Lee Ritchey and High Speed Digital Design

May 22, 2018

Download this episode (right click and save)

 

Overview: When Lee Ritchey “got through launching things to the moon” his career took off (in Silicon Valley...before it was Silicon Valley!) and he is now widely regarded as one of the premier authorities on high speed PCB and system design. He is the founder of Leading Edge and author of Right the First Time.

 

 

Show Highlights:

  • Lee started as a microwave engineer who designed chips that went up on the Apollo
  • ICs and “faking” logic
  • High Speed design courses first offered at Berkeley.
  • He wrote the books to make the students happy and provide the coursework that didn’t exist.
  • High speed signal path losses - how do we control skew? Where does it come from?
  • And what’s the answer? Spread glass.

Links and Resources:

Lee Ritchey on Linkedin

Right the First Time

Lee’s Digital Library

Lee Ritchey’s Presentation at AltiumLive 2017 in Munich

SI forum - an email forum that is a very good resource for SI questions.

  • All one needs to do is send an email to: si-list-freelists.org
  • Type subscribe in the subject line to become a member.

Transcript:

Hi everyone, this is Judy Warner. Welcome back to the OnTrack Podcast. If you would please subscribe, and let us know what you'd like to hear more about here on OnTrack.

Today, another amazing guest Lee Ritchey, who truly needs no introduction. But if you haven't met Lee before - Lee is considered to be one of the industry's premier authorities on high-speed PCB and system design. He's the founder and president of Speeding Edge - an engineering consulting and training company, some of you have read his book. He's author of 'Right the First Time' and he has a very illustrious, amazing background, and we also had the privilege of having Lee speak last year at AltiumLive in Munich, so I'm delighted to have this conversation with Lee. Not too long after Design Con, so I know he'll have some great wisdom to share.

So, before Lee and I get started, also please connect with me on LinkedIn or on Twitter @AltiumJudy and Altium is also on LinkedIn, Twitter, and Facebook. Lee I'm going to start with a very high-level question. First of all, welcome. It's good to see you again.

Thanks for the invite.

Always my pleasure.

So, you're known as the high speed authority in our industry, but how did you get there? How did you, out of all the paths you could have taken sort of in your technology field, how did you end up kind of going down this rabbit hole?

I started out as a microwave engineer on the Apollo program, which as you probably know, was a long, long time ago. And the company I went to work for was in Silicon Valley, when it was not yet Silicon Valley. Well, after we got through launching things to the moon, NASA decided, well we're done with that, and lots of us - like sixty thousand of us - had to go find something else to do.

Wow.

Well I was in Silicon Valley and integrated circuits were starting to be a big play and the big jobs were designing things with integrated circuits digitally. And so I interviewed for a job designing equipment for testing digital integrated circuits, and got the job. And said, oh now I’ve got to go learn something about logic. So I went to bookstores and got books and I faked it, and that's how I got got my start. And of course, since I was already in the microwave end of things, transmission lines were already part of that, and that was what you had to be good at if you wanted to use ECL, and the high speed computers back then were ECL.

That's how I switched from microwave to what everybody called digital, and for a long time digital was slow enough so you could pretend it was digital, but I never did and always designed on transmission lines and so, as the speed went up all around the industry, everybody needed to learn how to do something with this high speed stuff. And I had a design company from '82 to '92 where we designed pretty much all the early work for Sun, Silicon Graphics, Cray - people like that, and invariably I would get a new client and the engineers knew nothing about high-speed design. So I'd spend two days teaching them the basics so that we could design their board. Out of which grew the courses that I do now. Which I first began to offer at UC Berkeley, and the complaint after every course was: there's no book, there's no textbook, and that's where the books that you mentioned, came from as I  had to write those to make the students happy at UC Berkeley. And in 1999, we decided that we didn't like working for companies so we started Speeding Edge.

Okay.

That's it, that's how I got here by - almost by accident.

