1 00:00:00,120 --> 00:00:02,460 The following content is provided under a Creative 2 00:00:02,460 --> 00:00:03,880 Commons license. 3 00:00:03,880 --> 00:00:06,090 Your support will help MIT OpenCourseWare 4 00:00:06,090 --> 00:00:10,180 continue to offer high-quality educational resources for free. 5 00:00:10,180 --> 00:00:12,720 To make a donation or to view additional materials 6 00:00:12,720 --> 00:00:16,680 from hundreds of MIT courses, visit MIT OpenCourseWare 7 00:00:16,680 --> 00:00:17,880 at ocw.mit.edu. 8 00:00:25,340 --> 00:00:28,360 OLIVIER DE WECK: So let's get going on system integration 9 00:00:28,360 --> 00:00:29,590 and interface management. 10 00:00:29,590 --> 00:00:34,075 So we're essentially moving up the right side of the V. 11 00:00:34,075 --> 00:00:37,300 And the idea is all the details have been designed. 12 00:00:37,300 --> 00:00:40,090 The concept has been selected. 13 00:00:40,090 --> 00:00:42,850 We know what we're doing, but now we have to re-integrate 14 00:00:42,850 --> 00:00:44,930 the system and make sure it works-- 15 00:00:44,930 --> 00:00:47,260 it performs the functions that are intended 16 00:00:47,260 --> 00:00:49,780 and satisfies the requirements. 17 00:00:49,780 --> 00:00:52,730 And so what I'm going to cover today is three things. 18 00:00:52,730 --> 00:00:54,660 First of all, why do we care? 19 00:00:54,660 --> 00:00:57,300 Why is interface management important? 20 00:00:57,300 --> 00:00:58,330 Why is it important? 21 00:00:58,330 --> 00:00:59,840 There's two main reasons. 22 00:00:59,840 --> 00:01:02,190 One is because a lot of failures originate-- 23 00:01:02,190 --> 00:01:05,200 so a lot of failures originate at interfaces. 24 00:01:05,200 --> 00:01:07,960 And the other reason is because in order 25 00:01:07,960 --> 00:01:10,240 to design and manufacture a lot of systems, 26 00:01:10,240 --> 00:01:13,810 we need to work with partners and suppliers. 27 00:01:13,810 --> 00:01:16,540 And I'll talk about interface management, sort of the details 28 00:01:16,540 --> 00:01:18,600 of it, the types of interfaces. 29 00:01:18,600 --> 00:01:22,570 I'll Introduce the DSM, the Design Structure Matrix, 30 00:01:22,570 --> 00:01:23,980 as an important method. 31 00:01:23,980 --> 00:01:30,250 And then we'll talk about ICDs, Interface Control Documents. 32 00:01:30,250 --> 00:01:32,630 They're a very big deal in practice. 33 00:01:32,630 --> 00:01:34,900 And then finally, I'll talk about system integration-- 34 00:01:34,900 --> 00:01:37,840 in particular, the sequence which you integrate systems, 35 00:01:37,840 --> 00:01:40,960 and then the role of standards. 36 00:01:40,960 --> 00:01:44,890 All right, so here, the point that I want to make here 37 00:01:44,890 --> 00:01:47,870 is when you talk about interfaces, 38 00:01:47,870 --> 00:01:49,840 the first thing you need to be clear is, is it 39 00:01:49,840 --> 00:01:53,370 an internal interface or an external interface 40 00:01:53,370 --> 00:01:54,650 of the system? 41 00:01:54,650 --> 00:01:58,860 And so I think you remember this chart here that has-- 42 00:01:58,860 --> 00:02:00,650 is this you guys-- 43 00:02:00,650 --> 00:02:03,550 that has a system boundary that's defined. 44 00:02:03,550 --> 00:02:05,260 So the system boundary, in this case, 45 00:02:05,260 --> 00:02:08,530 we have this digital camera here. 46 00:02:08,530 --> 00:02:12,220 And the internal interfaces are essentially the interfaces 47 00:02:12,220 --> 00:02:14,860 that are within this boundary. 48 00:02:14,860 --> 00:02:17,500 So those are internal interfaces within the camera, 49 00:02:17,500 --> 00:02:20,650 within the peripherals, that are included in the system 50 00:02:20,650 --> 00:02:21,710 boundary. 51 00:02:21,710 --> 00:02:23,560 And then we have external interfaces, 52 00:02:23,560 --> 00:02:27,250 which would be the interfaces between the system boundary 53 00:02:27,250 --> 00:02:29,020 and anything that's on the outside 54 00:02:29,020 --> 00:02:31,120 that you don't design or control directly. 55 00:02:31,120 --> 00:02:35,650 So that's an important distinction. 56 00:02:35,650 --> 00:02:39,250 The next point I want to make is that you 57 00:02:39,250 --> 00:02:44,170 have to really carefully think about interfaces as seeds 58 00:02:44,170 --> 00:02:46,480 of potential system failures. 59 00:02:46,480 --> 00:02:50,930 And so I have three examples here for you. 60 00:02:50,930 --> 00:02:53,110 So the first one is-- 61 00:02:53,110 --> 00:02:54,390 this is actually-- 62 00:02:54,390 --> 00:02:58,060 Katya, sorry, this is from Russia, from 2007. 63 00:02:58,060 --> 00:02:59,620 This is traffic. 