Building Complex Boards with Special Guest Daniel Beeker | Trace Talks EP 1

WEBVTT
00:00:06.844 --> 00:00:13.204
I'm Amit, the PCB guy. We're starting a new podcast for all the electrical engineers
00:00:13.204 --> 00:00:20.004
and PCB designers out there, really focusing on how to design and manufacture a better circuit board.
00:00:20.524 --> 00:00:25.304
And today, for the first ever podcast, we have these amazing guests who have
00:00:25.304 --> 00:00:28.004
a wealth of knowledge. So we have Steve here.
00:00:28.204 --> 00:00:35.624
He's a part of our R&D team and our manufacturing process engineering team to
00:00:35.624 --> 00:00:37.224
make fantastic circuit boards,
00:00:37.384 --> 00:00:43.804
and to make things that really you would think are not possible to make as well. I'm Steve Carney.
00:00:44.184 --> 00:00:48.364
I'm the research and development manager at Sierra Circuits.
00:00:48.624 --> 00:00:54.244
I've been with the company almost 15 years, though my primary background is
00:00:54.244 --> 00:01:01.064
the mechanical side of it, drilling, laser, lamination, lithography part of it.
00:01:01.064 --> 00:01:05.184
And then we also have Dan, who's, you know, just been in the industry forever
00:01:05.184 --> 00:01:10.344
and is, I would say, made his way into the, you know, signal integrity experts.
00:01:10.444 --> 00:01:14.404
So we're super excited to have both of you here for our very first podcast.
00:01:15.024 --> 00:01:18.844
I'm Dan Beaker, and I'm a technical director at NXP Semiconductor.
00:01:19.484 --> 00:01:25.284
I've been doing this type of work starting with microcontroller,
00:01:25.504 --> 00:01:27.104
microprocessor development systems.
00:01:27.104 --> 00:01:35.364
I started at Motorola in 1980 and worked my way from the tools group into the
00:01:35.364 --> 00:01:41.644
strategic marketing research group and then moved to the field as an applications engineer in 1987.
00:01:42.324 --> 00:01:48.444
I went from Phoenix to Detroit and have been supporting automotive customers ever since.
00:01:49.644 --> 00:01:56.324
And this will be my 43rd year at the Motorola Freescale, now NXP company.
00:01:56.324 --> 00:01:58.864
It's been a wonderful journey. Uh,
00:01:59.508 --> 00:02:04.708
I'm still excited about what I do every day, and it's because of my great luck
00:02:04.708 --> 00:02:10.508
in becoming associated with Ralph Morrison and with other wonderful experts
00:02:10.508 --> 00:02:14.528
and friends like we have here today in this talk.
00:02:14.708 --> 00:02:17.548
So I'm super excited to have you guys here with us.
00:02:18.028 --> 00:02:23.648
We are at the amazing Levi's Stadium, and it's very apropos since the Niners
00:02:23.648 --> 00:02:24.868
are going to the Super Bowl.
00:02:25.168 --> 00:02:27.528
Also, Design Con is happening across the street.
00:02:28.268 --> 00:02:31.968
Dan, why don't you talk to us a little bit about DesignCon, what you're doing
00:02:31.968 --> 00:02:34.828
there, and what your class is about. Yeah, thank you.
00:02:35.468 --> 00:02:37.608
Thanks for letting me be here. This is pretty exciting.
00:02:39.433 --> 00:02:40.953
Yesterday, I got to teach my
00:02:40.953 --> 00:02:45.033
first workshop for DesignCon. I was really excited to be able to do that.
00:02:45.193 --> 00:02:52.433
And the focus was on a field-based design philosophy as applied to power distribution systems.
00:02:52.693 --> 00:02:58.533
We had probably 40, 45 people in there. And the response was, this is mind-blowing.
00:02:58.813 --> 00:03:04.933
We never thought of things that way. And then a comment from the lady in charge
00:03:04.933 --> 00:03:10.573
of checking out the attendees in, she said, I listened to your talk and I could
00:03:10.573 --> 00:03:12.013
understand what you're talking about.
00:03:12.193 --> 00:03:17.553
So that's really important to me, to be able to convert the high level physics
00:03:17.553 --> 00:03:23.893
that people are afraid of into the real rules and the real ideas that they can
00:03:23.893 --> 00:03:26.133
embrace and use in their everyday jobs.
00:03:26.233 --> 00:03:30.893
And that was the goal that I was given as I studied with Ralph Morrison.
00:03:31.433 --> 00:03:34.573
And I think that the mission is going very well.
00:03:34.873 --> 00:03:38.353
And I am teaching a class today on high-density interconnect design,
00:03:38.973 --> 00:03:43.433
focusing on the board stack and the vertical transition vias because that seems
00:03:43.433 --> 00:03:45.733
to be the part most people are having trouble with.
00:03:45.933 --> 00:03:53.413
And then I'm also lucky enough to be part of a panel discussing the sensor proliferation
00:03:53.413 --> 00:04:00.353
as it goes into electric vehicles and the challenges that those pose to both the vehicle.
00:04:02.416 --> 00:04:07.276
Range because you're using a whole lot more over the electricity and the reliability
00:04:07.276 --> 00:04:08.756
and the software complexity.
00:04:08.996 --> 00:04:10.816
So it's a nice approach to the system.
00:04:11.356 --> 00:04:14.216
That's amazing. Yeah, let's break some of that down, Dan. Okay.
00:04:14.496 --> 00:04:21.616
So first things first, how would you say the electromagnetic field physics plays
00:04:21.616 --> 00:04:26.016
a role for a PCB designer and how should they understand that and how should
00:04:26.016 --> 00:04:28.076
it affect their work, their layout?
00:04:28.256 --> 00:04:31.156
Their design? That's a really good question. And that's
00:04:31.156 --> 00:04:34.656
one of the foundation issues that
00:04:34.656 --> 00:04:38.276
have led to the industry accepting failure
00:04:38.276 --> 00:04:42.376
of emc as part of the status quo and we're
00:04:42.376 --> 00:04:47.896
basically taught that the electricity you know which is the really the energy
00:04:47.896 --> 00:04:53.516
that we use in the circuit boards is something that travels in in the wire and
00:04:53.516 --> 00:05:01.576
that perception has drawn a number of design guidelines that have evolved over time,
00:05:01.856 --> 00:05:03.616
starting with the older technology,
00:05:03.996 --> 00:05:11.736
the HMOS, HCMOS, where making those types of rules and choices for your design
00:05:11.736 --> 00:05:15.096
philosophy weren't so bad they made it fail.
00:05:15.596 --> 00:05:23.416
But with today's technology, we have ICs that switch in 60 picoseconds is not
00:05:23.416 --> 00:05:25.536
an unusually fast switching event.
00:05:26.276 --> 00:05:30.436
And then you really have to start focusing on the reality.
00:05:30.956 --> 00:05:34.736
The energy we use is electromagnetic field energy.
00:05:35.816 --> 00:05:39.236
Electromagnetic field energy does not travel in a wire.
00:05:39.496 --> 00:05:43.816
It travels in the space between conductors, transmission lines.
00:05:43.956 --> 00:05:46.816
And that goes for everything from DC to blue light.
00:05:48.318 --> 00:05:53.198
The philosophy needs to be changed so that when they make choices in the layout,
00:05:53.418 --> 00:05:58.878
that they're building the plumbing, the waveguides for the energy to travel in.
00:05:59.058 --> 00:06:03.618
And that has to start from the very beginning, from the board stack into placement
00:06:03.618 --> 00:06:07.038
of components and into the actual routing.
00:06:07.258 --> 00:06:13.018
And with that approach, it changes the whole landscape. When you use a field-based
00:06:13.018 --> 00:06:18.898
approach for the layout, their chances of success are significantly higher.
00:06:19.338 --> 00:06:24.258
And that's what I'm trying to help people understand, especially the PCB design
00:06:24.258 --> 00:06:30.658
guys, because they're the ones that have to put the, they design the copper.
00:06:30.798 --> 00:06:34.618
And this is what I want to change the focus. You're not designing the copper.
