Power Integrity | How to Improve Power Integrity in Your Design

Judy Warner:
Hi Heidi, hi Steve, thanks so much for joining today on the podcast. It's great to see you both.

Steve:
Nice to be here.

Heidi L. Barnes:
Great to be here with you.

Judy Warner:
Well, why don't you take just a quick moment and tell our viewers and listeners about who you are in case they don't know you.

Steve:
Ladies first.

Heidi L. Barnes:
Thank you, Steve. And it's a pleasure to be here with Steve. He's the one that got me involved with Power Integrity, and now I am actually the product owner for Power Integrity with Keysight's PathWave EDA simulation tools. And yeah, it's been a great privilege to be working with Steve and to be doing this podcast today. And Steve, your turn.

Steve:
how far we've come since 2015 when we first started working together, and it's been a blast, so I'm having fun. I'm Steve Sandler, Managing Director at Pico Test, and we specialize in, mostly in measurements that have to do with power electronics, power integrity, and everything between those.

Judy Warner:
Okay, so that was the easy part. Now I'm gonna ask

Heidi L. Barnes:
I'm gonna

Judy Warner:
you

Heidi L. Barnes:
go.

Judy Warner:
a question. I'm sure you've heard it many times, but I don't know that our audience has heard either of you say that. So how would you define power integrity?

Steve:
That's an easy one because it's in almost all of my keynotes. Power integrity is the science of getting power to devices that it likes. And that means different things for different kinds of devices, and some devices are more sensitive than others. But our job in power integrity is to make sure that we've optimized the power for the devices that we're powering so that they can work at their peak performance.

Judy Warner:
I like that,

Heidi L. Barnes:
And I'd

Judy Warner:
Heidi.

Heidi L. Barnes:
kind of like to add to that. Steve, very simple. I work with a lot of high-speed digital applications. We talk about power delivery. Power integrity is the delivery of power to high-speed digital loads. The one that I've more recently come up with is power delivery is not DC. It's AC. So DC resistance IR drop is important. But power integrity is really about the impedance involved in delivering power to a load, because the load is not DC, it's AC, it's a dynamic, highly dynamic current. And for that matter, that's why EM simulation is so important. We need to use an EM simulator to get that impedance right and really look at how that power is being delivered to a dynamic load.

Steve:
Good point.

Judy Warner:
I think that's a great explanation. So, Heidi, you're at Keysight. Steve, you and I have talked many times and I know

Heidi L. Barnes:
Thanks for watching!

Judy Warner:
that you happen to be an ADS expert that you just basically jumped in many years ago. So what does it take to simulate power integrity? You touched a little bit on that, Heidi, but what else would you say that it really takes to get good accurate measurements?

Steve:
you know, it's so much easier than it used to be. You're right, I kind of jumped in 2009, actually 2008, I think. Oh my God, my first ADS simulator. And even since then, it's gotten so much easier. The interfaces are much more intuitive. So it used to be very complicated to, you know, get all the settings for meshing and momentum settings and all that stuff. And now most of that is automated. So, you know, there's a little bit. of a learning curve to learn the interface of the tool and where the knobs are. Other than that, it's really very easy to use. You import the printed circuit board, you define where the power supplies are, you define where the loads are, and the rest is pretty much automatic. In fact, even now, you know, ADS can help you pick the right capacitors. It can help you tell where to put the capacitors. It's really, you

Judy Warner:
Mm-hmm.

Steve:
know, almost intuitive, I think. And it keeps getting better. So every year it seems like it has both more capability and also becomes easier to use. That's a great combination. And

Judy Warner:
Yeah,

Steve:
yeah,

Judy Warner:
it

Steve:
I am

Judy Warner:
is.

Steve:
a certified expert now and I'm

Heidi L. Barnes:
Yes.

Steve:
very proud of that.

Heidi L. Barnes:
And Keysight's very proud to have you as our certified expert for power integrity.

Steve:
Thanks.

Judy Warner:
Well, early on, I think it'd be an interesting application to talk about Steve. I remember, I'll never forget the first time I interviewed you and didn't really know much about your background. And the first thing that comes out of your mouth is you casually start talking about modeling the power on the International Space Station.