64 00:02:59,620 --> 00:03:01,115 This is one of the big-- 65 00:03:01,115 --> 00:03:05,830 not [RUSSIAN],, but prospekt, the six lanes 66 00:03:05,830 --> 00:03:07,750 or eight lanes on one side, and there's 67 00:03:07,750 --> 00:03:09,880 a crossroad coming in here. 68 00:03:09,880 --> 00:03:12,010 This is a traffic intersection. 69 00:03:12,010 --> 00:03:13,660 And you can see traffic is completely 70 00:03:13,660 --> 00:03:15,500 blocked on this side. 71 00:03:15,500 --> 00:03:19,180 So traffic engineers think a lot about interfaces, 72 00:03:19,180 --> 00:03:23,140 which are mainly your intersections, the roundabouts. 73 00:03:23,140 --> 00:03:26,980 Interfaces are really where a traffic system makes or breaks 74 00:03:26,980 --> 00:03:28,400 the traffic system. 75 00:03:28,400 --> 00:03:29,380 So that's a bottleneck. 76 00:03:29,380 --> 00:03:32,380 An interface is a bottleneck, basically. 77 00:03:32,380 --> 00:03:34,390 The second example here in the middle, this 78 00:03:34,390 --> 00:03:38,590 is actually an example from MIT from an undergraduate class 79 00:03:38,590 --> 00:03:43,330 in design where the students were designing an airfoil. 80 00:03:43,330 --> 00:03:46,080 So this has been produced with a foam cutter. 81 00:03:46,080 --> 00:03:48,550 And you have a spar that goes through here. 82 00:03:48,550 --> 00:03:52,690 And this was intended for the MIT Formula SAE race car 83 00:03:52,690 --> 00:03:56,080 to create downward pressure on the rear axle. 84 00:03:56,080 --> 00:03:57,690 They tested it in the wind tunnel, 85 00:03:57,690 --> 00:04:01,670 and it worked well at 20 miles per hour, 40 miles per hour. 86 00:04:01,670 --> 00:04:05,290 And then at 60 miles per hour, there was a failure, 87 00:04:05,290 --> 00:04:08,620 and the airfoil just basically disintegrated 88 00:04:08,620 --> 00:04:10,720 and was carried off into the wind tunnel 89 00:04:10,720 --> 00:04:13,060 and completely shredded by the propeller. 90 00:04:13,060 --> 00:04:14,480 Why did that happen? 91 00:04:14,480 --> 00:04:17,830 Well, they thought about the shear loads. 92 00:04:17,830 --> 00:04:19,329 So when you have an airfoil, there's 93 00:04:19,329 --> 00:04:23,710 a force that's acting orthogonally to the airfoil. 94 00:04:23,710 --> 00:04:25,520 So they thought about that. 95 00:04:25,520 --> 00:04:29,710 The problem is there's also a torque, like a pitching moment, 96 00:04:29,710 --> 00:04:31,300 a torque that's generated. 97 00:04:31,300 --> 00:04:34,750 And the higher the air speed is, the higher the torque. 98 00:04:34,750 --> 00:04:36,820 They didn't think-- there was no real good way 99 00:04:36,820 --> 00:04:41,680 to react against the torque that was being generated. 100 00:04:41,680 --> 00:04:44,450 So at that high speed, the airfoil, all of a sudden, 101 00:04:44,450 --> 00:04:49,790 tilted and created a lot of drag and basically was destroyed. 102 00:04:49,790 --> 00:04:54,740 So the airfoil itself was OK, but the interface was not. 103 00:04:54,740 --> 00:04:58,570 The third example here is a software example or a software 104 00:04:58,570 --> 00:05:00,580 interface, very famous. 105 00:05:00,580 --> 00:05:03,760 This was the launch failure of the Ariane 501. 106 00:05:03,760 --> 00:05:06,760 The very first launch of the Ariane 5 107 00:05:06,760 --> 00:05:10,330 basically led to an auto-destruct of the rocket 108 00:05:10,330 --> 00:05:11,710 upon ascent. 109 00:05:11,710 --> 00:05:13,920 There was an accident investigation. 110 00:05:13,920 --> 00:05:18,160 And I'm just quoting you here from the key passage, which 111 00:05:18,160 --> 00:05:23,470 said that "the active initial reference system transmitted 112 00:05:23,470 --> 00:05:26,280 essentially diagnostic information to the launcher's 113 00:05:26,280 --> 00:05:27,780 main computer." 114 00:05:27,780 --> 00:05:30,480 And so the main computer was interpreting this information 115 00:05:30,480 --> 00:05:34,110 as though the rocket has gone off course and trying 116 00:05:34,110 --> 00:05:36,840 to correct for it when, in fact, it hadn't. 117 00:05:36,840 --> 00:05:40,140 It was actually physically going exactly the right trajectory, 118 00:05:40,140 --> 00:05:41,620 but it was misinterpreted. 119 00:05:41,620 --> 00:05:46,200 And this is also a very famous failure due to software reuse. 120 00:05:46,200 --> 00:05:50,610 So three examples of interfaces-- bottlenecks 121 00:05:50,610 --> 00:05:53,970 in traffic, structural failures, and erroneous function 122 00:05:53,970 --> 00:05:55,740 calls in software. 123 00:05:55,740 --> 00:05:59,920 And so be careful about interfaces. 124 00:05:59,920 --> 00:06:03,960 The second reason we care is working with manufacturers 125 00:06:03,960 --> 00:06:05,300 and suppliers. 