Well I always say, most of us got here by accident. I mean some EEs take a nice, clean, straight path, but even they don't take that straight path., I know I didn't end up here on purpose either. But that's interesting, I didn't know that's how your book came to be. So the last time I saw you, you were actually speaking at DesignCon. And how often are you teaching these days actually ?

Most of my classes are private and I would guess about every two months or so.

Okay.

There are two-day, three, and day classes so, that's about as often as I want to do that. I don't know if you've lectured for 14 hours in a row, but it kind of wears out.

I don't think I know anything about anything enough to talk for 14 hours, maybe raising kids I don't know - like I don't think I know about anything to talk that long so now we - now we know for sure, you're way smarter than me.

Oh no, I just came from a different path that's all.

Well you and I were talking recently, preparing for this call and we sort of went down this path talking about PCI Express. So can you kind of talk a little bit about the evolution of speed and the extremely acute curve that we've taken in the last couple of years?

All right, well maybe let's start with PCI itself, which is the Bus architecture that is in - has been - in all the personal computers that you can buy, that's what PCI stands for: Personal Computer Interface, and it was a parallel Bus that you might have seven or eight plug-in cards, all on this Bus. That CPU could talk to any of those, any time, and then originally there was - it clocked at 33 megahertz, and it wasn't too long before the CPUs got faster than the Bus, and all of a sudden we were - what we call - IO-bound. We couldn't get any more performance out of a PC, because the Bus was too slow. So we upped the speed to 66 megahertz, and then a hundred, and for a lot of reasons they're too complicated for today, we couldn't go past 100 and that block limited how fast you could make a personal computer. And so we realized the architecture had to change. And the reason is that if the CPU can talk to any spot in the backplane at any time - to do that really fast, you have to have really, really short connections and that was not realistic. So we turned to an architecture that actually is old. The difference was signaling protocol, that is in PCI Express, has its origins way back with IBM. I was using it in '74, where we would connect two boxes to each other, where we couldn't do that with a parallel Bus, because the the noise in the background was too high. And so that's not a new technique adjust that early on, as you know, the guts of a computer is a parallel architecture meaning lots of bits switching in parallel, and the differential links that we're talking about here are serial.

So at each end, you had to go from a parallel Bus to a serial stream on the line, the other end - go back. And at that time, the serializers and deserializers were extremely complex and expensive. So the only reason you'd ever do that, is there were - if you were stuck, you couldn't do it any other way. Well as we'd gotten to where we have a billion transistors in an IC, these serializers and deserializers are what we call basically free. So all of a sudden it doesn't cost much to go from parallel to serial and back. And the advantage of that is, you can - you can drive... Well let me start from - in a parallel Bus were either series or parallel terminated - if you're lucky, you can drive that at 2 gigabits per second, that's very hard to do. With a serial Bus - we can drive them and we are right now driving them at 32 gigabits per second, which you could never do any other way, and this is how we're getting all the performance we need in the internet. Everyplace else is with these serial lengths and that's what PCI Express is. We switch from a parallel Bus to a serial Bus, to allow us to go faster.

Well, when you have serial links there, you can only have a driver and a CI receiver on the same net. So how is it the CPU's going to talk to six or seven devices like it was doing with the old parallel Bus? And the answer is, we have to have a switch chip somewhere so that we can switch between the CPU of whatever we want to talk to. Well early on, those were expensive chips, so we only use PCI Express in real high-end PCs like a gamer would buy. But we've now integrated those switch circuits right in the CPU so it's not an extra part to buy.

Oh okay.

So it's everywhere. So pretty much everywhere we've got PCI Express, well in itself it's not really all that big a deal because the early PCI Express was - well depends on your point of view - as fast as 500 megabits per second on the line and that's not special, to these terms. The rub is, we have started to go up the performance curve where we've got Gen 1, Gen 2, Gen 3, and so on. And Gen 3, which is it just about around the corner for everybody - it's 8 gigabit per second. That's not slow, and we start to see things that we could ignore at lower rates.

