Overview: Charles Pfeil and Judy take a look back on the days of manual design, before exploring today’s very different eCAD competitive landscape and the tools that take designers to the next level. Join Altium’s Judy Warner and Charles Pfeil for a conversation on PCB design, autorouters and the future of software design tools.
- Introducing the iconic and award-winning Charles Pfeil
- Tape ups and touch ups with red pen precision
- Artistry not OCD, PCB design for form and function
- Charles on joining Altium and Activeroute
- “You know we’re paid by the hour, right?!”
- Where are we going to be in 10 years?
Links and Resources:
Charles Pfeil and Happy Holden at AltiumLive
See all of Charles Pfeil’s articles on EDN here.
Hey everybody, Judy Warner here, Director of Community Engagement for Altium. Welcome to the OnTrack Podcast. If this is your first
time listening, we're glad to have you. If you're returning, we're happy to have you back.
The OnTrack podcast is produced every week, so please add the OnTrack podcast to your favorite RSS feeds or iTunes. You can also follow me on LinkedIn and Twitter @AltiumJudy, and be sure to follow Altium on LinkedIn, Twitter, and Facebook.
So let's get into the show. Well, I have a treat for you today. I have a guest, Charles Pfeil, who is, he won't say it but I will, he's iconic in the industry and we're gonna talk a little bit about the history of routing.
So, I want to tell you a little bit about Charles, forgive me for reading here for a moment.
Charles has been in the PCB industry for over 50 years as a designer and owner of a service bureau and has worked in marketing and engineering management at Racal Redac, ASI, Cadence, PADS, VeriBest, and Mentor Graphics, so pretty much everybody, and now Altium. He was inducted into the PCB Design Hall of Fame through UP Media in 2013 for his contribution to PCB design software.
Charles has been working at Altium since 2015 as a Senior Product Manager. So, welcome Charles, we always have a blast when we talk together, so thanks for doing this adventurous podcast with me.
You're welcome, glad to be here.
So, you are on the other side of the country from me right now.
I'm here in La Jolla as you know and I get to see you sometimes once a month, and you and I have had some interesting conversations here in the office, kind of a shared belief that we're sort of losing some of the history, right? Of our industry, particularly design, sort of where we've come from and where we're going, and I really value people like you, Happy Holden, Lee Ritchie, these people in the industry really that carry this whole history inside of them, so I'm always happy to have a chance to sort of glean some of that from you and you've told me some great stories, so thank you again for doing this.
So, how I met Charles was when I was a journalist from iConnect 007 and I interviewed him to learn about ActiveRoute, which was a new product for Altium at the time, and just sort of casually at the end of the conversation I mentioned to Charles "So, what's your history?" not knowing who Charles was and then, out he comes with the history of PCB design.
So yeah. I felt like a knucklehead, but it ended up being a good article, so you were a good sport.
So let's start talking about where at least you and I remember things which is tape-up. So, tell us about your earliest memories of doing tape-ups and what that was like and sort of what some of the problems were.
Okay, well my first job, I worked inspecting Rubyliths, and at that time the way that people routed boards was to essentially cut away the things that they didn't want, so what was left was the routing.
So there's a process in which they used an exacto knife and my job was to look for scratches on the Rubylith and patch them up, and I would have to say from that to what we do today is a tremendous change, and it went through an evolution where for manual design that the methods changed, the materials changed, and also the complexity of the design changed, and the same thing has happened once we started working with computers to design, right?
The materials change in the sense of the board materials. The methods change in that there's constantly an evolution of how much the software is able to do to help the designer, but just a note on the manual design and where this term 'artwork' came from. That, the interconnect would be done by drawing it, right?
Usually with red and blue pencils and most of the designs are two layers, and you would try and interconnect them and, you know, erase a line if it doesn't work out, but generally, you know, the type of components back then were DIP high C's if you had any at all, and it was, it was fairly simple and you could draw with red and blue pencils what you wanted and then you would either use black tape or red and blue tape in order to create an artwork that would be used in fabrication in order to get the design made.
Well, there were many different styles when you get to the artwork side. Some people like to use the black tape and create curves and arcs on everything, and some would not care about concentric arts, others, like me, would be very OCD and want to make every single corner perfect and make it look like some kind of art. It's not really art, but it's just the implication is that the person who's doing it is really caring about how it looks.
Absolutely, which by the way, one question I want ask you I know the answer to but I think the listeners will appreciate, what were you touching up that Rubylith with? Just a red pen.
