Russell Hoffman ("Host"), High Tech Today
Carolyn Spivak ("CS"), Chip Designer
Host: ...My guest today is Carolyn Spivak. She's a chip designer for a major company, and we're going to be talking about what a chip designer does, and, maybe, what you would want to learn if you want to get into the field of chip design and how the field has changed. Carolyn, welcome to the show!
CS: Thanks a lot, Russell.
Host: And you've been in the chip design industry for quite a while now, right?
CS: That's right, more than fifteen years.
Host: More than 15 years. You were telling me before we went on the air, that in 'the biz' people track how long you've been in the industry by how close the tolerances are. Why not start by telling us a little bit about that aspect of it, because I think everybody understands that chips are basically electronic circuits placed on silicon. So why not tell us a little bit about what you do, and what the tolerances were when YOU got into the industry.
CS: Ah, I'd be giving away my age if I did that!
Host: (Laughs) Only to those in the know!
CS: (Laughs) Right. I guess people have the general idea that as chip technologies get more advanced, the width of lines that you can make on chips that you use to build up the transistors and the circuits, get--the widths of those lines get smaller and smaller. And, we usually talk about them in terms of design rules.
When I started doing chip design, I was working in five micron design rule technology, which, let me say, is very very small, but right now, we've gone an order of magnitude smaller than that, so there's really just a huge increase in the amount of stuff that you can pack onto these circuits, over that time span.
Host: And you can actually put the, I guess, probably [something like] the square of the tolerance, in terms of the number of elements on a chip?
CS: Right, it gets interesting as you go down, because what determines what you can pack in is sort of a--it's a reasonably complex function of the different kinds of things that you're putting on a device. You may put down conductors which are basically just metal wires close to each other. Those have one kind of tolerance. You also want to put transistors down. And there are rules about design--how close you can design two transistors next to each other.
And so it's not always real straight forward. But basically, any time you can make finer design rules, the amount of circuitry, the amount of functionality that you can pack onto a piece of silicon the same size, really goes up dramatically.
Host: In the industry that you're in specifically, why not tell us a little bit about--you're doing Gate Arrays, I think you had said?
CS: Right. The group that I'm with does, actually sort of a sub-branch of a sub-branch of chip design. Probably the chips that most people hear about are standard product chips like microprocessors and memories and a lot of the stuff that goes into PC's and so forth. And, as you can imagine they sell, you know, companies who make those sell very large quantities of those, and they sell them to different users who use them in different kinds of ways. In other words, the same microprocessor or the same memory can go into different kinds of PC's or different kinds of hardware systems.
There's another whole area of chip design, which is called a kind of an ugly name, but it's called application-specific chip design, and there basically the idea is that you're designing a chip for a specific application. In other words somebody might have, oh, I don't know, any piece of hardware that you might think of. A modem or something, where they want to put--they want you to make a chip just for them, that they can put into that modem and that will do the job they need, but somebody else with a different kind of modem design might use a different kind of chip.
The kind of chip design that I'm involved in is sort of right in the middle, in that the chips we make can be programmed by the people who use them, to do different things. Now, in a way, microprocessors do that. Software's really a way to program microprocessors to do different kinds of things. But, the devices that we make, which are called Field Programmable Gate Arrays, basically allow users a greater level of control over what's going on on the silicon, than programming microprocessors with software does.
Host: More control? How is that?
CS: Sure. Because in the kind of chips that we make--in a microprocessor, the conductors are all laid down, the transistors are all laid down, and when you write software that gets compiled into code, you're controlling in a sense, what comes out of the device, but the paths that your signals take through the device, if you want think of it that way, can only go really in certain ways. The ways that the architect of the microprocessor designed it. Whereas, with the chips that we design, if you're the person who's using them, you can actually sit down on your workstation and kind of draw almost, what you want the interconnections on the device to look like, and how you want each little piece of logic on the chip to behave. That's really all under your control. And you then send the device the information that it needs to, really, program itself that way.