The one we're going to talk about today is not lost - it's the thing we called skew. Skew is the fact that the two sides of a differential pair don't arrive at the receiver at the same time - and some numbers on this, the most common until recently - most common data rate in switches and routers for the internet, was 10 gigabits per second. Where one bit is 100 picoseconds. And I did a test board in 2013 at DesignCon where we discovered 62 picoseconds of error in a path which is almost an entire data bit at 10 gigabits per second - which destroys the work. The link does not work and of course, we've got Gen 4 coming in at 16 gigabits and Gen 5 at 32 - where 32 of the bit period is only 30 picoseconds. So that error I just talked about, is two whole bits which means, nothing's gonna work right? So the question is, where does this skew come from?

So is that, Lee let me interrupt you for just a moment. Is that what you - the course a you taught this year at DesignCon, was that your focus?

Yes well the bootcamp, I call it 'Getting to 32 gigabits per second' which covers a number of things, skew being one of them.

Okay.

Of course the first worry almost everybody had, is loss. So we had this flurry of activity to make low loss materials and smooth Cochrane on and on like that. At the same time we were doing that, I see manufacturers figured out a way to improve their circuits so that loss is not a player anymore - not a big deal anymore. For example, the latest Vertex I guess their Vertex 8 or 9 from Xanax, at 28 gigabits per second, the Lincoln tolerates 38 GB of loss. Meaning that we start out with say a thousand millivolts and we wind up with four at the receiver, and it still works, so all the drive to have the world's best, lowest loss laminate is not a player anymore. Skew is, skew's killing everybody. All these laminates from people like Rogers and - you mentioned one earlier - I can't remember.

Taconic.

Taconic are simply not necessary - not necessary. At any rate, skew is just - and that was the theme of DesignCon this year. It is how do we control this bloody thing? So my impression - I guess that comes up next - is where's it come from?

Yeah, where does it come from Lee?

Well if the two sides were different to pair with different links, that would be, I think obviously, one way that can happen.

Right.

But pretty much everybody else had to design physical links to a few mills, so that tends not to be what the problem is. The problem is - and these are what we call micro defects or micro effects - the glass cloth in laminates you know, on average has a pitch between threads of about 16 mils - between 16 to 20 mils - Traces are 4 or 5 mils wide, so there's a huge difference between the width of a trace and the spacing of those glass fibers. Well, laminate is a mixture of glass and resin, and the dielectric constant of glass is on around 6 and the dielectric constant of resins is less than 3. That means that the lower the dielectric constant, the faster the signal is going to go.

That's right.

So if I had one side of a differential pair on the glass, and on the other one in between, there are two different speeds. That's where the problem comes from.

Now along those lines, I was just talking to Chris Hunrath from Insulectro, and he was talking about spread glass. What do you think?

That's what the answer is.

It is?

Okay cuz you know I've been here at Altium a little over a year and I guess I missed the spread glass thing, but I'm like, that actually sounds like it makes sense.

Well I and my colleagues have been the drivers of spread glass.

Really? Tell us about that.

Well so we found - I've got to confess - we found out by accident,

That's how all good inventions are found right?

Yes, and it had to do with - let me think about when that was - about 2005, we were trying to improve the uniform distribution of glass by using two plies of thin glass, hoping that they would sort of average out. And I had a fabricator in Oregon who says: you know, if you use a single ply of 33:13, you could save some money. So we built the test board and by chance that was really spread very nicely and so we had no skew problems. So all of a sudden, we thought we've solved all the world's problems by just using this glass. And then we built our test board from a different weaver.

So and this weaver spread the glass in one direction but not the other?

Yeah.

So, if you were to get that DesignCon paper you'd discover that we had really good skew one way.

One direction - - [laughter].

