WEBVTT 00:00:00.000 --> 00:00:02.430 The following content is provided under a Creative 00:00:02.430 --> 00:00:03.730 Commons license. 00:00:03.730 --> 00:00:06.030 Your support will help MIT OpenCourseWare 00:00:06.030 --> 00:00:10.060 continue to offer high-quality educational resources for free. 00:00:10.060 --> 00:00:12.690 To make a donation or to view additional materials 00:00:12.690 --> 00:00:16.560 from hundreds of MIT courses, visit MIT OpenCourseWare 00:00:16.560 --> 00:00:17.904 at ocw.mit.edu. 00:00:22.987 --> 00:00:25.920 DUANE BONING: So I want to welcome everybody here 00:00:25.920 --> 00:00:28.110 in Singapore-- or there in Singapore, 00:00:28.110 --> 00:00:29.850 and here in Cambridge. 00:00:29.850 --> 00:00:31.020 My name is Duane Boning. 00:00:31.020 --> 00:00:34.380 I'll be one of the two lecturers for our class. 00:00:34.380 --> 00:00:37.050 We'll get into some of those logistics 00:00:37.050 --> 00:00:38.760 and details in a minute, but we've 00:00:38.760 --> 00:00:40.470 got some preliminary things. 00:00:40.470 --> 00:00:45.510 Since we've started videotaping you already, 00:00:45.510 --> 00:00:49.620 one thing you need to know is there is a video recording 00:00:49.620 --> 00:00:50.580 policy. 00:00:50.580 --> 00:00:54.630 Pretty much everybody here will be videotaped. 00:00:54.630 --> 00:00:57.360 If you don't want to be videotaped, 00:00:57.360 --> 00:01:00.780 there are a couple of seats here on the right. 00:01:00.780 --> 00:01:02.310 If you do want to be videotaped, you 00:01:02.310 --> 00:01:05.550 can come and be accessible and viewable. 00:01:05.550 --> 00:01:08.970 Right at the front, you can see there's these little red tabs 00:01:08.970 --> 00:01:10.410 on the desk. 00:01:10.410 --> 00:01:12.330 That's in the blind spot of the video. 00:01:15.750 --> 00:01:17.910 I think some of these were being handed out 00:01:17.910 --> 00:01:22.380 as you were coming in, but if you didn't get them, 00:01:22.380 --> 00:01:25.365 here's another copy of that videotaping policy. 00:01:29.400 --> 00:01:34.590 So what we do is most of these lectures are, of course, live, 00:01:34.590 --> 00:01:37.950 video linked with Singapore. 00:01:37.950 --> 00:01:40.110 There are occasions in the calendar 00:01:40.110 --> 00:01:43.830 where it's a holiday in one place or the other. 00:01:43.830 --> 00:01:46.380 When it's a holiday here, we don't meet. 00:01:46.380 --> 00:01:49.170 When it's a whole holiday in Singapore, but not a holiday 00:01:49.170 --> 00:01:51.930 here, we will still meet here in Cambridge, 00:01:51.930 --> 00:01:55.920 but the video version of the local lecture 00:01:55.920 --> 00:02:00.210 will be available on the website so that folks in Singapore 00:02:00.210 --> 00:02:04.330 can catch up on that. 00:02:04.330 --> 00:02:06.720 And in fact, we understand the first one of those 00:02:06.720 --> 00:02:08.669 is this coming Thursday. 00:02:08.669 --> 00:02:10.860 Is everybody ready for your holiday? 00:02:10.860 --> 00:02:14.866 It's been a long, hard term already, right? 00:02:14.866 --> 00:02:16.780 Is that right, Singapore? 00:02:16.780 --> 00:02:18.113 OK. 00:02:18.113 --> 00:02:19.530 Yeah, that's the Chinese New Year, 00:02:19.530 --> 00:02:22.380 so there will be no class meeting in Singapore 00:02:22.380 --> 00:02:23.520 on Thursday evening. 00:02:26.410 --> 00:02:27.190 Let's see. 00:02:29.870 --> 00:02:31.910 OK. 00:02:31.910 --> 00:02:39.140 Has everybody found their way to the Stellar website already? 00:02:39.140 --> 00:02:43.430 Anybody who has not seen the Stellar website for this class? 00:02:45.950 --> 00:02:50.080 I'm asking in part because, if you are not actually 00:02:50.080 --> 00:02:52.450 registered-- let me go to the Stellar site here-- 00:02:56.150 --> 00:03:00.170 oh, this may not be viewable in Singapore. 00:03:00.170 --> 00:03:03.350 This is accessible only to those registered in the class, 00:03:03.350 --> 00:03:05.060 I believe. 00:03:05.060 --> 00:03:08.300 So if you're having trouble getting to the website, 00:03:08.300 --> 00:03:11.900 please email me right away, and we will add you to the roster, 00:03:11.900 --> 00:03:14.810 even before you get registered. 00:03:14.810 --> 00:03:16.400 It will be important, because all 00:03:16.400 --> 00:03:19.840 of the materials for the class are 00:03:19.840 --> 00:03:22.480 deployed through the website. 00:03:22.480 --> 00:03:26.440 In fact, it'll be very rare when we have any kind of handouts 00:03:26.440 --> 00:03:28.180 to give in class. 00:03:28.180 --> 00:03:34.840 Lecture notes, reading material, reading back-up, assignments, 00:03:34.840 --> 00:03:37.240 weekly problem sets-- all of that material 00:03:37.240 --> 00:03:38.200 is through the website. 00:03:42.840 --> 00:03:44.462 Now, what I will typically do-- 00:03:44.462 --> 00:03:45.920 let me see if I can get back here-- 00:03:49.170 --> 00:03:54.960 usually I will try to have the PowerPoint slides loaded 00:03:54.960 --> 00:03:58.600 a day before the lecture. 00:03:58.600 --> 00:04:02.320 One thing that I noticed in a few past years is sometimes 00:04:02.320 --> 00:04:06.300 little groups of students would work together 00:04:06.300 --> 00:04:11.340 and delegate somebody to print out the lecture slides 00:04:11.340 --> 00:04:14.560 and bring a set of copies for their little group. 00:04:14.560 --> 00:04:17.160 So for example, the LFM students would often 00:04:17.160 --> 00:04:21.180 have one person printing out that morning or the night 00:04:21.180 --> 00:04:22.620 before, if they were available. 00:04:22.620 --> 00:04:24.210 So you guys can self-organize. 00:04:24.210 --> 00:04:26.970 If you actually like having any kind of print-out 00:04:26.970 --> 00:04:32.250 to take notes on a copy of the PowerPoint slides, 00:04:32.250 --> 00:04:35.730 feel free to print those out or self-organize 00:04:35.730 --> 00:04:38.970 to share that duty. 00:04:38.970 --> 00:04:43.320 Again, we're not going to be making copies available 00:04:43.320 --> 00:04:46.960 directly in the class. 00:04:46.960 --> 00:04:51.390 So what I want to do, first off, is talk a little bit 00:04:51.390 --> 00:04:56.262 about the class and the assumed prerequisites for the class, 00:04:56.262 --> 00:04:57.720 and then we'll come back and I want 00:04:57.720 --> 00:05:00.120 to get us to know each other. 00:05:00.120 --> 00:05:03.150 I know many of you have had some classes together. 00:05:03.150 --> 00:05:06.580 Others of you are perhaps new to each other, 00:05:06.580 --> 00:05:09.040 so want to share a little bit of experiences 00:05:09.040 --> 00:05:12.810 so we know what the basin of experience is-- 00:05:12.810 --> 00:05:16.980 because this class is meant to be fairly interactive. 00:05:16.980 --> 00:05:20.790 It is a manufacturing process control class, 00:05:20.790 --> 00:05:24.150 and one of the best things about it is many of the students 00:05:24.150 --> 00:05:26.942 in the class have an awful lot of experience. 00:05:26.942 --> 00:05:28.650 For example, the students and the leaders 00:05:28.650 --> 00:05:31.860 for manufacturing program, LFM, are typically 00:05:31.860 --> 00:05:35.500 coming from several years of industrial experience. 00:05:35.500 --> 00:05:38.640 Similarly, students have had a lot 00:05:38.640 --> 00:05:43.080 of experience, whether it be doing experimental work 00:05:43.080 --> 00:05:44.670 with processes in-- 00:05:44.670 --> 00:05:47.950 as part of their program, or what have you. 00:05:47.950 --> 00:05:51.450 So it is meant to be interactive. 00:05:51.450 --> 00:05:53.620 We depend on you sharing your experiences, 00:05:53.620 --> 00:05:56.580 your insights, your doubts. 00:05:56.580 --> 00:06:00.150 In many cases, those can be often the most interesting part 00:06:00.150 --> 00:06:00.970 of the subject-- 00:06:00.970 --> 00:06:02.930 say, I tried that-- didn't work. 00:06:02.930 --> 00:06:06.540 Here's why-- because a lot of the classes about tools 00:06:06.540 --> 00:06:12.120 and experiences for dealing with manufacturing processes. 00:06:12.120 --> 00:06:14.860 And then we'll get into a little bit more of the details, 00:06:14.860 --> 00:06:19.770 some of the course schedule, and some of the other logistics. 00:06:19.770 --> 00:06:22.680 Jumping ahead a little bit here, there's 00:06:22.680 --> 00:06:28.110 kind of a long chain listed formally on the course 00:06:28.110 --> 00:06:30.270 catalog for prerequisites-- 00:06:30.270 --> 00:06:36.270 says things like 2.008, or 2.810, or 6152J, 00:06:36.270 --> 00:06:42.240 and then also a couple of other subjects like 6041 or 15064J. 00:06:42.240 --> 00:06:46.350 What we intend to convey with these prerequisites is, number 00:06:46.350 --> 00:06:49.200 one, they're not specific hard prerequisites 00:06:49.200 --> 00:06:50.460 for those subjects. 00:06:50.460 --> 00:06:55.650 What we mean to convey is you should have hopefully 00:06:55.650 --> 00:07:00.160 some substantial experience with some manufacturing process. 00:07:00.160 --> 00:07:03.750 So that's things like 2810, 6152J. 00:07:03.750 --> 00:07:08.850 So 2810 is essentially the manufacturing physics subject-- 00:07:08.850 --> 00:07:11.715 manufacturing processes subject in course two. 00:07:11.715 --> 00:07:13.530 I recognize some of you for that. 00:07:13.530 --> 00:07:20.000 I think I gave a guest lecturer in that subject 00:07:20.000 --> 00:07:22.800 on semiconductor manufacturing. 00:07:22.800 --> 00:07:24.860 So that would be one example. 00:07:24.860 --> 00:07:29.540 6152J is a course six, electrical engineering computer 00:07:29.540 --> 00:07:33.020 science subject on microfabrication-- 00:07:33.020 --> 00:07:37.550 not only semiconductor, but also MEMS microfabrication. 00:07:37.550 --> 00:07:40.070 And so that would be another example through coursework, 00:07:40.070 --> 00:07:42.590 where you might have gotten that kind of experience. 00:07:42.590 --> 00:07:45.500 And then similarly, you might have actual true 00:07:45.500 --> 00:07:47.160 industrial experience. 00:07:47.160 --> 00:07:51.800 So the goal here is actually to just go deep enough 00:07:51.800 --> 00:07:55.670 or build on a deep enough knowledge of some process 00:07:55.670 --> 00:07:57.350 technology. 00:07:57.350 --> 00:08:01.010 This semester of the subject, I will actually 00:08:01.010 --> 00:08:06.440 be using semiconductor process technology a little bit 00:08:06.440 --> 00:08:09.380 more than we have some semesters in the past 00:08:09.380 --> 00:08:13.220 to illustrate, and demonstrate, and to get 00:08:13.220 --> 00:08:16.040 some practice with some of the concepts 00:08:16.040 --> 00:08:17.390 that we're dealing with here. 00:08:17.390 --> 00:08:19.400 But many of the problems will also 00:08:19.400 --> 00:08:23.000 deal with macroscopic processes, from metal 00:08:23.000 --> 00:08:29.910 bending to forming and other sorts of processes. 00:08:29.910 --> 00:08:33.559 A little later, we expect Dave Hardt, who is the collector. 00:08:33.559 --> 00:08:34.370 He'll be here. 00:08:34.370 --> 00:08:37.789 And his background especially emphasizes 00:08:37.789 --> 00:08:43.047 the panoply of mechanical and broad based process 00:08:43.047 --> 00:08:45.380 technologies, whereas my experience is a little bit more 00:08:45.380 --> 00:08:48.110 solidly in microfabrication-- 00:08:48.110 --> 00:08:50.930 both in semiconductor, and more recently, 00:08:50.930 --> 00:08:55.520 in micro-electromechanical systems, or MEMS fabrication. 