Pang: This is a second interview with Jim Yurchenco, May 25, 2000, in an office somewhere in the IDEO complex.... Frankly, I'm lost.
I want to begin with a question about the Xerox mouse. What sorts of problems did it have that needed to be solved to make a mouse that could be mass-producible mouse?
Yurchenco: Well, the Xerox mouse that we examined was, I suspect, one of a relatively small number of actual physical mice that had been made. It had a couple problems.
The first was the way in which the ball was actually mounted and tracked on the table. They had a relatively small ball, and it was installed in a gimballing mechanism that required some bearings and some high-precision parts. These added cost and complexity, and caused reliability problems. They were operating under the assumption that the ball had to be pushed down onto the table, but one of the things we discovered in our research-- and this became fundamental to the design of all mice-- was that the ball could be held in contact with the table by gravity. It doesn't need to push, it can float.
The second was that the Xerox mouse was doing its encoding by means of mechanical commutators: they had little wires that brushed onto a commutator similar to what you'd find on a motor. It was expensive to do this: you needed two sets of these commutators, there were potential wear problems, and certainly reliability problems, in an environment in which dirt could be easily picked up.
Now, you think, motors have commutators and they work fine. Well, yes, true, but the amount of friction in the system in a motor is much less important than the amount of friction of a mouse, where you just have a ball that's picking up rotational energy from contact with the table. So the commutators had to be very delicate, and very lightly forced. By going away from a mechanical commutator to an optoelectronic system, we eliminated almost all the friction in the mouse. That also made the manufacturing simpler and reduced the cost.
One of my experiences later on with mice, when I designed the Microsoft mouse, was that their supplier, Alps Electric, actually had developed a very low-friction, relatively low-cost commutator which they were using in their mice. Alps was a company that specialized in building potentiometers, and the technology they had and the knowledge they had in-house that allowed them to do this; but they were the only people I've ever seen since the Apple mouse came onto the marketplace that has gone away from the optoelectronic encoder and used a commutator for this type of mouse design.
But basically, everybody builds them with optoelectronic encoders: they're reliable, they're inexpensive, they're easy to clean, they have relatively loose tolerances, and they're very, very easy to manufacture.
Pang: Can you tell me how you came up with the design for the ribcage?
Yurchenco: Well, ultimately it was a game of "Where do you put in 3D space certain components that have to be related to each other in certain relationships?" You start out and you know you have to have a ball and two encoders that have to be on the horizontal axis of the ball, and 90 degrees apart from each other. You needed an idler wheel that was at 45 degrees to those encoders. So you have these four objects in space that you have to package.
There are also a bunch of relationships that are established that driven by the resolution you want on the mouse-- how many dots per inch from tracking that you get. You start out with a certain size for the ball and encoder wheel, so there's a geometric relationship between the diameter of the ball and the diameter of the rollers on the encoder shaft. You also have the number of slots on the encoder wheel, and the size of the beam that the LED is emitting that passes through these slots and is going to the phototransistors.
The physical process is that you put the known objects into a drawing, and then you start building structure around. Eventually you end up with something that holds all these things in place. Then you have to say, okay, how am I going to make this structure?
We chose injection molding process because it's a very repeatable process, it gives you parts that are good enough precision for the kinds of things we needed to do here, and it's very low cost. Once you build the tool, the cost is a small amount of plastic resin, plus a small amount of time in an injection molding machine. That allows you build very, very low-cost parts. The parts are durable enough and reliable enough for the application, for the usage they're going to see, the environment they'll be in, and the lifetime of the part; and they're lightweight.
So they meet all the criteria you need to build this thing. It's why you see so much injection molded plastic in so many consumer products: it's very versatile process, and it allows you to create shapes that are complex that you can't get any other way with a low-cost process. So in some ways it's almost obvious. It's like "Why is a wheel round?" Because it works that way. In this case it's something a designer almost doesn't question, because it's such an obvious choice. If you remember The Graduate, and the advice the older man gave Dustin Hoffman: "Plastics." Well, there's some truth to that.
So injection molded plastic allows you do a lot of things. Now, the down side of IMP, as with any process of this nature, is you have to build a mold that can create the shapes. That mold is basically a block of steel, in which you cut out holes that you fill with a molten plastic that then hardens into the shape you want. All well and good. However, there are many shapes that are very difficult to create in a mold for technical reasons, and in the end you have to be able to get that part out of the mold. You need to be able to open the tool, be able to eject the part, so you can close it and make another one. And that requires design features in the part that you wouldn't normally need for functionality of the part, but you need for manufacturability of the part.
That's the interesting task for a designer of a plastic part. You can build a plastic part of whatever shape and size and complexity they want, but how do you mold it? Melding the two-- the functional design of the part, and the manufacturing design of the part-- is where the skill of a designer comes in. That's where being able to manipulate shapes, and being able to visualize how things are going to happen, becomes a great benefit in being able to design a part that's manufacturable. So you spend a lot of time moving lines and surfaces around to enable yourself not only to have them functional, but to be able to make them in high volume.