126 00:06:05,300 --> 00:06:07,620 So the picture here is essentially 127 00:06:07,620 --> 00:06:11,280 the Dreamliner, the Boeing 787. 128 00:06:11,280 --> 00:06:14,250 And you can see here the different colors essentially 129 00:06:14,250 --> 00:06:15,870 represent the different countries 130 00:06:15,870 --> 00:06:19,140 or the different suppliers that are providing different parts 131 00:06:19,140 --> 00:06:20,670 of the airplane. 132 00:06:20,670 --> 00:06:23,555 Now, does anybody-- let me ask this question to MIT, 133 00:06:23,555 --> 00:06:27,090 see if we can hear them now. 134 00:06:27,090 --> 00:06:29,810 No? 135 00:06:29,810 --> 00:06:33,270 Well, maybe we can use the chat. 136 00:06:33,270 --> 00:06:36,540 What's different between the Dreamliner-- 137 00:06:36,540 --> 00:06:39,000 and you could have seen a similar picture, say, 138 00:06:39,000 --> 00:06:42,300 for the 737 or prior generation aircraft. 139 00:06:42,300 --> 00:06:45,960 So what's the real difference here with the interfaces, 140 00:06:45,960 --> 00:06:46,785 with the 787? 141 00:06:49,650 --> 00:06:52,500 AUDIENCE: [INAUDIBLE] 142 00:06:52,500 --> 00:06:54,960 OLIVIER DE WECK: Yeah, that's true. 143 00:06:54,960 --> 00:07:00,120 Composites is one thing, but that's a structural thing. 144 00:07:00,120 --> 00:07:01,770 I have something else in mind. 145 00:07:01,770 --> 00:07:06,390 There's some other fundamental change that was made in the 787 146 00:07:06,390 --> 00:07:07,500 beyond composites. 147 00:07:07,500 --> 00:07:08,966 AUDIENCE: [INAUDIBLE] 148 00:07:08,966 --> 00:07:09,882 OLIVIER DE WECK: Nope. 149 00:07:09,882 --> 00:07:13,640 AUDIENCE: [INAUDIBLE] 150 00:07:13,640 --> 00:07:16,140 OLIVIER DE WECK: This has to do with the way the airplane is 151 00:07:16,140 --> 00:07:18,272 assembled, with assembly. 152 00:07:18,272 --> 00:07:21,440 AUDIENCE: [INAUDIBLE] 153 00:07:21,440 --> 00:07:23,760 OLIVIER DE WECK: Yes, yes, exactly. 154 00:07:23,760 --> 00:07:26,620 So the supply chain is very different. 155 00:07:26,620 --> 00:07:28,620 And I guess I lost my-- 156 00:07:32,388 --> 00:07:35,270 hm, did I reverse this? 157 00:07:35,270 --> 00:07:45,360 So the point I want to make is that in the 787, 158 00:07:45,360 --> 00:07:48,530 the modules are essentially pre-outfitted. 159 00:07:48,530 --> 00:07:51,500 Like, all the systems, like the electrical system, 160 00:07:51,500 --> 00:07:54,700 the hydraulics, everything is already installed 161 00:07:54,700 --> 00:07:57,520 by the time the module arrives at the assembly 162 00:07:57,520 --> 00:08:00,610 plant in Everett and close to Seattle. 163 00:08:00,610 --> 00:08:03,340 Before, you got the structural sections, but then 164 00:08:03,340 --> 00:08:07,390 all the electricals, all the inside was done on site. 165 00:08:07,390 --> 00:08:10,840 So the modules are pre-outfitted with all the subsystems. 166 00:08:10,840 --> 00:08:14,590 And the final assembly itself is very short, only a few days. 167 00:08:14,590 --> 00:08:17,920 So a big, big step, and that's exactly right. 168 00:08:17,920 --> 00:08:20,690 That's the supply chain impact. 169 00:08:20,690 --> 00:08:25,480 OK, so then the final point here is 170 00:08:25,480 --> 00:08:26,980 this is essentially a launch vehicle 171 00:08:26,980 --> 00:08:28,690 design, a launch vehicle that's being 172 00:08:28,690 --> 00:08:31,944 designed at NASA, the SLS. 173 00:08:31,944 --> 00:08:33,610 And if you think about a launch vehicle, 174 00:08:33,610 --> 00:08:37,450 you have the first stage, second stage, third stage, payload, 175 00:08:37,450 --> 00:08:39,522 the fairing. 176 00:08:39,522 --> 00:08:40,990 Actually, a lot of these fairings 177 00:08:40,990 --> 00:08:43,390 are made here in Switzerland. 178 00:08:43,390 --> 00:08:46,770 And defining these interfaces very clearly 179 00:08:46,770 --> 00:08:49,420 is critical to reduce complexity. 180 00:08:49,420 --> 00:08:51,190 Understanding the different types, which 181 00:08:51,190 --> 00:08:54,160 we're going to get into next-- 182 00:08:54,160 --> 00:08:57,280 identifying the interfaces is a way to reduce risk. 183 00:08:57,280 --> 00:08:59,470 And I just talked about these problems 184 00:08:59,470 --> 00:09:01,720 originating at interfaces. 185 00:09:01,720 --> 00:09:04,570 And so hopefully, it's pretty clear now 186 00:09:04,570 --> 00:09:07,410 that this is a critical issue. 187 00:09:07,410 --> 00:09:09,070 So what I'd like to do now-- and we'll 188 00:09:09,070 --> 00:09:11,830 try to troubleshoot the audio issues in the meantime-- 189 00:09:11,830 --> 00:09:15,310 is for you to turn to your partner exercise 190 00:09:15,310 --> 00:09:18,490 and share with each other, what is 191 00:09:18,490 --> 00:09:20,530 an instance in your past experience, 192 00:09:20,530 --> 00:09:24,340 maybe during an internship or in a project that you worked on, 193 00:09:24,340 --> 00:09:27,360 where interface definition and management was critical? 