probes work a lot better than others and some of the common misconceptions people have aren't necessarily true. Sometimes a basic electrical high-voltage diff probe can work pretty well.

Sometimes not. It really depends a lot on the probe. So let's bear down and look at our first example here, and we're going to focus first on looking at really rather low-frequency signals. This is from a pretty old variable frequency drive.

It's a pretty high-voltage DC bus, using silicon transistors and power devices, and it's vintage 1990. So these are not really fast rise times. And because they're not really fast, it's easy to see the impact of noise, bandwidth, and frequency response differences between some older and newer probes, both from Teledyne LeCroy and from Tektronix, a player in this market as well. So the first thing we're going to do... here's where we are in the agenda.

We're looking at this particular example here.

This is the equipment we're using. It's one of our eight-channel, 12-bit, 1 GHz oscilloscopes.

We have an older high-voltage differential probe, the ADP305, and a newer one. The performance characteristics are quite a bit different between these two even though they look very similar.

And then the same is true for Tektronix. They have a newer one and an older one, kind of the same vintage as ours.

They have a little bit more bandwidth on this one. We'll take a look at what that means. And then we've got this pretty old drive.

It works. It's 500 ns rise times. These are pretty slow, but it's a good source to look at basic performance. It really helps us isolate some simple things.

These are the two Teledyne LeCroy probes.

You might be familiar with this one. It's an older blue one. It's pretty lightweight. This is the black one, it's heavy, feels very substantial.

It just so happens this one is a much better performer than this one. This one's pretty old and it was probably fine for its time, but its time was 20-plus years ago. So, the first thing we're going to do is look at the lower-side gate drive, which is not the most challenging measurement. That's the green dot indicated here on the drive. We're just starting out at this just to kind of give ourselves a reference. This isn't as challenging a measurement. So by showing what the probes look like on this side, it helps us understand what the differences might be when we look at them on the high-voltage side on the upper side.

So we're going to start off. We're going to look at the lower-side gate. We're going to look at the newer LeCroy high-voltage diff probe in yellow and the older one in magenta. And we've just taken a longer acquisition and we've zoomed in on here in the bottom.

And, we've got some parameters on and whatnot. And I've zoomed in even further here. You can see the highlight of this white box is shown in this lower trace here, and you can look carefully. You see some differences in the signal.

It's not super apparent. So if we turn on some parameters, you see some slight differences. Again, not a huge difference here. I mean, it's really pretty small.

There are some differences in the amplitude, and again, we're looking at a gate drive.

The HVD probe is a 1% accuracy probe versus this 2% accuracy probe and a little bit more bandwidth in the yellow newer probe than the magenta older probe, which means there's a little bit better rise time even though the signal is rather slow compared to the bandwidth of the probe. You get a little bit of the impact there. Now if you really zoom in on this base, this is where some of the things we talked about in part one really come alive.

And so you see the magenta, the older probe, kind of varies around the zero level a lot, and there's a little bit more transient pickup. This transient here is real. It's in the signal. It's coming from the switching on the upper side, most likely.

I'm 99% sure that's where it's coming from. And then you see this kind of non-flatness of this trace here. That could be due to two things. It could be due to the CMRR performance of the probe not being good at lower frequencies, not being as good as the newer yellow-colored trace probe here.

It could also be due to the frequency flatness of the magnitude response. A little bit of fractions of a dB can really make a difference here. So if we're varying from DC to a few MHz, and it's not a really flat frequency probe response, you'll see this kind of variation in the response. Whereas the newer probe is nice and flat.

It's kind of what you'd expect and a little less noisy as well.

It's maybe two-thirds of the noise of the Magenta probe, the old probe.

So that's really easy to see.

And we see the same kind of things on the top. As it's rising up here, we see a much more consistent clean shape on the new probe and something a little bit different on the old probe here. Again, a little bit of peaking, a little bit of bouncing above the line, and then kind of settling back down again. And we're going to go into a little bit more detail about that.

And again, a little bit more immunity here. Better common-mode rejection ratio means we're not picking up as much of this transient, which is floating around in the air, so to speak, and being picked up by the antenna, which is the leads. I mean, we have to remember, just going back a few slides, these are our probes. These are two pretty long leads.

And if you don't connect them very well in the circuit, you create a big loop. They're going to be a big antenna. So just a quick note, when I did these tests, I kept these two leads very parallel to each other, which is a good practice. Sometimes you even tape them together with electrical tape so they don't open up and develop a big loop, and that helps.

And then some probes are designed a little bit better so it minimizes the pickup, both inside the amplifier here with the CMRR and also in how we can attenuate some of the interference coming here at the input. So just a note there that the new probe does a better job of that, which you can see in this transient pickup yellow trace being a lot lower.

Alright. Just kind of keeping advancing. When we look at the base after the fall time, we again see the same kinds of things we saw before in the base before the rise. So we won't dwell on that too much.

And just before the fall, we see the same kind of things we saw just after the rise. So really, nothing different here. It's kind of consistent pretty much everywhere you look. Now let's take a look at a competitive probe.

And the first takeaway I want you to have before I go any further is, there's nothing in this presentation that's saying any of these Tektronix probes are bad probes. In fact, I think the P5205, which is their vintage 1990s kind of probe, is better than our ADP305. Thankfully, we've moved beyond our ADP305.

This THDP0200, it's a pretty good probe.

And so not saying anything bad about that. Now when we look at these two probes, we can see they have some of the same things. The newer probe, a little less noisy than the older probe.

Which you would expect. Newer things tend to be a little quieter, less noise.

And so we're going to use the newer probe as kind of the reference standard going forward for all the Tektronix compares.

And if we look on the left here, it's Teledyne LeCroy versus Tektronix. I've squeezed and scrunched the screenshots a little bit, so it kind of matches up in the same shape and size. And you look at something really quickly and you say, wow, that Tektronix probe is quite a bit more noisy.