Heidi L. Barnes:
I'm going to go ahead and turn it off.

Judy Warner:
So for our listeners and our audience, why don't you touch on that? How? sort of what took you down this career path and your experience on the space station.

Steve:
You know, I think we get most places accidentally and... And, you know, get led through experiences that land us somewhere. And I happened to be working for a company that specialized in satellite electronics. I've been in satellite electronics almost since the 1970s. And they happened to be the major subcontractor for the space station. And they asked me to consult on the space station, which I was happy to do. And so I spent a lot of time in Palo Alto in those days. But one day I overheard a conversation between the subcontractor NASA and they were trying to decide how you could do the worst-case analysis of something that was as big as the space station. Nobody ever done it and it was around Christmas time. So I went home and I came up with a business plan to build the company to do the worst-case analysis of the space station and I submitted it as an unsolicited proposal and I won. You know, overnight I had a company and I had seven people working on space station. We spent six years models I think I think I delivered a little over 300,000 pages of analysis in six years.

Judy Warner:
Unbelievable.

Steve:
And it's crazy to this day I can't tell you if winning that contract was a good thing or a bad thing but I think it was probably a good thing.

Heidi L. Barnes:
I think

Judy Warner:
Well,

Heidi L. Barnes:
most of us would agree it was a good thing.

Judy Warner:
yeah. Well, it makes for good stories on podcasts.

Steve:
I just overheard people saying that they didn't know how to do this. I was thinking, well, it's not that hard.

Judy Warner:
says Steve Sandler.

Steve:
Alright.

Judy Warner:
So I heard you say one thing is that what was interesting about that is that in regards to that application you had to be right, right? And and

Steve:
Yeah.

Judy Warner:
it's it seems like you know knowing that what kind of pressure did it put on you and what did you learn and grow

Steve:
Well,

Judy Warner:
through?

Steve:
we learned a lot. Yeah, we learned a lot. We studied a lot. And I would say that wasn't really anything new. You know, I've been in the satellite business a long time. And the mentality always was that if we're wrong, really bad things happen. And people point to things like the Challenger. And, you know, we remember when things don't really go well. So the motto was be right. The very first thing we did when I started the company to do the analysis of the space station was to figure out why things are wrong. and what makes them go wrong. And it was a very interesting process. We spent about a month and a half doing that. We found a few things. One thing we found is that most errors occurred during data entry. And so like if you have a

Judy Warner:
Hmm.

Steve:
hundred engineers in a roll typing in the tolerances for a point one microfarad capacitor, they tend to type in different numbers and they don't always get it consistent. And so one of the first things we did in Space Station is to create databases that were locked and protected databases so that you couldn't type in those numbers. They had to be retrieved from a database. And so long as everybody pointed to the right part number, everybody got the same answer. That was one thing that we instituted. things too. For example, every piece of paper that left AEI Systems was reviewed by at least two people and that was true even if I wrote it. There were two people that read everything that I wrote to see if we could make sure that they were right. Every simulation that we did was tied to a measurement of something. So we never took anything for granted, even a data sheet, whatever it was. We had a measurement that proved that our nominal was right. Because what's the point of doing, you know, worst case of Monte Carlos of something that's not nominally right. We had to be nominally right. And, and so we, we think did things like that also. And that's actually become a lot more sophisticated over the years. We used to, you know, just put the simulation and the measurement on the same page and you'd look at them. sophisticated tools that put them into the same screens and rescale them and things like that so we can actually see even minute differences. But the rule always was be right. And it's really fascinating when you look at the world of satellite systems today, and you look at the SpaceX's and it's mostly just experiments, right? And everybody says, wow, look at

Heidi L. Barnes:
Thanks

Steve:
that,

Heidi L. Barnes:
for

Steve:
they

Heidi L. Barnes:
watching!

Steve:
landed vertically and it did. But if you look statistically at their performance, it's relatively poor. successful at this point, which is somewhere around one or one and a half sigma. And in satellite systems we're looking for six sigma. So, yeah, it might be pretty good by industry standards, but certainly not by space standards.