So it's a real different mind set than a microprocessor. A microprocessor--you write software, the architecture's already very strictly defined, and somebody's built up a compiler that lets you compile into that. But if your application is for something that microprocessors aren't so good at doing--and there are lots of those things--you're kind of out of luck. And, before Field Programmable Gate Arrays came along, your only choice at that point was to do traditional kinds of application specific chip design, and the problem with that is that it takes a really long time from the time that you tell you chip vendor exactly what you want, to the time that you get your chips back. That's a really long development process. And by putting the programming power into the hands of the end user, what Field Programmable Gate Arrays have really done is to--to really shorten up the design cycles for some of those things.
Host: Are there also, perhaps, speed advantages, um, to something that's designed with a field programmable gate array versus a CPU? Lower chip count, which would make it cheaper, perhaps it would run faster also, as well as being easier to develop?
CS: It really depends on the application. Certainly--it wouldn't be fair not to mention that what you pay for the kind of flexibility that a field programmable gate array gives you is speed as compared to an application specific circuit. But that doesn't mean that--you may not be able to achieve much higher speed than you would have with a microprocessor. It really depends on what your application is.
Host: ...[break]... My guest is Carolyn Spivak, a chip designer, who designs field programmable gate arrays, and we're talking about the technology of chip design, which is so important to every high tech device ever manufactured, and we're finding what this level of production is like. So Carolyn, why not tell us a little bit about what an average day is like for you. What does the process of chip design entail?
CS: All right. Well, there's really a whole spectrum of things that come into it, and it runs all the way from trying to figure out what customers want to see in new devices, and of course we have marketing folks who help us out with that, all the way through to the extreme other end of the cycle which is, keeping an eye on what's going on in our factory, and making sure that production of our devices is going smoothly, and everything is running the way it should in the factory, because there's a real--you know, even though we're called chip designers, our point is to make chips that our company can sell and that will work for our customers.
And so, even though we're called chip designers, a big piece of the design job is making sure that what we've designed is manufacturable! And I'd say, from the ten thousand foot viewpoint, really all our design activities after we figure out what we think customers want, are aimed at making sure that we can deliver devices that meet those needs. In other words, that we'll be able to make them in the factory. And the kinds of things that go on in the chip design, first a very high level view of what a customer wants. And then, maybe either a high level capture of that kind of functionality, or a logic design capture, kind of at a block level. There are fairly standard blocks that logic designers use. And for us, the circuits that we make are very very dense. Probably among the densest in the industry. And so a lot of effort goes into working with layout engineers to make sure that things are packed as tightly as they can be in the layout--but not too tightly! Because, again, any manufacturing process has a certain tolerance, and you can't go beyond that tolerance and expect to reliably make a product.
Host: Well, talking about reliability, have there been large changes--or small changes, or any changes in the reliability of chips since you've entered the industry, and how do you measure reliability? I mean, it seems to me as a user of computers all the time, they basically work. When one finally--if a card breaks in a computer you throw the card out and get a new card. How do you, at your level, and that would be I think where the rubber meets the road, or where the silicon meets the sand--in terms of reliability, how do you program, or how do you design for, how do you measure, whether or not you've achieved a certain level of reliability?
CS: It's--It's a good question. There are a lot of very formal ways to do it, because reliability is of course one of the things that equipment manufacturers worry about the most.
I guess the first thing I ought to say is that as chip technologies have allowed putting more functionality onto a single piece of silicon, reliability has increased hugely, because--the insides of chips are really, really reliable. What's unreliable is solder connections. So, as the amount of function that you've been able to put on a chip increases you have fewer and fewer solder connections in your system because more and more of that goes inside the device. And that's exactly why, you know, that there have been these kinds of advances. At the chip level we do standard, fairly boring things like heating devices up and letting them run for long periods of time, under very hot conditions, and making sure that they last a long time under those kinds of conditions.
The kind of reliability I was talking about is actually not for the end user, although that's real important, but for being able to make sure that the part--that the natural variations in production processes, don't cause any problems in the part. In other words, it's sort of this idea that you want your design to be real robust, so that when you put it in production--any production process has variation in it, and you need to make sure that your design takes that into account.
That kind of reliability is important not just from a cost point of view, but also it's important because customers depend on shipments, and if a semiconductor vendor has yield problems, what happens is that a customer doesn't get their part, so that's a real important part of the process for us.