00:08:55.520 --> 00:08:57.140 So that's one of the prerequisites. 00:08:57.140 --> 00:09:00.410 Now, the other prerequisite that we kind of assume-- 00:09:00.410 --> 00:09:02.930 and if you don't have, you might be doing a little bit 00:09:02.930 --> 00:09:05.120 of quick background reading-- 00:09:05.120 --> 00:09:08.870 is indicated by things like 6041 or 15064, 00:09:08.870 --> 00:09:13.970 and that is some basic probability and statistics. 00:09:13.970 --> 00:09:16.910 We'll actually be developing quite a bit of the statistical 00:09:16.910 --> 00:09:18.770 machinery, so if you don't-- 00:09:18.770 --> 00:09:22.550 haven't had a full subject in that, not to worry-- 00:09:22.550 --> 00:09:25.690 I recognize many of you also from-- 00:09:25.690 --> 00:09:29.900 is it 2.853, 2.854? 00:09:29.900 --> 00:09:31.220 Get that subject right-- 00:09:31.220 --> 00:09:34.380 that's Stan Gershwin's manufacturing process? 00:09:34.380 --> 00:09:34.880 What is it? 00:09:34.880 --> 00:09:38.330 Manufacturing systems-- manufacturing 00:09:38.330 --> 00:09:42.260 systems, where we had a two or three-lecture sequence 00:09:42.260 --> 00:09:46.550 on basic statistics and statistical modeling. 00:09:46.550 --> 00:09:49.230 We'll actually revisit some of that, 00:09:49.230 --> 00:09:51.170 so you'll see some of that again, but then 00:09:51.170 --> 00:09:54.110 be applying that to problems of manufacturing. 00:09:58.840 --> 00:10:00.520 I'll go over this detail, and then we'll 00:10:00.520 --> 00:10:05.260 come back to get some introductions from you guys. 00:10:05.260 --> 00:10:09.400 This semester we actually have two required textbooks. 00:10:09.400 --> 00:10:13.720 The great thing is a new book has come out. 00:10:13.720 --> 00:10:17.200 It came out nominally 2006, but it was only available 00:10:17.200 --> 00:10:18.610 starting last year. 00:10:18.610 --> 00:10:23.140 And that's a book by Gary May and Costas Spanos. 00:10:23.140 --> 00:10:24.390 Gary May is at Georgia Tech. 00:10:24.390 --> 00:10:29.470 Costas Spanos at UC Berkeley, and they had developed a course 00:10:29.470 --> 00:10:32.680 on semiconductor process control that 00:10:32.680 --> 00:10:36.190 was very close to a subject in the dim, dark past 00:10:36.190 --> 00:10:39.700 that I used to teach dedicated entirely to semiconductor 00:10:39.700 --> 00:10:41.030 process control. 00:10:41.030 --> 00:10:43.090 So finally, we have a book for that. 00:10:43.090 --> 00:10:47.590 It's a nice condensed book, and you all should try to get that, 00:10:47.590 --> 00:10:49.480 because we're going to have readings 00:10:49.480 --> 00:10:52.210 from this book starting pretty quickly. 00:10:52.210 --> 00:10:55.120 So you might have to go to Amazon, 00:10:55.120 --> 00:10:57.310 do two-day shipping or something like that. 00:10:57.310 --> 00:11:00.772 If somebody has Amazon Prime, you 00:11:00.772 --> 00:11:02.230 can use that and get free shipping. 00:11:02.230 --> 00:11:04.570 I don't know if any of you do that. 00:11:04.570 --> 00:11:08.380 The other book is Introduction to Statistical Quality Control. 00:11:08.380 --> 00:11:11.080 This is more broad based. 00:11:11.080 --> 00:11:14.050 May and Spanos illustrate everything the semiconductor 00:11:14.050 --> 00:11:17.050 process technology. 00:11:17.050 --> 00:11:22.210 Montgomery is basically a statistical process control, 00:11:22.210 --> 00:11:25.000 a little bit of design of experiment. 00:11:25.000 --> 00:11:27.400 It really emphasizes the statistical machinery 00:11:27.400 --> 00:11:29.750 that we're going to be using. 00:11:29.750 --> 00:11:32.680 So you are expected to either go and get 00:11:32.680 --> 00:11:34.780 both of those-- both of these books 00:11:34.780 --> 00:11:38.170 have been put on reserve in the MIT library, 00:11:38.170 --> 00:11:41.800 so if you don't want to spend the money, you can, I guess, 00:11:41.800 --> 00:11:44.440 do the painful thing of trying to go to the library 00:11:44.440 --> 00:11:45.790 and read there. 00:11:49.670 --> 00:11:52.250 At least the Montgomery book is a great book 00:11:52.250 --> 00:11:55.020 to have on your bookshelf long term. 00:11:55.020 --> 00:11:58.190 It'll be a great reference book throughout your career. 00:11:58.190 --> 00:12:00.200 The semiconductor process control-- certainly, 00:12:00.200 --> 00:12:02.600 if you're going in the semiconductor area, that's 00:12:02.600 --> 00:12:05.170 also very useful. 00:12:05.170 --> 00:12:06.760 OK, so the grading in the class-- 00:12:06.760 --> 00:12:10.240 the way this works is we will have-- 00:12:10.240 --> 00:12:13.060 I think it's something like eight weekly problem sets, 00:12:13.060 --> 00:12:15.402 primarily in the first 2/3 of the class. 00:12:15.402 --> 00:12:17.110 And then, in the last third of the class, 00:12:17.110 --> 00:12:23.090 we stop the problem sets and you'll go into a team project. 00:12:23.090 --> 00:12:25.090 The problem sets will comprise about 40% 00:12:25.090 --> 00:12:26.620 of the grade in the class. 00:12:26.620 --> 00:12:29.200 We'll have two quizzes-- 00:12:29.200 --> 00:12:33.520 one about a little bit more than a third of the way through, 00:12:33.520 --> 00:12:38.570 another one a little bit more than 2/3 of the way through. 00:12:38.570 --> 00:12:42.950 Those are about 40% of the grade in the class. 00:12:42.950 --> 00:12:45.590 The good thing about having those quizzes during the term 00:12:45.590 --> 00:12:49.930 is we do not have a final exam. 00:12:49.930 --> 00:12:51.310 Anybody heartbroken about that? 00:12:51.310 --> 00:12:52.810 [APPLAUSE] 00:12:52.810 --> 00:12:53.950 OK. 00:12:53.950 --> 00:12:57.070 However, the reason we don't is that, 00:12:57.070 --> 00:12:59.560 in that last third quarter of the class, 00:12:59.560 --> 00:13:01.870 we switched gears a little bit and ask 00:13:01.870 --> 00:13:05.440 you to start applying some of the techniques and tools that 00:13:05.440 --> 00:13:07.600 will have been talked about through most 00:13:07.600 --> 00:13:09.922 of the semester in a team project. 00:13:09.922 --> 00:13:11.380 I'll talk more about what that team 00:13:11.380 --> 00:13:15.110 project is a little bit later. 00:13:15.110 --> 00:13:20.880 So that's meant to be a fairly substantial effort. 00:13:20.880 --> 00:13:24.600 The team project is going to be a team project. 00:13:24.600 --> 00:13:30.630 It's going to be typically three-student teams. 00:13:30.630 --> 00:13:35.250 And of course, that should be very collaborative in nature. 00:13:35.250 --> 00:13:38.310 In contrast to that, the other assignments 00:13:38.310 --> 00:13:43.170 are meant to be individual efforts. 00:13:43.170 --> 00:13:46.230 Now, working on the problem set, I 00:13:46.230 --> 00:13:49.410 think it's great to interact, talk with each other 00:13:49.410 --> 00:13:52.230 to understand the problems, even to bounce off 00:13:52.230 --> 00:13:56.190 possible lines of attack or solution approaches, 00:13:56.190 --> 00:13:57.390 but where we mean-- 00:13:57.390 --> 00:14:00.130 especially on the assignments-- the weekly assignments, 00:14:00.130 --> 00:14:02.160 the problems sets-- is everybody should do 00:14:02.160 --> 00:14:04.170 their own independent write-up. 00:14:04.170 --> 00:14:08.130 I find that writing and explaining not just 00:14:08.130 --> 00:14:10.560 writing down the answer number, by the way. 00:14:10.560 --> 00:14:13.890 We will expect actual articulation 00:14:13.890 --> 00:14:16.770 of your thought process. 00:14:16.770 --> 00:14:20.620 That should be an individual effort. 00:14:20.620 --> 00:14:24.170 Now, we'll come back to that the assignments a little bit. 00:14:24.170 --> 00:14:26.860 We'll try to emphasize this, perhaps, as well a little bit 00:14:26.860 --> 00:14:29.050 more on the assignments themselves. 00:14:29.050 --> 00:14:30.610 The perspective I'd kind of like you 00:14:30.610 --> 00:14:33.700 to take on these weekly problem sets 00:14:33.700 --> 00:14:38.740 is not that you're trying to come up with the answer number 00:14:38.740 --> 00:14:43.310 like you might on an exam or something like that, 00:14:43.310 --> 00:14:46.510 where you just needed the number, but instead, 00:14:46.510 --> 00:14:52.600 a little bit of what I think of as a practicing manufacturing 00:14:52.600 --> 00:14:55.060 perspective-- 00:14:55.060 --> 00:14:57.100 I couldn't call it an industrial perspective 00:14:57.100 --> 00:14:59.890 or an academic perspective, but the key thing 00:14:59.890 --> 00:15:01.360 that you have to do, and will have 00:15:01.360 --> 00:15:03.175 to do throughout your career is convince. 00:15:05.680 --> 00:15:09.040 So you have to convince us that the answer 00:15:09.040 --> 00:15:14.500 that you're putting down makes sense, why it's reasonable. 00:15:14.500 --> 00:15:17.020 So you need to articulate the thought process, 00:15:17.020 --> 00:15:19.370 not just come up with answers. 00:15:19.370 --> 00:15:21.370 And that's part of the reason why we really 00:15:21.370 --> 00:15:26.123 want that to be an individual effort, an individual write-up. 00:15:26.123 --> 00:15:27.540 And of course, everybody's already 00:15:27.540 --> 00:15:31.780 gotten to the course URL. 00:15:31.780 --> 00:15:35.680 So going backwards a step here, let me introduce-- 00:15:35.680 --> 00:15:37.900 I hoped that Dave would be here by now-- 00:15:37.900 --> 00:15:40.810 must be having a very difficult decision point 00:15:40.810 --> 00:15:44.930 and in the voting booth there-- who to vote for. 00:15:44.930 --> 00:15:46.240 Again, I'm Duane Boning. 00:15:46.240 --> 00:15:49.000 I'm a professor of electrical engineering and computer 00:15:49.000 --> 00:15:50.620 science. 00:15:50.620 --> 00:15:51.250 Dave Hardt. 00:15:51.250 --> 00:15:52.630 Will be here a little bit later. 00:15:52.630 --> 00:15:56.170 Let me introduce for you the teaching assistant, 00:15:56.170 --> 00:15:58.270 Hayden Taylor. 00:15:58.270 --> 00:15:59.450 We do have microphones here. 00:15:59.450 --> 00:16:01.120 Why don't you say a couple of words? 00:16:01.255 --> 00:16:02.338 HAYDEN TAYLOR: [INAUDIBLE] 00:16:02.338 --> 00:16:03.508 DUANE BONING: Yeah-- good. 00:16:03.508 --> 00:16:04.300 There's the camera. 00:16:04.407 --> 00:16:05.740 HAYDEN TAYLOR: Hello, Singapore. 00:16:05.740 --> 00:16:07.930 I'm Hayden. 00:16:07.930 --> 00:16:09.260 I'm in electrical engineering. 00:16:09.260 --> 00:16:12.430 I work with Duane on manufacturing approaches 00:16:12.430 --> 00:16:14.650 for microfluidic devices. 00:16:14.650 --> 00:16:16.540 We work a lot with polymers. 00:16:16.540 --> 00:16:19.630 And I'm excited to work with you all this time. 00:16:19.630 --> 00:16:22.840 Feel free to email me at any time. 00:16:22.840 --> 00:16:25.910 I'll be organizing office hours, and I'll 00:16:25.910 --> 00:16:29.680 announce that on the website when the time has come. 00:16:29.680 --> 00:16:32.050 For those of you in Singapore, I suppose 00:16:32.050 --> 00:16:33.820 we'll have to have virtual office hours. 