00:06:34.798 --> 00:06:39.638
You're designing the spaces, the transmission lines. The copper is what you
00:06:39.638 --> 00:06:44.138
use to set the boundaries to direct the field energy from where it's coming
00:06:44.138 --> 00:06:46.698
from to where you want it to be.
00:06:46.898 --> 00:06:51.318
And that's done by making sure that where it has to go, based on the laws of
00:06:51.318 --> 00:06:56.458
physics, that it will go where you want it to be because you've created that pathway.
00:06:56.818 --> 00:07:01.578
Yeah, so I mean, I think that's amazing and it debunks some misconceptions about
00:07:01.578 --> 00:07:07.038
PCB layout. And I think that's, you know, that's a great message for any audience
00:07:07.038 --> 00:07:09.098
that's serious about what they're doing.
00:07:09.478 --> 00:07:13.198
So that's fantastic. So let's dive in a little bit deeper.
00:07:14.238 --> 00:07:18.778
You know, in terms of, you know, if I'm a PCB layout person and I have to, you know,
00:07:19.578 --> 00:07:24.098
pick component placements and, you know, design a stack up, like how would you,
00:07:24.298 --> 00:07:30.818
given this kind of way of thinking, how can you give some specifics about component placements?
00:07:31.738 --> 00:07:35.238
Or, you know, what you wouldn't do or stack-ups that, you know,
00:07:35.398 --> 00:07:36.638
that you wouldn't, you know,
00:07:37.678 --> 00:07:41.438
I would say, mis-designed stack-ups that could lead to issues?
00:07:41.618 --> 00:07:47.758
Well, it starts with being able to accept the very basic principles that manage
00:07:47.758 --> 00:07:50.358
or define the way fields move.
00:07:51.258 --> 00:07:58.498
I found that there are three simple rules that are based on the concept of threes.
00:07:58.838 --> 00:08:04.198
You know, to be able to contain field, you need two discrete conductors separated by a space.
00:08:04.618 --> 00:08:11.978
So that's the first triplet. to build a system, you only get three components.
00:08:12.158 --> 00:08:17.118
You get the conductor, you get the dielectric, and you get a switch.
00:08:18.218 --> 00:08:19.878
There's really nothing else in
00:08:19.878 --> 00:08:24.958
the universe. The most complicated microprocessor is a giant switch array.
00:08:25.238 --> 00:08:30.518
And when the switches turn on and off, they don't have current flowing through them in a loop.
00:08:30.678 --> 00:08:36.278
They take a space that has energy stored in it and connect that space to another
00:08:36.278 --> 00:08:42.118
space where there's no energy and the energy then flows into that space.
00:08:42.358 --> 00:08:45.838
And then the third triplet is that there's only three things you can do with
00:08:45.838 --> 00:08:46.938
electromagnetic field.
00:08:47.138 --> 00:08:52.118
You can store it, you can move it, or you can convert it to kinetic energy.
00:08:52.998 --> 00:08:56.698
This is it. That's the entire electronics industry.
00:08:56.698 --> 00:09:03.778
Our job is to be able to take those three sets of three rules and use them to
00:09:03.778 --> 00:09:10.578
manage the way the field moves because we are in the business of building products that generate,
00:09:10.858 --> 00:09:15.238
manage, and consume electromagnetic field energy.
00:09:15.238 --> 00:09:17.558
So that should be the guiding light.
00:09:17.698 --> 00:09:24.278
And once the designers take that to heart, then the decisions they make about
00:09:24.278 --> 00:09:30.178
how they place things are focused on the managing of the energy in those spaces.
00:09:30.258 --> 00:09:32.998
So that drives the board stack. You need to have...
00:09:35.202 --> 00:09:39.962
Ground, which is one of the conductors. I call that the continuous conductor,
00:09:39.962 --> 00:09:46.222
which has to be connected from the source of the energy to the load and not
00:09:46.222 --> 00:09:50.082
be broken. It has to be a continuous piece of copper.
00:09:50.422 --> 00:09:57.182
Next to that is a continuous dielectric. That is right next to the continuous conductor.
00:09:57.442 --> 00:10:02.142
So two of the three elements of moving and controlling field are defined.
00:10:02.322 --> 00:10:07.682
And what's left is what I call the switch conductor, which starts with the other
00:10:07.682 --> 00:10:13.342
half of the waveguide that's bringing the energy onto the board, so that's the power in,
00:10:13.522 --> 00:10:18.502
goes to a switch that probably is a voltage regulator, which serves to be able
00:10:18.502 --> 00:10:24.342
to direct the energy from one transmission line or one part of the space to
00:10:24.342 --> 00:10:28.002
a new space, and then that gets distributed across the circuit board.
00:10:28.122 --> 00:10:31.282
And when you're done, when all the switches are closed, you should be able to
00:10:31.282 --> 00:10:36.242
see a complete containment of the dielectric on two sides, one by the continuous
00:10:36.242 --> 00:10:40.722
conductor or what we call ground, and the other is the switch conductor with all the switches sort.
00:10:40.982 --> 00:10:46.522
So then you've got a pipe that's completely condosed. And that is what drives your board stack.
00:10:46.682 --> 00:10:52.942
So you want to have a ground plane as your focal point, and then you have signal
00:10:52.942 --> 00:10:55.562
and power layers, whatever you have on opposite sides of that.
00:10:55.702 --> 00:11:03.302
So you have two dielectrics that are already defined by having the two elements
00:11:03.302 --> 00:11:08.002
of containing and moving field, which is two discrete conductors that direct
00:11:08.002 --> 00:11:09.782
the energy in that dielectric space.
00:11:10.142 --> 00:11:15.662
And then that's what drives the board stack up. And then that drives the routing of the signals.
00:11:16.937 --> 00:11:20.737
You have to think about when you use a component, that component replaces what
00:11:20.737 --> 00:11:25.617
would have been just a piece of copper so that the field has to enter into the
00:11:25.617 --> 00:11:27.877
switch and then back into it.
00:11:27.997 --> 00:11:31.877
So aware of the dielectric, where the energy is at, where it's moving at,
00:11:31.997 --> 00:11:37.557
and controlling that whole movement of field in those three-dimensional spaces is the key.
00:11:38.317 --> 00:11:41.677
Okay, wow, that's amazing. Thank you for listening to our podcast.
00:11:41.857 --> 00:11:46.097
The podcast is for you electrical engineers and PCB designers out there to learn
00:11:46.097 --> 00:11:51.857
from. And we also have an amazing discussion forum that we recently launched called Sierra Connect.
00:11:52.297 --> 00:11:58.057
So go there right now, post your questions, and industry experts will respond to your questions.
00:11:58.057 --> 00:12:02.417
It's an amazing resource that you should take advantage of. So let me ask about
00:12:02.417 --> 00:12:06.077
some practical, like, techniques for, you know,
00:12:06.757 --> 00:12:11.357
basically using, you know, ground and shielding to, you know,
00:12:11.417 --> 00:12:17.617
help a designer, let's say, be EMC compliant first time, first time design.
00:12:17.937 --> 00:12:23.057
Can you give some thoughts around that? Well, it's the idea of not necessarily
00:12:23.057 --> 00:12:28.057
having shields, but having first discrete spaces for each signal.
00:12:28.637 --> 00:12:31.897
And that's so that there's a place for them to be.
00:12:32.077 --> 00:12:35.337
I mean, we don't have coaxial interconnections on the circuit boards.
00:12:35.417 --> 00:12:41.917
So we have to make sure that we have at least a two-sided waveguide for the energy to move through.
00:12:42.936 --> 00:12:46.716
That's the first thing you do. And then one of the rules that I follow is I
00:12:46.716 --> 00:12:48.696
follow a minimum etch approach.
00:12:48.856 --> 00:12:52.376
So if I paid for copper, I'm not giving it back unless I have to.
00:12:52.616 --> 00:12:59.996
So I will flood all of the unused spaces with ground and then use vias to stitch that ground together.
00:12:59.996 --> 00:13:06.956
That gives me added robustness, especially to ESD and other types of interference,
00:13:06.956 --> 00:13:13.856
because I provide more surface area for these impulses of energy that are going
00:13:13.856 --> 00:13:21.196
to cause problems and distribute them across a surface that's not used for signal movement.
00:13:21.196 --> 00:13:24.576
So on the outer layers especially, I want to have copper out there.