194 00:09:34,040 --> 00:09:37,630 AUDIENCE: Some of us went this summer to Russia, to Moscow, 195 00:09:37,630 --> 00:09:41,380 to follow a summer camp on system engineering and space 196 00:09:41,380 --> 00:09:42,420 engineering. 197 00:09:42,420 --> 00:09:43,960 And we had a project there where we 198 00:09:43,960 --> 00:09:47,110 had to design a mission to deep space. 199 00:09:47,110 --> 00:09:49,750 And everyone was assigned to a group. 200 00:09:49,750 --> 00:09:53,200 The groups were around five to seven people, 201 00:09:53,200 --> 00:09:55,630 and there were five to seven groups. 202 00:09:55,630 --> 00:09:59,800 And each group had a subsystem to design, 203 00:09:59,800 --> 00:10:02,740 but no group was assigned to system engineering. 204 00:10:02,740 --> 00:10:04,990 So actually, there was no one to make sure 205 00:10:04,990 --> 00:10:06,590 that all the interfaces were OK. 206 00:10:06,590 --> 00:10:09,580 So we had to go to each other group 207 00:10:09,580 --> 00:10:10,840 and ask for the interfaces. 208 00:10:10,840 --> 00:10:14,020 And then when we had to go back, it was a mess. 209 00:10:14,020 --> 00:10:16,985 And at the end, every group had something good. 210 00:10:16,985 --> 00:10:19,910 But when we tried to put all the systems together, 211 00:10:19,910 --> 00:10:26,270 it was a big mess and nothing was working. 212 00:10:26,270 --> 00:10:29,510 OLIVIER DE WECK: That's a pitch for systems engineering. 213 00:10:29,510 --> 00:10:30,850 So that's a good point. 214 00:10:30,850 --> 00:10:33,055 So even if you negotiate bilateral interfaces 215 00:10:33,055 --> 00:10:35,560 amongst subsystems, there's no guarantee 216 00:10:35,560 --> 00:10:38,110 that it will work at the system level. 217 00:10:38,110 --> 00:10:42,750 OK, can we get from MIT? 218 00:10:42,750 --> 00:10:45,227 Can somebody give us what you discussed? 219 00:10:53,520 --> 00:10:55,830 AUDIENCE: Is this the mic working? 220 00:10:55,830 --> 00:10:58,620 OLIVIER DE WECK: Yeah, now we can hear you. 221 00:10:58,620 --> 00:11:01,380 AUDIENCE: Great. 222 00:11:01,380 --> 00:11:04,680 So my grad school project here at MIT 223 00:11:04,680 --> 00:11:08,220 is a secondary payload on a big NASA mission. 224 00:11:08,220 --> 00:11:12,605 And as a secondary payload, the interface control 225 00:11:12,605 --> 00:11:14,020 is a really big deal. 226 00:11:14,020 --> 00:11:15,659 So we're getting all of our power 227 00:11:15,659 --> 00:11:17,700 and all of our communication from the spacecraft, 228 00:11:17,700 --> 00:11:21,300 and we're also getting a heater circuit from them. 229 00:11:21,300 --> 00:11:24,042 So any change is a big deal. 230 00:11:24,042 --> 00:11:25,500 We had to change our heater circuit 231 00:11:25,500 --> 00:11:27,291 and we had to go through a lot of paperwork 232 00:11:27,291 --> 00:11:28,860 just to splice a wire. 233 00:11:28,860 --> 00:11:34,300 So no issue with it, but it's not as easy as it sounds, 234 00:11:34,300 --> 00:11:38,010 and there's a lot of control over it, I would say. 235 00:11:43,794 --> 00:11:45,960 OLIVIER DE WECK: OK, maybe another example from MIT. 236 00:11:50,430 --> 00:11:51,109 Somebody else? 237 00:11:51,109 --> 00:11:52,150 AUDIENCE: Does this work? 238 00:11:52,150 --> 00:11:53,671 OLIVIER DE WECK: Yeah. 239 00:11:53,671 --> 00:11:54,420 AUDIENCE: So the-- 240 00:11:54,420 --> 00:11:56,753 OLIVIER DE WECK: We've got it fixed out now, by the way. 241 00:11:56,753 --> 00:11:59,575 I think this is all working now, so good. 242 00:11:59,575 --> 00:12:01,200 AUDIENCE: The last project I was a part 243 00:12:01,200 --> 00:12:04,870 of was building an aircraft for a Red Bull Flugtag. 244 00:12:04,870 --> 00:12:09,620 So it was like a giant glider that we'd push off a platform. 245 00:12:09,620 --> 00:12:11,760 But we were kind of building it in a hurry. 246 00:12:11,760 --> 00:12:15,120 The time constraints were pretty quick, 247 00:12:15,120 --> 00:12:17,280 and we didn't really have all the requirements 248 00:12:17,280 --> 00:12:20,320 until a month or two before the competition. 249 00:12:20,320 --> 00:12:22,470 So we winged a lot of the design and construction. 250 00:12:22,470 --> 00:12:25,770 And so we thought about how to build the wing 251 00:12:25,770 --> 00:12:28,960 and how to build the empennage and how to build the cockpit, 252 00:12:28,960 --> 00:12:31,750 but we didn't think about how to put them all together. 253 00:12:31,750 --> 00:12:34,800 And so it turned out just being wrapped with a lot of carbon 254 00:12:34,800 --> 00:12:37,290 fiber and epoxied. 