Heidi L. Barnes:
I was going to add out of I think your aerospace work and working in that industry, you came up, as you mentioned, you do simulation and measurement. And I think that's one of the great things about working with you and also, you know, learning to do those power integrity measurements, learning to do the simulations that are needed. But one of the things that came out of that work in the aerospace industry was the non-invasive stability margin measurement and methodology. which allows you to determine or check, validate the stability of a power supply

Steve:
So, there's the question

Heidi L. Barnes:
for

Steve:
of how

Heidi L. Barnes:
an aerospace

Steve:
to do it. And

Heidi L. Barnes:
application

Steve:
I think that's the key for it. If

Heidi L. Barnes:
and

Steve:
you

Heidi L. Barnes:
do

Steve:
want

Heidi L. Barnes:
it

Steve:
to

Heidi L. Barnes:
non-invasively.

Steve:
do it, not the easy

Heidi L. Barnes:
You

Steve:
way,

Heidi L. Barnes:
don't

Steve:
but

Heidi L. Barnes:
have

Steve:
you don't

Heidi L. Barnes:
to open

Steve:
have to.

Heidi L. Barnes:
up

Steve:
But it's

Heidi L. Barnes:
the

Steve:
a challenge,

Heidi L. Barnes:
control

Steve:
and

Heidi L. Barnes:
loop.

Steve:
I think

Heidi L. Barnes:
And

Steve:
you

Heidi L. Barnes:
I

Steve:
learn

Heidi L. Barnes:
think you

Steve:
very quickly.

Heidi L. Barnes:
learned very early on that in the practical world, you can't always access that feedback control loop of a power supply. And so NISM is a wonderful thing and very excited. Steve just released with Keysight. video on non-invasive stability, margins, and measurement and simulation.

Steve:
Yeah, it's really interesting how that came about. You know, we wanted to get these measurements, but they weren't really available on a circuit board. So, you know, I happened to be up in Palo Alto and NASA was there and I said, you guys happen to have an X-Acto knife. And they said, why? And I said, like to cut these boards up a little bit so I can make access for the control loops. And of course everybody flipped out about as much as you would expect they would, right? I know they weren't gonna give me an X-Acto knife. So, but that's actually how that came about is that I realized we needed some sort of a way to make these measurements without having to cut the traces on the circuit board. And it's been pretty good, you know, it's been around now for almost a dozen years, I think it is, and it's adopted pretty widely across the aerospace industry and even at NASA. And so we're pretty proud of that. But I think that, you know, one of the things that progresses, if you look at how we do our measurements and how we do our simulations today versus how we did them back in 1990, when I was doing Space Station, things have evolved so much. And I go back and I read those reports now and, you know, all the assumptions were very carefully written, you know, to make sure that we said exactly what we did and exactly what we didn't do. And almost every one of them has this note that says it doesn't include any printed circuit board effects. And today I would say, why bother, right?

Heidi L. Barnes:
I'm

Steve:
I

Heidi L. Barnes:
going

Steve:
mean,

Heidi L. Barnes:
to

Steve:
it's

Heidi L. Barnes:
go to

Steve:
all

Heidi L. Barnes:
bed.

Steve:
printed circuit board effects. And I still see that occasionally in reports from AEI that, you know, to worry about the

Judy Warner:
Yeah.

Steve:
printed circuit board and it's amazing how low in frequency printed circuit boards get in the way.

Judy Warner:
Yeah, I remember you saying that to me one time. And for me, coming from the board industry, it felt so lowly, you know, compared to some of the high science things you do. And I'm like, hey, boards,

Heidi L. Barnes:
Thanks for watching!

Judy Warner:
I'm relevant again. This is awesome. Cause

Steve:
Yeah.

Judy Warner:
I

Heidi L. Barnes:
Hahaha

Judy Warner:
know something about circuit boards, but when you get up to those speeds and frequencies, like you said, everything matters. So that's a bit of... sort of where you've been in the past. That's power integrity past. Where would you say we are today? What are some of the challenges? What are some of the drivers of the technology?