00:16:33.820 --> 00:16:36.100 I'm very happy to talk over Skype or something 00:16:36.100 --> 00:16:39.651 like that, if you want, so feel free to drop me an email. 00:16:39.651 --> 00:16:40.151 Thanks. 00:16:44.005 --> 00:16:45.380 DUANE BONING: And Hayden just got 00:16:45.380 --> 00:16:49.077 back a little over a week ago-- is that right-- from Singapore. 00:16:49.077 --> 00:16:50.600 HAYDEN TAYLOR: That's right. 00:16:50.600 --> 00:16:52.517 DUANE BONING: [? We ?] spent a couple of weeks 00:16:52.517 --> 00:16:54.920 there for some of the SMA program things. 00:16:54.920 --> 00:16:57.350 The core secretary will be my assistant 00:16:57.350 --> 00:17:03.500 over near my office in EECS headquarters in building 38. 00:17:03.500 --> 00:17:05.359 And that's [? Charlene Blake. ?] So if you 00:17:05.359 --> 00:17:07.790 have some administrative questions, 00:17:07.790 --> 00:17:12.140 your best bet is probably emailing Hayden and me 00:17:12.140 --> 00:17:13.460 directly. 00:17:13.460 --> 00:17:16.790 I'll be the lead lecturer this term. 00:17:16.790 --> 00:17:18.890 Normally, Dave Hardt is the lead lecturer. 00:17:18.890 --> 00:17:23.569 He's actually nominally on sabbatical this term, but-- 00:17:23.569 --> 00:17:26.930 so I'll be taking care of more of the administrative things, 00:17:26.930 --> 00:17:30.300 and he'll get to give fun lectures. 00:17:30.300 --> 00:17:33.260 So if you have administrative things, please-- 00:17:33.260 --> 00:17:36.500 probably best to contact either me or Hayden directly. 00:17:36.500 --> 00:17:39.920 If you have questions or need to drop off, 00:17:39.920 --> 00:17:44.030 for example, late problem sets, that 00:17:44.030 --> 00:17:46.830 can be done with [? Charlene. ?] 00:17:46.830 --> 00:17:49.260 And we'll have to have to let [? Charlene ?] know 00:17:49.260 --> 00:17:52.900 that she really outdid herself on the breakfast here. 00:17:52.900 --> 00:17:55.620 So at any time, as soon as you're done with one, 00:17:55.620 --> 00:17:57.150 we have lots more to eat, so feel 00:17:57.150 --> 00:18:00.570 free to wander over and grab a little bit more. 00:18:03.320 --> 00:18:05.480 OK, we'll dive back into a little bit more 00:18:05.480 --> 00:18:08.450 about the course, but I thought it would be a little bit 00:18:08.450 --> 00:18:13.370 fun to just quickly go around and hear a sentence or two 00:18:13.370 --> 00:18:14.300 about-- 00:18:14.300 --> 00:18:17.420 and I'm particularly interested in what kind 00:18:17.420 --> 00:18:21.500 of process background, or manufacturing background, 00:18:21.500 --> 00:18:24.140 or technology background you have-- 00:18:24.140 --> 00:18:27.230 or if you don't have much background, what kind of areas 00:18:27.230 --> 00:18:32.430 you're especially interested in, or potentially interested in. 00:18:32.430 --> 00:18:37.610 So we'll do that quickly here on the Cambridge end first. 00:18:37.610 --> 00:18:41.250 I know many of you, again, have been taking classes together, 00:18:41.250 --> 00:18:43.930 so you know each other better than I you. 00:18:43.930 --> 00:18:47.060 I recognize some faces here a little bit. 00:18:47.060 --> 00:18:50.540 But this is part of my way to get to you as well as for you 00:18:50.540 --> 00:18:52.910 guys to get to know each other. 00:19:03.520 --> 00:19:07.243 OK, let me go through a few of the other logistics, 00:19:07.243 --> 00:19:09.160 and then we'll start getting into a little bit 00:19:09.160 --> 00:19:13.340 of introductory material. 00:19:13.340 --> 00:19:15.340 I think we've already done some of that. 00:19:18.390 --> 00:19:20.813 Let me come back to the team projects, 00:19:20.813 --> 00:19:22.230 because I think it'll be important 00:19:22.230 --> 00:19:25.440 for you folks to be starting to think very 00:19:25.440 --> 00:19:31.770 early about possible projects and possible team formation, 00:19:31.770 --> 00:19:35.400 because that will help-- as you're learning some 00:19:35.400 --> 00:19:38.880 of the tools and techniques in these different kinds 00:19:38.880 --> 00:19:43.360 of topical areas, you can pay more attention, 00:19:43.360 --> 00:19:45.210 because you'll know, OK, I really 00:19:45.210 --> 00:19:47.820 want to think about process diagnosis 00:19:47.820 --> 00:19:50.830 for this particular project, or what have you. 00:19:50.830 --> 00:19:53.310 So the typical kinds of topics-- 00:19:53.310 --> 00:19:56.400 generically, the topics in these team projects 00:19:56.400 --> 00:20:00.060 involve things like process diagnosis; 00:20:00.060 --> 00:20:02.490 process improvement, often with a little bit 00:20:02.490 --> 00:20:07.170 of statistical process control, detection, 00:20:07.170 --> 00:20:11.550 and debug kinds of things; process optimization 00:20:11.550 --> 00:20:13.230 and robustness-- 00:20:13.230 --> 00:20:15.930 things like use of design of experiments 00:20:15.930 --> 00:20:19.290 to characterize the process, and then 00:20:19.290 --> 00:20:23.460 seek to optimize some outputs, as well as minimize 00:20:23.460 --> 00:20:26.910 some variations and get to a robust process; 00:20:26.910 --> 00:20:29.910 and then a number of more advanced applications, 00:20:29.910 --> 00:20:32.850 perhaps dealing with things like yield modeling or defect 00:20:32.850 --> 00:20:34.020 modeling-- 00:20:34.020 --> 00:20:36.750 these sorts of projects. 00:20:36.750 --> 00:20:39.270 And the basic expectation in these team projects 00:20:39.270 --> 00:20:41.040 is you'll need to learn a little bit 00:20:41.040 --> 00:20:44.820 more background on the basic process 00:20:44.820 --> 00:20:47.550 and what the problem is. 00:20:47.550 --> 00:20:53.670 A really interesting aspect of these projects is it's best-- 00:20:53.670 --> 00:20:56.540 it really works nicely if we can tap 00:20:56.540 --> 00:21:00.800 into a nice rich set of existing data, 00:21:00.800 --> 00:21:03.590 or in some cases, even generation of new data 00:21:03.590 --> 00:21:06.043 arising out of either your past experience 00:21:06.043 --> 00:21:07.085 or your current research. 00:21:10.340 --> 00:21:13.940 So for example, one of the, say, typical three-team members 00:21:13.940 --> 00:21:16.730 might have access to-- 00:21:16.730 --> 00:21:21.560 hopefully only public, shareable-- nothing secret-- 00:21:21.560 --> 00:21:24.740 public data from their work experience. 00:21:28.130 --> 00:21:31.280 Some of you may have generated a big set of data 00:21:31.280 --> 00:21:34.140 for your master's thesis, or this or that, 00:21:34.140 --> 00:21:38.090 and we're only able to look at this aspect of the problem, 00:21:38.090 --> 00:21:39.920 because that was the key question, 00:21:39.920 --> 00:21:41.900 but it was sort of nagging at you. 00:21:41.900 --> 00:21:43.820 There's all this other rich data that you 00:21:43.820 --> 00:21:47.210 would have thought it interesting to look 00:21:47.210 --> 00:21:49.430 at from other perspectives-- 00:21:49.430 --> 00:21:55.070 things like building very simple response surface models for. 00:21:55.070 --> 00:21:58.000 But you didn't really have the chance to go down that path. 00:21:58.000 --> 00:22:02.030 So keep your mind open, thinking ahead to the project 00:22:02.030 --> 00:22:04.340 as we go out-- go throughout the term, 00:22:04.340 --> 00:22:06.650 and say, oh, that's an interesting topic. 00:22:06.650 --> 00:22:08.510 That's an interesting technique. 00:22:08.510 --> 00:22:10.910 That reminds me of a very interesting data 00:22:10.910 --> 00:22:14.750 set that I came across back a year ago or whatever-- 00:22:14.750 --> 00:22:16.940 because that might form the basis for a very 00:22:16.940 --> 00:22:18.090 interesting team project. 00:22:21.560 --> 00:22:24.180 The output of this are two parts. 00:22:24.180 --> 00:22:27.510 There will be a world presentation. 00:22:27.510 --> 00:22:29.440 So the team will present-- 00:22:29.440 --> 00:22:31.790 in fact, the last two class periods, 00:22:31.790 --> 00:22:34.820 the very last week of class, will be dedicated 00:22:34.820 --> 00:22:36.680 to these team presentations. 00:22:36.680 --> 00:22:39.530 And then there will also be a project report. 00:22:39.530 --> 00:22:41.480 Unlike the problem sets, where everybody 00:22:41.480 --> 00:22:44.540 writes their own separate problem set, 00:22:44.540 --> 00:22:47.630 the team writes the project report. 00:22:47.630 --> 00:22:50.660 So that's one document that comes together from the team. 00:22:53.390 --> 00:22:58.535 Good-- so let me pop back. 00:22:58.535 --> 00:23:00.410 We're not going to ask you who you voted for, 00:23:00.410 --> 00:23:04.120 but we figured it was a tough decision, because it 00:23:04.120 --> 00:23:07.102 was taking a little while. 00:23:07.102 --> 00:23:08.810 We've already done all the introductions, 00:23:08.810 --> 00:23:11.655 so now you, fortunately, know most everybody. 00:23:11.655 --> 00:23:13.405 DAVID HARDT: I hope everybody knows that I 00:23:13.405 --> 00:23:16.403 got dressed for the occasion. 00:23:16.403 --> 00:23:17.320 What am I [INAUDIBLE]? 00:23:17.320 --> 00:23:19.612 DUANE BONING: You have to wear a tie to vote in the US. 00:23:19.612 --> 00:23:21.000 You may not know that. 00:23:21.000 --> 00:23:22.390 DAVID HARDT: For those of you who don't already know me, 00:23:22.390 --> 00:23:23.200 I'm David Hardt. 00:23:23.200 --> 00:23:25.660 And I'm actually on sabbatical this semester, which 00:23:25.660 --> 00:23:27.250 means I don't teach any classes and I 00:23:27.250 --> 00:23:28.420 don't sit on any committees. 00:23:28.420 --> 00:23:29.860 DUANE BONING: Why do you wear a tie if you're on sabbatical? 00:23:29.860 --> 00:23:31.660 DAVID HARDT: So I'm here today to sit on a committee 00:23:31.660 --> 00:23:32.535 and do some teaching. 00:23:32.535 --> 00:23:34.150 Yeah. 00:23:34.150 --> 00:23:37.450 But I welcome you all to probably the best 00:23:37.450 --> 00:23:39.190 class at MIT-- 00:23:39.190 --> 00:23:43.312 and hello to all of our students there in Singapore as well. 00:23:43.312 --> 00:23:44.770 I hope you're enjoying the weather. 00:23:44.770 --> 00:23:46.240 It's beautiful here. 00:23:46.240 --> 00:23:48.130 And the only reason I'm late is that I 00:23:48.130 --> 00:23:50.800 had to drive the other way to get to the polling station, 00:23:50.800 --> 00:23:52.640 and I got trapped by a huge traffic jam. 00:23:52.640 --> 00:23:55.600 So I'm a dedicated voter. 00:23:55.600 --> 00:23:56.850 DUANE BONING: Thank you, Dave. 00:24:03.000 --> 00:24:05.895 OK, so let's get a little bit into the-- 00:24:05.895 --> 00:24:07.770 an overview of what we're going to be dealing 00:24:07.770 --> 00:24:10.560 with throughout the semester, and then dive 00:24:10.560 --> 00:24:14.790 in a little bit on some concepts and basic background 00:24:14.790 --> 00:24:16.180 and terminology. 00:24:16.180 --> 00:24:20.100 So some of the key ideas that we're after in this subject 00:24:20.100 --> 00:24:25.610 is really dealing with basically the problems in manufacturing. 00:24:25.610 --> 00:24:28.790 And the biggest problem of all is really 00:24:28.790 --> 00:24:32.710 dealing with quality and manufacturing variation. 