255 00:12:37,290 --> 00:12:40,280 And in some cases, we used wood. 256 00:12:40,280 --> 00:12:44,670 And it was a very heavy and poor-looking assembly once all 257 00:12:44,670 --> 00:12:45,510 that was together. 258 00:12:45,510 --> 00:12:48,340 Even though each part individually looked great, 259 00:12:48,340 --> 00:12:50,280 they didn't really come together perfectly. 260 00:12:50,280 --> 00:12:54,160 So we'll change that next year, but could have thought more. 261 00:12:54,160 --> 00:12:55,590 OLIVIER DE WECK: OK, good. 262 00:12:55,590 --> 00:12:57,240 Great example. 263 00:12:57,240 --> 00:12:59,580 All right, so hopefully, this is sort 264 00:12:59,580 --> 00:13:03,840 of motivation that, in a sense, a system is 265 00:13:03,840 --> 00:13:07,582 the functions you want, the components, and the interfaces. 266 00:13:07,582 --> 00:13:08,790 You've got to have all three. 267 00:13:08,790 --> 00:13:10,260 If you only have great components 268 00:13:10,260 --> 00:13:12,360 but you don't understand the interfaces, 269 00:13:12,360 --> 00:13:16,550 you stick them together, I'm not sure what you're going to get. 270 00:13:16,550 --> 00:13:17,661 Yes, do you want to-- 271 00:13:17,661 --> 00:13:23,070 AUDIENCE: [INAUDIBLE] Is it also important the interfaces 272 00:13:23,070 --> 00:13:26,384 with testing facilities? 273 00:13:26,384 --> 00:13:27,800 OLIVIER DE WECK: Yeah, absolutely. 274 00:13:27,800 --> 00:13:31,790 So testability requirements, how you're going to test it, 275 00:13:31,790 --> 00:13:32,730 is critical. 276 00:13:32,730 --> 00:13:34,890 Also, ground support equipment-- you know, 277 00:13:34,890 --> 00:13:36,980 like maintenance equipment. 278 00:13:36,980 --> 00:13:39,300 So not just the interfaces, the system 279 00:13:39,300 --> 00:13:42,540 when you're operating it, but during testing, 280 00:13:42,540 --> 00:13:44,010 during maintenance. 281 00:13:44,010 --> 00:13:47,700 And those are often forgotten as well, so great point. 282 00:13:47,700 --> 00:13:50,550 So let's keep moving here. 283 00:13:50,550 --> 00:13:53,520 I want to talk about the types of interfaces, 284 00:13:53,520 --> 00:13:56,280 and then we'll get to DSM. 285 00:13:56,280 --> 00:14:01,350 So here's a set of examples of interfaces that we encounter. 286 00:14:01,350 --> 00:14:07,410 Let's see if we can get the slides back up here. 287 00:14:07,410 --> 00:14:10,860 So essentially, here's some examples. 288 00:14:10,860 --> 00:14:14,250 In the upper left, we have a valve, and then 289 00:14:14,250 --> 00:14:16,320 a tank on the right side. 290 00:14:16,320 --> 00:14:19,110 And we essentially have mass flow 291 00:14:19,110 --> 00:14:25,290 from the valve to the tank, and m dot, dm dt. 292 00:14:25,290 --> 00:14:26,310 We have a rocket. 293 00:14:26,310 --> 00:14:28,920 If you think about the rocket interface with the payload, 294 00:14:28,920 --> 00:14:30,510 it's transferring momentum, right? 295 00:14:30,510 --> 00:14:33,660 The rocket provides thrust, force F, 296 00:14:33,660 --> 00:14:36,150 for some amount of time, delta-t, 297 00:14:36,150 --> 00:14:38,580 and that's the momentum imparted. 298 00:14:38,580 --> 00:14:43,170 We have a heat exchanger that is maybe 299 00:14:43,170 --> 00:14:46,110 giving heat off to the air, the surrounding air. 300 00:14:46,110 --> 00:14:48,978 So we have a heat flux, q dot. 301 00:14:48,978 --> 00:14:51,690 We have solar cells, PV cells, that 302 00:14:51,690 --> 00:14:55,950 are producing electrical power, voltage times current. 303 00:14:55,950 --> 00:14:59,700 And this power flux is being sent to a battery where 304 00:14:59,700 --> 00:15:01,420 the energy is stored. 305 00:15:01,420 --> 00:15:04,914 So these are all pretty obvious examples. 306 00:15:04,914 --> 00:15:06,330 And then on the right side, things 307 00:15:06,330 --> 00:15:07,830 get maybe a little bit less obvious. 308 00:15:10,470 --> 00:15:15,060 You go to a URL and you download an HTML file into your browser. 309 00:15:15,060 --> 00:15:15,990 That's an information. 310 00:15:15,990 --> 00:15:18,150 You're getting data. 311 00:15:18,150 --> 00:15:21,390 You have a motion sensor attached, say, to your house, 312 00:15:21,390 --> 00:15:23,190 and it triggers an alarm. 313 00:15:23,190 --> 00:15:27,210 That's a command, different from data. 314 00:15:27,210 --> 00:15:29,700 NPR is National Public Radio. 315 00:15:29,700 --> 00:15:34,470 This is kind of the high-quality radio news reporting. 316 00:15:34,470 --> 00:15:36,360 And you listen to this. 317 00:15:36,360 --> 00:15:38,670 So there's a cognitive interface. 318 00:15:38,670 --> 00:15:45,160 And then if any of you have had any psychotherapy, 319 00:15:45,160 --> 00:15:48,210 this is also an interface between the patient 320 00:15:48,210 --> 00:15:49,320 and the psychotherapist. 