And then there's a lot of technical details in a molded part, like controlling the tolerances. If you have tight tolerances, where you place them in the part has an effect on how they'll come out, and the length of various features has an effect on what those tolerances are. So there are these other aspects of a design that come into play when you're designing and injection-molded part that makes it a much more difficult process than it sounds at first. You don't just make a shape and fill it with plastic.
The challenge in the case of the ribcage was that there were a lot of very small features that had to be crammed into a very small space, and building a mold to do that was complex. Nobody had actually done this before, so it was never completely clear that it would work when you put it together, so there was this element of risk. And you can model and make samples all you want, but you never know whether it's really going to work until you mold the part. Fortunately, we had a good toolmaker, Vic Renden at Micromold. He built a beautiful tool, and the parts worked right out of the chute-- first shots out of the mold worked.
Since then the technology has improved, people's understanding of the technology has improved, it's better now that it was however many years ago it was. Better designers have attacked the mouse since me and refined it, so current mice are even simpler, although you can still see their roots in the ribcage. Most of those parts are integrated right into the base of the mouse; it's no longer a stand-alone part. It takes more complicated tooling, which costs more, and they take longer to build; but it saves money, and right now mice are commodity products, so you get rid of a part, you save money.
Pang: When you were working on this, was this work that took place only on paper, or were you working with physical models as well?
Yurchenco: We had built some conceptual models, using similar materials or whatever we could get our hands on, that demonstrated that the physical principles we were going to use actually worked. But the actual layout was something I created in my mind, built it on paper, then handed it to Bud. He made the model, and he proved-- the model proved-- that it worked.
Pang: So a lot of the details were worked out on paper.
Yurchenco: There all worked out on paper. In the end, everything in that design basically was worked out on paper, without any modeling. This was before CAD was a commonly-used tool, so it was done strictly as a paper design. A lot of calculations of snaps and things like that, but mostly it was done in the brain, and then out the pencil.
Pang: And at that time, how much of your time would have spent doing the calculation?
Yurchenco: Very little. Probably less than five percent of my spent was spent calculating, the rest was moving shapes around, trying things out-- creating a shape in your brain, laying it out on paper in sketch form, and seeing if it will work. You can also see ways to improve it: if you move this wall here you can do this, and if you move this wall then this part comes out better over here.
It kind of grows up as a whole. You work on one section and paste it in, you work on another section and paste it in, then you put the two together and you see how they interrelate; then you change them, and they interrelate better. You gradually build this whole thing: it isn't like you build this one corner and it's done, and you build this second corner and it's done. There's a constant shuffling back and forth throughout the thing, and it grows as a whole until the whole thing is done.
There was also, I have to admit, I had a sort of aesthetic interest in mind. I was really concerned with what the thing looked like. So I spent time making shapes that were pleasing to me, and-- I hope-- integrated well with each other and didn't look completely random. I did have some interest in the way the final product looked, even though most people would never see it. I think that desire probably came from my background in fine arts and sculpture, that I'd studied at Stanford and previously. So my interest carried over there.
Pang: What would define a pleasing structure?
Yurchenco: Absolutely personal. It was like, beauty is in the eye of the beholder. To me, I liked the way the shapes worked together. But it wasn't accidental: there was a lot of deliberateness in there, I think is what I'm trying to get at.
Pang: Were there any cases when you would choose a design that was more pleasing, but wouldn't work quite as well as something uglier?
Yurchenco: No, I don't think so. Part of the definition of an aesthetically pleasing design-- for an engineer-- is how well does it work? So there was an effort to make both those things happen. I couldn't in fairness to my client compromise the design for aesthetic reasons; but you have a lot of choice in the way you do things. So I didn't feel like I had to compromise.
Pang: Once you passed this on to Bud to build the model, it took him about six weeks, as he recalled--
Yurchenco: I don't remember the time frame--
Pang: Did the design change in the course of building the model, as a result of finding something that was too difficult to machine?
Yurchenco: Well, making a machined model is a very different process that making a molded part. Designing and building a tool is a very different process from designing and building a plastic model: in one case you're working in the negative, in the other case you're working in the positive. Bud was working in the positive, and making it out of a piece of plastic. So he had to be able to get into areas to machine that would be very different than you do when you're building a tool. When you're building a tool, you're working in steel, but you're also working in the negative: even though the tools may be similar, the processes are extremely different.