And it is, in fact, the exact same signal. I cross-triggered the Tektronix scope from the LeCroy scope, so it's the exact same signal. And you might say, why is that so much more noisy? And this is really operator error, the operator being me.

And I had this operating in 500x mode to measure a gate drive, which means it was using 10 times the attenuation as the Teledyne LeCroy probe, which autoranges down to 50x attenuation. So my mistake, I forgot. This is also a 12-bit scope on the left, which helps, and an 8-bit scope here on the right. At the time I took this acquisition, that was the best Tektronix could do in terms of vertical resolution.

So this is really pilot error, but it's just cautionary too. Make sure if your probe does not autorange and you're measuring a small signal, make sure you put it in the best possible attenuation. In this case, I should have put this probe here from Tek in a 50x attenuation.

Okay. So, but, yeah, going back really quickly, you see, otherwise, it's performing pretty much the same. Let's say it's comparable quality in terms of immunity for this particular variable frequency signal.

Okay. Now let's look at the upper drives, the upper data side. That's a little more challenging here.

It's tougher. There's no doubt about it. Now, what makes it a little bit easier in this case is the signal speed is pretty slow, but it's still an almost 700 V DC bus. So it's a lot of voltage switching, and that generates a lot of interference and also other common-mode issues.

So, again, we're looking at the old LeCroy in magenta and the new LeCroy probe in yellow. And what jumps out now is I seem to have this big bump here. I seem to have a lot of overshoot happening here with this ADP305, a lot more than I have with my newer HVD3106. So that deserves some exploring.

So let's zoom in on that. And one of the contributors to this, and we touched on this in part one, is this probe, the ADP305, does what almost everybody did at that time, which was peak the frequency response as the magnitude, or as the bandwidth, reached the rated bandwidth of about 100 MHz. And everybody, as best as I know, did this at that time. So we were not without sin.

But what that does is it pushes the rise time up because now you have a boost, basically, of 6 dB in what the signal really is at these very high frequencies, and the very highest frequencies are represented as the rising edge. So it kind of creates this mistaken impression that my signal actually has a ton of overshoot, but it doesn't. This is a better representation ? the HVD ? which has very little peaking, maybe one or 1.5 dB, a very minimal amount.

So, that's really what we're seeing here.

It could also be caused by common-mode rejection ratio not being as good or maybe the probe loading being impacting a little bit and not quite as good as the new one. So it's hard to unwind all these things individually, but more than likely, that is mostly due to the peaking of the probe.

And here, you see on the high side, you see a little bit more difficulty in the old probe kind of rejecting the switching from elsewhere in the circuit or interference from elsewhere in the circuit. You really see that magnified here on a magenta trace. And also this flatness issue, which we've talked about previously, same kind of thing.

Now if you look at the comparison between the Tek on the right and the LeCroy on the left, and again, a couple of things pop out here. Again, my fault, it's in 50x. So we have to ignore the extra noise here that's operator error, my operator error.

Otherwise, ignore the fatter trace, the noise. You see the same kind of resistance to immunity here on the signal. That's a pretty good probe. It's about the same amplitude, maybe a little bit more poking up here, a little not quite as good CMRR, which is impacting the signal.

So, otherwise, pretty good. But when you zoom in here, here's a couple of things going on. Number one is, if you look in the Tektronix manual, this is that peaking that I was talking about, and they actually represent in their manual. Hey. We're going to have an infinitely fast step. It's basically going to have this overshoot, and it's about a 15% overshoot here. And you're seeing that a bit here.

And, also, this could be due as well, and say, probably not the best operating practice in the world, but I kind of got the signal off the screen here. This is the original channel acquisition, and I've kind of, like, offset it below the screen and then punched the signal up. So this could be something to do with the overdrive recovery of the oscilloscope and the probe itself for all I know. So it's really hard to know. We don't see that here.

So it's really hard to know what it did to cause this, but it's definitely a different signal shape, and we can see that here. And it could be due to a couple of different factors. Otherwise, it's fairly well maintained except for that little bit more overshoot.

So what are the conclusions we draw here? It's not an infinite number of different tests, but a couple of things really stand out is that if we separate the speed of the source, meaning using a source that doesn't have the fastest rise times in the world, then some things are a little easier to jump out. Just the raw immunity from interference, which is represented by the probe CMRR, the noise of the probe, any peaking that might exist in the frequency response to the probe. And so you can see that I would generally say that older differential probes, 20, 25-year-old designs, they're just not as good.

They weren't necessarily terrible, but they were the best that we could do at that time, all of us, every vendor. And for some vendors, the cheapest probes are still the best they can do. So, I think you just have to make sure that if these are critical measurements for you, use a newer generation probe. Invest a little bit of money, use a newer generation of probe, and you're going to get better performance.

And these performances are really pretty apparent in terms of overshoot, rise time, and immunity to noise and things like that. So that's kind of the big takeaway of this first comparison.

Alright. So let's move on to the second comparison here.

This one is now we're looking at a lighting supply. It's an upper-side gate drive signal only. It's about a 500 V DC bus.

The rise time is quite a bit faster compared to what we looked at before, about 70 ns, whereas the last one was only 500 ns.

And so we're looking at a bunch of different high-voltage probe topologies all from Teledyne LeCroy and different vintages and models of products we've offered over the years. This is where we are in the agenda.

Just as a reminder, this is the equipment we used in this comparison. So another, this time a four-channel, 12-bit oscilloscope, also with 1 GHz. So well more than the bandwidth of the probes. Then we've got a bunch of different probes, all plus or minus 100 MHz bandwidth. I didn't list the bandwidth for this DA1855A, but that's 100 MHz as well.

And we're just going to look at how all these compare. And here we're throwing in, at the time of this when this was done about eight, nine years ago, this was the first of its kind optical probe, first ever, from a major oscilloscope manufacturer equipped to work with an oscilloscope. So kind of a one of a kind at that moment in time. And then we're at a customer site looking at their half-bridge LED driver with these characteristics.