00:24:32.710 --> 00:24:36.770 Other aspects of manufacturing are often 00:24:36.770 --> 00:24:42.050 very particular to the specific process technology that's 00:24:42.050 --> 00:24:45.920 arising, but there's a lot of very generic issues 00:24:45.920 --> 00:24:49.580 with variation, and the control, and elimination, 00:24:49.580 --> 00:24:52.550 and modeling of the process, and the variation, 00:24:52.550 --> 00:24:54.230 and ways to reduce it. 00:24:54.230 --> 00:24:58.310 So in some sense, dealing with variation 00:24:58.310 --> 00:25:01.880 is one of the key themes in the subject. 00:25:01.880 --> 00:25:04.520 And what we're going to start with in the first couple 00:25:04.520 --> 00:25:07.970 of lectures is basically dive in, get a little bit 00:25:07.970 --> 00:25:11.030 of process physics background. 00:25:11.030 --> 00:25:13.820 On Thursday, we'll dive in a little bit more 00:25:13.820 --> 00:25:17.060 on semiconductor fabrication, and then, on Tuesday, we'll 00:25:17.060 --> 00:25:19.280 see a variety-- 00:25:19.280 --> 00:25:25.230 Dave will talk about a variety of forging and other process 00:25:25.230 --> 00:25:28.970 examples that give a wider perspective on manufacturing, 00:25:28.970 --> 00:25:32.180 and the physics that are at work, and where 00:25:32.180 --> 00:25:38.610 variation naturally, inherently arises in those processes. 00:25:38.610 --> 00:25:41.900 Then, well, once we've got a little bit of a feel for where 00:25:41.900 --> 00:25:43.970 variation comes from, we want to dive 00:25:43.970 --> 00:25:47.750 into some of the techniques for understanding that and dealing 00:25:47.750 --> 00:25:48.350 with it. 00:25:48.350 --> 00:25:50.990 And that really gets us into statistical models. 00:25:54.950 --> 00:25:58.310 And we'll talk in great length about the contrast 00:25:58.310 --> 00:26:01.700 between physical models of the process based 00:26:01.700 --> 00:26:07.400 on detailed understanding of the specific mechanisms 00:26:07.400 --> 00:26:12.080 at work versus empirical modeling of the process based 00:26:12.080 --> 00:26:13.310 on data-- 00:26:13.310 --> 00:26:16.100 data and observations. 00:26:16.100 --> 00:26:20.810 The wonderful thing about data is it's real 00:26:20.810 --> 00:26:24.740 and it encompasses all of both the ideal behavior 00:26:24.740 --> 00:26:26.840 and the non-ideal behavior. 00:26:26.840 --> 00:26:29.000 And a lot of what we've dealing with, again, 00:26:29.000 --> 00:26:32.000 is the problems, really trying to understand 00:26:32.000 --> 00:26:37.760 where variation is coming from, what its characteristics are. 00:26:37.760 --> 00:26:41.790 If we can start to get a model for some of that variation, 00:26:41.790 --> 00:26:43.280 now we've got a handle that we can 00:26:43.280 --> 00:26:46.270 use to try to eliminate it-- 00:26:46.270 --> 00:26:48.640 either eliminate it up front or control it 00:26:48.640 --> 00:26:51.930 and compensate for it. 00:26:51.930 --> 00:26:54.540 Some of the techniques that we deal with for dealing with, 00:26:54.540 --> 00:26:57.120 again, that data is building effects models, 00:26:57.120 --> 00:26:58.680 designed experiments. 00:26:58.680 --> 00:27:02.010 You may have run across DOE, design of experiments. 00:27:02.010 --> 00:27:04.110 It's a form of statistical technique, 00:27:04.110 --> 00:27:07.560 and we will be learning the details of that 00:27:07.560 --> 00:27:10.410 and applying some of those approaches 00:27:10.410 --> 00:27:14.190 with basic input-output data sets and data models, 00:27:14.190 --> 00:27:19.010 both in specifying the right set of experiments 00:27:19.010 --> 00:27:22.670 to perform to very efficiently sample sample 00:27:22.670 --> 00:27:26.290 our process, and then the ways to model that. 00:27:26.290 --> 00:27:28.040 And perhaps one of the most important ways 00:27:28.040 --> 00:27:31.280 of using those empirical models that we get of the process 00:27:31.280 --> 00:27:34.820 is to improve the process, do process optimization, really 00:27:34.820 --> 00:27:38.960 looking for improved operating points that have reduced 00:27:38.960 --> 00:27:43.550 sensitivity to variation, that really improve the robustness, 00:27:43.550 --> 00:27:46.170 as well as meet multiple objectives. 00:27:46.170 --> 00:27:48.860 So we'll do a little bit of optimization-- multi-objective 00:27:48.860 --> 00:27:52.240 optimization coverage as well. 00:27:55.180 --> 00:27:58.440 So let's talk a little bit about, what are the-- 00:27:58.440 --> 00:28:05.220 oops-- what are the goals of manufacturing process control 00:28:05.220 --> 00:28:06.540 and manufacturing processes? 00:28:06.540 --> 00:28:08.460 What are some of the key characteristics you 00:28:08.460 --> 00:28:11.843 would have of a good process? 00:28:11.843 --> 00:28:13.260 What are the things you would like 00:28:13.260 --> 00:28:15.960 to either minimize or optimize? 00:28:15.960 --> 00:28:21.200 So I already gave you an advanced look at one. 00:28:21.200 --> 00:28:22.250 I'll throw out one. 00:28:22.250 --> 00:28:24.230 Of course, we're very often interested 00:28:24.230 --> 00:28:28.190 in minimizing the cost, and cost rises 00:28:28.190 --> 00:28:29.640 in a lot of different ways. 00:28:29.640 --> 00:28:31.400 What are some of the ways that cost arise? 00:28:31.400 --> 00:28:34.897 There's the inherent materials coming in. 00:28:34.897 --> 00:28:37.355 What are some of the other costs associated with a process? 00:28:40.010 --> 00:28:41.960 Anybody-- Singapore as well-- 00:28:41.960 --> 00:28:42.820 AUDIENCE: Labor. 00:28:42.820 --> 00:28:44.410 DUANE BONING: Labor, yeah. 00:28:44.410 --> 00:28:46.350 AUDIENCE: [INAUDIBLE] overhead, support. 00:28:46.350 --> 00:28:48.660 DUANE BONING: Overhead, support. 00:28:48.660 --> 00:28:49.710 AUDIENCE: Time. 00:28:49.710 --> 00:28:51.540 DUANE BONING: Time. 00:28:51.540 --> 00:28:53.490 Yeah, in fact time process-- 00:28:53.490 --> 00:28:55.710 time is certainly a characteristic 00:28:55.710 --> 00:28:57.480 that we might want to optimize for. 00:28:57.480 --> 00:28:59.490 That would be a typical characteristic. 00:28:59.490 --> 00:29:02.310 Cost has many of these flavors. 00:29:02.310 --> 00:29:03.670 Yes? 00:29:03.670 --> 00:29:04.980 AUDIENCE: Setup costs-- 00:29:04.980 --> 00:29:06.840 DUANE BONING: Setup costs-- 00:29:06.840 --> 00:29:12.420 so setup costs are an interesting one. 00:29:12.420 --> 00:29:14.850 In various times, we'll touch on some 00:29:14.850 --> 00:29:16.830 of these operational issues. 00:29:16.830 --> 00:29:19.950 Many of you have had [? 2853. ?] We said-- 00:29:19.950 --> 00:29:22.830 which was really about manufacturing systems 00:29:22.830 --> 00:29:26.970 and things like scheduling, throughput-- 00:29:26.970 --> 00:29:28.870 those kinds of issues. 00:29:28.870 --> 00:29:32.590 And so time is a little bit of that as well-- 00:29:32.590 --> 00:29:36.110 not just the inherent manufacturing, 00:29:36.110 --> 00:29:39.140 physical action on the product time, 00:29:39.140 --> 00:29:42.650 but also transport time-- all of these kinds of other issues. 00:29:42.650 --> 00:29:47.720 This subject will not be dealing so much with logistics. 00:29:47.720 --> 00:29:50.180 It'll really be much closer to the things 00:29:50.180 --> 00:29:53.400 that impact directly the product itself. 00:29:53.400 --> 00:29:59.010 But certainly, setup time can have an interaction-- 00:29:59.010 --> 00:30:02.330 processes setup in general can have a strong impact 00:30:02.330 --> 00:30:03.320 on processed quality. 00:30:06.550 --> 00:30:10.540 Very often, for example, in semiconductor processing-- 00:30:10.540 --> 00:30:13.870 maybe this will resonate with your experiences 00:30:13.870 --> 00:30:15.190 in other processes-- 00:30:15.190 --> 00:30:19.510 if we change substantially the process setup 00:30:19.510 --> 00:30:22.320 in a particular piece of equipment, 00:30:22.320 --> 00:30:25.440 the first few wafers will behave very differently 00:30:25.440 --> 00:30:28.200 before the equipment gets to a stable state. 00:30:28.200 --> 00:30:30.150 Sometimes we even run dummy wafers 00:30:30.150 --> 00:30:32.370 in order to equilibrate that setup. 00:30:32.370 --> 00:30:33.930 So setup is a good one. 00:30:33.930 --> 00:30:36.930 Some more-- what are other things? 00:30:36.930 --> 00:30:37.760 I'm sorry-- 00:30:37.760 --> 00:30:38.640 AUDIENCE: [INAUDIBLE] 00:30:38.640 --> 00:30:39.840 DUANE BONING: [INAUDIBLE]. 00:30:39.840 --> 00:30:41.610 Certainly. 00:30:41.610 --> 00:30:42.510 AUDIENCE: Equipment. 00:30:42.510 --> 00:30:44.550 DUANE BONING: Equipment, yep. 00:30:44.550 --> 00:30:48.120 We want to use the equipment as efficiently as possible, 00:30:48.120 --> 00:30:49.200 and maintain it. 00:30:49.200 --> 00:30:51.720 AUDIENCE: [INAUDIBLE] a good machine or a bad machine-- 00:30:51.720 --> 00:30:54.180 DUANE BONING: Good machines and bad machines-- absolutely. 00:30:54.180 --> 00:30:56.610 And in fact, probably the first third 00:30:56.610 --> 00:30:59.790 of the class, the classic statistical process control, 00:30:59.790 --> 00:31:02.610 will be all about detecting whether the equipment is 00:31:02.610 --> 00:31:03.660 behaving as it should. 00:31:03.660 --> 00:31:08.425 Is it in a good state or in a bad state? 00:31:08.425 --> 00:31:10.033 AUDIENCE: [INAUDIBLE] 00:31:10.033 --> 00:31:11.950 DUANE BONING: So here we had quality control-- 00:31:11.950 --> 00:31:12.825 AUDIENCE: And rework. 00:31:12.825 --> 00:31:16.480 DUANE BONING: --and rework, yeah. 00:31:16.480 --> 00:31:17.200 AUDIENCE: Energy. 00:31:17.200 --> 00:31:19.150 DUANE BONING: Energy, yeah. 00:31:19.150 --> 00:31:21.370 Yes, that's a good point. 00:31:21.370 --> 00:31:27.760 Actually, I think, dealing with energy, as well 00:31:27.760 --> 00:31:32.230 as the consumption of materials and the output of materials, 00:31:32.230 --> 00:31:33.940 is becoming a more and more interesting 00:31:33.940 --> 00:31:37.760 aspect of manufacturing processes. 00:31:37.760 --> 00:31:38.260 Let's see. 00:31:38.260 --> 00:31:38.760 When was it? 00:31:38.760 --> 00:31:43.360 Sunday night-- just to share another anecdote-- 00:31:43.360 --> 00:31:48.610 I was serving on a panel at a big semiconductor conference. 00:31:48.610 --> 00:31:51.140 Actually, it's the semiconductor-- 00:31:51.140 --> 00:31:53.050 biggest circuit design conference 00:31:53.050 --> 00:31:54.400 in the semiconductor industry. 00:31:54.400 --> 00:31:58.870 And I was on a panel all about environmental or green 00:31:58.870 --> 00:32:00.130 manufacturing. 00:32:00.130 --> 00:32:03.340 And so the reduction of energy usage 00:32:03.340 --> 00:32:07.690 is becoming a very big deal, as well as the output of things-- 00:32:07.690 --> 00:32:11.740 in semiconductor side, output of global warming gases and so 00:32:11.740 --> 00:32:13.190 on-- the reduction of those. 00:32:13.190 --> 00:32:15.700 So that's a good point. 00:32:15.700 --> 00:32:20.020 By the way, I was on a panel at 7:30 on Sunday evening. 00:32:23.060 --> 00:32:26.330 People here-- what happened on Sunday? 00:32:26.330 --> 00:32:27.980 That was Super Bowl here. 00:32:27.980 --> 00:32:31.670 I was not very happy to have to attend the panelist 00:32:31.670 --> 00:32:34.550 dinner the second half of the Super Bowl, 00:32:34.550 --> 00:32:39.270 but as it turned out, I'm very glad I missed it. 00:32:39.270 --> 00:32:42.590 OK, so we've hit some of these other ideas here. 00:32:42.590 --> 00:32:44.840 Here's one that nobody mentioned. 