And so back to square one. Why aren't people spreading glass? So we got to digging around - and it was for laser drilling of blind vias.

Because, if you think about it, if you have like the classic 4 mil core was called 1080 glass. If you look at that, the glass bundles are round or not spread out, there are big voids between. Well so, if the guy drilling a laser drilled blind via wanted to get rid of the glass, he'd set the intensity to burn the glass and then go tear right through the backstop directly right, and so the laser drilling industry is the one responsible for spreading glass.

Like you said, complete accident.

It is, and they don't care about signal integrity, they care about laser drilling - so I and a guy named Scott 'Hindiga' of Cisco, started going to IPC's sessions on 'Standards for the Last Week' with the intention of getting some standards for how you spread the glass. Well it was got a whole lot like herding cats.

[laughter]

What an IPC committee being like herding cats? I don't know what you're talking about Lee [laughter]

Yeah, so around the table we had five or six weavers and they would not tell us how they wove their glass.

Because it was proprietary?

Yes, so he couldn't come up with the standard, and the only standard that the laser drilling people had was, you take the section of this cloth and they put it in the chamber and see how - put compressed air on one side and see how leaky it was. That's it - that's all there is today. If it was leaky, let's make it an X, and it was good. That would not be good enough for what we're doing here. So we are not done solving this problem, and there's about seven or eight different ways that people approach it. Now if you were at DesignCon this year, you'd discover two papers were presented by Cisco where, when they built the boards at five degrees to the weave, they got the best skew results. So that's how they're...

- Wow, yeah.

Can you imagine what that happens to you in a fab shop if you say pop this artwork on their at 5 degrees?

[Laughter]

-well I remember hearing about - from a colleague, he was an EE - they were actually at some point, because of the glass crossing and there being those bundles, they were actually starting to do it at - basically laying the prepreg at a diagonal. Do you follow what I'm saying?

Oh yes.

-do you remember that? That was kind of going on for a while to see if that would help, and I think it did help a little bit but again, the expense. There's so much loss of materials doing that, that it didn't make sense, or there was the trend of making that weave super, super tight so at least it was consistent, even if it was lossy. So, I feel like we've been going at this from a lot of angles - but hearing from Chris Hunrath, at Insulectro, it sounds like some people are really moving towards that spread glass and getting some good results.

Oh, that's my choice, but if you were to get the PCI Express Design Guide from Intel, they would tell you one of two things: you route all the signals at fifteen degrees through the X and Y-axis, or route them X and Y, and then you have the fabricator rotate the artwork 15 degrees on the panel.

Yeah.

That's in the standard for PCI Express.

I had no idea that was in the standard. I've heard about it, but that being - kind of anecdotally, but I didn't know it was actually written in the standard.

Yup it's in the standard from Intel, and you can - if you imagine a backplane where you have a regular array of pins for connectors and so forth - there is no way to route it at a 15 degree angle, because it's constrained by the pin array that's X and Y - so that's not a choice. So that leaves you with only the choice of popping the thing at an angle on the panel and then - you've been around enough to know the fab shop's gonna look at you like you have lost your mind.

Yeah.

Well we're not doing that, we're not doing that. We found some weaves that we know are well enough controlled that we're succeeding without that.

Okay good, that's good news. So this spread glass - so that's helping with the skew, you're saying, among other things?

It solves the problem.

It does?

Yeah.

That's amazing.

The rub is, you've got to be very careful who the weaver is.

Well can't you spec in a certain -

- What if I tell you that I had two weavers with the same stuff? That's the problem. When you say 33:13; there's no standard, that just means there's X in this direction Y in that direction, that's all that means.

Are there are prepreg providers that are - it sounds like there's prepreg providers that are doing it the way that you prefer. Or maybe other high speed...

There are, there are.

So do tell or can you tell?

Well the Doosan material I mentioned to you before we started this, is one of them.

Okay.

And a couple of Isola materials are okay but nothing else is.