321 00:15:49,320 --> 00:15:51,630 This is an effective interface. 322 00:15:51,630 --> 00:15:53,680 So these are very different. 323 00:15:53,680 --> 00:15:55,590 And you can say, well, there's hundreds 324 00:15:55,590 --> 00:15:57,670 of types of interfaces. 325 00:15:57,670 --> 00:15:59,130 Well, no. 326 00:15:59,130 --> 00:16:02,520 The argument here is all interfaces 327 00:16:02,520 --> 00:16:07,210 can be reduced or described by these four canonical types. 328 00:16:07,210 --> 00:16:10,090 The first one is physical connection. 329 00:16:10,090 --> 00:16:13,125 So A connects to B. B connects to A. 330 00:16:13,125 --> 00:16:15,570 A physical connection typically implies 331 00:16:15,570 --> 00:16:19,260 that there's a force or torque that 332 00:16:19,260 --> 00:16:21,380 can be transmitted across the interface, 333 00:16:21,380 --> 00:16:23,490 and it's always symmetrical. 334 00:16:23,490 --> 00:16:27,220 And then we have energy, mass, and information flows. 335 00:16:27,220 --> 00:16:29,130 And I'll give you some examples of each 336 00:16:29,130 --> 00:16:30,810 of these types of interfaces. 337 00:16:30,810 --> 00:16:33,630 And these energy, mass and information flows 338 00:16:33,630 --> 00:16:35,310 typically are asymmetric. 339 00:16:35,310 --> 00:16:37,740 It flows from one to the other. 340 00:16:37,740 --> 00:16:41,120 Physical connection is always symmetric. 341 00:16:41,120 --> 00:16:44,660 So let's go through these, and I'll give you examples. 342 00:16:44,660 --> 00:16:46,560 And then we'll see what can we do with it. 343 00:16:46,560 --> 00:16:51,440 So here's physical connection examples-- 344 00:16:51,440 --> 00:16:53,370 rollers, brake pads. 345 00:16:53,370 --> 00:16:59,260 A finger touching a touch screen is a physical interface. 346 00:16:59,260 --> 00:17:01,670 The interface can be reversible. 347 00:17:01,670 --> 00:17:04,339 In other words, the connection is temporary-- 348 00:17:04,339 --> 00:17:06,950 and examples of that would be electrical connectors, 349 00:17:06,950 --> 00:17:10,890 USB ports, latch mechanisms, nuts and bolts-- 350 00:17:10,890 --> 00:17:14,839 or permanently connected, such as in-- so one of you 351 00:17:14,839 --> 00:17:19,099 mentioned riveting, spot welding, fusing. 352 00:17:19,099 --> 00:17:21,349 And a fun question that comes up, 353 00:17:21,349 --> 00:17:23,240 well, I'm a software engineer. 354 00:17:23,240 --> 00:17:26,240 What is physical connection mean for software? 355 00:17:26,240 --> 00:17:29,250 Well, it's a kind of tricky question. 356 00:17:29,250 --> 00:17:32,870 I think the closest analog we have is compiling. 357 00:17:32,870 --> 00:17:35,510 If you have pieces of source code 358 00:17:35,510 --> 00:17:38,660 and you compile your program, then you've 359 00:17:38,660 --> 00:17:43,100 essentially fused this together into something new, right? 360 00:17:43,100 --> 00:17:45,660 So I'll show you the example. 361 00:17:45,660 --> 00:17:47,420 So how do we describe it? 362 00:17:47,420 --> 00:17:49,550 Well, here's an OPM model of-- 363 00:17:49,550 --> 00:17:51,740 I'll give you a couple examples from xerography. 364 00:17:51,740 --> 00:17:54,530 So we have a main motor here, number one, 365 00:17:54,530 --> 00:17:58,460 and then two different clutches, number two and number three. 366 00:17:58,460 --> 00:18:01,070 And the process linking them is engaging, right? 367 00:18:01,070 --> 00:18:03,890 The main motor engages through the clutch. 368 00:18:03,890 --> 00:18:06,650 We can hide the process, and then 369 00:18:06,650 --> 00:18:10,380 we just have this bi-directional interface between one and two 370 00:18:10,380 --> 00:18:11,750 and one and three. 371 00:18:11,750 --> 00:18:15,260 And then if we show this as a mini-DSM here, one 372 00:18:15,260 --> 00:18:17,960 connected to two, one connected to three, 373 00:18:17,960 --> 00:18:20,690 but two and three do not connect directly with each other. 374 00:18:20,690 --> 00:18:23,190 They don't have a direct physical interface. 375 00:18:23,190 --> 00:18:25,770 And you can see that because those cells are empty. 376 00:18:25,770 --> 00:18:28,820 So physical connection implies symmetry 377 00:18:28,820 --> 00:18:32,700 in the design structure matrix that we can build up. 378 00:18:32,700 --> 00:18:33,736 Yes, [INAUDIBLE]? 379 00:18:33,736 --> 00:18:34,787 AUDIENCE: [INAUDIBLE] 380 00:18:34,787 --> 00:18:35,870 OLIVIER DE WECK: Oh, DSM-- 381 00:18:35,870 --> 00:18:38,450 so DSM means Design Structure Matrix. 382 00:18:38,450 --> 00:18:40,410 And I will introduce that in a minute. 