00:13:24.596 --> 00:13:28.576
So when I get a bubble of field on it, that field gets spread out across the
00:13:28.576 --> 00:13:34.776
flood and has a lower impact on the traces because it's all about surface area.
00:13:34.876 --> 00:13:38.156
The greater the surface area, the happier the field is to be there.
00:13:38.156 --> 00:13:42.156
And that helps to reduce the impact of this high-energy impulse,
00:13:42.336 --> 00:13:48.556
maybe 15,000 or 20,000 volts, to cause the circuit to either malfunction or
00:13:48.556 --> 00:13:51.436
even worse is to have the system be damaged.
00:13:51.636 --> 00:13:54.556
So those are kind of the things that I would do. Okay.
00:13:55.416 --> 00:13:59.376
Again, the dielectric awareness, you need to make sure that if you're in the
00:13:59.376 --> 00:14:02.896
dielectric between layer one and two and you want to go down in the board stack,
00:14:03.116 --> 00:14:06.136
that you have to have a pipe to take it down through the board stack.
00:14:06.136 --> 00:14:12.576
And one of the most common failures I see is people will use a signal via and
00:14:12.576 --> 00:14:15.756
forget that they have to connect the dielectric and the Z-axis.
00:14:15.876 --> 00:14:19.876
And they hope that it gets there. It's like if you have a water faucet in the
00:14:19.876 --> 00:14:25.056
basement and you turn the water on and you want it to go to the upstairs and you don't have a pipe.
00:14:25.376 --> 00:14:28.636
Well, it gets there eventually, but you're not going to like the results.
00:14:28.696 --> 00:14:30.556
And that's exactly how a field behaves.
00:14:30.756 --> 00:14:33.696
The energy will get there eventually, but you're not going to like it.
00:14:33.696 --> 00:14:37.776
There's going to be problems with single integrity, problems with radiator susceptibility.
00:14:38.036 --> 00:14:41.876
All those kind of things happen when you just don't follow the simple rule.
00:14:43.062 --> 00:14:48.102
Two discrete conductors separated by a space. Okay. Well, that's great.
00:14:48.102 --> 00:14:49.882
That's good practical advice. It's all about the space.
00:14:51.102 --> 00:14:55.542
Yes. And didn't your daughter compose a song about that? Well,
00:14:55.582 --> 00:14:57.302
she's my performing artist.
00:14:57.462 --> 00:14:59.722
So she plays the guitar and sings.
00:15:00.122 --> 00:15:03.142
And about, this was, I think, the
00:15:03.142 --> 00:15:08.762
first year that I was actually going to present at PCB West until 2015.
00:15:09.362 --> 00:15:15.562
And I wanted to have a hook to get people to have some different perspective.
00:15:15.782 --> 00:15:19.942
Because it took me almost 10 years before Ralph Morrison finally broke through
00:15:19.942 --> 00:15:25.002
my brain and had me stop thinking about circuit theory and wires and current
00:15:25.002 --> 00:15:26.302
loops and return currents.
00:15:26.682 --> 00:15:32.242
And I said, I need to find a way so these people don't take 10 years to absorb this.
00:15:32.322 --> 00:15:36.842
I want them to be able to be brainwashed into getting it in the one or two hours
00:15:36.842 --> 00:15:38.722
that I have their focus for.
00:15:38.942 --> 00:15:43.662
So I heard my children were listening to Megan Trainor's All About That Bass.
00:15:44.662 --> 00:15:48.902
And I'm like, you know, bass and space rhyme. I think I'm going to take a stab
00:15:48.902 --> 00:15:52.922
at trying to write words that will apply to my classes.
00:15:53.242 --> 00:15:57.042
And so I sat in my office with my headphones and listened to that song over
00:15:57.042 --> 00:16:01.002
and over again and writing my It's All About the Space song.
00:16:01.002 --> 00:16:05.742
And the focus is try and get people to think about the idea that the energy
00:16:05.742 --> 00:16:08.982
moves in spaces and it's not the wires.
00:16:09.522 --> 00:16:14.342
And it came out to be pretty cool. I was going to have both my daughters sing
00:16:14.342 --> 00:16:18.022
it, but there was never a chance when they were coordinating at the same time.
00:16:18.182 --> 00:16:24.342
And the week before I had to leave for PCB West that year, I said, I need the song.
00:16:24.522 --> 00:16:28.702
You've got to go take care of it. So my daughter went upstairs with her iPhone
00:16:28.702 --> 00:16:30.722
and played my song for me.
00:16:30.882 --> 00:16:34.062
And the rest is history. I use it every time I.
00:16:35.140 --> 00:16:40.120
Present. And it's been pretty well received. The goal is that people wake up
00:16:40.120 --> 00:16:42.680
in the middle of the night screaming, it's all about the space.
00:16:43.500 --> 00:16:49.500
And then I've achieved my goal. That's great. That's one that was very endearing as well.
00:16:49.940 --> 00:16:55.300
So I think you've done a great job of setting the foundation.
00:16:55.680 --> 00:17:01.100
And it seems like Ralph Morrison played a big role as your mentor.
00:17:01.100 --> 00:17:04.160
Do you want to talk a little bit about that and just, you know,
00:17:04.280 --> 00:17:10.140
how Ralph Morrison, you know, changed your perspective on field wave,
00:17:10.320 --> 00:17:12.180
you know, physics of electronics?
00:17:12.520 --> 00:17:20.500
It was life-changing to say the least. The way I met Ralph was a series of very fortunate accidents.
00:17:20.980 --> 00:17:26.580
And there were some events that led to me being asked to put together a panel
00:17:26.580 --> 00:17:29.860
of experts for one of the company conferences for customers.
00:17:29.860 --> 00:17:35.840
And like any good pretend EMC guy, I had a really nice library of books.
00:17:36.040 --> 00:17:38.980
So I had books from Howard Johnson.
00:17:39.600 --> 00:17:44.080
I had books from Dr. Archambault. I had Lee Ritchie stuff.
00:17:44.380 --> 00:17:50.940
I had Henry Ott, the Kenan books. And so I started calling the authors because
00:17:50.940 --> 00:17:53.980
I committed to making this happen.
00:17:54.300 --> 00:17:59.660
And all of them were going to a conference in the UK.
00:17:59.860 --> 00:18:02.640
There was an international EMC consortium happening.
00:18:03.840 --> 00:18:08.240
And like everybody else, you have to have Ralph Morrison's Grounding and Shielding book.
00:18:08.340 --> 00:18:13.320
That was the one you had to see. So I had this beautiful set of books that I rarely looked at.
00:18:13.480 --> 00:18:15.880
And I called and left a message for Ralph.
00:18:16.120 --> 00:18:18.940
And he called me back and says, who are you and why are you calling?
00:18:19.060 --> 00:18:21.960
And I'm like, well, I'm Dan Beaker from Freescale.
00:18:22.789 --> 00:18:27.729
We're having a, I'm having a panel of experts and I'd like you to be part of it. And he agreed.
00:18:28.029 --> 00:18:30.609
And so we started talking about it. I found out he was a physicist.
00:18:30.849 --> 00:18:34.869
I'm like, oh, this is great. You can be my first speaker because you can establish
00:18:34.869 --> 00:18:38.809
the basis for the, for the talk. You'll talk about EM physics.
00:18:39.049 --> 00:18:43.209
And so we had the, I had several experts there and I'm not going to name names
00:18:43.209 --> 00:18:47.129
at this moment, but Ralph was first and the rest of them started doing their talk.
00:18:47.349 --> 00:18:50.549
By the time they got done, he was so mad he could hardly stand still.
00:18:50.829 --> 00:18:53.629
He was bouncing up and down going they've got it wrong they've got
00:18:53.629 --> 00:18:56.669
it wrong that's not how it works i know so ralph
00:18:56.669 --> 00:19:00.089
and he was probably his mid-80s by then already because
00:19:00.089 --> 00:19:03.649
i dragged him out of retirement and he's bouncing up and down i said ralph we
00:19:03.649 --> 00:19:06.889
you got to calm down let's go have lunch and you can tell me about it and he
00:19:06.889 --> 00:19:12.149
started his first lecture to me on on physics and fields and i'm like this really
00:19:12.149 --> 00:19:17.109
makes sense because everything else was was uh confusing especially going to
00:19:17.109 --> 00:19:18.969
different single integrity classes,
00:19:19.309 --> 00:19:22.109
there was always contradiction in what they were speaking about.