So here are the four probes. These two are familiar from the last compare. I'm putting in two new ones here. This looks like a box of an amplifier, and it is, basically. It's a differential amplifier with an attached probe pair.

The reason this is kind of unique is this probe pair is like a matched set. And by matching the performance of these two sides of the differential measurement, then basically, you can achieve a very high CMRR at the input to the two terminals of this device. And so, this was a very popular item back in the late nineties and in the 2000s. It was pretty much the premier measurement system. So you've got to compare here. It's interesting to show it.

The new solutions, I'll say, on the right here are pretty much better than this. And so if you have one of these old things, it's got its place. It did some things really, really well, but also, we've moved on. We've got some better things now.

So we'll show you some data on that. So this is the circuit we're measuring. It looks something like this in general.

It's an AC input or the power factor correction circuit, a half-bridge. We're looking at the upper-side gate drive right here.

And that's floating. Obviously, it's floating at the DC bus voltage when we're doing these measurements. So most people would say, well, electrical probes aren't going to work there, and you're going to need an optical probe. That's just common sense. So let's just test that assumption and see how we do. We've got a lot of comparisons here.

So I'm going to just kind of give you this ability to go back and look at this, but we're first going to look at a few probes and isolated an optical probe against this old differential amplifier. And then we're going to look at a bunch of probes when they're connected in the circuit simultaneously, which you really should not do. And so we're going to demonstrate why you don't want to do that, how we're going to impact the loading of the circuit by doing that, and also how an old probe impacts the loading of the circuit as well. And then we're going to look at each probe alone in the circuit.

Basically, our four probes here. One, two, three, four. Alone in the circuit, you can see, hey. This is how they perform, and you can draw your own conclusion here.

It's a pretty good study. So this is that first comparison I talked about.

And here we have in magenta, we have that 60 MHz optical probe. And then we also have the, in tan. This is a memory store trace. So this wasn't probed simultaneously.

These probes were not in the circuit at the same time. That's why the edges don't line up. So you just have to kind of take for faith that it's going to be roughly or should be roughly the same signal shape. We just happen to capture them at different points in time.

And what you see pretty quickly here is you see that this DA1855A, it did okay for its time, but it's got a lot of loading in the circuit here. The zero level isn't really well established. It's got a bunch of peaking due to loading more than likely.

It's picking up more transients due to a poor CMRR.

So it's not the greatest thing in the world. It was okay for its day.

But the optical probe is really the best reference, and that's what we're seeing here. It looks pretty faithful to what we expect that gate drive to look like.

So if we advance the slide, now we're going to do that same comparison. Now we're looking at both probes in the circuit at the same time, which you don't want to do. This is not a recommendation. This is a warning.

Right? So we had that last slide, which basically said, look how good this magenta trace looked. Right? Nice and clean and crisp.

And now we just put that instead of on a memory trace, we put the DA1855A in the circuit at the same test point at the same time, and look how the signal as measured by the optical probe degrades. It degrades quite a bit because now this other probe is loading the circuit. So this is just a cautionary tale that you should never connect two probes in the circuit at the same test point at the same time without expecting something like that to happen. And it doesn't matter what probe you're connecting.

Here, I've now connected the HVD3106 with the optical probe at the same time. They're both doing the same thing. So, it's not so great, and so don't do this.

You just never want to do that. And then to take it to the extreme, the oldest probe, I stuck that in the circuit, and that was the worst of all in terms of how it measured the signal and what it did with loading when the other probes were connected as well. Now my DA1855A is a memory trace, and I'm looking at the HVD.

I'm sorry. I'm looking at the DA1855A and the ADP305 in the circuit at the same time, and it's kind of like the worst possible scenario. So, again, just cautionary tales. Don't do this. So now let's look at the four probes in isolation, only one of them in the circuit at a time. And what you see here is, this is the old differential probe, the ADP305, the 25-year-old one, and, it's borderline okay, I guess. No one would want this probe if they knew this is how it performed.

You've got this transient here that's above the Miller plateau. Never a great idea.

You don't want that transient to be confused with something that might cause the device to switch or something. So, this is not great. We'd all agree this is really not what we want. And most of this performance here is due to the excessive electrical loading of this particular probe and poor CMRR and also, that's really electrical loading and poor CMRR.

So when you look at that DA1855A, that's the box amplifier with the probe pair. That doesn't look too much different, actually.

So, again, we have the same issues with this transient that's picked up. It's really not real. It's not that high. It's a little high. It's definitely not above the Miller plateau.

So but otherwise, it looks pretty much the same. So these are the two oldest probes and you compare, and it's not so great. Then we look at the newest high-voltage differential probe, and now we can start to look at this and say, that's not too bad. And this is where a lot of people just write off high-voltage differential probes, these electrical rectangular-shaped probes with the long leads.

They say they're no good. I can't use them. But that's not really true. The better ones are really not too bad.

And this one shows a pretty good signal shape. Yeah. It's not perfect.

It's nice and clean electrically.

Our transient here is nicely suppressed by the common-mode rejection ratio of this probe, which is actually really high. So it's about 65 dB at 1 MHz.

And so it's really doing a pretty good job of suppressing that transient, and it's just below the Miller plateau. So this isn't too bad a result. And kind of the best one at all is the optical probe.

And this is kind of a case study for, I don't love the noise on this. Optical probe topologies should generally have more noise for sure, but the simple reality is they can provide a really good shape reference. And so having an optical probe in your lab, even if you don't have a lot of them, if you just have one to use as a shape reference is not too bad an idea. It's actually a pretty good idea. And so, really, that's the takeaway here.

The old probes and technologies are really being superseded by some new technologies with better and more consistent performance.

So optical technologies, generally, in general, better CMRR, less circuit loading. I should have indicated more bandwidth for some of the more expensive newer ones, but they generally have more noise. You just can't get around it. It's how it is.