383 00:18:40,410 --> 00:18:42,950 We've briefly mentioned it before in the class, 384 00:18:42,950 --> 00:18:46,500 but I'll get into it in some detail. 385 00:18:46,500 --> 00:18:48,780 Here's some more picture examples 386 00:18:48,780 --> 00:18:50,490 of physical connections-- 387 00:18:50,490 --> 00:18:56,370 so a welding joint, irreversible, reversible, nuts 388 00:18:56,370 --> 00:19:00,360 and bolts, RJ45, an electrical connector. 389 00:19:00,360 --> 00:19:01,470 And then here's your-- 390 00:19:01,470 --> 00:19:03,000 hey, I'm an old guy. 391 00:19:03,000 --> 00:19:04,520 I learned Fortran. 392 00:19:04,520 --> 00:19:07,245 Well, Basic was my first language, and then Fortran. 393 00:19:07,245 --> 00:19:08,820 I did a lot of Fortran. 394 00:19:08,820 --> 00:19:10,960 You guys probably don't even know what that is. 395 00:19:10,960 --> 00:19:12,380 But Fortran-- right, [INAUDIBLE]?? 396 00:19:12,380 --> 00:19:14,840 Fortran's kind of a big deal still in some-- 397 00:19:14,840 --> 00:19:15,980 AUDIENCE: [INAUDIBLE]. 398 00:19:15,980 --> 00:19:16,860 OLIVIER DE WECK: OK. 399 00:19:16,860 --> 00:19:20,610 So basically the way this works is you have your source code, 400 00:19:20,610 --> 00:19:22,140 which are .f files. 401 00:19:22,140 --> 00:19:26,370 These are basically Ascii files, right? 402 00:19:26,370 --> 00:19:29,430 And then first, you transform them into binary files, 403 00:19:29,430 --> 00:19:30,960 object files. 404 00:19:30,960 --> 00:19:34,810 And then these get merged together into an executable. 405 00:19:34,810 --> 00:19:38,370 So that's the process of compiling. 406 00:19:38,370 --> 00:19:41,820 Any questions about physical connections, 407 00:19:41,820 --> 00:19:46,170 this type of interface, physical connection? 408 00:19:46,170 --> 00:19:47,430 Is it clear? 409 00:19:47,430 --> 00:19:48,680 OK, any questions at MIT? 410 00:19:53,710 --> 00:19:55,870 Can you still hear me? 411 00:19:55,870 --> 00:19:57,647 AUDIENCE: Yes, we're good. 412 00:19:57,647 --> 00:19:58,730 OLIVIER DE WECK: OK, good. 413 00:19:58,730 --> 00:20:01,130 So let's move on to energy flows. 414 00:20:01,130 --> 00:20:05,690 Energy flows are present when there's a net exchange of work 415 00:20:05,690 --> 00:20:07,400 between two components. 416 00:20:07,400 --> 00:20:12,720 So power is flowing across the interface dw dt, right-- 417 00:20:12,720 --> 00:20:15,710 joules per second, watts. 418 00:20:15,710 --> 00:20:18,530 And the energy, this energy flow, this power flow, 419 00:20:18,530 --> 00:20:20,270 can take different forms-- 420 00:20:20,270 --> 00:20:23,880 electrical power, which is very common in many products. 421 00:20:23,880 --> 00:20:25,940 And so the electrical power usually 422 00:20:25,940 --> 00:20:30,410 comes either as DC, Direct Current, or AC, Alternating 423 00:20:30,410 --> 00:20:31,620 Current. 424 00:20:31,620 --> 00:20:34,460 And if you know a little bit about the history 425 00:20:34,460 --> 00:20:38,270 of electrical systems and networks, there was a huge war. 426 00:20:38,270 --> 00:20:40,310 It's called the AC/DC wars. 427 00:20:40,310 --> 00:20:41,390 It's not the band. 428 00:20:41,390 --> 00:20:42,740 This is not the rock music. 429 00:20:42,740 --> 00:20:45,800 This is the war between Edison, who 430 00:20:45,800 --> 00:20:48,860 is a big proponent of direct current, 431 00:20:48,860 --> 00:20:52,790 and Tesla and Westinghouse, who were really pushing 432 00:20:52,790 --> 00:20:56,820 AC for big power distribution. 433 00:20:56,820 --> 00:20:59,200 So we won in the end? 434 00:20:59,200 --> 00:21:00,320 AC won. 435 00:21:00,320 --> 00:21:02,535 It looked initially like Edison was really-- 436 00:21:02,535 --> 00:21:04,035 and there's still, in New York City, 437 00:21:04,035 --> 00:21:06,350 there, for a long time, and other places, 438 00:21:06,350 --> 00:21:08,540 you still had DC power systems. 439 00:21:08,540 --> 00:21:12,770 The problem with DC power is there's 440 00:21:12,770 --> 00:21:14,960 huge losses with distance. 441 00:21:14,960 --> 00:21:18,020 And with AC, you can go to very high voltages 442 00:21:18,020 --> 00:21:20,840 and transmit power with relatively little loss 443 00:21:20,840 --> 00:21:22,220 over large distance. 444 00:21:22,220 --> 00:21:26,090 That's the main reason that AC won out, essentially. 445 00:21:26,090 --> 00:21:28,430 The thing that's important here-- and power, of course, 446 00:21:28,430 --> 00:21:30,560 is current times voltage. 447 00:21:30,560 --> 00:21:35,420 The thing that's important is that these voltage levels-- 448 00:21:35,420 --> 00:21:38,540 and in the case of AC power also, the frequency 449 00:21:38,540 --> 00:21:41,420 is very important, 50 or 60 hertz. 