00:19:22.349 --> 00:19:26.069
And looking backwards, it was, none of that was malicious as they were talking
00:19:26.069 --> 00:19:27.789
about what worked for them, their experience.
00:19:28.149 --> 00:19:32.469
But it wasn't always based on the science. And so I hired Ralph to come and
00:19:32.469 --> 00:19:38.049
teach classes for me and we became friends and he decided he was going to make me his apprentice.
00:19:38.149 --> 00:19:40.869
He was going to teach me field physics or kill me trying.
00:19:41.629 --> 00:19:46.709
And we spent as much time as we possibly could whenever I would come to PCD West.
00:19:46.969 --> 00:19:50.269
I would spend a few days with Ralph. And a few times, I actually got to fly
00:19:50.269 --> 00:19:51.969
out here and stay at his house.
00:19:51.989 --> 00:19:55.369
And we would go up into his little guest house with a whiteboard.
00:19:55.389 --> 00:19:58.289
And he's telling me all the realities of physics.
00:19:58.509 --> 00:20:03.869
And then I would make my comments back. And he was almost the ruler on the knuckles
00:20:03.869 --> 00:20:08.009
when I'd say the wrong thing, because he was very keen on language and making
00:20:08.009 --> 00:20:09.989
sure that I used the proper words.
00:20:10.129 --> 00:20:13.549
And he explained in great detail why what I said was wrong.
00:20:14.129 --> 00:20:17.169
And we went through that relationship and then,
00:20:18.018 --> 00:20:22.178
i was still going to pcb west and other signal integrity conferences just to
00:20:22.178 --> 00:20:26.158
try to get some knowledge and having classes with some of the other people that
00:20:26.158 --> 00:20:32.158
were known in the industry and and trying to decipher what was real and uh.
00:20:33.700 --> 00:20:37.680
I finally got to the point where I felt comfortable with teaching it.
00:20:38.080 --> 00:20:45.580
And one of the sessions I did was in Austin for the company's applications and design community.
00:20:45.860 --> 00:20:49.820
And Ralph was in the audience because I was there trying to get him to help
00:20:49.820 --> 00:20:53.260
convince us that fields were real and EM physics was important.
00:20:53.720 --> 00:21:00.480
And he walked up to me after I'd been in my presentation and I knew I was in trouble.
00:21:00.620 --> 00:21:03.320
I thought I'd said something wrong. And he looked up and says,
00:21:03.440 --> 00:21:05.000
Dan, I think you finally got it.
00:21:05.140 --> 00:21:10.000
So that's one of the most important and exciting achievements of my career.
00:21:10.420 --> 00:21:17.160
But taking his knowledge and using it in my everyday designs and working with
00:21:17.160 --> 00:21:23.040
customers to help them understand the physics and how the behavior is important
00:21:23.040 --> 00:21:25.380
and how using your product,
00:21:25.480 --> 00:21:28.420
the PC board structure, to direct the energy,
00:21:28.680 --> 00:21:32.840
I can work on anything now. You know, it changed my whole perspective.
00:21:33.120 --> 00:21:37.040
I'm not afraid of, you know, electric vehicle inverter designs,
00:21:37.600 --> 00:21:40.440
high speed network controllers, all that.
00:21:40.620 --> 00:21:43.620
The rules are the same. All that changes is the geometry.
00:21:43.920 --> 00:21:46.080
So the only math I need is algebra.
00:21:46.940 --> 00:21:50.280
The structures are predefined and it's very simple.
00:21:50.520 --> 00:21:56.460
Once you stop being confused and scared by the hairy math, the calculus that
00:21:56.460 --> 00:22:00.640
everybody presents in the school, and start looking at things from the very
00:22:00.640 --> 00:22:04.860
simple behaviors that are really driven by, you know, physical structures.
00:22:05.120 --> 00:22:10.240
And that part was probably the most difficult for me to, as a digital engineer.
00:22:10.440 --> 00:22:16.120
I mean, I worked on 68020 stuff. I thought I was Superman. and I didn't care
00:22:16.120 --> 00:22:18.120
about geometries or fields.
00:22:18.300 --> 00:22:24.140
But once I understood that it was energy, it did travel in the dielectric and that,
00:22:25.004 --> 00:22:29.344
It was all about creating the spaces that would direct the energy to each other.
00:22:29.644 --> 00:22:33.044
So that changed everything. That's great. No, that's very endearing.
00:22:33.224 --> 00:22:36.084
It's fantastic. It's always good to have a mentor like that. That's amazing.
00:22:37.544 --> 00:22:44.184
Do you have an example of how the energy misbehavior in a board caused the board not to function well?
00:22:44.344 --> 00:22:46.344
Do you have any kind of thoughts or examples on that?
00:22:46.684 --> 00:22:49.484
If not, that's fine. But I thought I might as well ask and see.
00:22:49.524 --> 00:22:54.684
Well, every time there's an EMC failure, it's because they failed to control
00:22:54.684 --> 00:22:57.004
the electromagnetic energy properly.
00:22:57.484 --> 00:23:01.544
And one of the most common ones have to do with power supply designs.
00:23:01.804 --> 00:23:06.724
Almost all of the radiated emissions issues that I've worked on are because
00:23:06.724 --> 00:23:13.724
of there is an improper understanding of delivery of energy to the ICs.
00:23:13.864 --> 00:23:21.364
And we were taught before that if you have any types of radiated emissions problems,
00:23:21.364 --> 00:23:24.884
It's the output, the switching event that causes the problem.
00:23:25.004 --> 00:23:27.564
And what I've discovered is it's not the switching event.
00:23:27.824 --> 00:23:32.924
It's the opposite event. This is when you turn on a switch and the energy moves
00:23:32.924 --> 00:23:38.244
into that pipe, there's never a continuous impedance.
00:23:38.444 --> 00:23:43.324
So it's like you have two different size hoses and you have a big hose and a little hose.
00:23:43.424 --> 00:23:49.144
When the big hose turns on, the little hose drops pressure and that's the reduction of voltage.
00:23:49.684 --> 00:23:55.744
So you get a drop in voltage that's the same frequency, same edge as the output.
00:23:55.944 --> 00:23:59.744
So if it's a nanosecond switch, if you get a nanosecond, I call it depletion
00:23:59.744 --> 00:24:02.784
wave, going upstream looking for the output.
00:24:03.678 --> 00:24:09.158
Source of energy. And if you don't have the right amount of energy at the right
00:24:09.158 --> 00:24:14.498
place on the board, then this nanosecond event can find a quarter wavelength
00:24:14.498 --> 00:24:16.678
and find its way to then radiate.
00:24:16.878 --> 00:24:21.378
And most commonly is they will not have followed the one dialectic rule because
00:24:21.378 --> 00:24:23.078
they don't think power is important,
00:24:23.418 --> 00:24:26.458
clocks are important, that they will separate power from ground.
00:24:26.738 --> 00:24:31.098
And that causes a discontinuity that will allow sometimes the energy to,
00:24:31.258 --> 00:24:32.898
the depletion energy, that wave,
00:24:33.098 --> 00:24:37.678
because it's creating a magnetic field too, and that magnetic field that's relative
00:24:37.678 --> 00:24:43.358
to the changing voltage is what causes a changing voltage and finds its way into space.
00:24:45.359 --> 00:24:50.219
That happens everywhere. Well, I mean, that was a great explanation.
00:24:50.559 --> 00:24:54.739
And I hope people, you know, really kind of digest that and understand that
00:24:54.739 --> 00:24:59.239
Sierra Circuits has been working very hard for electrical engineers and PCB
00:24:59.239 --> 00:25:02.859
designers like yourself on our engineering tools.
00:25:03.199 --> 00:25:07.219
These are engineering tools to help you design faster. We want to reduce your
00:25:07.219 --> 00:25:10.019
design time and get the design right the first time.
00:25:10.179 --> 00:25:14.579
So our top tools on our website, number one is the PCB stack-up planner.