00:32:44.840 --> 00:32:50.180 Flexibility is also very key in manufacturing processes 00:32:50.180 --> 00:32:50.750 as well. 00:32:50.750 --> 00:32:54.170 We're increasingly finding that the same manufacturing 00:32:54.170 --> 00:32:57.650 line needs to be adaptable rapidly 00:32:57.650 --> 00:33:03.510 to be able to deal with a bigger product mix than ever before. 00:33:03.510 --> 00:33:06.570 So some of the focus in this subject, 00:33:06.570 --> 00:33:09.830 again, is going to really be on the processes, and variation, 00:33:09.830 --> 00:33:12.560 and quality in the processes. 00:33:12.560 --> 00:33:18.740 A little bit bigger emphasis is going to be on unit processes. 00:33:18.740 --> 00:33:23.030 The entire aggregation of the overall sequence of product 00:33:23.030 --> 00:33:25.040 to make the overall-- 00:33:25.040 --> 00:33:28.460 or sequence of processes, aggregation of the unit 00:33:28.460 --> 00:33:32.510 processes to make the entire process line, and process flow, 00:33:32.510 --> 00:33:36.120 and the final product is important. 00:33:36.120 --> 00:33:39.020 We'll touch on some of these kinds of things of stack up 00:33:39.020 --> 00:33:42.470 of quality across multiple processes, 00:33:42.470 --> 00:33:47.240 but we'll be-- end up emphasizing developing 00:33:47.240 --> 00:33:51.230 techniques like statistical process control, design 00:33:51.230 --> 00:33:54.740 of experiments and optimization, with a little bit more 00:33:54.740 --> 00:33:58.910 of an emphasis on the unit operations. 00:33:58.910 --> 00:34:05.350 Part of the reason is more and more for really high-quality 00:34:05.350 --> 00:34:06.520 manufacturing-- 00:34:06.520 --> 00:34:08.980 the recognition is you can't simply 00:34:08.980 --> 00:34:12.550 inspect or hope to control at the end of the overall process. 00:34:12.550 --> 00:34:18.040 You have to have every unit process along the way running 00:34:18.040 --> 00:34:23.179 as effectively and at the highest quality possible. 00:34:23.179 --> 00:34:25.900 So again, the number one emphasis 00:34:25.900 --> 00:34:28.600 here is going to be dealing with maximizing 00:34:28.600 --> 00:34:31.988 quality, conformance to specifications, and so on. 00:34:31.988 --> 00:34:34.030 And some of these other things that we've already 00:34:34.030 --> 00:34:35.440 talked about here-- 00:34:35.440 --> 00:34:37.840 things like improving throughput, improving 00:34:37.840 --> 00:34:40.600 flexibility, reducing cost-- 00:34:40.600 --> 00:34:44.199 those are going to be secondary. 00:34:44.199 --> 00:34:47.020 There will be interactions with some of these topics 00:34:47.020 --> 00:34:49.060 that you have run into or might run 00:34:49.060 --> 00:34:51.190 into another in other subjects. 00:34:51.190 --> 00:34:57.060 And actually, a very interesting area of research-- 00:34:57.060 --> 00:34:59.970 in fact, an active area of research of Stan Gershwin, 00:34:59.970 --> 00:35:02.550 who some of you had in 2853-- 00:35:02.550 --> 00:35:06.450 is this interaction between quality and quantity-- 00:35:06.450 --> 00:35:09.570 that is, typical operations and how 00:35:09.570 --> 00:35:15.390 you manage the sequencing of parts through the line-- 00:35:15.390 --> 00:35:19.800 the interaction between throughput and quality-- 00:35:19.800 --> 00:35:23.430 for example, dealing with issues of setup, 00:35:23.430 --> 00:35:27.750 dealing with issues of slight degradations in quality 00:35:27.750 --> 00:35:31.440 and how that might impact things like scheduled maintenance 00:35:31.440 --> 00:35:33.060 or unscheduled maintenance. 00:35:33.060 --> 00:35:34.950 So it's a very interesting topic. 00:35:34.950 --> 00:35:37.680 We'll occasionally, again, overlap slightly with that, 00:35:37.680 --> 00:35:42.480 but our number one emphasis here is dealing with the process, 00:35:42.480 --> 00:35:44.160 understanding and modeling the process, 00:35:44.160 --> 00:35:47.770 and dealing with quality and variation. 00:35:47.770 --> 00:35:52.000 So here's a few typical process control problems. 00:35:52.000 --> 00:35:53.880 I'll share a couple of these. 00:35:53.880 --> 00:35:57.240 And just to wake up Dave, on the next slide, 00:35:57.240 --> 00:35:59.010 we'll let him say a few words about some 00:35:59.010 --> 00:36:01.860 of the other examples. 00:36:01.860 --> 00:36:04.500 So here's a typical example. 00:36:04.500 --> 00:36:08.220 Probably, in fact, the number one problem-- 00:36:08.220 --> 00:36:12.090 or number one challenge in semiconductor manufacturing 00:36:12.090 --> 00:36:15.570 is the minimum feature size is very, very 00:36:15.570 --> 00:36:18.780 small-- becoming smaller with each technology generation. 00:36:18.780 --> 00:36:21.090 The minimum feature size is typically 00:36:21.090 --> 00:36:25.890 the channel length of an individual MOS transistor. 00:36:25.890 --> 00:36:30.690 That's often the most critical parameter, and it varies. 00:36:30.690 --> 00:36:32.940 With current technology, we're down 00:36:32.940 --> 00:36:35.880 in heavy-duty manufacturing at 90 nanometers. 00:36:35.880 --> 00:36:38.430 Some of the leading edge manufacturers 00:36:38.430 --> 00:36:43.740 have gate or channel patterned lengths of 65 nanometer. 00:36:43.740 --> 00:36:46.660 Occasionally, you'll hear those terms thrown out. 00:36:46.660 --> 00:36:50.730 What that means is the minimum feature size being patterned 00:36:50.730 --> 00:36:55.460 is about 65 nanometers wide. 00:36:55.460 --> 00:36:56.480 That's small. 00:36:56.480 --> 00:36:57.590 That's tiny. 00:36:57.590 --> 00:37:00.830 And as we continue to scale, that variability 00:37:00.830 --> 00:37:07.400 in the width of that is becoming tougher and tougher to achieve. 00:37:07.400 --> 00:37:13.550 So very often, there will be challenges on the manufacturing 00:37:13.550 --> 00:37:16.070 line to detect-- 00:37:16.070 --> 00:37:18.500 monitor, detect, and then compensate 00:37:18.500 --> 00:37:20.030 for those kinds of variability. 00:37:22.820 --> 00:37:25.670 An emerging technology-- here's an interesting one 00:37:25.670 --> 00:37:28.020 that connects up with, for example, 00:37:28.020 --> 00:37:31.190 some of the Singapore MIT Alliance research efforts. 00:37:31.190 --> 00:37:36.960 A DNA diagnostic chip might have uneven flow channels. 00:37:36.960 --> 00:37:41.960 So for example, here, these are often microfluidic devices 00:37:41.960 --> 00:37:46.190 manufactured through embossing processes, where 00:37:46.190 --> 00:37:48.770 one might be using a polymer and trying 00:37:48.770 --> 00:37:52.730 to create channels that might be 10 microns wide, 10 00:37:52.730 --> 00:37:54.080 microns deep-- 00:37:54.080 --> 00:37:58.280 and maybe, in some places, very, very narrow, for example, 00:37:58.280 --> 00:37:59.810 or very, very shallow-- 00:37:59.810 --> 00:38:02.840 maybe 20 to 40 nanometers deep. 00:38:02.840 --> 00:38:05.480 I mention those because those are actually 00:38:05.480 --> 00:38:09.440 real numbers for some of the researchships 00:38:09.440 --> 00:38:15.720 that are being developed here at MIT in electrical engineering, 00:38:15.720 --> 00:38:18.410 mechanical engineering, and elsewhere-- very interesting 00:38:18.410 --> 00:38:22.040 properties, when one gets down to that small dimension. 00:38:22.040 --> 00:38:23.630 And some of those properties depend 00:38:23.630 --> 00:38:28.400 critically on those dimensions, as you can imagine. 00:38:28.400 --> 00:38:31.220 So in fact, the manufacturing process problem 00:38:31.220 --> 00:38:34.970 is, how do you get processes that 00:38:34.970 --> 00:38:36.600 work in a controllable fashion? 00:38:36.600 --> 00:38:43.220 How do you inspect and detect feature sizes, 00:38:43.220 --> 00:38:47.300 geometric parameters, roughness of the geometry, 00:38:47.300 --> 00:38:51.950 as well as some other material properties in order 00:38:51.950 --> 00:38:54.530 to achieve the optimal functionality 00:38:54.530 --> 00:38:56.630 that one might need-- 00:38:56.630 --> 00:39:03.250 for example, controlling the flow of individual DNA from one 00:39:03.250 --> 00:39:08.890 place to another through these incredibly tiny, tiny channels? 00:39:08.890 --> 00:39:10.630 What's another example? 00:39:10.630 --> 00:39:13.130 Somebody mentioned assembly. 00:39:13.130 --> 00:39:16.480 Assembly is a manufacturing process, 00:39:16.480 --> 00:39:22.040 and the worst example of this is that toys that come 00:39:22.040 --> 00:39:24.980 disassembled-- they never fit, at least when 00:39:24.980 --> 00:39:29.180 I try to assemble them at home. 00:39:29.180 --> 00:39:36.350 Geometric fit and how those imperfections can build up, 00:39:36.350 --> 00:39:37.430 even if-- 00:39:37.430 --> 00:39:41.150 an interesting aspect-- even if a few parts 00:39:41.150 --> 00:39:44.600 are close to nominal when you try to put them together, 00:39:44.600 --> 00:39:49.918 sometimes the aggregate doesn't work out so well. 00:39:49.918 --> 00:39:51.460 And then there may be other example-- 00:39:51.460 --> 00:39:55.515 for example, very high-density electrical connectors. 00:39:55.515 --> 00:39:57.940 And I think, Dave, you had some example-- 00:39:57.940 --> 00:39:59.503 this comes out of an LFM thesis, or? 00:39:59.503 --> 00:40:00.920 DAVID HARDT: No, this was actually 00:40:00.920 --> 00:40:03.690 research we did [INAUDIBLE] a number of years ago. 00:40:03.690 --> 00:40:05.970 And it's actually just like-- 00:40:05.970 --> 00:40:07.470 it's actually a trend with all these 00:40:07.470 --> 00:40:11.333 that [INAUDIBLE] for years. 00:40:11.333 --> 00:40:13.420 Of course, [INAUDIBLE]. 00:40:13.420 --> 00:40:17.490 DUANE BONING: Can you guys in Singapore hear Dave? 00:40:17.490 --> 00:40:18.000 No? 00:40:18.000 --> 00:40:19.208 AUDIENCE: No, we cannot hear. 00:40:19.208 --> 00:40:20.317 AUDIENCE: [INAUDIBLE] 00:40:20.317 --> 00:40:24.120 DAVID HARDT: [INAUDIBLE] mic's not working. 00:40:24.120 --> 00:40:28.200 Just because this is sort of a theme-- 00:40:28.200 --> 00:40:30.990 you can imagine electrical connectors-- big heavy ones, 00:40:30.990 --> 00:40:34.560 like used in cars or in old electronics back in the tube 00:40:34.560 --> 00:40:37.740 days, where you could grab hold of some of these connectors 00:40:37.740 --> 00:40:38.940 with your hand-- 00:40:38.940 --> 00:40:42.250 just the metal part itself. 00:40:42.250 --> 00:40:45.300 Now, as electronics has become more and more dense, 00:40:45.300 --> 00:40:47.130 as things have become smaller and smaller, 00:40:47.130 --> 00:40:50.310 the requirement for higher and higher density of connectors-- 00:40:50.310 --> 00:40:53.803 if not just to handle these high-density chips, 00:40:53.803 --> 00:40:56.220 but just look at the back of your computer-- some of those 00:40:56.220 --> 00:40:59.100 connectors there-- particularly the interconnect type 00:40:59.100 --> 00:41:00.870 of connectors. 00:41:00.870 --> 00:41:02.460 Well, the industry that made that 00:41:02.460 --> 00:41:06.090 made it the same way for years, and it worked fine. 00:41:06.090 --> 00:41:10.860 And it's a great example of how variability is actually 00:41:10.860 --> 00:41:12.340 a relative term. 00:41:12.340 --> 00:41:16.440 And so the variability was relatively small. 