That's a good hint, it's a good hint, look I mean I'm doing this podcast hoping to have a takeaway - so I don't just bring up all the problems and then say have a nice day thank you for sharing that. Right that's the goal like: yeah I found out what works, good luck! No.

So I have been seeing this word and this thing skews, so thanks for sharing that. One thing I could see as a potential problem - and tell me if I'm right or not - since that speed curve has risen so acutely, it seems like, the people who weren't previously doing high-speed design must be getting pulled into that space whether they want to go there or not right ? That - would that be a correct statement?

That's true yes.

When you and I were talking before this call about - let's talk about resources. About where these designers that are coming into this space I mean - speak just a moment about DesignCon, cuz I know you're pretty passionate about that show, and particularly giving out really good information?

DesignCon is the only conference I know where the level of information you need in this area exists. It's where everyone who has done research, or has studies and that sort of thing that are advancing the state of the art, that's where the papers get presented. It used to be - that was it. There were no tutorials, no education, that sort of thing. But over the last four or five years, we've added several things. This year we had three all-day boot camps on topics that matter to people who are trying to get on top of things. I did one title ‘Getting to 32 gigabits per second’ which dealt with all these topics. Intel did a three-hour on 'what is this PCI Express and what do you have to worry about' for people who have not seen it before and if it was five years ago, you might see there was nothing there for a board designer.

Now you would say it's the place you go for a board designer. There was a time when the PCB West was, but that has - I've been keeping track of that for a while - not offering the kinds of things you need for the these topics and I'm not sure why. I certainly have been talking to people who run it saying that you've got to offer tutorials, you used to do that. We used to offer stuff for engineers and they quit doing that because well, the guy who was running it was a board designer and he considered design an art, and their art dropped the stuff that appealed to engineering. Now, that stuff has to be learned by the designers.

It does, and as an old board designer person, I had to learn it from the board manufacturing side because I didn't realize - because I had left the industry for a while and come back, that things had sped up so much, that board designers all of a sudden weren't just dealing with: oh here's the specs, just adhere to the tolerances, do what the documentation says and have a nice day. There wasn't like now, high-speed board designers have to think about performance and all this wacky stuff. I mean the way we clean the board, the way we etch the board, the way we drill the board, everything can in a positive or negative way, affect the performance and it mortified me to think we got to a space where we could be completely IPC compliant and the board wouldn't perform as expected.

Yeah that's a good thing to observe and that is standards. By definition standards document the past by definition, and there are no standards group I can think of right now that has more behind the curve than IPC is, because they...

Why do you think that is? I have my suspicions, you're probably right actually, and you're more of an authority than me. But I'm just wondering why? What's caused that to happen?

Well who's driving it? Well, volunteers are driving it. Where do they come from? Well when IPC was at its prime, the standards committees were all staffed by engineers from aerospace companies.

Yeah true.

And the quality of the work was superb.

It's very true.

That's not true anymore - not true anymore, who goes to the IPC now? I don't, nothing there for me...

Why do you think the committee's aren't run by aerospace engineers anymore, or the Intels of the world, or Ciscos? Why why do you think it's not?

I wish I knew.

Yeah I really don't know either, I thought maybe you'd have some...

In the aerospace, that part of it, the aerospace contractors got out of the standards business and remember Jimmy Carter had a thing, The Commercial off-the-shelf Masters what... at any rate we're going.

Yeah - faster, cheaper...

Yes and the standards bodies that aerospace had, lost their money for example.

So they lost their funding to focus on that kind of thing.

I can see you probably heard of that. You know the last time it was updated? 1998.

[Laughter]

How outdated do you think that is?

And the update was to correct some spelling.

[laughter]

Oh my gosh, that's just plain sad, but they've come out with AS9100 and other things to replace it. I don't know the quality of those specs.

Well aerospace is on average about 15 years behind the industry now, it used to be the other way around.