But some of the high-quality electrical high-voltage differential probes can perform very well and at very reasonable cost. One-fourth the cost of an optical probe or less in some cases. And so you really need to consider them even if they're only 120 MHz or whatever. You should really consider that and take a look at what they are.

Give the newer ones a chance. Not all newer high-voltage differential electrical probes are good. I'll be blunt, but as we saw earlier, that Tektronix THDP0200, not a bad probe. The HVD class of probes from Teledyne LeCroy, really good.

So, give these probes a chance. And you probably want to invest in at least a single optical probe in your lab of some bandwidth so you can help establish a reference wave shape. And then you can compare your electrical probes and understand how well they're performing. Again, never connect two probes to the same test point at the same time because you're going to get these signal aberrations. And worst case, it may cause the circuit to misbehave, which you definitely don't want.

Alright. Let's look at another example of some upper-side gate drive signals. Now we're looking at a very low-voltage DC bus, and we're looking at GaN. So pretty fast rise times, about 10 ns or so.

So, we're at this place right here. This is what we call example five, and that's where we are in the agenda. This is the equipment we're using. It looks pretty much the same as the last list, same oscilloscope.

These two probes, the high-voltage optical probe, 60 MHz from, 2016 vintage. Let's throw a passive probe in the mix.

We're only doing that for the lower-side gate drive capture to use as a reference signal.

I don't recommend, Teledyne LeCroy does not recommend, your use of low-voltage passive probes in power electronic circuits. There are a lot of ways things can go wrong for you.

In this case, it's 20 V, so lower risk, for sure. But you should follow your lab rules and common sense guidelines for what you're doing with low-voltage probes in high-voltage circuits. A lot can go wrong. The DUT is a customer-supplied cascaded H-bridge, which is a GaN-loaded device with about a 10 ns rise time. Again, a 20 V DC bus. These are the list of probes. It looks pretty much the same as the previous list, except this one's different here, and let's get to the compare.

Again, now we're looking at a cascaded H-bridge, it's MOSFETs, it's GaN. We're looking at the upper-side gate drive. Here are the differences.

We've got one probe connected on each leg of this cascade H-bridge. So I got the old ADP305 on one leg, the HVFO103 on another leg, and the HVD3106 on the third leg. And then we have the passive probe sitting down here on the low side.

So and it's only 20 V, so you don't have this wide voltage range switching, but it's 20 V that's switching very fast.

So this is our list of comparisons.

First, we're going to use this passive probe to establish kind of like a reference signal shape, and we're going to look at the ADP at the same time looking at the upper-side gate drive and understand how well that's doing. And then we're going to look at some of the other probes, all connected at the same time, again, looking at different legs. And we're going to look at no DC bus voltage applied with no circuit load, then DC voltage applied, but no load, zero amps. And then the fourth comparison is DC voltage applied and a load.

Alright? So we're going to look at those three different examples, and they're all kind of labeled in the slide. I think you'll be able to follow along. So this magenta trace is that passive probe on the low side, and it looks really nice.

It's well defined, well behaved, got a nice wave shape here. And when we look at the upper side, for the ADP305, not surprisingly, we're seeing the same problems we saw before. So we see a lot of overshoot here, which we don't like.

We see kind of an unstable one level, again, due to CMRR.

Issues probably are this kind of not a zero level. It's bouncing around a bunch again, probably due to the CMRR more than anything else in this case, just because the switching speed is so fast. I'm guessing that. It's not so much a frequency response flatness issue from DC to a MHz or anything. It's really CMRR, poor CMRR performance-related issues there. So this was all predictable based on what we know about that blue probe, the ADP305. Now we look at our three newer probes compared to the passive probe reference two newer probes, the optical probe in yellow, the differential probe, high-voltage differential probe in green compared to the passive probe low-side reference.

And look, we say, okay.

The optical probe has less bandwidth, 60 MHz versus 120, and guess what? It's measuring a rise time about a nanosecond slower than the 120 MHz electrical probe. In this case, again, no load, no voltage applied. Right? So the easiest test case. And the green probe, the high-voltage differential probe is very favorable compared to the optical probe.

Now things change a little bit when we put a voltage on, and here's something I don't know why, but with no voltage, I'm sorry, with voltage but no load, what we're seeing is something pretty different. Now I took off my passive probe because now I can't have that in the circuit. I had it on the high side. I think so I took it off here.

And I'm picking up a lot of interference here. And you can just see with the electrical probe, the green probe, it's picking up some interference. And so there's something very fast happening in the circuit. I don't know what it is.

But it's probably something in the circuit that switches super fast during no load, and it's causing that external interference. And it's getting picked up on the electrical probe where you see the optical probe really suppresses it pretty well.

And so that's the big advantage of the optical probe. They just don't have these direct electrical couplings and long leads, and they can do a lot better in some of these situations.

Now you see it does get better when we apply a load. This emitting source must slow down somehow, and it's not as challenging then for the electrical probe to suppress that outside interference, and it gets better. So this is why having an optical probe, I think, is really a pretty good idea just to kind of understand, like, here's what I would expect to see, and get a good compare. And, again, newer probes of the same type, meaning high-voltage differential, newer versus older, really night and day in terms of performance.

I'd say choosing the right probe topology for the measurement is less critical when the bus voltage is very low, but it might become more critical if your rise time starts to get super fast at the same time. But even then, you saw the optical probe was a little bit better in some cases. That's because of their better CMRR and very low loading at the tips.

But in this particular case, this isn't true. Today, they have optical probes. We have optical probes up to 1 GHz, but this one only had 60 MHz of bandwidth, which is pretty marginal for that kind of rise time signal.

The disadvantage of optical probes is the really high-bandwidth ones cost a lot more than the high-voltage differential electrical probes. A factor of two or 3x for equivalent or more bandwidth. So it's a lot more money.