00:25:14.579 --> 00:25:19.719
Knowing that you have a good stack up right away for your design.
00:25:19.959 --> 00:25:24.499
Number two is the bomb checker. It'll do basic scrubbing, make sure your ref
00:25:24.499 --> 00:25:29.059
deses are good, your MPNs are good, the MPN matches the description.
00:25:29.499 --> 00:25:33.039
You know, all these are amazing features of the bomb checker.
00:25:33.199 --> 00:25:39.179
We also have an impedance tool, which is based on Maxwell's equations.
00:25:39.599 --> 00:25:44.359
And these are all for free for the PCB designer and electrical engineer.
00:25:44.579 --> 00:25:46.739
You know, please go check them out. It's all for you.
00:25:47.099 --> 00:25:50.399
So, as you mentioned, Sierra Circuits is a PCB manufacturer.
00:25:50.399 --> 00:25:53.579
We have one of our PCB manufacturing experts here.
00:25:54.199 --> 00:25:58.799
He's been working on kind of a new way of building circuit boards.
00:25:59.819 --> 00:26:06.039
You want to talk a little bit about that? Yes. So, we've been working on what
00:26:06.039 --> 00:26:12.739
is now referred to as ultra high density, which is defined as sub 2 mil line in space and below.
00:26:13.339 --> 00:26:18.219
So where most people have taken the approach where they're putting the traces
00:26:18.219 --> 00:26:22.259
on the surface and trying to make the fine circuitry that way,
00:26:22.859 --> 00:26:26.319
we go the opposite direction.
00:26:26.319 --> 00:26:32.219
We actually trench into the dielectric and then fill plate to form the traces.
00:26:32.779 --> 00:26:35.279
So that has two major advantages.
00:26:36.039 --> 00:26:43.559
First of all, when you put, say, a one mil line on the surface and you want
00:26:43.559 --> 00:26:49.679
an equivalent to a half ounce copper, you've essentially got almost a square feature.
00:26:50.400 --> 00:26:55.360
So, when you go to laminate it and you put pressure on it and squeeze the resin
00:26:55.360 --> 00:27:00.260
between the traces, it puts a lot of pressure and those features move around.
00:27:00.260 --> 00:27:01.900
We found that out the hard way.
00:27:02.640 --> 00:27:09.460
So, if you embed the traces into the laminate, you don't have any registration issues.
00:27:09.780 --> 00:27:12.180
Everything stays exactly where it needs to be.
00:27:13.180 --> 00:27:20.580
The other big advantage is, you know, when you look at even the semi-additive
00:27:20.580 --> 00:27:25.020
processes where they're going with super thin base coppers, you still have to etch it.
00:27:25.700 --> 00:27:30.820
So when you have a, say, a one mil line and you have that etch exposed,
00:27:30.820 --> 00:27:32.820
you're still going to get some etching.
00:27:32.980 --> 00:27:36.600
You're going to get some little mouse bites and you're going to get some degradation.
00:27:37.000 --> 00:27:41.520
It's going to, you're going to lose consistency of that trace line.
00:27:41.520 --> 00:27:47.260
So, by essentially forming a trench and putting the copper in that,
00:27:47.640 --> 00:27:52.920
then we're protecting the side of the trace from any etchant.
00:27:53.100 --> 00:27:58.520
So, we get a super consistent trace width all the way through however long we have to go.
00:27:59.840 --> 00:28:01.740
So, those are the two big advantages.
00:28:03.780 --> 00:28:07.640
Another thing that, you know, there's a number of things that we can do with
00:28:07.640 --> 00:28:10.220
it that we're just now starting to look at.
00:28:11.520 --> 00:28:19.600
You don't really need termination pads through holes, things like that.
00:28:19.820 --> 00:28:24.680
Essentially, what we can do with it is that you can cross one trace,
00:28:24.680 --> 00:28:31.340
and rather than drill a hole, you can just trench down to that next layer and connect that way.
00:28:31.340 --> 00:28:36.280
So you don't need all the holes and all the vias and all the landing pads and
00:28:36.280 --> 00:28:44.220
all of that. so that's essentially it so sounds exciting yeah,
00:28:45.368 --> 00:28:49.548
Yeah, so we're getting down, I don't know if you can see this,
00:28:49.868 --> 00:28:54.528
this is an example of a one mil, line, mil and a half space.
00:28:55.768 --> 00:29:00.708
We're getting into some super high density materials.
00:29:01.728 --> 00:29:04.748
I'll show up for the audience. Yeah. Super high density.
00:29:06.628 --> 00:29:09.768
So anyway, yeah, that's it.
00:29:11.568 --> 00:29:16.068
Just real brief. What do you think how that's going to kind of impact and,
00:29:16.068 --> 00:29:21.268
you know, design in general and the physics of, you know, the electromagnetic
00:29:21.268 --> 00:29:22.788
fields that we've been talking about?
00:29:23.008 --> 00:29:29.888
I think it's pretty exciting. I was privileged to be part of the original announcement of this product.
00:29:30.448 --> 00:29:39.548
And there are a lot of opportunities to do things in the laminate layer so you don't waste that space.
00:29:39.548 --> 00:29:45.548
So my first thought when I heard about this was you've got a one mil wide conductor
00:29:45.548 --> 00:29:49.368
that's three mils deep and they can place them two mils apart.
00:29:49.648 --> 00:29:53.748
And if you run an impedance calculator on that, you've got a 47 ohm trace.
00:29:55.640 --> 00:30:01.760
In free space. So, you can have a single layer of controlled impedance transmission
00:30:01.760 --> 00:30:04.380
line, and you can repeat that orderly.
00:30:04.480 --> 00:30:09.620
You can do, you know, a signal, a ground, and a signal, and put two traces with
00:30:09.620 --> 00:30:13.020
47 ohms in a space of about seven mils.
00:30:13.320 --> 00:30:17.660
And that's pretty exciting. And especially with the challenges of the new ICs,
00:30:17.800 --> 00:30:23.000
we're going to 0.4, 0.2, 0.3 pitch components.
00:30:23.320 --> 00:30:28.260
How do you connect to those things? You know, people use HDI or ultra-HDI,
00:30:28.540 --> 00:30:31.160
but that's really difficult to implement.
00:30:31.440 --> 00:30:35.900
And I think that being able to do this type of routing, especially when you
00:30:35.900 --> 00:30:40.160
can punch a hole and drop some copper from one layer to the other,
00:30:40.340 --> 00:30:44.100
that without having to have a via and then make these types of connections.
00:30:44.380 --> 00:30:48.480
And there are other things I'm thinking about is because I'm getting all kinds
00:30:48.480 --> 00:30:55.280
of ideas after hearing what Mike just said, is that they, I think if you look
00:30:55.280 --> 00:30:56.920
at this, I think this is a way,
00:30:57.240 --> 00:30:59.700
one, of reducing the impact of the weed.
00:31:00.140 --> 00:31:06.760
So the idea of having different signals travel at different effective propagation times,
00:31:07.000 --> 00:31:11.420
which is a real issue for some of the very high-speed buses that I've taken
00:31:11.420 --> 00:31:15.200
a number of classes where you're looking at how do you overcome the impact of
00:31:15.200 --> 00:31:22.840
the varying dielectric properties as you go from the resin to the glass to the resin to the glass.
00:31:23.060 --> 00:31:28.920
And from trace to trace, that variation can cause significant issues with meeting
00:31:28.920 --> 00:31:29.960
the timing requirements.
00:31:30.100 --> 00:31:35.460
I think that being able to do this, you reduce the impact of that type of issue.
00:31:35.680 --> 00:31:42.120
And I think this is kind of a way to extend the upper bandwidth of the traditional
00:31:42.120 --> 00:31:47.540
FR4 material without having to go to some of the more exotic materials.
00:31:47.540 --> 00:31:52.360
I think that there's some great potential for that, and it makes you be able
00:31:52.360 --> 00:31:55.660
to develop a very high-performance product.
00:31:56.918 --> 00:32:04.258
And keep the cost down. And I think from what I've been told is this process
00:32:04.258 --> 00:32:12.278
is reasonably cost competitive to traditional etch or removal technologies,
00:32:12.278 --> 00:32:20.918
certainly in order of the cost of a blind or buried via type approach that you
00:32:20.918 --> 00:32:25.858
can probably get higher densities with similar costs to a traditional board.