00:41:16.440 --> 00:41:19.140 It was small relative to the characteristic dimension 00:41:19.140 --> 00:41:20.100 that you needed. 00:41:20.100 --> 00:41:24.330 But as that shrunk down, these major manufacturers 00:41:24.330 --> 00:41:26.100 found out that they really couldn't 00:41:26.100 --> 00:41:28.373 use their old technology to meet this. 00:41:28.373 --> 00:41:30.540 Of course, they went through a lot of the procedures 00:41:30.540 --> 00:41:33.180 we're going to go through and they improved things, 00:41:33.180 --> 00:41:34.920 but they couldn't meet those dimensions, 00:41:34.920 --> 00:41:38.772 and had to make a technology change as well. 00:41:38.772 --> 00:41:40.230 DUANE BONING: This is my sneaky way 00:41:40.230 --> 00:41:42.900 to get an extra [INAUDIBLE]. 00:41:42.900 --> 00:41:45.660 DAVID HARDT: Also, on this minimum feature size-- 00:41:45.660 --> 00:41:47.370 and I'd ask Duane to comment on this-- 00:41:47.370 --> 00:41:51.210 I was once with a gentleman from IBM Research 00:41:51.210 --> 00:41:53.280 who told me they were just now-- this was maybe 00:41:53.280 --> 00:41:54.930 five, six years ago-- just now getting 00:41:54.930 --> 00:41:56.280 into some of the advanced techniques 00:41:56.280 --> 00:41:58.860 that we're going to talk about near the end of the term here. 00:41:58.860 --> 00:42:00.390 And I was a little bit surprised, 00:42:00.390 --> 00:42:03.540 but then he pointed out, back when the feature 00:42:03.540 --> 00:42:05.730 size on the chips was large-- 00:42:05.730 --> 00:42:09.240 I don't know what large was-- micron scale or several 00:42:09.240 --> 00:42:10.380 microns-- 00:42:10.380 --> 00:42:14.340 the inherent variability of the processes was fine-- 00:42:14.340 --> 00:42:15.910 fine to manufacture these things. 00:42:15.910 --> 00:42:17.850 But when you get down to the nanometer range, 00:42:17.850 --> 00:42:22.120 all of a sudden, you have to look at some other techniques. 00:42:22.120 --> 00:42:26.370 So there are these timeless things, 00:42:26.370 --> 00:42:28.710 like toys that don't fit together 00:42:28.710 --> 00:42:31.410 and other things related to larger dimensions, 00:42:31.410 --> 00:42:34.470 but then this idea of the ever-shrinking mechanical 00:42:34.470 --> 00:42:38.010 dimensions of things leads to an ever-increasing importance 00:42:38.010 --> 00:42:40.683 of process control. 00:42:40.683 --> 00:42:41.850 So our future is guaranteed. 00:42:41.850 --> 00:42:44.220 DUANE BONING: I'll skip ahead, because maybe [? not ?] 00:42:44.220 --> 00:42:47.250 going through all of these, but if there's another one or two 00:42:47.250 --> 00:42:49.708 of these from your experience that you wanted [INAUDIBLE].. 00:42:49.708 --> 00:42:54.930 DAVID HARDT: Well, these are three large-scale issues. 00:42:54.930 --> 00:42:56.790 This is usually where I ask if anybody here 00:42:56.790 --> 00:43:00.337 works for Boeing or any other airframe manufacturer. 00:43:00.337 --> 00:43:01.920 DUANE BONING: We have some automotive. 00:43:01.920 --> 00:43:03.180 We have some [INAUDIBLE]. 00:43:03.180 --> 00:43:04.380 DAVID HARDT: Yeah. 00:43:04.380 --> 00:43:06.720 The automotive actually does a lot better 00:43:06.720 --> 00:43:09.120 on variation control, because they have-- 00:43:09.120 --> 00:43:12.450 as you'll see, it helps to have some volume 00:43:12.450 --> 00:43:15.330 and to be able to learn and to get your processes 00:43:15.330 --> 00:43:16.320 in steady state. 00:43:16.320 --> 00:43:19.800 Things like airframes are still sort of one-off. 00:43:19.800 --> 00:43:21.805 And because of the large dimension there, 00:43:21.805 --> 00:43:24.180 it's really not an unpleasant-- it's not a pleasant thing 00:43:24.180 --> 00:43:27.443 to watch two halves of an airplane being put together-- 00:43:27.443 --> 00:43:29.860 although they've gotten a lot better than they used to be. 00:43:29.860 --> 00:43:33.720 What was it they used to say-- that the number of thousands 00:43:33.720 --> 00:43:36.600 of pounds [INAUDIBLE] flying in a 747? 00:43:36.600 --> 00:43:39.092 [INAUDIBLE] we're also going to talk later-- 00:43:39.092 --> 00:43:40.800 actually, I'm not sure we will this term, 00:43:40.800 --> 00:43:43.200 but there's an interesting LFM thesis 00:43:43.200 --> 00:43:45.708 done on these plastic throttle bodies for fuel injection. 00:43:45.708 --> 00:43:47.250 This is a little bit different issue. 00:43:47.250 --> 00:43:50.690 This was an entirely different way of making a classical part. 00:43:50.690 --> 00:43:54.390 It had been made for years out of aluminum-- 00:43:54.390 --> 00:43:56.670 machined aluminum-- well-defined process, 00:43:56.670 --> 00:43:58.525 well-understood process control. 00:43:58.525 --> 00:44:00.150 Then, for good and sufficient reasons-- 00:44:00.150 --> 00:44:02.370 mainly weight and production economy-- 00:44:02.370 --> 00:44:05.220 they went to injection molded parts, 00:44:05.220 --> 00:44:07.170 and variation became a big problem. 00:44:07.170 --> 00:44:09.990 And how to get rid of it was a-- was and still 00:44:09.990 --> 00:44:11.050 is an interesting issue. 00:44:11.050 --> 00:44:15.900 So a number of things come up, often because of change-- 00:44:15.900 --> 00:44:18.500 not always just because of things being in stasis. 00:44:18.500 --> 00:44:19.500 DUANE BONING: Thank you. 00:44:24.120 --> 00:44:29.070 So again, some of these points that Dave 00:44:29.070 --> 00:44:32.492 made about, for example, the question or example you posed 00:44:32.492 --> 00:44:34.950 on semiconductor manufacturing we'll talk a little bit more 00:44:34.950 --> 00:44:36.450 about on Thursday, because it was 00:44:36.450 --> 00:44:38.130 a very interesting transformation 00:44:38.130 --> 00:44:42.960 from defect-oriented problems as the real yield and quality 00:44:42.960 --> 00:44:46.170 limiter to dimensional control. 00:44:46.170 --> 00:44:54.020 And essentially, parametric variation is the killer now. 00:44:54.020 --> 00:44:56.760 OK, so one of the things-- 00:44:56.760 --> 00:45:00.960 this connects up again also to a point that Dave just made-- 00:45:00.960 --> 00:45:05.820 is that there is an evolution of manufacturing process control 00:45:05.820 --> 00:45:07.920 many industries had gone through. 00:45:07.920 --> 00:45:10.710 And you will often see a trajectory 00:45:10.710 --> 00:45:15.570 where an industry may actually start with a process that 00:45:15.570 --> 00:45:18.270 is not all that stable, and they essentially 00:45:18.270 --> 00:45:21.390 have to inspect every single part that comes out-- 00:45:21.390 --> 00:45:25.950 100% inspection may have relatively high scrap rates-- 00:45:25.950 --> 00:45:28.860 even a few percent, for example-- 00:45:28.860 --> 00:45:35.330 with low throughput, and correspondingly high cost. 00:45:35.330 --> 00:45:38.300 So this might be thought to be a characteristic 00:45:38.300 --> 00:45:43.430 of immature processes, but it's also 00:45:43.430 --> 00:45:48.200 a characteristic of incredibly complex processes. 00:45:48.200 --> 00:45:50.810 So you would really like to avoid that. 00:45:50.810 --> 00:45:53.510 Quick question for-- well, some of the guys from semiconductor 00:45:53.510 --> 00:45:54.590 know this. 00:45:54.590 --> 00:45:59.060 Do we have 100% inspection of semiconductor chips today? 00:46:02.830 --> 00:46:04.412 What do you think? 00:46:04.412 --> 00:46:05.870 Do you think every single chip that 00:46:05.870 --> 00:46:09.140 comes off the line gets inspected, measured, 00:46:09.140 --> 00:46:10.910 verified that it produces? 00:46:10.910 --> 00:46:13.910 Do we have high costs or-- 00:46:16.990 --> 00:46:20.060 somebody who knows the answer-- 00:46:20.060 --> 00:46:23.800 well, I think perhaps it depends on what the chip is. 00:46:23.800 --> 00:46:25.690 You think every single microprocessor 00:46:25.690 --> 00:46:30.940 chip gets inspected before it gets packaged and put in a PC? 00:46:30.940 --> 00:46:32.470 You bet. 00:46:32.470 --> 00:46:34.930 You bet. 00:46:34.930 --> 00:46:36.925 On the other hand-- 00:46:36.925 --> 00:46:37.800 AUDIENCE: [INAUDIBLE] 00:46:37.800 --> 00:46:39.425 DUANE BONING: Oh, absolutely. 00:46:46.322 --> 00:46:49.310 A microprocessor chip might sell for a couple hundred dollars, 00:46:49.310 --> 00:46:50.550 so it's worth it. 00:46:50.550 --> 00:46:53.720 Now, if you're talking about a semiconductor chip that 00:46:53.720 --> 00:46:57.770 is a discrete diode worth probably 00:46:57.770 --> 00:47:01.160 a fraction of a penny in terms of the manufacturing 00:47:01.160 --> 00:47:03.380 cost, yeah, you-- 00:47:03.380 --> 00:47:06.560 it's a much more simple and less complex thing, 00:47:06.560 --> 00:47:09.740 and you probably don't need 100% inspection there. 00:47:09.740 --> 00:47:14.330 But most integrated circuits, at the end of the day, 00:47:14.330 --> 00:47:16.630 still need inspection at the end. 00:47:16.630 --> 00:47:17.436 Yeah, Hayden? 00:47:17.936 --> 00:47:19.686 HAYDEN TAYLOR: I thought it might be worth 00:47:19.686 --> 00:47:21.465 remarking that I think-- 00:47:21.465 --> 00:47:22.590 DUANE BONING: Speak loudly. 00:47:22.590 --> 00:47:25.610 AUDIENCE: Can everyone hear me in Singapore? 00:47:25.610 --> 00:47:27.233 No. 00:47:27.233 --> 00:47:28.275 DUANE BONING: Just shout. 00:47:28.775 --> 00:47:29.610 HAYDEN TAYLOR: OK. 00:47:29.610 --> 00:47:33.060 I think Intel, when they test their chips, 00:47:33.060 --> 00:47:36.480 they determine what the maximum operating 00:47:36.480 --> 00:47:39.570 frequency of any given chip is, and they sort them 00:47:39.570 --> 00:47:42.460 so the chips that happen to operate at the faster end 00:47:42.460 --> 00:47:44.820 are put in more expensive machines, 00:47:44.820 --> 00:47:47.580 and those that are slower go in the cheaper machines. 00:47:47.580 --> 00:47:50.160 So they actually get some extra benefit 00:47:50.160 --> 00:47:53.077 from the testing process. 00:47:53.077 --> 00:47:53.910 DUANE BONING: Right. 00:47:53.910 --> 00:47:56.340 So based on the quality or the performance, 00:47:56.340 --> 00:48:01.000 they can bin the chips and get differential pricing. 00:48:01.000 --> 00:48:05.370 Of course, they would be most beneficial if all of the chips 00:48:05.370 --> 00:48:07.200 operated at the high end. 00:48:07.200 --> 00:48:12.240 And so in some sense, that's both a limitation 00:48:12.240 --> 00:48:14.040 and an advantage. 00:48:14.040 --> 00:48:17.970 You can get a premium for the highest performing. 00:48:17.970 --> 00:48:22.560 But it's a great example of the range of variation 00:48:22.560 --> 00:48:25.320 and the effective, ultimately, on the end performance. 00:48:29.380 --> 00:48:32.140 A next step typically in some of the evolution 00:48:32.140 --> 00:48:35.590 of these manufacturing processes involve-- 00:48:35.590 --> 00:48:40.900 whoops-- involve rework, both at the end of the line, 00:48:40.900 --> 00:48:44.110 but more often pushing down to unit processes-- 00:48:44.110 --> 00:48:46.810 where, if you could do a inspection, 00:48:46.810 --> 00:48:48.610 you still have limited control perhaps 00:48:48.610 --> 00:48:50.500 over the individual unit process, 00:48:50.500 --> 00:48:53.140 but you can inspect it, determine something went wrong. 