But again, they serve a place in the market. Again, this new HVD series high-voltage diff probe really performs pretty well. Even though it's rated for 1000 V plus of swing, if you run it in that 50x attenuation, it's in a range of, like, 28 V. So it's picking up this 20 V switching signal really pretty good. So they can surprise you, if you're using them just within the 50x attenuation range, they can do really, really well. And quite frankly, even this probe in the 500x range probably would've looked pretty good, from what we're seeing because the noise performance is really pretty low.

Okay. So the last example I think I've got here for myself is the cascaded H-bridge motor drive. This is a 400 V DC bus. And what I'm doing here is I'm comparing an instrument with isolated inputs versus an oscilloscope with high-voltage isolated probes.

Now when I was actively involved in product management, product marketing, a few years back, people were always asking me "give me an oscilloscope with isolated inputs". And they assumed it was just magically better, and they wouldn't have to worry about using special expensive probes. They could use passive probes and everything. And they really didn't understand some of the real limitations of those products in terms of bandwidth and shielding and other things.

So this is going to bring that alive for you. So this is where we are in the presentation and then Jon's going to pick up with example seven.

And again, the equipment list looks pretty much the same as what we had before, except now we're comparing against a Yokogawa DL850 ScopeCorder, which is lower bandwidth, still with 12 bits, and they have high-voltage isolated inputs, whereas we have high-voltage isolated probes with low-voltage inputs. So the difference here with the oscilloscope, the isolation is being done in the probe, and here, it's being done in the instrument, and the probe is not isolated, not even shielded, really. It's kind of like a passive probe.

And it's a customer-supplied device, silicon, not super fast, and a 400 V DC bus.

Alright. So here is just to compare the instrument topologies just to kind of bring this home to you in a simplified block diagram form. The instrument with high-voltage built-in isolation is here on the left, and the oscilloscope that's achieving isolation through high-voltage probes is here on the right. Now the first thing you notice, and look at this color code up here, the color code is green is usually better, red is usually not good, and this orange is, could be good, could be not good.

And so I've kind of indicated the various components or advantages or disadvantages of these topologies with this colored text. And so, generally, instruments with high-voltage isolation built in have lower bandwidth, and their inputs are specialized. So they don't have the adaptability that you have on the scope. They're for specific things only.

But it's built in. They have high-voltage isolation built into this front-end amplifier somehow.

The probes or the test leads, whatever you want to call them, are very inexpensive, but they have very high capacitance and long unshielded, not coaxial cables back to your DUT. So they're antennas. I mean, to be quite blunt, they're antennas.

And antennas are generally more problematic the higher frequency you go and less problematic if you're dealing with low frequency. So think back to 30 years ago before the days of drives and things like that, you had 50, 60 Hz utility sine waves kind of plugging into a power analyzer or something. Hey. In those days, you didn't have to worry about high capacitance, long unshielded leads. You weren't going to get a lot of interference from that stuff. But in the days of power devices, power switching drives, you've got to worry about that. So here, it looks a little bit different.

The disadvantages, you have more expensive probes. They cost more money. There's just no doubt about it. But they're shielded back to the equipment, which is good.

You get higher bandwidth and general-purpose capability. It costs more in general than it does over here.

Any amplifier is a low-voltage front end, so you have to have this isolation in the probe because it's not built to take more than, like, 5 V input, best case.

So the compare we're looking at is we're looking at one leg with the Yokogawa DL850 ScopeCorder, and we're looking at another leg with the Teledyne LeCroy scope. And I've kind of got the color coding of the traces you're going to see on the next screen here. So we've got passive probes on the low side. We're looking at some lower, not crazy high-voltage probe looking at the line output, and then we're looking at the upper-side gate drive with the probe.

And this is what we got. So this is the Yokogawa ScopeCorder on the left. This is the Teledyne LeCroy oscilloscope on the right. And what should instantly be noticed by you is this is the upper-side gate drive in this top view in both cases. Here, we're looking at that high-voltage optical probe, and I've got what I call a perfect signal.

Here, I'm just picking up ring, ringing, ringing, ringing, ringing. I mean, this is all due to antenna pickup and lack of rejection of the emitted signal in the environment. And so it's not good. And then if we look at the lower-side gate drive, again, we're picking up a lot of that ringing just radiated from the upper-side switching, so this isn't real. Oops.

You go back. This isn't real. It's being just picked up by this interference right here. And so none of these two things are doing a great job of suppressing the ring. It's working out pretty well here on the scope.

And the phase output voltage is a pretty low-frequency signal, all things being said.

That doesn't look too bad over here. You can see a little bit more ripple on the top here that you don't see over here.

Yeah. I could certainly live with that. That was all I saw, but you can see this isn't really a great instrument or a great setup for measuring very high-frequency signals, like gate drives, but it might work fine for things like phase outputs, where your frequency is a lot lower, so to speak, and you don't have some of the challenges of floating signals. So isolated inputs and instruments with isolated inputs are not necessarily better.

That just needs to be understood. The attraction of an isolated oscilloscope input channel, people wanting that, it's well understood. It's safe for the operator, for the DUT, for the instrument, but they don't do a good job of measuring high-speed signals. And the measurements are often compromised. The bandwidth is generally very low. You've got unshielded cables that are antennas. They pick up a lot of transients, and they also have a lot of capacitance in them as well, but they can do very well for low-bandwidth sensor or phase output signals for sure.

Okay. So I think we're running pretty much right on time, and this is where I'm going to hand it off to Jon. And, Jon, it looks to my notes like we're pretty much right on time here, so take it away.

Awesome. Thank you, Ken. So this example here is the last example that we'll show in this webinar in part two. We're in example seven, we're going to talk about device switching measurements.

Looks like having issues controlling the slide deck. Oh, there we go.

Okay. I'll take care of it for you.

Yeah. It just looks like there's a bit of a delay. I'll try one more time here. But, essentially, in this device switching example, we're going to look at a bunch of different probes measuring the same device switching. And, essentially, what we want to see is how these probes are impacting our measurement.