00:32:26.058 --> 00:32:29.898
And again, I think there are advantages not just for the smaller traces,
00:32:30.098 --> 00:32:34.758
but being able to increase the transmission speeds.
00:32:34.958 --> 00:32:40.898
And Mike was talking about earlier about how they were looking at ways to reduce
00:32:40.898 --> 00:32:45.378
the travel time because we have issues with trying to get signals from place to place.
00:32:45.538 --> 00:32:51.478
And one of the issues with high speed signals is that they need to have a specifically
00:32:51.478 --> 00:32:54.938
controlled impedance very quickly after they leave the driver.
00:32:55.218 --> 00:32:59.638
And then there's a limit to how far apart things can be before you have to start
00:32:59.638 --> 00:33:00.658
managing that impedance.
00:33:01.258 --> 00:33:08.758
With the faster transition times, then it basically extends the length that
00:33:08.758 --> 00:33:12.678
you can put things, the separation, before you have to care about controlled impedance.
00:33:12.938 --> 00:33:16.958
And the result from that will be that maybe there are a number of signals where
00:33:16.958 --> 00:33:21.978
we don't even won't care about that impedance because we are placing the two
00:33:21.978 --> 00:33:24.398
drivers now within a lump distance
00:33:24.398 --> 00:33:28.598
because the lump distance has gone from maybe an inch or two inches,
00:33:28.738 --> 00:33:33.058
but because we have a signal that travels faster, so it goes farther during
00:33:33.058 --> 00:33:35.818
that rise time, which is when the signal is turning on,
00:33:36.158 --> 00:33:39.458
that we can maybe get three or four inches. And that is going to be a big delay.
00:33:40.381 --> 00:33:48.401
Another thing I forgot to mention is we're doing a fully additive metallization in the trench.
00:33:49.041 --> 00:33:53.741
So we no longer have the dendrite structure of the copper foil to deal with.
00:33:54.081 --> 00:33:57.821
Because if you look at the semi-additive processes, even if you're putting a
00:33:57.821 --> 00:34:03.121
3-micron foil on the surface, you still have the dendrite structure to get it
00:34:03.121 --> 00:34:04.281
to stick to the laminate.
00:34:05.101 --> 00:34:08.901
So we don't have to work by embedding them.
00:34:09.081 --> 00:34:14.601
We don't have to worry about peel strength much anymore because it's not going
00:34:14.601 --> 00:34:16.021
to move once it's in there.
00:34:16.321 --> 00:34:20.301
So we don't have the dendrite structure. And that seems, yeah,
00:34:20.421 --> 00:34:26.981
so it's we've solved some of the problems that that have that are compounding
00:34:26.981 --> 00:34:28.361
with ultra high density.
00:34:29.301 --> 00:34:33.761
So, you know, they want to bump the signal speed up, but then they're going
00:34:33.761 --> 00:34:39.781
to super flat copper foils, low dendrite structures.
00:34:40.321 --> 00:34:47.601
And then that magnifies the problem of getting all these little traces to stick in the same spot.
00:34:47.821 --> 00:34:52.081
Yeah. And then it reduces losses because there's less interaction between the
00:34:52.081 --> 00:34:58.001
field and the surface of the copper with it being smoother. So you have a nicer surface area.
00:34:58.101 --> 00:35:01.781
So you have less of that energy converted into heat.
00:35:02.121 --> 00:35:06.741
But, you know, one of the three things you do with field is to convert it into kinetic energy.
00:35:06.921 --> 00:35:10.601
And it interacts with every molecule it comes in contact with.
00:35:12.232 --> 00:35:17.392
Copper or the dielectric. And by making it smoother, then you have the,
00:35:17.512 --> 00:35:22.492
it makes the field happier as it screams down through at the speed of light. Less bumpy.
00:35:23.012 --> 00:35:26.532
Well, that's pretty exciting. I can't wait to play with that technology.
00:35:27.072 --> 00:35:33.512
No, I think it is very exciting. And there's, it's very, I think of it as cross-functional, right?
00:35:33.632 --> 00:35:37.552
That we're changing the way manufacturing is happening, but that's absolutely
00:35:37.552 --> 00:35:41.732
going to change the way that an electrical engineer needs to think about their circuit as well.
00:35:42.272 --> 00:35:44.092
So lots to think about here.
00:35:44.652 --> 00:35:50.412
So Dan, I understand that, you know, you recently had a circuit board manufactured
00:35:50.412 --> 00:35:52.632
by Sierra Circuits, which, thank you very much.
00:35:52.792 --> 00:35:57.372
But would you like to talk a little bit about that, that circuit board and the experience?
00:35:57.712 --> 00:36:02.332
Yeah, I would love to. I was challenged with doing a, basically an Ethernet
00:36:02.332 --> 00:36:06.952
gathering board designed to work with radar sensors in automotive.
00:36:07.332 --> 00:36:12.572
And that was dubbed the Purple Box, And it was actually demoed at CES this year.
00:36:12.932 --> 00:36:22.852
And the challenges were a very small form factor and a very large component density.
00:36:23.732 --> 00:36:28.232
And the other thing was it needed to be done in a hurry, especially the first
00:36:28.232 --> 00:36:31.732
pass, because we needed to get prototypes built so they could start doing the
00:36:31.732 --> 00:36:38.072
software, because the software integration for an Ethernet control network is pretty challenging.
00:36:38.852 --> 00:36:44.332
And so I knew there was only one solution as I talked to Amit and said,
00:36:44.452 --> 00:36:49.532
I've got this board and can you fabricate it? And oh, by the way, can you assemble it?
00:36:49.672 --> 00:36:53.372
Because I don't think any of my traditional CMs are going to be able to handle
00:36:53.372 --> 00:36:57.332
this type of component density. And so we decided.
00:36:57.909 --> 00:36:59.669
We started that project, and we
00:36:59.669 --> 00:37:05.189
were able to get them to fabricate and assemble a number of these units.
00:37:05.949 --> 00:37:10.229
What, 15 days, I think, was the cycle time from start to finish.
00:37:10.369 --> 00:37:13.249
From start to finish, yeah. And all the boards worked.
00:37:14.309 --> 00:37:17.669
I used my philosophy to the design.
00:37:17.949 --> 00:37:25.269
It was functional, first thing out of the box, and we successfully demoed it.
00:37:25.269 --> 00:37:31.849
And it was so successful at CES that now I've been told by marketing, I need to go build.
00:37:31.949 --> 00:37:36.909
We're building 27 boards for me right now as fast as they possibly can because
00:37:36.909 --> 00:37:40.109
we have great demand for this particular application.
00:37:40.309 --> 00:37:42.449
And it's for autonomous vehicle.
00:37:42.669 --> 00:37:44.449
So very cutting edge stuff.
00:37:44.929 --> 00:37:49.089
We had challenges to use all of my techniques.
00:37:49.089 --> 00:37:53.469
But more importantly, as this board is intended to be a full reference design.
00:37:53.469 --> 00:38:02.089
So automotive components and automotive design rules and then my the plan will be to do full.
00:38:03.409 --> 00:38:08.329
Certification from an emc perspective and and do over temperature and the whole
00:38:08.329 --> 00:38:12.669
the whole by nine yards so that a customer can clone the design you know i want
00:38:12.669 --> 00:38:18.989
to be able to to look at this and do the right thing and i think we did a lot of very uh unique,
00:38:19.803 --> 00:38:24.763
approaches to the way we built the board. And I'm very excited about,
00:38:24.763 --> 00:38:27.203
you know, building more of them with Sierra.
00:38:27.583 --> 00:38:31.823
So. All right. Awesome. And I always plan and promise them up front that I'm
00:38:31.823 --> 00:38:32.623
going to make them crazy.
00:38:33.303 --> 00:38:38.683
I definitely do that. I always push my suppliers past the envelope and we develop
00:38:38.683 --> 00:38:42.043
new avenues together. And that's pretty exciting as well.
00:38:42.163 --> 00:38:45.143
And that's been something I've done with Sierra a couple of times.