Steve:
So there are so many and it's really an exciting time to be an engineer. And on the one hand, I wish I was a young engineer now with all these great tools that I have growing up. And then the other hand, I'm really glad I'm not a young engineer right now, the pressures are incredible. There's a lot happening, speeds keep getting faster. And, you know, 200 gigahertz stuff going on in millimeter wave. And, you know, I had a conversation with Al Navs just the other day. And it's like, you know, he starts at like 50 gigahertz. And so there's, you know, that craziness going on. Transceiver business is really, really booming, trying to keep Internet speeds up. And there's a lot of high speed work going on there. The focus on power has really been towards bigger processors. And you see it. AI is consuming tremendous. this horsepower

Judy Warner:
Mm.

Steve:
social media. In fact, I had a conversation just two weeks ago with an investor and he casually mentioned that the number one consumer for H100 Boardsman videos, TikTok. And that blew my mind.

Judy Warner:
Did you see that one coming you guys? I didn't.

Heidi L. Barnes:
Hahaha

Steve:
I did not, but all of these big companies, they're all creating these tremendous horsepower data centers, AI centers, HPC, and the requirements that that... puts out are incredible. You talk about printed circuit boards, and now we have people doing chips at 2,000 amps. I have one customer that's doing a chip at 5,000 amps. And to get that across a printed circuit board isn't so easy. And I think it was just yesterday we got an email telling us that our DesignCon paper for this year was accepted. We're doing a paper on the design, simulation, and validation of a 2,000 amp core power realm. And

Judy Warner:
Oh

Steve:
you think

Judy Warner:
my

Steve:
about

Judy Warner:
gosh.

Steve:
that, and it almost seems casual on the one hand, right? because we've been getting there for a long time. And then other people say, you know, it's crazy how far we've come. But I would say back in, I think it was 1976 or 1977, I had a part-time job at a power supply company while I was in high school. And we were building one volt, 2,000 amp power supplies for IBM back in the day. And what I would say is they were a lot bigger than they are now.

Judy Warner:
Yeah.

Steve:
Maybe now they're a lot more sophisticated, but. There's always been this fringe and what's happening I think is the fringe is moving closer. You know, it's almost mainstream now to be doing 500 or 1000 amps, which is kind of crazy to think about. And when you look at the ramifications of that, you know, it's not only developing a 5000 amp power supply and being able to support the impedance that it takes for that to do its thing at 10 gigahertz or whatever it's switching at. But you also have all of these signal integrity channels that somehow you have to not ball. while you're doing that and so containing EMI in sockets and trying to get power through the board and get the planes as short as you can get I mean getting power modules to the point that they fit it's all challenging migration to 48 volt power systems it sounds really easy and there's a lot of benefits to it higher efficiency lower conduction losses but I'm doing an EDI concession on the migration to 48 volts and one thing people don't think about is that we don't actually have the equipment and tools for it Power rail probes are really great, but they don't go up high enough to support data center and AI. They don't have the dynamic range for data center and AI. So, at two-port impedance, we've mastered that. We know exactly how to do it, but we don't know how to do it at 48 volts. And so, all of a sudden, the challenges that are placed on new measurement technologies and instruments are becoming really challenging, but also fun.

Judy Warner:
So Heidi,

Heidi L. Barnes:
Thanks for watching!

Judy Warner:
how, what, how, how are you going to do this? How do you develop tools?

Heidi L. Barnes:
Well, simulation's a little bit easier, because when I hit the wrong button, it doesn't melt down and burn up in front of me.

Steve:
That's true.

Heidi L. Barnes:
So actually, simulation's becoming very important, because you really do want to get it right the first time and not have dramatic failures or very expensive failures late in the design. So one of the things I think Steve brought up earlier was how the automation that we're seeing in the EDA industry is really helping. We have Power Integrity Pro, PI Pro, that's targeted at optimizing the EM simulator for simulating the power delivery network of a printed circuit board.

Steve:
Thank you.

Heidi L. Barnes:
And as Steve mentioned, it makes it really easy to identify the source of power, the VRM, or your connector input, identify your sinks. the switching DC to DC converters, and you're up and running with, because it recognizes

Steve:
Thank you.