Hey. Why don't I take over the slide advancement for you? There you go.

Ken. Right. Yeah. So as you could see, it's the last example in this webinar here. Next slide, Ken.

And the equipment that we're using for this example is a different oscilloscope that we've used in previous examples here. So we have the Teledyne LeCroy WaveRunner 8038HD, which is an eight-channel, 350 MHz oscilloscope with 12 bits of resolution.

You could see here the probes, four probes that we're going to be looking at. I'll talk a little bit about them in just a moment here. But, briefly, I just want to mention the DUT that we'll be evaluating is an Infineon CoolGaN device, which is a GaN MOSFET that has about a 6 ns rise time. We're going to evaluate this device as it is operating at a 500 V DC bus.

Next.

So these are the four probes that we're going to make this measurement on. The top two probes here should be familiar probes from Ken's example a few slides back, with the addition of two probes here in the bottom side of the grids. So we have another HVD probe, a high-voltage differential probe, which is essentially the same model that we have in the top right, just has about three times the bandwidth. And then we also have an isolated, optically isolated probe, which is a different architecture than the high-voltage differential probes. It has right around the same bandwidth as the high-voltage differential probe in the bottom left.

So next slide.

And a little background on the measurement that we're going to be making here. We are measuring VDS on a 500 V GaN device.

VDS is the drain-to-source voltage, and we're going to measure this on the high-side device in the half bridge.

We're going to use those probes. Next slide.

So specifically in this webinar in part two today, we're just going to look at high-voltage differential probes, electrically isolated ones. So we're not going to look at the optically isolated one in today's webinar. We're going to save that for part three in June, so make sure you tune in next month for that compare.

But these are the three probes we're going to be looking at and, you could see, when we use each of these to measure VDS on the 500 V GaN MOSFET, this is what we get. So each of these signals looks a little bit different than one another. And just to emphasize the difference between each of these probes' impact on the measurement, you could see here on this slide, I have overlaid all of these signals on top of each other. So you could really see the difference of each of these probes.

I've also enabled at the bottom here a bunch of parameters to show the difference in rise times and overshoot that you see in each of these probes.

And so what I want to do for this compare now is look at some common questions that you would have about the difference in the signal quality that we see comparing these three different probes.

So the first question is, why does this lowest bandwidth probe have so much overshoot compared to the newer high-voltage differential probes?

Another question is, why is the amplitude different for all three of the probes, based on the top amplitude?

Third, with a GaN device, I would expect to see a much different signal shape between a 100 MHz probe and something that has four times its bandwidth, but I don't. Why does the signal shape look pretty much the same on these two probes that have very different bandwidths?

Number four, why is the rise time of the lowest bandwidth probe the fastest rise time out of the three probes? Why does that make sense?

And five, we have the signal shape of the lowest bandwidth probe, the ADP.

Why is the signal shape after the step so much different than the other two probes? It looks to be much worse even though it has a pretty comparable bandwidth to the 120 MHz probe.

And the last question, why do all these signals have a timing difference between them? I would expect all of them to line up. Right? And so, we're going to attempt to answer all these questions here. And so just to visualize what these questions look like when we're visualizing the signals. So number one here at the top, you could see this major overshoot that we see in the lowest bandwidth probe.

And why might that be?

So Ken mentioned this in this webinar. We talked a little bit about it in part one as well. But this probe specifically to achieve 100 MHz of bandwidth, it has a highly peaked frequency response, which means that it amplifies high-frequency content incorrectly in order to achieve a higher bandwidth.

And so the benefit of this is that it's able to achieve this higher bandwidth, but this comes at the cost of incorrectly boosting high-frequency content, which manifests itself in the signal as overshoot. And so comparatively, the two newer high-voltage differential probes don't have this highly peaked frequency response in order to achieve its bandwidth, and so we don't see excessive overshoot.

Okay. The next question is about the amplitude of the different signals. So you could see at the bottom of the display, each of them have slightly different amplitudes. But really, we would consider, if you go to the next slide, Ken, we would consider all of these amplitudes to be the same because they're all within a 1% range of accuracy.

For example, the two newer high-voltage differential probes are both rated at 1% accuracy.

And seeing a 508 V and 502 V reading and measuring a 500 V signal is well within the accuracy range of these probes. Also, for the older probe that has twice as bad of accuracy compared to the two newer ones, this value of 497 is also within its accuracy rating, whether it be compared to the two newer probes or even to the 500 V measurement. So this is what we would expect. It's within spec.

The third question is, why do both of these probes that have very different bandwidths have very similar signal shape, given I'm looking at a GaN device with very fast rise times? Why does a probe that has a 120 MHz bandwidth have a signal shape that looks very similar to a probe that has a 400 MHz bandwidth?

So the way I want to answer this question is, a lot of times people making measurements on GaN assume that just because they're measuring with a wide bandgap device, they'll require a measurement system that is 1 GHz.

However, oftentimes what we see is that GaN is implemented with rise times in the single-digit nanoseconds and even 10 ns range. So even though this is faster than traditional silicon devices, this rise time range is not as fast as what the device is rated for. We see that it's implemented in designs slower than what it's rated for to better control dead time and EMI. And so this device that we're looking at has a 6 ns rise time, which translates to about a 60 MHz signal bandwidth.

And typically, what we like to tell users is that the measurement system should have three times the bandwidth of the signal bandwidth. And so in this case, we would recommend that the measurement system have 180 MHz of bandwidth.

And so, that being said, when looking at a 400 MHz probe and a 120 MHz probe, both of these probes are sufficiently higher than the signal bandwidth and so we have pretty similar rise times here.

The higher bandwidth probe, the 400 MHz probe has a little bit faster rise time, and that's because it can capture higher-frequency content because of its higher bandwidth. That's exactly what we would expect. And because it could capture higher-frequency content, we also see in that signal shape that there's a little bit of that peaking of higher-frequency content after the step.

That's what we would expect.