00:38:45.363 --> 00:38:50.443
If you haven't heard of Sierra Circuits, Sierra Circuits is a PCB manufacturer
00:38:50.443 --> 00:38:57.343
and assembler all-in-one, located in the Bay Area, right around all the innovation that's happening.
00:38:57.663 --> 00:39:03.363
And Sierra Circuits is capable of building everything from start to finish,
00:39:03.623 --> 00:39:09.823
from simple standard product to uber-complex, HDIs, flexes, rigid flex,
00:39:10.443 --> 00:39:11.783
high-speed applications.
00:39:12.403 --> 00:39:16.683
You know, anything that you can think of, we pretty much can build. And we do it quick.
00:39:17.303 --> 00:39:23.943
So, if you need to maintain your schedule and be on time, Sierra Circuits is your vendor of choice.
00:39:24.443 --> 00:39:27.903
Thank you so much. And so, that board had a blind via. Do you want to just take
00:39:27.903 --> 00:39:31.603
the audience through the basics of, you know, what it takes to do a blind via?
00:39:33.063 --> 00:39:38.983
So, this particular board actually had three separate drill files in it.
00:39:39.403 --> 00:39:43.243
So it had a field via.
00:39:44.244 --> 00:39:48.644
Program, and then it had a final through hole, and then it had a laser via.
00:39:50.044 --> 00:39:56.724
So the laser vias, those are relatively simple these days.
00:39:58.524 --> 00:40:04.984
Really what we've been at this type of work for almost 12 years now.
00:40:04.984 --> 00:40:14.864
So one of the things that to make all this happen, we have a control system.
00:40:14.884 --> 00:40:19.604
So we put your stack up into this software program.
00:40:19.884 --> 00:40:27.944
And then what happens from there is that program talks to our material movement program.
00:40:28.124 --> 00:40:31.704
So one of the things that when you get into microelectronics,
00:40:31.704 --> 00:40:35.204
you have to control all your material movement. Everything is constantly moving.
00:40:35.444 --> 00:40:40.644
So anytime it goes into a press, it moves. So if you have, say,
00:40:40.824 --> 00:40:46.524
a 12 by 18 panel, you put it in the press, it's going to shrink 20 mils in the
00:40:46.524 --> 00:40:49.264
12 axis and almost 30 mils in the 18.
00:40:49.804 --> 00:40:56.124
So here you've got a 5 mil laser VN, you're trying to land on a 9 mil pad,
00:40:56.124 --> 00:41:01.464
and you've got a 30 mil shrinkage, then you've got to be able to measure that
00:41:01.464 --> 00:41:03.644
and control it and know where you're going to end up.
00:41:03.724 --> 00:41:11.024
So, what happens is that the first program takes the stack up,
00:41:11.464 --> 00:41:15.744
it outputs the data to the exact system, which is all the material movement.
00:41:16.743 --> 00:41:24.283
It puts it into our CAM software, which starts to design all the design rules for it.
00:41:25.243 --> 00:41:32.483
And then it also puts that stack up into our program that controls all the drilling.
00:41:33.123 --> 00:41:38.243
So what that does is it looks at the stack up and the copper weight and the
00:41:38.243 --> 00:41:41.983
material type, and it automatically defines feeds and speeds.
00:41:41.983 --> 00:41:45.943
It's another thing is it writes the whole targeting structure.
00:41:46.823 --> 00:41:51.563
That's another thing that if you look at the way the board goes through the
00:41:51.563 --> 00:41:56.503
shop, you have all the upfront stuff, all the programming and all the targeting
00:41:56.503 --> 00:41:59.603
structure and all of this stuff. Then you have to be able to track that.
00:41:59.923 --> 00:42:08.423
So what happens is we start with the exact and the exact tells the exposure unit what to do.
00:42:08.603 --> 00:42:12.383
And then the exposure unit puts all the targets in. and then it goes to drilling,
00:42:12.423 --> 00:42:15.603
then we pick up and acquire all those targets with the x-ray drill,
00:42:15.843 --> 00:42:19.663
and then we put targets in for that, and then for mechanical drilling,
00:42:20.123 --> 00:42:23.443
we're all vision drilling, so we don't have the old,
00:42:24.083 --> 00:42:29.483
and the old style was you started with a zero point, and then you went out from there.
00:42:29.803 --> 00:42:33.463
What these machines do is they disconnect all that, so they go out and they
00:42:33.463 --> 00:42:37.883
look at a set of targets that we put in with the direct imager.
00:42:39.041 --> 00:42:42.261
And then it calculates the scale and all of that off of that,
00:42:42.321 --> 00:42:45.101
and then it drills the pattern. So it's almost free-forming.
00:42:45.681 --> 00:42:50.321
So it compensates for all of the changes in the geometry of the board as it
00:42:50.321 --> 00:42:51.321
goes through the process. Right.
00:42:51.981 --> 00:42:56.981
So, because when you go through sequential laminations, what happens is that
00:42:56.981 --> 00:43:00.021
each time you go through it, it shrinks more and more and more.
00:43:00.021 --> 00:43:05.201
So when you get another thing that's happening, like with your board,
00:43:05.481 --> 00:43:13.121
you had what we refer to as a SF file where we had a non-conductive filled vias
00:43:13.121 --> 00:43:15.581
for via and pad type stuff.
00:43:15.961 --> 00:43:20.701
So what happens there is that gets drilled and then it goes down the street
00:43:20.701 --> 00:43:23.301
and we have to plate it, fill it, planarize it.
00:43:23.481 --> 00:43:27.181
And that in turn moves everything around. So when it comes back.
00:43:28.247 --> 00:43:34.407
It's not the same panel size as when it went out, so then we have to reacquire that.
00:43:34.747 --> 00:43:40.227
So then when we get into sequential lamination, we have all the machines talking to each other.
00:43:40.467 --> 00:43:45.967
So what happens is, like in the laser, say you wanted to add another layer to that.
00:43:47.907 --> 00:43:52.627
So when we do the laser, the laser has the capability. Laser is probably one
00:43:52.627 --> 00:43:54.327
of the smartest machines we have.
00:43:55.087 --> 00:43:59.047
So that'll go out and look at the alignment targets and calculate the scale.
00:43:59.467 --> 00:44:03.047
It'll drill the holes, and while it's doing that, it will drill another set
00:44:03.047 --> 00:44:07.947
of targets that the direct imager looks at. So the whole thing repeats itself.
00:44:08.127 --> 00:44:12.227
So if you're getting into stack views and things like that, we're constantly
00:44:12.227 --> 00:44:18.027
dealing with material movement. We're constantly dealing with ways to reacquire
00:44:18.027 --> 00:44:21.447
that movement and find out exactly where it's doing.
00:44:22.427 --> 00:44:27.287
So another thing, too, is controlling the movement, material movement.
00:44:27.547 --> 00:44:32.167
So we have a self-learning system. So when we look at your stack,
00:44:32.447 --> 00:44:38.867
the system automatically puts out predicted scale values for the inner layers.
00:44:39.067 --> 00:44:44.547
And these are based on material type, copper weights, how much copper you're
00:44:44.547 --> 00:44:48.647
going to have, because like a plain layer will move differently than a signal
00:44:48.647 --> 00:44:51.147
layer. So it calculates all this stuff.
00:44:52.140 --> 00:44:58.740
And then that image is output to the LDI, and then that in turn prints targets
00:44:58.740 --> 00:45:01.960
that when we get back to the x-ray drill,
00:45:02.200 --> 00:45:06.460
we go through and we measure each one of those layers, and we can calculate
00:45:06.460 --> 00:45:09.200
what the movement does, and then it's self-learning.
00:45:09.480 --> 00:45:14.760
So whenever we output a file with a prediction, it comes back through,
00:45:14.840 --> 00:45:18.300
we look at it at the x-ray drill, and then it's constantly updating.
00:45:19.280 --> 00:45:28.320
So right now we have roughly 200,000 data points in our exact software because
00:45:28.320 --> 00:45:29.780
it's looking at all kinds of stuff.
00:45:29.980 --> 00:45:33.400
And then it tells you what it likes and what it doesn't like.
00:45:33.960 --> 00:45:36.900
It actually spits out a layer map for drilling.