So the problems that Bud faced were very different than the problems the tool-maker faced. Bud, for instance, was able to make individual little pieces and glue them together; the tool-maker could make individual inserts for the mold, and screw them all together. But Bud could work with tools to cut plastic fairly quickly, while the tool-maker was working with steel, which is much harder to cut; but he could work with something called an EDM machine, with which you can create carbon positives, and burn holes in the steel. So in Bud's case it's very difficult to measure what you're doing, particularly with the kinds of tools you have; so it's almost got to be an act of faith that once you've started you're going to get it right.
But no, there weren't any significant design changes that occurred based on the model.
Pang: So it sounds like in this process all the hard work is at the front end: once you get the tool made, turning the parts out doesn't require nearly as much skilled labor.
Yurchenco: It doesn't require any skilled labor at all. Injection molding those parts is done automatically: the machine sits there and spits out parts. There's a molding cycle, the press closes, fills with plastic, cools for a period, the press opens, the part falls out onto a conveyor belt, conveyor belt carries it to a box, and every now and then someone comes around and takes the box away. It's as simple as that.
I'm oversimplifying the problems of injection molding, but basically there's very little skill once you've got the press set up and it's producing parts.
Pang: There was one other thing you and Rickson had talked about, which was keeping an eye on how these parts could be assembled.
Yurchenco: Yeah, you always have to think about how you're going to put these parts together, so that's part of the design: can the person who's going to assemble this get the parts together, can they physically get their fingers in, can you pass part B past part A once it's in there? You have to think about the assembly sequence when you're doing this. But that's just part of the basic design process: no matter what you design, you're going to have that problem. In a way, I probably should bring it up, but it's intrinsic: being able to assemble a product makes a big difference in whether it's successful or not. If it assembles easily that means there are fewer errors, it costs less money, and it's more reliable. So you take that into account.
Obviously, you're dealing in this case with rather small parts, so you need to have assemblers who have some dexterity and good eyesight and small fingers and so forth. You still try to make it as simple as possible from their point of view. You also try to make it difficult to assemble things in the wrong way. So you key parts, you put features on parts whose only purpose is to aid in assembly or aid in orientation, or you make parts so you can assemble them both ways and they still work. You can't always succeed, but designers do their best to prevent those kinds of errors from occurring.
Pang: How did the labor on the project break down, and how did the members of the group work together?
Yurchenco: Jim Sachs was doing the electronic design, and I've had my memory jogged, and Douglas was doing the industrial design-- the appearance of the product. So in terms of building the enclosure for the mouse, Douglas had to decide what the shape would look like, and I would work back and forth with him and say, "Okay, well I need a little more room here for this," and he would change his shape slightly, or he would push back and say, "I don't want to change that shape, change your shape." So there's always this tension between the mechanical designer and the industrial designer.
But I would start by giving him a rough envelope of how big the components are, and in his case he would come back to me and say, "I want this thing to be about this big inside the hand." So we would sort of establish a rough boundary about where we're working. That's typical of the process for any product design. You're going to design an object and it'll have an appropriate size. Working with an industrial designer, you'll establish what that size is. That size has to contain all the functional parts. Now if that's too big, it may be an impossible product: if the parts in the mouse were the size of a shoe box, it wouldn't have been a very useful product. So there are limits to how far those boundaries can be pushed; but you work together on those things.
So Douglas had final responsibility for the outer shape, and I had final responsibility for how the parts inside worked, and how the parts were put together: how was the cord put in, what the bottom is, what's the door, and so forth. Jim had to determine what electronic parts had to go in there to enable the mouse to function; he worked out the circuits, and then we worked together to place those parts on the circuit board so his electronics would not be in the way of the parts I needed. So we would agree on a rough circuit board shape, and I would say, "This is my space, and this is your space," and he would see if he could do his layout inside "his" space. And if there were issues, we'd go back and negotiate.
Again, it's a very typical part of the design process. In this case it was a little more closely linked only in that I actually had some electrical components as part of the ribcage assembly-- the phototransistors and the LEDs-- and so we had to decide on those physical components, and I had to know their size and shape in order to fit them inside the plastic, because they were physically contained within the ribcage, and then went down onto the circuit board. Other than that, there was nothing atypical about that process: it gets done all the time.
Pang: Do you recall where the name ribcage came from?
Yurchenco: It probably came from the fact that we were building a mouse, and we started naming the parts anatomically. At some point the drawings actually had anatomical names for all the parts, but Apple made us take them off. I had at one point the words "Ribcage, Rodent" as the title on the drawing on the ribcage. Apple wasn't amused, which I thought was kind of sad.
Pang: Was that a typical practice, to give nicknames to things you're working on?
Yurchenco: Oh, yeah, we give nicknames to parts or designs or products all the time. It helps you remember them, because you get tired of calling something "Bracket A" or "Bracket B" or "Left Support Block." So it'll resemble some object, and someone will say it, and everyone else will agree, "Oh, yeah, that's what that is!" Sometimes the names get based on how we feel about the client, so they're not always for public consumption. But it's really typical to come up with those names, at least in this office.