But, again, because both of these probes have more bandwidth than the signal bandwidth, the overshoot that we see on both of these probes is basically the same.

A 60 MHz signal bandwidth means that this overshoot that we see is probably an artifact of the device, not an artifact of either of these probes that has sufficiently higher bandwidth than the signal bandwidth.

Alright. The next question is about the rise time of these three signals. It seems to be that the lowest bandwidth probe, the 100 MHz probe, has the fastest rise time.

Why is that?

This ties back to what we were talking about with this probe's high-frequency peaking in its frequency response.

This high-frequency peaking artificially makes the rise time seem better than what it truly is, because of how the high-frequency components are amplified in its frequency response. So if you recall, the foundations of a square wave are essentially the odd harmonic sine waves summations that are in the signal. And if the magnitude of the higher-frequency harmonics are incorrectly amplified and the summation of that content is going yes. It's going to be faster, but it's going to be inaccurate because it was incorrectly boosted.

And so long story short, the rise time that we see out of this probe is an artifact of the probe and its high-frequency peaking rather than peaking in the signal itself.

So if you really want to get an accurate rise time measurement out of your probe, you want to make sure it has sufficiently high bandwidth. In this case, the HVD3220, which has 400 MHz of bandwidth, is going to be the best probe to get the job done.

Alright. Another question here is about the signal shape of the lowest bandwidth probe. You could see there's quite a lot going on after the step.

What exactly is this wacky waveform that we see after the step and why does it look so much different than the other two probes which have kind of similar bandwidth to this probe.

So really there are two things that we see in this probe after the step.

We're seeing high-frequency content right after the step and we see low-frequency behavior about 50, 60 ns after the step. And so this high-frequency response that we see or this high-frequency ringing that we see right after the step is due to the low CMRR of the probe. So for example, this lowest bandwidth probe, the ADP305, has 30 dB common-mode rejection at 1 MHz.

And comparatively, the two newer high-voltage differential probes have 60 dB CMRR at 1 MHz, which translates to a thousand times better than the older probe.

So that is why we're seeing that high-frequency ringing.

When it comes to the low-frequency behavior, we see about 60 ns after the step. This could be due to a variety of reasons. One of which could be due to variation in the frequency response in the probe. And so, what you'll learn is that very small variations in the frequency response can make a pretty big difference in the signal amplitude. So for example, a 0.5 dB frequency response variation could translate to a 5% amplitude difference. And so when you're looking at a 500 V signal, this 0.5 dB of frequency response variation will turn out to be 25 V. So you can really see how these small variations in frequency response could have a huge impact on your measurement.

The low-frequency behavior could also be due to probe loading, but it's hard to tell exactly where in the probe that's coming from. Typically, what we like to tell people is, you get what you pay for with these high-voltage differential probes and their ability to probe load the circuit.

Alright. And the last question here is about the signal skew between these three signals looks to be varying by 2 ns from each other. And the question may be, why do I see this timing difference between the three signals? Why are they not lining up as I would expect them to?

Well, the short answer for this is that these three signals were acquired in different acquisitions and so I wouldn't expect them to arrive at the same time as they're being triggered when we line them up to each other. However, even if they were captured at the same time, I wouldn't expect them to line up perfectly and that's because of the propagation delay in each probe. Propagation delay is the time it takes to transmit the signal through the probe's cable. And for high-voltage probes, this is typically in the teens of nanoseconds, 15 to 20 ns.

But you're going to see a little bit smaller variation in probe propagation delay when it comes to the same model class. So you could tell that we're looking at two newer high-voltage differential probes, same model class and there's only a skew there of single-digit ns.

But still there there's some variation there, and this is true for optical probes as well. We'll see in the example in the part three webinar in June. But essentially, you can't get rid of propagation delay off the bat. You have to calibrate it with a simple deskew method.

We're not going to talk about deskewing in this webinar. We'll talk about it in part three in June, especially when we're talking about making switching loss calculations. So make sure to tune in for that come June.

But what I'll do now is I will pass it over to Ken to conclude today's webinar.

Okay. I'll do a quick summary here of everything we covered today, all the different examples. And I'd say, first thing, never connect two probes to the same test point in the circuit at the same time. And we showed you plenty of examples where that just doesn't lead to good results.

And knowing the signature, so to speak, of your probe can really help you understand if the anomaly that you're seeing on the signal is in your circuit or if it was added by the probe. So it's a good thing to get to know your probe a little better and how it performs. An optical probe may give you a nice reference compare.

So that's a good thing that you may not need the optical probe. The electrical probe may work fine, but it's nice to use it as the reference compare for making all your electrical measurements.

Definitely, newer high-voltage isolated electrical differential probes perform better than old ones, but new doesn't also mean good. There are plenty of new bad probes. They generally cost a lot less money. And so be careful. You generally get what you pay for there.

And be careful about instruments with isolated inputs. It sounds great, but generally not a good solution for high-frequency signals. But if you're doing, like, power analyzer type measurements and stuff, it's great. I mean, they work great for that kind of thing.

The one upside is they're safe to use. So if you have not challenging low-bandwidth signals, then instruments with isolated inputs may be the way to go. And also, just lastly, you may not need as much bandwidth as you think. GaN doesn't always mean 1 GHz.

So choose wisely. You might be able to save a little bit of money and still get a good quality measurement.

Okay. So, a couple of other things. If you go to teledynelecroy.com/events, it has all of our listings of our webinars. You filter and sort and whatnot and find good resources there.

These are some on-demand webinars that are related to power topics you might find of interest. I won't read them all, but there's some good stuff there in that events page.

You can visit this page. There's a lot of good resources here on three-phase power and motor analysis, which is pretty cool.

This page has a lot of good resources for wide bandgap analysis, some examples, and information, everything else about what we have to help you out.

This power probes page has a kind of a good summary and a great selection guide that you kind of saw the roots of this in part one. So you can kind of do that online, help you pick the right probe.