Host: ...[break]... Why not tell us a little bit about what kind of a mind set it takes to get into this industry, and what are some good courses that someone might take, or what sort of interests people might have to begin with, that would be natural--lead them into something like chip design?
CS: Well there's a--like we were talking about during the break, there's really a bunch of different kinds of skill sets that people have in the industry. The classic chip designer is probably someone with an electrical engineering background. Most of the people that come into the field have either bachelor's or master's degrees in electrical engineering. And that involves a lot of math courses. For people who are young, I'd say the most critical thing is to take all the math courses you can, in junior high and in high school. I'm laughing a little because that's really hackney, but it's real important for people, because if you don't take those early math courses, you shut yourself off from a lot of engineering careers, not just electrical engineering.
So, there's sort of the classic electrical engineering path. But there are also a lot of people who specialize in other areas. One of them is layout. Where I work, we have people who are specialists in layout, and who might enjoy drafting kinds of things, or a lot of them come from a semi-graphics background. And then, some of the people who are design engineers--I'm hopeless at trying to arrange things dimensionally--that's just not a talent I have, so I really depend on people who have those kinds of skills, to do that piece of it.
There are also people who come in through programming kinds of backgrounds. And I'd say the general mind set--you have to enjoy debugging things. You have to--you have to really enjoy going in, and making things work, and finding out what's really wrong. The kind of people that we look for when we hire are ones who have real good problem solving skills, and a deep technical background--in some area. In a lot of cases now, there are pretty good college programs that include VLSI design.
Host: The equipment that you're working on is pretty expensive. You mentioned that there are really good college courses, but I would think that it must be quite a shock when you get out of college and get into the real world, because of the cost of the--chip fabricating takes a billion dollar company to do that.
CS: Right. Well, a lot of the schools have arrangements, there's actually a consortium, where students can make up their design using inexpensive or shareware kinds of CAD tools, and then send it off to [the consortium] and actually get parts fabricated. So, for some of the--actually a lot of the electrical engineering schools can actually send designs in and it's--of course it's good for students to be able to actually get stuff made. It's also hard to make a real chip in a semester, I guess! (Laughs)
But, even for students that don't have the opportunity to do that kind of thing, there's an awful lot of CAD programs floating around, and as PC's have gotten more and more powerful, you used to only be able to do CAD for integrated circuits on really expensive workstations. But now there's a lot of pretty good CAD programs that run on PC's and so that's certainly helped out in the academic environment a lot.
Host: Okay, and the last question: Where do you see this technology going, in the near term, say five years, and in the long term, say, twenty years? And if you want to tell me where you think it's going to be in 200 years, we're all ears!
CS: (Laughs) Well, in five years, I think it's basically going to be more of the same. In other words, the amount--things will continue, electronics will continue to get cheaper, we'll continue to be able to put more and more stuff onto less and less silicon, and that will translate for consumers into being able to buy less and less expensive versions of fancy electronic equipment.
Host: Which will be more powerful and more reliable.
CS: Which will be more powerful and more reliable. So that's kind of an easy one. The further viewpoint, there's a--there are always people who say, oh, in that kind of time frame, "all the technology will be optical" or in that kind of time frame, we'll get a dramatically different integrated circuit technology. And that's quite possible. You read some of these things about optical switching, optical computers and so forth, and we may, in that kind of time frame, we may be dealing with fairly different kinds of technology. But at the same time, you know, there have been a lot of technologies which are already kind of "just down the road" and I think that particularly the manufacturing experience we have in the current silicon technology is going to keep a lot of what consumers see in that technology, even in a fairly long term kind of view.
Host: Okay, well, I'd like to thank you very much for being on the show today. My guest has been Carolyn Spivak. She's a chip designer who designs field programmable gate arrays. And this has been a very interesting talk. It's interesting to hear what the lowest level of all the high tech devices, how they're built and what's going on. This has been High Tech Today with your host, Russell Hoffman...Thanks for listening. Bye Bye!
Please distribute this document IN ITS ENTIRETY (as a raw HTML file or printed document). Please link to it rather than placing it on another server. Thank you.
The Animated Software Company
Last modified March 27th, 1997.
Webwiz: Russell D. Hoffman
Copyright (c) Russell D. Hoffman