Yeah, a red pen. I would just, because I had a magnifying glass to find the scratches 'cause with an exacto knife they would be really small, you know nicks.
And, so I'd go through looking at it with a magnifying glass and then use a red pen to fill it in so that there's no place with the light to shine through, because the way they were doing it back then, if enough light shined through then there'd be a hole in the mask for etching the board. Yep. Which could, what would that-
Well, you'd have a- -well, it could have an open, well you'll have an open if the scratch went all the way across, right.
Yeah. Yeah, yeah. So anyways, continue on. I just-
Yeah. It's okay.
-it's just a funny thing, like, when I think of how much precision we have now, to think of us touching up artwork with red pens just cracks me up.
Yeah, yeah, and when using black tape I think it was common to start out with arcs and you'd do that by kind of using the knife to create a little cinch in the corner, but eventually arcs became the less productive method, let's put it that way, and so you would start making 45-degree corners by just cutting maybe three quarters of the tape and folding it over to create a corner, and so then the art was to make sure that everything was right on the grid, and when you did a 45 it'd be perfectly 45, and it got kind of excessive, but when computers came.
That's where it changed, and I do have an experience in that area when my service bureau switched to computer design as opposed to manual design.
So you were right in that transition then, you went on your service bureau right from Rubylith black tape right into computer-aided design.
Yeah, from tape-ups in 1978, from tape-ups to computer design, and that was a Racal-Redec system, and I lost all my customers because the only customers who wanted the computer design were people who already had it themselves. Most people who didn't have computer systems, they didn't trust it. We would give them plotted artwork and they couldn't edit it, so as a service bureau, I had to go out and find customers who not only used computers but Racal-Redec computers.
So was Racal-Redec's price point, was there a barrier to entry there, kind of for that kind of system for OEMs?
Sure. So, I bought two used systems, you know with a basic vector graphics monitor, PDP1134 with the software. I bought two used systems for, I think, my memory is not great on this, but I think it was around $125,000.
Holy cow, that's a lot.
It was a lot. For used.
It was an investment and it paid off over time. I did increase my price per IC by a factor of 10.
-which helps, you know, because you gotta pay for the apartment, but that's what happened in the industry.
So, routing in a computer system was different because with the computer of course you're effectively laying down the edge as you route as opposed to going through a stage and sketching the, you know, red and blue pencils
-but, so interactive routing was really, click-click click-click at every corner, okay? And if you wanted to make a change then you would either delete those and reroute it or some more sophisticated systems allows you to move the track around, and it was either that or autorouting and at that time autorouting, well it hasn't changed much but autorouting was primarily just an X&Y; exercise and adding a lot of vias. In a sense it was similar to the way that I see routing. I see routing doesn't care about vias and they just, whenever you got to make a change in direction you just pop a via and go, and with printed circuit board autorouters that's what they would do. They didn't care about how many vias you added and really vias, the only problem with vias back then was the cost of drilling them.
Right. At that time there was, wasn't.
Yeah, it was nominal.
It, yeah, well the cost was still there, but in terms of vias affecting the high-speed behavior-
Right, uh-huh. -there weren't high-speed designs. There were companies of course who were doing that, but vias didn't become a cost for signaling performance.
Right. So, people still didn't like vias. They could see with their mind that, "Hey I could route that without vias" so they would rather just go ahead and route it without vias, and so even though autorouters, every company tried to come out with the ultimate autorouter, and it could never match what the designer was thinking or what the designer really wanted as a final result, and I'd say generally that's been the problem with autorouting.
Too many vias, not really ending up with a route path that the designer had in his mind, and also the most common thing I hear is that it would take more time to clean it up than it would to just route it manually. Now I remember back in my early fabrication days knowing designers that would autoroute a portion, say 60% of the board they would autoroute, and then they'd do all the rest by hand.
Was that because of what you're talking about or was that just the speed to design at the time?
That would be a method. There's so many different ways of approaching it, but a method would be to route the critical signals first manually, and then just give it to the autorouter to finish everything else, because that would save time.
Now, autorouters evolved. The early autorouters really didn't care about anything but is there an open channel and is there space for a via, right? Then it evolved over time where there is a recognition that, well what can we do to try and eliminate vias, and I say eliminate because they would route with vias and then they'd have a pass to remove vias if they can, right? So, but it really wasn't fundamentally in the algorithm of the router to try and route without vias. Now, a number of autorouters gave the designer control, like CCT with a DO file where you could, you could put costs on the router to try and minimize vias, but it was difficult.