Yeah, it's just really sad but I think that speaks to a lot of the way that politics have been run unfortunately, and the way things are getting funded.

Yes aerospace doesn't drive technology anymore.

Yeah, that's a crazy thing to think - to say out loud - I don't know, for you and I who have been around a little while. So, before we get too far off track, so DesignCon is definitely one place. Now because Design Con, if you pay to go to DesignCon, you can get all the proceedings. What can you do if you didn't go to DesignCon?

Well you can, for a hundred bucks, buy all the proceedings for a given year.

So someone could go on their website right now and for a hundred bucks buy...

All the papers that were presented this year correct.

That's amazing. Okay well that's a really good resource. Obviously your book which - because we were gonna talk - I went on your website and I noticed you're having some kind of fire sale on - I don't know if it's part one or something - but 'Right the First Time' - it looked like you were selling it and then it sounds like you now have a digital library of things you've published over the years?

That's true, and back on the topic of the books. We have two volumes and there was really gonna be one but - I don't know if you've written books or not - but you start out with great enthusiasm and this long list of topics. Then I had a deadline which was a Berkeley class, and I was only halfway through my list. And so I said: okay, this is volume one, next year we'll do volume two and what you probably don't know is, our books are printed in color. You have to have that in order to illustrate a lot of the things that matter. No technical publisher will publish in color, none of them.

Really, I didn't know about that I know it's expensive, but I didn't know that they wouldn't do it in color.

Yeah, so we were so focused, we formed our own publishing company and we went to a printer and said: we want to print this book. You see, there's a little secret about publishing.

Bonus material!

You have to give them a check for all the books before they turn the press on.

Oh boy that's expensive.

So for each of these two lines I wrote a check for $50,000.

Wow!

Crossed my fingers that someone would buy them, otherwise my garage would be full of books. Well, we sold out of volume 1 and volume 2 came along it sold faster than volume 1 did because it was a pull from volume 1. I am just debating, do I want to write another check for like $50,000 to get to print more volume one's, and the answer is no. So it became an eBook .

Oh ok.

So if you go on the website, when you buy you get them both, one's an e-book and the other is a hard copy.

Ok smart.

Then 200 books from now they'll both be eBook.

[laughter]

You got smarter.

My garage will be empty

[laughter].

I thought it essential that the books be in color.

That makes sense because some of those diagrams you can't distinguish between certain things without color being present.

No the color's for this industry. And so we are publishers and we sell our own books and people pay. I may be where I need to write volume through 3 and there needs to be a volume 3, to cover the things you and I just talked about. But I have told my friends - if I start talking about writing another book - their job is to slap me around until I get rid of the idea.

[laughter]

Well I'm not gonna slap you because this speed curve's going ahead now, we need to learn about skew and stuff. So - and who else Lee Ritchey, is gonna write that book?

I have actually written another book, working newsletters and articles that are on that - and someone should put them together but it's not me.

You know, we should get our friend Barry Matties to do that for you. Maybe he will.

You saw my distraction behind me which is the model railroad and that's more important.

Okay well I'm going to ask you about that, so I'm putting a hold on that subject, but let's cover a couple more quick things. One: you and I had an interesting conversation about circuit board manufacturers that are capable of doing good high-speed work and I made a comment and you corrected me right away because I said: well you know, board manufacturing hasn't changed that much in North America over ... blah, blah, blah. And you said, no that's not true. And I said except for places like TTM and you said which TTM?

Yes.

So tell me what you meant by that and fill our listeners in you know, of that conversation you and I had, cuz I thought it was very valuable actually.

Well when you say TTM, you're really talking about a dozen or so fab shops, which were acquired one by one. All have different capabilities for different markets, and if you're not careful and you get a bid from some - say TTM - they'll choose the fab shop that has the most capacity at the moment, which may not have the skills you need. And so I learned the hard way, you have to know what their capabilities are and when you give them the order, you say what plant is allowed to build the board.