00:48:53.140 --> 00:48:56.170 And if there is a possibility of fixing it then and there, 00:48:56.170 --> 00:48:59.710 now you've got a big advantage. 00:48:59.710 --> 00:49:03.460 We already talked about also high durability at changeover, 00:49:03.460 --> 00:49:06.055 and that's an example we already talked about, 00:49:06.055 --> 00:49:10.180 about how cost rate and flexibility are linked. 00:49:10.180 --> 00:49:14.440 Of course, all of these-- if you can get quality up, 00:49:14.440 --> 00:49:17.020 get to higher and higher yield-- 00:49:17.020 --> 00:49:20.620 that has a dramatic effect, perhaps the strongest effect 00:49:20.620 --> 00:49:25.720 on overall throughput and the optimal output 00:49:25.720 --> 00:49:27.520 from your manufacturing plant-- 00:49:27.520 --> 00:49:30.580 especially in those industries where yields may not 00:49:30.580 --> 00:49:33.700 be all that high. 00:49:33.700 --> 00:49:36.940 For example, in many of the semiconductor processes, 00:49:36.940 --> 00:49:41.500 if you have 90% yield on a very complex microprocessor, 00:49:41.500 --> 00:49:43.080 you're doing pretty well. 00:49:43.080 --> 00:49:44.930 That's not great, is it? 00:49:44.930 --> 00:49:50.950 Can you imagine 90% output of-- 00:49:50.950 --> 00:49:56.580 90% functioning output of syringe needles 00:49:56.580 --> 00:50:00.270 that we saw, where 10% of them were defective in some way? 00:50:00.270 --> 00:50:02.400 That wouldn't be very satisfying, I think. 00:50:04.930 --> 00:50:08.000 OK, so one of the things that we're 00:50:08.000 --> 00:50:11.420 going to do a little bit in the next couple of lectures-- 00:50:11.420 --> 00:50:13.910 and I'm going to talk about a little bit generically 00:50:13.910 --> 00:50:17.240 to set the stage for that last 10 minutes or so of class-- 00:50:17.240 --> 00:50:22.550 is characteristics of manufacturing processes-- 00:50:22.550 --> 00:50:25.520 in fact, something of a taxonomy that Dave 00:50:25.520 --> 00:50:30.590 will cover on Thursday based on some 00:50:30.590 --> 00:50:34.850 of the inherent characteristics of the physical action 00:50:34.850 --> 00:50:37.460 of manufacturing processes. 00:50:37.460 --> 00:50:39.110 And we've already chatted a little bit, 00:50:39.110 --> 00:50:42.680 and we'll see more as we dive into those processes 00:50:42.680 --> 00:50:46.500 about why they don't always work correctly. 00:50:46.500 --> 00:50:49.760 So here I'm going to try to get a little bit of the terminology 00:50:49.760 --> 00:50:53.330 out and give you a little bit of a framework for thinking 00:50:53.330 --> 00:50:56.240 about the processes, and then some reflection 00:50:56.240 --> 00:50:59.720 on the different segments of the subject 00:50:59.720 --> 00:51:03.290 that we mentioned are coming up, and how 00:51:03.290 --> 00:51:08.910 they relate to a very-high level generic view of the process. 00:51:08.910 --> 00:51:10.860 So the key idea in a process, of course, 00:51:10.860 --> 00:51:15.830 is that we have some part that we want to produce at the end. 00:51:15.830 --> 00:51:18.680 We do that by working on a work piece. 00:51:18.680 --> 00:51:22.970 There is some physical material that we are transforming 00:51:22.970 --> 00:51:25.100 in some important way. 00:51:25.100 --> 00:51:26.930 We are changing the geometry. 00:51:26.930 --> 00:51:30.950 We are changing the material properties of that part. 00:51:30.950 --> 00:51:32.250 How do we do that? 00:51:32.250 --> 00:51:35.520 Well, we have to subjected to some process, 00:51:35.520 --> 00:51:42.040 and we do that through equipment, and in some cases, 00:51:42.040 --> 00:51:43.600 parts of the equipment-- things like 00:51:43.600 --> 00:51:48.310 tooling that come in contact with the part. 00:51:48.310 --> 00:51:53.170 So some examples of broadly, equipment-- and here, 00:51:53.170 --> 00:51:55.960 equipment we may even rise a little bit 00:51:55.960 --> 00:52:01.990 as both the physical machinery and the process environment 00:52:01.990 --> 00:52:05.620 that that machinery generates around the work piece. 00:52:05.620 --> 00:52:08.810 So examples might be an an etch bath. 00:52:08.810 --> 00:52:13.460 So that is a liquid chemical environment 00:52:13.460 --> 00:52:15.080 that might be done under the control 00:52:15.080 --> 00:52:18.870 of a piece of equipment. 00:52:18.870 --> 00:52:21.930 Other equipment that many of you may have come in contact with-- 00:52:21.930 --> 00:52:26.490 things like injection molders, lathes, drop presses-- 00:52:26.490 --> 00:52:30.060 and the whole goal of those are to act on some workpiece. 00:52:30.060 --> 00:52:33.300 So we may have silicon with certain layers already coded 00:52:33.300 --> 00:52:34.020 on them. 00:52:34.020 --> 00:52:36.990 We may have feedstock, things like plastic pellets 00:52:36.990 --> 00:52:39.120 or [INAUDIBLE], that come in. 00:52:39.120 --> 00:52:40.260 We may have sheet metal. 00:52:40.260 --> 00:52:41.970 Sheet metal already may have gone 00:52:41.970 --> 00:52:44.040 through an awful lot of processing, 00:52:44.040 --> 00:52:46.080 and we're looking at a unit step that's 00:52:46.080 --> 00:52:49.810 doing some additional processing on that. 00:52:49.810 --> 00:52:55.560 And then the output can either be a finished part 00:52:55.560 --> 00:53:00.090 direct for use or a part that is in some intermediate stage that 00:53:00.090 --> 00:53:03.630 will, again, go and be used in some larger manufacturing 00:53:03.630 --> 00:53:04.540 process. 00:53:04.540 --> 00:53:06.420 So it might be a shaft or a hood or something 00:53:06.420 --> 00:53:08.940 like that that's going to be further assembled, 00:53:08.940 --> 00:53:10.440 or an IC chip-- 00:53:10.440 --> 00:53:12.750 which, from the semiconductor fab, 00:53:12.750 --> 00:53:14.400 that's the finished product. 00:53:14.400 --> 00:53:17.130 But from somebody-- can't remember who said they worked 00:53:17.130 --> 00:53:19.080 at Dell-- 00:53:19.080 --> 00:53:22.470 the IC chip-- that's not a finished product. 00:53:22.470 --> 00:53:25.680 The PC and the assembly-- 00:53:25.680 --> 00:53:28.390 the packaging and assembly onto boards and the assembly 00:53:28.390 --> 00:53:31.920 of boards into the [? back plane ?] and so on 00:53:31.920 --> 00:53:35.980 are an important manufacturing process. 00:53:35.980 --> 00:53:40.560 So by definition, we're thinking of a manufacturing process 00:53:40.560 --> 00:53:45.240 as a change or the sequence of changes in some work piece 00:53:45.240 --> 00:53:46.510 material. 00:53:46.510 --> 00:53:48.390 The easiest way to conceptualize this 00:53:48.390 --> 00:53:51.180 is that it's going to be a change in geometry. 00:53:51.180 --> 00:53:53.610 We're building up some part. 00:53:53.610 --> 00:53:56.790 But equally important is a change 00:53:56.790 --> 00:54:00.810 in some material properties, some constitutive properties 00:54:00.810 --> 00:54:02.410 of that structure. 00:54:02.410 --> 00:54:04.410 We'll see lots of examples of that 00:54:04.410 --> 00:54:06.540 in semiconductor fabrication. 00:54:06.540 --> 00:54:09.750 Here I want to give you a kind of a generic conceptual 00:54:09.750 --> 00:54:11.580 semiconductor process model. 00:54:11.580 --> 00:54:13.710 And actually, I want to mention a little bit-- 00:54:13.710 --> 00:54:17.520 this ties into the history of the evolution of this subject. 00:54:17.520 --> 00:54:21.420 Back-- oh, God-- it's getting to be 18 years ago already-- 00:54:24.060 --> 00:54:27.540 when we were starting to look at ways to formalize discussion 00:54:27.540 --> 00:54:31.200 of semiconductor processing, we came up with-- 00:54:31.200 --> 00:54:34.140 myself and some other co-workers here at MIT 00:54:34.140 --> 00:54:37.080 came up with some terminology for this conceptual 00:54:37.080 --> 00:54:39.000 semiconductor process model. 00:54:39.000 --> 00:54:41.280 And we talked about it in terms of states 00:54:41.280 --> 00:54:43.830 and transformations of states, these states 00:54:43.830 --> 00:54:47.610 being the geometric state and the constitutive properties 00:54:47.610 --> 00:54:51.300 of the work piece In semiconductor fabrication, 00:54:51.300 --> 00:54:53.760 that's typically the wafer. 00:54:53.760 --> 00:54:56.010 And around that, one is generating 00:54:56.010 --> 00:54:59.040 a process or some environment around the wafer. 00:54:59.040 --> 00:55:03.060 That may be gases, temperature environment, 00:55:03.060 --> 00:55:09.300 other kinds of ways of generating and directing 00:55:09.300 --> 00:55:13.750 energy and material at the surface of the wafer. 00:55:13.750 --> 00:55:17.500 The machine itself is a key controlling parameter for how 00:55:17.500 --> 00:55:20.920 one generates that environment, and the rest of the facility 00:55:20.920 --> 00:55:25.390 also impacts that environment-- the feed material coming in, 00:55:25.390 --> 00:55:29.560 the very, very high purity chemicals, for example. 00:55:29.560 --> 00:55:31.480 Now, what's important about the machine 00:55:31.480 --> 00:55:35.980 is that the operator only has limited access. 00:55:35.980 --> 00:55:39.700 The operator, which may be human or may be an automation system, 00:55:39.700 --> 00:55:44.860 has only limited access to that equipment and the facility 00:55:44.860 --> 00:55:48.860 through some settings-- some knob settings, if you will, 00:55:48.860 --> 00:55:51.370 that you can perform on the equipment. 00:55:51.370 --> 00:55:56.080 But you also have other directional sensing capability. 00:55:56.080 --> 00:55:59.650 You have some number of readings that are telling you indirectly 00:55:59.650 --> 00:56:01.930 things about the machine-- 00:56:01.930 --> 00:56:05.440 machine states, sensor states, thermocouple states, 00:56:05.440 --> 00:56:08.260 which are telling you things about the wafer environment-- 00:56:08.260 --> 00:56:11.110 and then, indirectly-- if you're lucky, directly 00:56:11.110 --> 00:56:12.985 things about the state of the work 00:56:12.985 --> 00:56:18.190 piece itself, telling you things like the thickness of the film 00:56:18.190 --> 00:56:20.230 being grown on the wafer. 00:56:20.230 --> 00:56:24.220 And then we had a term, which is really kind 00:56:24.220 --> 00:56:28.120 generic to semiconductor processing, of a recipe that 00:56:28.120 --> 00:56:32.980 basically defined the settings, and perhaps even could 00:56:32.980 --> 00:56:35.203 be generalized to a control algorithm, 00:56:35.203 --> 00:56:37.120 if you will-- which might be a little bit more 00:56:37.120 --> 00:56:40.570 of a generic terminology-- for dealing 00:56:40.570 --> 00:56:42.880 with responses to the readings. 00:56:42.880 --> 00:56:46.180 So now you can think about things like real-time control-- 00:56:46.180 --> 00:56:48.940 or run by run control, which we'll talk about later-- 00:56:48.940 --> 00:56:52.390 that looks at how one determines what the settings should 00:56:52.390 --> 00:56:55.180 be on the equipment. 00:56:55.180 --> 00:56:59.050 Now, we came up with a slightly simplified version 00:56:59.050 --> 00:57:01.930 of the semiconductor process model just focused 00:57:01.930 --> 00:57:05.590 on the wafer, the process, and the settings. 