You really had to be an expert and I went to a number of companies where they had one person who was the router specialist and the designers would place the board and then send it to the router specialist who would do the automatic routing and that person would know how to set it up and it would be pretty good, but when it came back there still had to be a quite a bit of editing.
Who was, who created the best sort of autorouter of that time, and what kind of time frame are we talking about now, as we're moving forward?
Yeah, I would say, well it depends on who you talk to, okay, because-
Everybody has their favorites, right?
-everybody has their favorite.
So I would say though, obviously the Cooper Chan Technology, CCT. Their autorouter was the favorite, especially among Mentor, Cadence, PADS. Those companies all had interfaces to it and relied upon to do their autorouting, but once CCT was bought by Cadence for a measly four hundred and thirty or four hundred and fifty million dollars- Whoa. -whoa. All the other companies, Mentor, Very Best, and also PADS felt "whoops we can't go down this road" because, as usually happens in a competitive environment, those who have a relationship with one of their competitors, it's possible that that relationship could diminish over time.
Right of course, and we have to remind listeners right now that at that time PADS was not part of Mentor. Correct.
Correct. Neither was Very Best.
Very Best, right, so it was a whole different environment. So Very Best, they had autorouting capability and it was different. It was, it had some different data structures and different algorithms and actually that was one of the primary reasons why Mentor bought Very Best, was for that autorouting capability. Oh, okay interesting. See, I love, I mean, people that are around our ages know this but so many now that are, I don't know maybe 40 and under don't. You know that how, this used to be a, there used to be more than three or four players in this space, right?
Sure, sure. And we kind of all, ate each other up until we have what we have today. I would say the 90's and early 2000's there were so many mergers and acquisitions.
Yeah. Unbelievable-Unbelievable. -and you know, I went from Intergraph to PADS to Very Best to Mentor, right? And then now I'm at Altium, and looking at the evolution and the tools over that time the autorouter was a key thing, but realistically designers still preferred to route it just by themselves and manually, right?
It's true. It's, it really has a lot to do with what the designer thinks about when they're placing the board.
They know that a certain bus needs to go in a certain direction on a certain layer, and to set up those kind of controls and cost adjustments for an autorouter is a lot of work and it requires a lot of knowledge about the core capabilities. So, I think routing in general after, yeah after CCT was bought started turning in a different direction where rather than just autorouting, let's see what we can do to give the designer the kind of control with interactive routing that they want, and increase their performance.
What can we do to increase the performance?
Right, and is, tell me again now is this around the time that you joined Mentor's team?
Yeah, I joined Mentor in 1998.
Okay, so late 90's.
Yeah, it was-
And you helped, and you helped architect that vision right?
Yeah, I was part of it. I mean-
Part of it right, it's a team.
-it's always a lot of people involved, and I was involved in it, and I always tried to put my designer perspective into it, right? And I think that's my contribution, is not only being able to work with the development people but to be able to communicate the knowledge and expectation of a designer so that the software that we develop is something that a designer would appreciate.
Right, you had a designer-
Not that.you had a designer brain so you knew, which by the way, as you, as you alluded to in the beginning, is an artistic brain, really. It's, there's a lot of creativity there and an appreciation for not only function but also aesthetics and to put things in really neat, tight patterns, or.
Yeah, you call it OCD but there's that.
-there is that.
You know, this is a topic that has come up quite a bit in the last 10 years and I like to say the following, that yeah there's that artistic side but ultimately I believe it's about efficiency. And in the context of CAD design, efficiency means that the routing is as direct as required, of course if you're doing some kind of, you know, tuning on it it's not going to be direct, you're adding lights to it, but you want it to be as direct as possible. You want to manage many more things than we had to manage, you know back in the 60's and 70's, right?
Now you have to manage high-speed requirements, so you don't necessarily want to have everything tightly packed in parallel. I remember I was demonstrating some ActiveRoute capabilities to a high-speed product manager, probably a little over a year ago.
And there was a bug in the routing where it was, it was kind of messy, you know and the high-speed engineer says, "Oh that's great!" Right? Because it would totally eliminate cross-talk, you know? Exactly.
Yeah, so there has to be the balance and that fundamentally, that is the job of the designer.
How do you figure out the compromises, right, to achieve the right balance where you fulfill manufacturing, you fulfill high-speed requirements, and ultimately the design works, right?