You can load our software from our scope onto your PC, and it's free, just with simple registration.

And if you have a scope, then you can copy over LabNotebooks and reproduce everything you do in your lab right at your office desk, for instance. So that's pretty cool too.

And we hope you join us for the next webinar. We're going to talk a little bit more about specialized high-voltage optical probes and 60 V common-mode probes and some kind of, like, fringe, more fringe applications.

So with that, I'll turn it over to Hilary, and we'll see if we have any questions.

Thanks, Ken. Thanks, Jon. Thanks, everyone, for attending. If you do have questions, please go ahead and submit them into the question section of the control panel.

Question one, if I'm going to read it two different ways. Alternative to compare two signals even if we can't connect the probes at the same time, from slide 118, meaning how could they compare two signals if the CMRR is affected when two points are measured at the same time?

So I'm frantically trying to get back to that slide, and I'd say the right thing to do is kind of what we did here.

I'm going to go back one more.

Let me see. We have a memory trace here somewhere, and that's what we're doing here. So for instance, this orange trace. I basically took one shot, and I stored that in the memory.

We call it a memory. If it's a Tek scope, they call it a reference trace. I'm not sure what it's called on the Keysight, but everybody has something like that. So you can kind of store it in memory and then recall and display it and then disconnect that probe, put your other probe on, take a different shot.

And the only problem with that is you're going to see something that's like this here where it's not the exact same acquisition, but you can kind of assume it's mostly the same signal. It just might be in a slightly different width, and you can draw some pretty good conclusions. So that's a great method to use to compare two probes without connecting them in the circuit simultaneously or put them on two different legs, other like a cascaded H-bridge or a full bridge. That's another way to do it.

Okay. The next question, what is the easiest way to tell interference from poor probe common-mode rejection apart from other probe or environmental behaviors?

It's hard to unwind all these things. I've just gotten better at it through looking at a lot of different examples. I'd say that the CMRR performance, it really pays to look at the curve of CMRR versus frequency to kind of know what you're getting, and then you have to understand something about the speed of the rise time of your signal, and equate that to a frequency that might be radiated, switching frequency. And just make sure the probe is performing reasonably well in that area.

In part one we gave an example where you could measure the real-world impact of CMRR by connecting leads together and doing a short and then connecting the signal and the circuit, comparing it to an optical probe. So there are some examples we had in part one that helped you kind of isolate what the CMRR impact might be and correlate that back to a specification for the probe. But it's hard, quite frankly, it's hard to unwind some of these things.

Here in this one, just generally, the lower-frequency events like you see here are generally loading issues, and then things like this are generally CMRR issues. And that's a decent rule of thumb, but it's not perfect.

It's hard to unwind everything from each other.

Okay. Next question.

Why do manufacturers peak their high-voltage differential probe so much even though it's known to cause issues?

My question is, why?

Because people want more bandwidth, and they don't know any better. No one asked questions. And so I've literally had discussions with customers where they've looked at our newer probe where we chose not to make that choice because we felt people wanted a very faithful, nice square signal edge.

And customers just simply wouldn't believe that the old thing they had, that they've been using all these years, that showed all this overshoot on their signal, wasn't the real right signal and that ours was. I had many layers of arguments with many different customers where I really had to kind of make this case and say, no. You've got to look at your manual. Look at what the frequency response is or put in a step and measure it at low voltage and look at what the probe's doing and compare that to what you might see with a passive probe.

And they come around, but it's just what everybody did back in the day. And the reason they did it was because it was really hard to achieve 100 MHz for those probes, and so you had to boost them. As you got close to 100 MHz, you had to peak them a lot or else they would have a 60, 70 MHz probe, and nobody wanted to buy that. They wanted to buy a 100 MHz probe.

Now I think designs have gotten better.

We can achieve 120 MHz or 400 MHz without the excessive peaking. If you look at that Tek probe that I showed earlier in one of the compares, what's interesting about that is that its rated at 160 MHz.

But in some cases, the rise time you're measuring with that is not as fast (as our 120 MHz probe) and which is really interesting. In this particular signal, it's a little bit faster, but for having so much more bandwidth, 160 or 200 MHz depending on the range, it's not really getting the benefit of that. So I think, in that case, they wanted a probe probably that had more bandwidth than anybody else's probe. So they want to rate it at 200 MHz or 160 MHz.

So again, they boosted it near that 160, 200 MHz point. And they get the overshoot, but they get more bandwidth. So if people are only concerned about bandwidth, they might think that's a good thing without really knowing the downside of what they're getting with that extra bandwidth. So it's yeah.

I mean, we don't do it anymore. If others do, you really kind of have to ask them why, but probably it's just a trade-off between bandwidth and another performance characteristic that people don't understand they're they're being traded off for, so to speak. I hope ? that was a long answer. I hope that made sense.

And then the final question, is there an app note, do we have an app note that explains bandwidth equals 0.35 over rise time?

Oh, I'm sure we do somewhere. Jon did a coffee break on this, either in 2023 or 2024, that we could probably send you a link to. But it described that whole, that kind of rule of thumb. And it's not a hard and fast rule of thumb. There's a lot of caveats about it, but it holds pretty well.

And so there are a couple of webinars. Maybe Jon wants to dig up those links and get them to Hilary so, Hilary, you can forward them to the person asking the question. I think they'd find those links helpful.

Yeah. So I'll definitely forward that. And, in that webinar as well, we talk about those caveats as to when we deviate from the 0.35 coefficient.

Great. Thanks, Jon. Okay. So I think that's it. If any of the attendees decide they have any other questions, you can go ahead and email them to me.

I believe you've all received multiple emails from me, hilary.lustig@teledyne.com, and I'll make sure that we answer them. I want to thank you all for joining. I hope you join us for part three on June 25th. And in the meantime, if you have any measurement needs or questions that pop up, please don't hesitate. Have a great day everyone, and thanks again, Jon and Ken.

We'll see you next time.

Thank you all.