Heidi L. Barnes:
the nets, it can find components, makes it easy to add models. So you can be up and running in a matter of minutes and figure out what's the impedance of my power delivery network. And then the thing I think that I learned a lot from Steve was you don't just stop there with the EM model. You have to look at the end-to-end PI ecosystem.

Steve:
Yeah, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so, so

Heidi L. Barnes:
So a lot of our collaboration. and most recently with Ben Dannen of Northrop Grumman, and then also a lot of our work with Jack Carroll of Xilinx

Steve:
I'm sorry.

Heidi L. Barnes:
has been looking at how do we model that power supply with the state space average model that's been developed by Steve? How do we model the dynamic load, and how do we bring that all together into the simulation of a PI ecosystem and look at the actual transient steady state noise ripple on the power rails? The one thing I should add most recently, and I think Steve just mentioned, we have the world of the supercomputers and these massive designs requiring thousands of amps. And that's one challenge with thermal and very low impedance down to the micro ohms. But we also have the AI world or the autonomous vehicle world where you're seeing such a high density of electronics that they have to play nice. together.

Judy Warner:
Yeah.

Heidi L. Barnes:
The noise from one power rail can't contaminate the other power rail, and yet they're all connected because they're all coming from the main, you know, one source of power. They're all using sort of the same ground. It's this giant interconnected web of power and ground, power delivery network type thing. And

Steve:
and

Heidi L. Barnes:
you can't

Steve:
you

Heidi L. Barnes:
let

Steve:
can't

Heidi L. Barnes:
conducted

Steve:
let them take

Heidi L. Barnes:
EMI

Steve:
your

Heidi L. Barnes:
get

Steve:
money

Heidi L. Barnes:
from one

Steve:
from

Heidi L. Barnes:
system

Steve:
one person

Heidi L. Barnes:
to

Steve:
to

Heidi L. Barnes:
another.

Steve:
another.

Heidi L. Barnes:
And that's one of the things with the automotive industry, they have some of the strictest standards for conducted

Steve:
Yeah.

Heidi L. Barnes:
EMI and radiated EMI. And so that's one of the things we've done recently in our Power Integrity EM tool. We have an analysis where you get everything working with a DC IR drop, a simple, you know, a setup you can understand. And then the automation is you can just copy to a conducted EMI analysis where it will set up the ground reference plane, just like in a conducted EMI measurement. And it does the differential setup of the ports. And it adds an actual switching model, a simplified switching model for your DC to DC converter, which is oftentimes the source of the EMI noise. And then it runs your simulation for you. So there's a lot of automation there that really helps the power integrity engineer get up and running and doing some pre-layout design work before they actually start fabricating hardware.

Judy Warner:
Well...

Steve:
Yeah, and some of the leading advances really are in the automotive industry now. I

Judy Warner:
Mm-hmm.

Steve:
think they're paving the way in simulation. But simulation is also becoming a lot more accepted. I remember even 20 years ago people were talking about garbage in, garbage out, and models are useless. It's not really quite the same these days. You do need to get the models right. against measurements, but it's not so hard to create good accurate models today. And the automotive industry is really pushing it hard.

Judy Warner:
Well, that's always the benefit, right? Is that we go out, we get out over our skis a little bit technologically, but it's like the rising tide raises all boats, right? Then we

Steve:
Yes.

Judy Warner:
gotta fix the whole ecosystem around it, right?

Steve:
Yep.

Judy Warner:
And I guess that's a good thing. So it's...

Steve:
I think it's a good thing. Yeah. And so now I think the next limit is the semiconductor companies that actually make these power modules that really don't understand too much about power integrity of their applications. So, you know, I think we did a paper design con with Texas Instrument last year. This year, we're including monolithic power systems in our design con paper. I'm working with analog devices for the supercomputer conference in Denver in November. And so I think that the semiconductor companies are coming around and they're starting. to realize that they need to get their act together because they're I mean they're the ones that are at the very front end of the power delivery network and they don't quite understand what it means to get that right but they're at least starting to engage now and hopefully by engaging they'll learn what they need to know and they'll start helping us get what we need and as a result you know everything gets better it gets better for everybody I think that's I think that's how it happens you're just like you said you know over their skis then we play catch up because we realized we made a mess and we got to go clean up our mess. I