And if it comes out looking where everything is nice and neat, that's a bonus because the next person who has to make edits on it - it'll be much easier, right? Right.
And yeah, frankly most of time or many times if it's not a high-speed net, those geometries don't matter so much, but if it doesn't take any extra time to make it nice and neat what's the problem with it?
Yeah, I get what you're saying, yeah.
There's sort of some elegance about the design, and like you said, many boards go into respins and you're not always doing that on your own work, so that totally makes sense.
So, you joined Altium what two years ago about, approximately?
Yeah, in May of 2016, excuse me, 2015. 2015, and so you've been working on ActiveRoute here, so what, you know, you've always throughout your whole career been pushing for trying to make things better, what are you trying to accomplish sort of on the work you're working on now, on the ActiveRoute?
Yeah, so it is one of a number of things that I have worked on, but I would say most of my time has been spent on ActiveRoute, and the intent of ActiveRoute is to give the designer another tool in their interactive tool set, okay? It's not for everything, but the intent is that when you have a group of signals that you want to route together, whether it be, you know, a bus or just individual nets that start and end in approximately the same place, I want to give the designer the ability to just, through a couple simple clicks, have those routes the way that he's thinking, okay? And what is the designer thinking? That's the key. But I would say that, you know, it, with most boards is pretty clear. You see the connections, that's where they start and end, and I'm looking for a tool that can allow that designer to just select those connections and have it route on the layers that he wants them on, ability to spread them out, ability to pack them together. These little techniques are there as options so that ultimately it looks like he would do manually. It's not a replacement for interactive routing, it's instead an accelerator.
Oh, okay. By the way, along the lines of accelerator, I was talking to somebody, I have to share this story with you, I was talking to a designer and he was telling me that he was sitting in your session at AltiumLive where I think you were showing ActiveRoute, and he said someone in the room said, "You know we're paid by the hour, right?" because it was making it more efficient, so I think that's a compliment to you Charles, to the work you're doing.
There is that trade-off, but that's been there forever, that designers fear being replaced by the computer, right?
Right, so that's gonna be my next question, just, so get ready about, to talk to me about AI and designers being replaced.
Okay, well, we can go there now.
Okay, you ready?
Yeah, I'm ready.
So, I know there is this fear or I've heard there's fears of designers being replaced by AI or computers and all that. So, what say you, Charles Pfeil?
Well, I say that it's going to be a while, all right? If you can imagine all the energy and all the time and talent that has been focused on auto routing over the years since, you know you have companies like ASI in the 60's, late 60's had automatic routing for IBM, okay on mainframes-
Oh I didn't know that.
- and, you know, programming it with punch cards. So, from there until now the ability to get autorouting to succeed has failed all right?
And this is just-
Or has it gotten us there since the 1960's, right?
We're still not, you know. It's not there, and generally designers have rejected it and it's not because they're afraid they're going to lose their job.
And, besides that, like I say usually, you don't sell these tools directly to the designer, you sell it to their manager, and if they can get the job done faster they're all for it okay?
But the problem is that it hasn't been adequate, the autorouting capability. It's a tough, tough problem. Every design is different, right? It's different.
And- And it's very, and because of that it's such a complex brain task, really.
Right, and there are standard circuits, of course you know, standard interfaces and memory circuits that need to be routed the same way, but then the placement's different, the board size is bigger or smaller or a different aspect ratio or it's flex or, you know, could be a lot of different variables.
There's many, many more variables right now, too. So, to fully automate that is going to take a completely different perspective for design I believe, and it's going to take a huge investment, and the interesting thing is okay in ten years, let’s suppose somebody started on-
You know, are you reading my mind? My next question was gonna be where are we gonna be in ten years Charles? So-
Yeah, exactly. Let's suppose somebody started a project to use AI and that's in the, in the context of doing machine learning and analyzing databases and building decision trees, and even if you got that working over five of those ten years, a lot of the things will change. They'll evolve and this is one reason why software vendors can't keep up with the technology.
There's always something new, always something new and on our list of enhancements to work on, it's always the latest technology, right?
And there's only so much we can do so we have to pick and choose which technology we're gonna focus on with each release.
Well, and, I was just talking to Ben Jordan here about this.
Well, until I came here, since I came from the board and assembly side, I had no real appreciation for the time and discipline it takes for development and for coding that software.