00:57:05.590 --> 00:57:10.352 And I want to give you a real quick example of this. 00:57:10.352 --> 00:57:12.310 Actually, I'll probably skip this and come back 00:57:12.310 --> 00:57:13.880 to this next time. 00:57:13.880 --> 00:57:18.700 This is looking at examples for one particular process dealing 00:57:18.700 --> 00:57:20.420 with oxidation. 00:57:20.420 --> 00:57:22.360 Now, what was interesting is we came up 00:57:22.360 --> 00:57:24.110 with this generic process model. 00:57:24.110 --> 00:57:26.350 This was early '90s. 00:57:26.350 --> 00:57:28.390 When I came back to MIT, I started teaching 00:57:28.390 --> 00:57:31.330 a subject on semiconductor process 00:57:31.330 --> 00:57:34.420 control dealing with statistical process modeling, 00:57:34.420 --> 00:57:37.810 yield, design of experiments, and so on. 00:57:37.810 --> 00:57:39.640 And then I got involved in some research 00:57:39.640 --> 00:57:42.370 that overlapped with Dave Hardt and some others. 00:57:42.370 --> 00:57:45.610 And he had a course, and-- 00:57:45.610 --> 00:57:47.890 on manufacturing process control, 00:57:47.890 --> 00:57:52.990 and he had a process model for control circa 1995. 00:57:52.990 --> 00:57:55.780 Maybe you had come up with it even earlier. 00:57:55.780 --> 00:57:59.470 And one of the readings that you need to grab off of the website 00:57:59.470 --> 00:58:03.580 is an overview of manufacturing processes that 00:58:03.580 --> 00:58:08.020 is circa 1995, '96, where the terminology is almost 00:58:08.020 --> 00:58:10.430 exactly the same. 00:58:10.430 --> 00:58:14.290 And in fact, the overlap was astonishing. 00:58:14.290 --> 00:58:16.810 The tools and techniques that I was teaching 00:58:16.810 --> 00:58:19.600 in my separate subject and Dave was teaching in his subject 00:58:19.600 --> 00:58:23.350 was about 80% the same, and then just the process details 00:58:23.350 --> 00:58:24.670 kind of changed. 00:58:24.670 --> 00:58:26.950 So in fact, we merged the subject, 00:58:26.950 --> 00:58:30.520 and that's how this subject came about. 00:58:30.520 --> 00:58:34.120 And again, the key ideas here are there are controls. 00:58:34.120 --> 00:58:36.130 There is the operator, and some settings, 00:58:36.130 --> 00:58:38.320 and the equipment, material-- 00:58:38.320 --> 00:58:41.170 that we're all about changing the geometry and property 00:58:41.170 --> 00:58:44.680 through the action of the process environment, 00:58:44.680 --> 00:58:46.457 or these energy states. 00:58:46.457 --> 00:58:48.040 Now, in this class, we're really going 00:58:48.040 --> 00:58:50.770 to be focused a lot on the overall process 00:58:50.770 --> 00:58:53.500 outputs [INAUDIBLE] processes, and in some cases, 00:58:53.500 --> 00:58:55.790 the aggregation of those. 00:58:55.790 --> 00:58:58.240 And just to get a little bit of terminology 00:58:58.240 --> 00:59:03.700 as a prelude for both Thursday and next Tuesday, 00:59:03.700 --> 00:59:07.570 one can think about a vector y of characteristics 00:59:07.570 --> 00:59:10.000 of the product that are important-- 00:59:10.000 --> 00:59:13.000 shape parameters, particular thicknesses, 00:59:13.000 --> 00:59:16.150 perhaps material properties-- like index of refraction 00:59:16.150 --> 00:59:17.710 of thin films-- 00:59:17.710 --> 00:59:22.630 the conductivity of particular layers, as well as transistor 00:59:22.630 --> 00:59:24.920 characteristics, for example. 00:59:24.920 --> 00:59:28.900 And that is a function of the process parameters, 00:59:28.900 --> 00:59:31.840 both controls that one can change 00:59:31.840 --> 00:59:37.040 and other parameters that may be fixed and set for the process. 00:59:37.040 --> 00:59:39.610 So a key question is, what are these alphas? 00:59:39.610 --> 00:59:41.350 What are these process parameters? 00:59:44.990 --> 00:59:46.880 And how do we go about controlling those? 00:59:46.880 --> 00:59:51.660 So we'll touch on this picture a little bit more next time, 00:59:51.660 --> 00:59:56.990 but what I want to skip to here is a-- 00:59:56.990 --> 01:00:01.550 we'll hear about these different characterizations or taxonomy 01:00:01.550 --> 01:00:02.810 of the process next time. 01:00:05.570 --> 01:00:08.900 I want to give you one last description 01:00:08.900 --> 01:00:15.590 building on this very simplified mathematical terminology 01:00:15.590 --> 01:00:23.510 at least for the process, which is we can go in and-- 01:00:23.510 --> 01:00:24.750 here we go. 01:00:24.750 --> 01:00:25.580 Here we go. 01:00:25.580 --> 01:00:28.970 We can split out some of those process parameters 01:00:28.970 --> 01:00:32.420 as those things that we have control over 01:00:32.420 --> 01:00:34.940 as inputs to the process. 01:00:34.940 --> 01:00:39.560 And now, if we do just a very, very simple first-order Taylor 01:00:39.560 --> 01:00:43.230 expansion, we get some nice insight 01:00:43.230 --> 01:00:47.650 into where variation comes in a typical process. 01:00:47.650 --> 01:00:52.070 So this is the very simple variation equation 01:00:52.070 --> 01:00:53.600 that, in some sense, helps structure 01:00:53.600 --> 01:00:56.270 what we're going to be doing through the rest of the term. 01:00:56.270 --> 01:01:00.320 One can think about deviations in the output 01:01:00.320 --> 01:01:03.950 deltas in those characteristics, and we 01:01:03.950 --> 01:01:07.100 can expand that in terms of some of the alpha parameters 01:01:07.100 --> 01:01:11.300 and some of the controllable parameters. 01:01:11.300 --> 01:01:18.670 So this dy d alpha-- that's sensitivity to delta a alphas. 01:01:18.670 --> 01:01:19.600 What's delta alphas? 01:01:19.600 --> 01:01:23.780 Those are disturbances that you would prefer were not there. 01:01:23.780 --> 01:01:26.320 They're not things that you intentionally are changing. 01:01:26.320 --> 01:01:30.320 Those are inherent disturbances in the process. 01:01:30.320 --> 01:01:34.300 So there's one component that deals with disturbances. 01:01:34.300 --> 01:01:36.670 And then there's also intentional changes 01:01:36.670 --> 01:01:41.020 we might have, when we want to or intentionally make a change 01:01:41.020 --> 01:01:43.190 to the control inputs. 01:01:43.190 --> 01:01:47.030 And so what's interesting is the strategies that we might use 01:01:47.030 --> 01:01:51.070 for minimizing this delta y. 01:01:51.070 --> 01:01:54.280 And in some sense-- 01:01:54.280 --> 01:01:57.470 let's see. 01:01:57.470 --> 01:02:00.470 Let me get to a nice summary here. 01:02:00.470 --> 01:02:01.730 Here we go. 01:02:01.730 --> 01:02:06.350 In some sense, the goal of the first third of the class-- 01:02:06.350 --> 01:02:09.170 the techniques we'll be talking about our statistical process 01:02:09.170 --> 01:02:12.530 control, which is all about detecting 01:02:12.530 --> 01:02:15.920 these inherent deviations, these disturbances, 01:02:15.920 --> 01:02:18.770 and seeking to minimize them. 01:02:18.770 --> 01:02:21.410 Then we can talk about process optimization, 01:02:21.410 --> 01:02:23.510 things like design of experiments, 01:02:23.510 --> 01:02:26.810 where we're basically trying to build a model for y-- 01:02:26.810 --> 01:02:29.240 the output-- as a function of process 01:02:29.240 --> 01:02:33.560 parameters, and seek to minimize the sensitivity 01:02:33.560 --> 01:02:37.610 to those disturbances, to have as robust a process as 01:02:37.610 --> 01:02:39.020 possible. 01:02:39.020 --> 01:02:41.180 And then finally, we can also think 01:02:41.180 --> 01:02:45.590 about active modes of process control, where you actually 01:02:45.590 --> 01:02:48.500 manipulate some of the control parameters, 01:02:48.500 --> 01:02:51.830 perhaps in response to observed deviations, 01:02:51.830 --> 01:02:55.280 to ultimately minimize or counteract, 01:02:55.280 --> 01:02:59.600 through feedback control, the effects of those disturbances. 01:02:59.600 --> 01:03:03.560 So that in the nutshell is the map 01:03:03.560 --> 01:03:06.740 for what we're going to be doing through the rest of the term. 01:03:06.740 --> 01:03:09.920 We're going to be dealing, first off, with statistical process 01:03:09.920 --> 01:03:12.380 control and the statistical background for that. 01:03:12.380 --> 01:03:14.270 We're then going to be modeling the process, 01:03:14.270 --> 01:03:17.270 building up design of experiments, response surface 01:03:17.270 --> 01:03:20.420 modeling technologies for modeling 01:03:20.420 --> 01:03:22.580 and optimizing the process, and then 01:03:22.580 --> 01:03:25.460 thirdly, we'll talk about some basic strategies 01:03:25.460 --> 01:03:28.970 for feedback control to compensate 01:03:28.970 --> 01:03:31.370 for some of these processes. 01:03:31.370 --> 01:03:34.060 So with that, we'll leave you. 01:03:34.060 --> 01:03:37.030 Folks in Singapore, we'll see you again next week. 01:03:37.030 --> 01:03:44.570 Again, do catch the videotape of Thursday's lecture. 01:03:44.570 --> 01:03:46.790 Please go to the website. 01:03:46.790 --> 01:03:50.840 This lecture is on there, but also there are two-- 01:03:50.840 --> 01:03:54.020 well, there's one reading that you should grab 01:03:54.020 --> 01:03:55.670 right away, which is about a-- 01:03:55.670 --> 01:03:59.150 was is it-- six or seven pages process overview 01:03:59.150 --> 01:04:03.500 which defines some of this terminology in written form 01:04:03.500 --> 01:04:05.180 that we've talked about here. 01:04:05.180 --> 01:04:07.940 And I will also post what the reading assignment 01:04:07.940 --> 01:04:10.330 is that you can get started on. 01:04:10.330 --> 01:04:12.290 I don't think it's posted up there yet, 01:04:12.290 --> 01:04:14.300 but go out and try to get these books, 01:04:14.300 --> 01:04:17.870 because there will be-- the reading assignment is basically 01:04:17.870 --> 01:04:21.380 start on the first two chapters, which are very talkative 01:04:21.380 --> 01:04:24.900 overview of manufacturing processes. 01:04:24.900 --> 01:04:26.840 So we'll get you started on those. 01:04:26.840 --> 01:04:28.340 One quick question, Dave-- 01:04:28.340 --> 01:04:30.020 is there any information on-- 01:04:30.020 --> 01:04:33.088 is there a pro seminar when that starts up, 01:04:33.088 --> 01:04:34.130 that you want to mention? 01:04:34.130 --> 01:04:34.880 DAVID HARDT: Yeah. 01:04:34.880 --> 01:04:37.130 2.888 is already-- there's a Stellar site on it. 01:04:37.130 --> 01:04:40.616 I think I've sent most of the preregistered students notes 01:04:40.616 --> 01:04:44.712 on this, but it starts a week [INAUDIBLE].. 01:04:44.712 --> 01:04:45.420 DUANE BONING: OK. 01:04:45.420 --> 01:04:47.310 So we'll have more information on that. 01:04:47.310 --> 01:04:52.680 Many of you, like those enrolled in the mechanical engineering, 01:04:52.680 --> 01:04:54.850 MEng program, will know about this already, 01:04:54.850 --> 01:04:58.680 but this series of seminars from folks from industry and so on 01:04:58.680 --> 01:05:02.930 will quite likely be of interest to other people in the class. 01:05:02.930 --> 01:05:06.720 So we'll start alerting you about those evening seminars 01:05:06.720 --> 01:05:09.130 here as well. 01:05:09.130 --> 01:05:13.310 So welcome to the class, and we'll see you on Thursday.