And they'll accept that request and send it to the best location?

They won't get my business if they don't.

I am writing the check you will send it here.

Yes, yes and if you don't do that you won't get paid.

Yeah that totally makes sense. Well you and I talked about Stafford Springs which is a board shop, one of TTM's facilities I've always wanted to go through and it sounds like you've been through and that sounds like at least one of the locations that is capable of doing those high-speed designs?

That's right, and another one is in Hillsboro Oregon. It was originally called Merricks.

Oh yeah I remember Merricks.

Yeah and of course if I'm building small volumes it's my - my choice is a little place down in Orange called MEI.

It's not MEI anymore.

No it's got another name.

It's Summit Technologies which I agree they - I know Jerry Partita there - I hope to have him on this podcast actually because I think they've done a real good job there.

Yeah so you're sort of testing out something that we all say explicitly. If you're designing for this kind of space you need to be in direct conversation with the engineer at the fab shop, so that you don't make decisions that are not realistic. So I always have got that guy at the other end when I'm designing a new board.

I think that comes up again and again on this podcast by the way, people saying you have to talk to the key people at the board house which I totally agree with.

So let's talk about - I used to blog on Microwave Journal about what shops I thought - because I was working for Transline Technology, which is a really small board house in Orange County but they're quite good at RF and microwave. But I used to try, because I would see board houses say, sure yeah we do high-speed or the microwave because they've been built on say Rogers 4350 which processes much like FR4 -

- not really  

- well not exactly but close. It's pretty stable but then they would take an order for something that was PTFE because they built some 4350, and then they'd fail and say, sorry we tried to build three times and we failed. And I kept seeing this happen over and over again. So I started writing about what I thought people should know, what they should look for in a good board manufacturer, i.e. what percentage of their work is high speed.

So let's stop for just a second and clarify what high speed means.

Okay.

Because RF & microwave is not high speed with respect to the digital world.

Yes that's true.

Actually RF and microwave were simple compared to digital boards. I consider them to be trivial.

Yeah but some of the stuff you do on those for microwave boards are funky and weird.

That's not because they need to be like that. That's because RF engineers say things that are goofy

[laughter]

It is true because again why I started blogging is because RF engineers were starting to lay out their own boards which was not a good model.

Yeah, remember I'm an RF Engineer, and so most of the stuff you see people asking for on those RF and microwave boards is goofy; it's not good engineering, pure and simple. At any rate, so when we're talking about high-speed I'm talking about, cuz this is far away the majority, of things that must be digital.

Yes high speed digital, yeah.

Yeah and the people who can do that are good at laminating high layer type boards, and very rare in our microwave or high layer board count - almost never.

- Yeah this is true.

It is two different animals.

So if a guy says I'm a high speed fabricator and I make RF microwave boards, that's a different capability than what I need.

Yes it's true. So let's just talk about high-speed digital so you're right, somebody who can do high layer counts what would you look for laundry-list wise?

Well at the top of the list, is boards like mine. If you're not making boards like mine, you're gonna lean on me, and I'm not ready for that.

Yeah good point.

Yeah it really is, that's my first thing, you're making boards with the same class that I want you to builm, if the answer is no I'm gonna go someplace else.

And I would say, and how much of their work is like your work because if it's a really small percentage that would make me nervous too.

Yeah but it's my experience that if somebody's making high count boards, that's about all they're making.

Yeah it's true, they kind of - they kind of aim at that. And they're good at it, they're busy.

Yeah and they're busy and that's what makes them profitable actually. So and that also you will see in the equipment set.

There you go, exactly - exactly.

You won't see archaic… you'll see the most modern tools that gives them the precision and...

Yeah exactly, a really good vacuum lamination guy can supply and will build a twenty four layer board with ten mill vias and 12:1 aspect ratios.

Yeah. And so you've gotta go find fabricators who are in your sweet spot.

Yeah I agree with that.