It takes a lot of time and a lot of discipline to put in all these new features and we're so, sort of instant gratification, we're just like "oh, well, put it in your next release." Not so easy.
Yeah, it's just a fact of life, and our customers, they want these things and we want to give them to them, but it has to be an evolution.
So if a company decides okay we're going to use AI to automate the design process, sure they could analyze all the different types of designs and understand them and how to use them, but then when they start with a new design, how could an AI decide on its own what the board size should be? What the stack up should be? Whether or not to use micro vias? What are the critical signals? I guess it would have to be able to read a data sheet, that is hard enough for humans to read.
That aren't always accurate, may I add.
Yeah, that's true.
I mean, most of the time they are, but. I'm not saying it's not possible, it certainly is possible. But, it's going to be a long time.
And maybe, who knows, maybe MCM's will come back and we won't have printed circuit boards. There's always that possibility. Yeah, yeah.
We just don't know, and the, even, you know, printing circuit boards - where is that technology going to be in 10 years?
Exactly, yeah. 3D printing and there's, gosh, so much that they're grappling with, so who knows?
Yeah, and the core problem is that a company is not going to design a product, let me rephrase that, a innovative company is not going to design a product that uses all these standards and circuit technology of the past. They have to have something new and different.
Right. This is why Intel stays in business, and all the other chip manufacturers, because you know they need higher performance, they need more capabilities, and so it's a complete redesign except for like I say standard interface is a memory, that's not a redesign, until the next standard comes out, but those tend to last a couple years and then, you know, you look at memory DDR 2, 3 & 4 they just change over the years.
And the interesting thing about, I wanted to comment here, about memory in terms of the PCB designer, the DDR2 circuit was the most difficult to route because of the T branching. Then the DDR3 came with the fly-by signal technology, that was much easier for the designer to work with as long as you have some kind of electrical signals, and DDR4 is really just a minor evolution, significant in terms of performance over the DDR3, so who knows where that's going next. Will the next phase be like the change between DDR2 and 3? I don't know. They have the specs for DDR5 and I'm just unfamiliar with them. But, so technology will continue to change.
Maybe AI will be able to support a segment at a time, and I know that there's people in the industry who are talking about this and wanting to do something with it, but I'm unaware of a officially funded project to do it at this time.
I'm with you there, and I would think you would hear about it from your connections, I don't know. Maybe, we don't know. And would a company you know like Cadence or Altium or Mentor fund a project that they know is going 2 to 3 years and maybe, maybe not work?
It's a risk. A potential risk, so. Yeah, so I would say to all the designers who are concerned about losing their job, don't worry.
But what will be extremely important is that they make sure that they continue to be educated and understand the latest technology and what requirements those bring to the design process. Then you have your value. Yep.
And you won't lose your job.
And I would say to that, that almost through the OnTrack newsletter and now podcast I hear that message over and over and over again from leaders saying, "the key to my success was I continued to learn and evolve" and so it's been great to talk with you Charles I always learn so much from you.
I wanted to ask you one last question. I have this part of the podcast I like to call "Designer After Hours".
I've just noticed a lot of designers have a creative bent and usually have some interesting hobby, so what are maybe one or two of your favorite hobbies?
Well, I could list three, all right?
One is playing golf. I love to play golf. Second would be, photography, taking pictures, mostly landscape pictures.You can see behind me a couple landscapes, but I like to say that I prefer landscapes because they don't move very much. I've tried to do a few weddings and it was a total disaster, so I'm just not that, but and then third I do write a lot and I've, you know, written poetry and I really enjoyed doing that, so those are the three things.
Speaking of writing, I wanted to let our listeners know that Charles has written, which I'm delighted about my passion for history being preserved, how many parts are we now up to, ten? How many are we gonna finish at?
Well, actually 8 of them have been published in EDN magazine.
EDN Magazine, right, so there's a whole series of Charles' experience over all these years and goes into depth, so if you want to learn more please go to EDN.com and probably just search by Charles Pfeil and you'll find that. How many are you gonna, I know eight have been published, so how many- It'll be, eight have been published, I think it'll be twelve or thirteen in the end.
Twelve total, okay. Okay, well thanks again Charles, we've gotta run, but I could, I'm never tired of talking to you. You're a dear and thank you so much for your contribution to the industry, and it's always great to talk to you.
Well, that's all we have for today. Please remember to subscribe to the OnTrack podcast and follow us on social media, and remember to always stay OnTrack.
Click here to go to the Altium Designer Free Trial and Download Today.