00:45:37.380 --> 00:45:44.540
So if you do have a shift, we can actually scale to a signal layer.
00:45:45.300 --> 00:45:51.440
So we can go through and we can isolate what the tightest layer is and we can scale to that.
00:45:52.940 --> 00:45:58.480
When we're dealing with sub-assemblies, if we have three or four sub-assemblies,
00:45:58.520 --> 00:46:03.160
we can measure all of them and then lay them out on the table and calculate
00:46:03.160 --> 00:46:09.580
the scaling so we get a best fit for that whole set of sub-assemblies.
00:46:09.580 --> 00:46:14.140
It's got to have a huge impact on not just the yield, but the reliability of the product.
00:46:14.140 --> 00:46:18.100
As well so that's pretty exciting yeah we know exactly what we're dealing with
00:46:18.100 --> 00:46:20.040
and we don't yeah because we're
00:46:20.040 --> 00:46:25.820
all quick turns so we have to do this all up front um and it and it's,
00:46:26.660 --> 00:46:30.880
it's something that you have to build so we've been working on this for like
00:46:30.880 --> 00:46:36.380
the last 12 years so what happens is that this stuff is so um.
00:46:38.087 --> 00:46:43.587
It's so process intensive. So how we run our presses, how we store our material,
00:46:44.007 --> 00:46:46.727
how we do all this stuff is unique to us.
00:46:46.767 --> 00:46:53.327
So if someone was to take our database for scaling, it wouldn't do them any good.
00:46:53.567 --> 00:47:01.167
So this is something that's evolved. And then we've gotten into a lot of strange stuff over the years.
00:47:01.427 --> 00:47:11.267
So a lot of mixed materials, things like that. So we have rules in our drilling control software.
00:47:12.187 --> 00:47:17.807
So it'll look at certain materials. We get a lot of mixed materials these days.
00:47:17.807 --> 00:47:24.007
So you'll have a piece of Rogers surrounded by a bunch of FR-4,
00:47:24.247 --> 00:47:26.527
and they'll stick all kinds of stuff in there now.
00:47:26.707 --> 00:47:32.767
So this is actually smart enough. It goes, okay, this material is the most difficult to drill.
00:47:33.347 --> 00:47:38.707
So it'll grab that information and it'll set up the machine to drill that specific material.
00:47:39.707 --> 00:47:47.927
So we just need a whole lot of continuity and consistency out of everything.
00:47:50.096 --> 00:47:53.156
That's why they worked the first time. Yeah, that's why they worked the first
00:47:53.156 --> 00:47:54.136
time. All of the boards worked.
00:47:54.376 --> 00:47:58.716
Yeah, and that's why we can build it as fast as we can because all this stuff is scripted.
00:47:58.816 --> 00:48:02.896
And we're really lucky we got a really smart automation guy.
00:48:04.216 --> 00:48:09.096
Because another thing, when you look at the alignment targets,
00:48:09.436 --> 00:48:13.996
again, we found out this the hard way, you know, years of trial and error.
00:48:13.996 --> 00:48:20.636
So if you're trying to measure movement for a specific core,
00:48:20.836 --> 00:48:26.736
so like I say, if you've got a plain area, it's not going to move much,
00:48:26.916 --> 00:48:28.976
where a signal layer will move a lot.
00:48:29.356 --> 00:48:37.156
So what has to happen to track that, because originally we put all those targeting,
00:48:37.396 --> 00:48:39.496
that targeting structure on the outside of the board.
00:48:39.776 --> 00:48:43.596
So all of that's moving exactly how we thought it was going to.
00:48:43.996 --> 00:48:47.256
Then you got this big chunk of copper in the center of it, and that didn't move at all.
00:48:48.616 --> 00:48:53.796
So we were able to automate it. So depending on what that layer looks like,
00:48:53.956 --> 00:48:58.296
we'll actually move those targets inboard and we'll measure the movement of
00:48:58.296 --> 00:49:00.796
the PAR, and we'll forget about the outside of it.
00:49:00.976 --> 00:49:06.856
How does that, the fact that one of my process requirements is I leave all the
00:49:06.856 --> 00:49:08.036
copper, I ground flood everything.
00:49:08.276 --> 00:49:13.296
Does that help improve the behavior of the so-called signal layers because there's
00:49:13.296 --> 00:49:16.056
more? surface, more copper in there. Yeah.
00:49:16.316 --> 00:49:19.716
So the more copper you have, the more stability you have.
00:49:19.876 --> 00:49:23.516
Okay. So that, again, I'm telling everybody again, when I preach this,
00:49:24.016 --> 00:49:27.516
is it helps improve the yields not just from the fact you use,
00:49:27.636 --> 00:49:33.076
in the edge process, you keep the fidelity of the edge higher because you consume less of the chemicals,
00:49:33.276 --> 00:49:37.956
but in this case, it helps improve the fidelity of the drills because things
00:49:37.956 --> 00:49:40.916
don't move around as much during the lamination process.
00:49:41.156 --> 00:49:43.356
It makes it more stable. Don't get rid of the copper.
00:49:43.656 --> 00:49:45.916
Keep your copper. Be stingy.
00:49:46.336 --> 00:49:50.776
Yeah, I know. I agree. Yeah. I mean, there's no use sending it down the drain, essentially.
00:49:51.870 --> 00:49:59.830
I mean, we recycle it, but, you know, it's better to leave it on, yeah.
00:50:00.870 --> 00:50:07.310
So it also helps when it comes to lamination because you've got more to—the
00:50:07.310 --> 00:50:13.090
resin will stick to the copper better than it will to the laminate to laminate.
00:50:13.690 --> 00:50:17.930
Yeah, sometimes when I'm walking the floor, I see funky designs where it's like
00:50:17.930 --> 00:50:22.010
a single trace going the length of a board. and I'm like, oh, dear.
00:50:23.090 --> 00:50:28.470
That gets to be interesting. Yeah. Well, you probably know this better than
00:50:28.470 --> 00:50:33.590
I do, but electricity is essentially lazy. So if you...
00:50:36.062 --> 00:50:40.622
If you put little tiny features inside of a plane area, then,
00:50:40.622 --> 00:50:44.582
you know, you start to plate it because you're asking the power to go underneath
00:50:44.582 --> 00:50:48.622
the resist and come up and then go back underneath and come up again and come
00:50:48.622 --> 00:50:49.722
up again and come up again.
00:50:49.722 --> 00:50:54.822
And it finds that first way out and then initializes in the plating bat and
00:50:54.822 --> 00:50:57.122
it goes, okay, I'm done. I'm not going to go any further.
00:50:57.422 --> 00:51:02.802
So, but anyway, we're addressing that whole set of problems now too.
00:51:02.802 --> 00:51:06.162
With so there's a lot of um it's
00:51:06.162 --> 00:51:09.662
pretty pretty cool right now there's a lot of um the way
00:51:09.662 --> 00:51:13.302
we're building going to be building boards is is definitely going to change
00:51:13.302 --> 00:51:18.822
in the next couple years nice well i hope to have you back and we can talk about
00:51:18.822 --> 00:51:23.782
that and you know i think this was it's always amazes me how cross-functional
00:51:23.782 --> 00:51:28.742
this industry is right from physics to to layout to So, you know,
00:51:29.182 --> 00:51:32.582
manufacturing, it all kind of plays on each other. There's a lot of science involved.
00:51:32.782 --> 00:51:35.102
There's a lot going on in the circuit board. Yeah.
00:51:35.422 --> 00:51:39.522
So I definitely appreciate you guys taking the time to, you know,
00:51:39.602 --> 00:51:44.642
meet with us, meet with the audience and share your wisdoms. So that's always great.
00:51:45.022 --> 00:51:47.382
Thank you. Just remember, it's all about the space.
00:51:48.902 --> 00:51:52.582
That's a good one. Thanks for tuning in to Sierra Circuit's podcast.
00:51:53.222 --> 00:51:59.142
We very much appreciate taking the time. and the goal is for you to learn a
00:51:59.142 --> 00:52:03.002
few things every time from the industry experts we have on.
00:52:03.222 --> 00:52:06.522
So stay tuned for future experts that we have for you.