And so - and of course - if you're in it for a little while you figure out there's about six choices in the US.

Yeah there's - it's true - there's not a lot and as you're speaking those are coming to mind I won't sell for anybody today but they are relatively easy to find.

Now - and now just before we finish that - six years ago, there were none in Asia because they were busy making consumer electronics.

Yeah I'm interested in your perspective on that, because I don't really know what that is - if it wasn't kind of cookie cutter consumer. So what's the state of the ability in China these days?

We've got as good a capability in Korea and China as there is in the US.

I guess that's good news from a price point standpoint?

Yeah and it's bad news for the American laminators though.

I know I'm an American and I feel whatever but we gotta - - it is what it is.

Don't forget not so long ago we made TVs.

I know we made a lot of stuff.

It's the nature of the business that we go where the low-cost labor is.

Yes that's true.

Okay, now back to your trains. So always at the end of the podcast here, which we are beginning to wrap up here, and thank you so much for your time. I always learn a lot from you Lee. You are working on - this part of the podcast I call 'designers after hours' because my observation has been, a lot of people that I know that are pretty smart engineers and designers, have neat hobbies after hours. So tell us about your after-hours fun?

Well the one that you can see in the background - there's a lot of railroad.

Let's see it, can you flip your screen that way without disrupting us?

Oh there it is.

Can you see it?

Yip I sure can.

And the one you can't see is I repair vacuum tube radios.

You do what?

Old, radios vacuum tube radios.

No way.

Yeah.

Just for fun - who still uses vacuum tube radios?

Well we're starting to get Wi-Fi guys who think that vacuum tubes are the way to go. But I started out making my spending money as a kid, fixing the radios. I've always been in there you just may know...

I know.

-so I'm just back to fixing old radios.

That's kind of a fun hobby, but vacuum tubes - like that cracks me up.

Yeah you probably didn't know you can buy new vacuum tubes?

I didn't - I didn't but I was talking to someone recently and I - it cracked me up because I remember doing this as a kid so, is - remember going to like the drugstore - the hardware store and you could test the vacuum tubes?

Yeah, come on over I've got a tester.

No way - somebody mentioned that

- I was like whoa! Like my childhood came flooding back going with my dad to the store and sticking in the vacuum tube testers that's funny.

[laughter]

Yeah, good times - good times.

Well we're about out of time Lee, so thank you again for your time. And I know we could go further and further but I will share the link to your information about your books, and I will also share the link you shared with me for the design concept people that download white papers, and if you think of anything else just let me know and we'll put it on the show notes here. So we don't leave people feeling hopeless. We get them registered for DesignCon and get some papers in their hand, and get some books in their hands so they could do their job better.

One last thing we talked about I oughta mention, if you can - if you do those - what do you call the thing we did in Munich?

Oh AltiumLIve.

Oh yeah if you keep doing those I would argue you should start offering these training courses.

Okay, all right you heard it here from the mouth of Lee Ritchey.

I have to go show this to the CEO later - proof!

It's been my pleasure.

Thanks so much Lee. And we will talk to you soon again. This has been Judy Warner with the OnTrack podcast. Thank you for joining us and thank you Lee Ritchey, have a good day. Have a good day. Bye.


 

 

Disclaimer: We respect the unique perspectives of all of our OnTrack podcasts guests. Therefore, we choose to offer their uncensored opinions in favor of full transparency. However, all opinions expressed are exclusively those of our guests and do not reflect the views of Altium or our employees.

 
Previous Article
John Watson, Senior PCB Designer at Legrand North America on Altium Nexus
John Watson, Senior PCB Designer at Legrand North America on Altium Nexus

Curious about Altium Nexus? Find out what the buzz is all about.

Next Article
What is PCB design? With Kelly Dack and Judy Warner
What is PCB design? With Kelly Dack and Judy Warner

What is design? Listen to the Altium Podcast featuring Judy Warner, today!