450 00:21:41,420 --> 00:21:43,640 This is very standardized, OK? 451 00:21:43,640 --> 00:21:49,130 So if you're going to design a satellite system, or a rover, 452 00:21:49,130 --> 00:21:53,930 or an automotive system, or a medical device or a consumer 453 00:21:53,930 --> 00:21:58,190 product, the voltage levels are pretty-- 454 00:21:58,190 --> 00:22:00,380 there's a few choices, but you can't just 455 00:22:00,380 --> 00:22:03,590 choose some arbitrary number, because your whole industry, 456 00:22:03,590 --> 00:22:06,500 your batteries, your connectors, everything is now 457 00:22:06,500 --> 00:22:08,420 pretty standardized. 458 00:22:08,420 --> 00:22:11,870 So for the Octanus 1, Rafael, what did you guys 459 00:22:11,870 --> 00:22:13,490 choose for your power system? 460 00:22:13,490 --> 00:22:14,882 What's your-- 461 00:22:14,882 --> 00:22:17,710 AUDIENCE: [INAUDIBLE] 3.3 volts. 462 00:22:17,710 --> 00:22:20,900 OLIVIER DE WECK: 3.3-volt bus across, OK. 463 00:22:20,900 --> 00:22:23,300 And then if you need more energy, 464 00:22:23,300 --> 00:22:25,240 you just have to add batteries. 465 00:22:25,240 --> 00:22:27,050 But that will essentially-- and you 466 00:22:27,050 --> 00:22:28,970 can do transformations as well. 467 00:22:28,970 --> 00:22:31,350 You can have different voltage levels in your system, 468 00:22:31,350 --> 00:22:33,980 but that adds to the complexity. 469 00:22:33,980 --> 00:22:37,100 Thermal flux is essentially heat flux, dq dt. 470 00:22:37,100 --> 00:22:41,330 There are three fundamental mechanisms of heat transfer-- 471 00:22:41,330 --> 00:22:44,520 conduction, convection, and radiation. 472 00:22:44,520 --> 00:22:48,210 In spacecraft, radiation is probably the most important 473 00:22:48,210 --> 00:22:50,270 because that's the only way to get rid 474 00:22:50,270 --> 00:22:52,820 of heat out of your spacecraft. 475 00:22:52,820 --> 00:22:57,130 And then you can do conduction and convection internally. 476 00:22:57,130 --> 00:23:00,360 RF power, so microwaves. 477 00:23:00,360 --> 00:23:02,130 Transmitting or transferring power 478 00:23:02,130 --> 00:23:05,820 through microwaves is becoming-- 479 00:23:05,820 --> 00:23:08,850 of course, we now had microwave ovens for a long time. 480 00:23:08,850 --> 00:23:12,600 But even doing power beaming over larger distances 481 00:23:12,600 --> 00:23:14,610 with microwaves is definitely something 482 00:23:14,610 --> 00:23:16,380 that's been demonstrated, and there's 483 00:23:16,380 --> 00:23:18,450 a lot of interest in it. 484 00:23:18,450 --> 00:23:20,880 Here also, we are pretty standard frequencies, so 485 00:23:20,880 --> 00:23:24,420 2.4 gigahertz, 5.8 gigahertz. 486 00:23:24,420 --> 00:23:27,610 You have a whole industry built around these standards. 487 00:23:27,610 --> 00:23:30,270 By the way, why the 2.4 gigahertz? 488 00:23:30,270 --> 00:23:32,010 Let's see if-- at MIT. 489 00:23:32,010 --> 00:23:34,147 This 2.4 gigahertz for microwaves, 490 00:23:34,147 --> 00:23:35,230 where does that come from? 491 00:23:42,650 --> 00:23:44,350 Anybody know where that comes from? 492 00:23:46,940 --> 00:23:49,532 AUDIENCE: Is that the frequency that can be transmitted easily 493 00:23:49,532 --> 00:23:50,490 through the atmosphere? 494 00:23:53,090 --> 00:23:55,460 OLIVIER DE WECK: Actually, no, it's not good. 495 00:23:55,460 --> 00:23:56,570 It's absorbed. 496 00:23:56,570 --> 00:23:58,950 It's absorbed in the atmosphere. 497 00:23:58,950 --> 00:24:00,440 It's the opposite. 498 00:24:00,440 --> 00:24:03,250 This 2.4 is very important because it's 499 00:24:03,250 --> 00:24:06,530 a resonant frequency of the water molecule. 500 00:24:06,530 --> 00:24:09,811 So your microwave oven will work at 2.4 gigahertz. 501 00:24:09,811 --> 00:24:10,310 Why? 502 00:24:10,310 --> 00:24:14,690 Because if you radiate microwaves at that frequency, 503 00:24:14,690 --> 00:24:17,510 it makes the water molecules vibrate because that's 504 00:24:17,510 --> 00:24:20,710 a resonant frequency, and you transfer a lot of heat 505 00:24:20,710 --> 00:24:24,920 or a lot of energy into your water molecules, which 506 00:24:24,920 --> 00:24:28,850 are the majority of your food. 507 00:24:28,850 --> 00:24:31,330 So it's actually really bad for-- 508 00:24:31,330 --> 00:24:33,470 you'd like to use 2.4 gigahertz, but if you 509 00:24:33,470 --> 00:24:37,160 try to use that frequency for atmospheric power transfer, 510 00:24:37,160 --> 00:24:39,510 you're right at the resonant mode. 511 00:24:39,510 --> 00:24:42,590 So if you have a high level of humidity in the air, 512 00:24:42,590 --> 00:24:44,450 you're going to lose a lot of that power, 513 00:24:44,450 --> 00:24:47,840 just heating the atmosphere up and not transmitting the power. 514 00:24:47,840 --> 00:24:49,430 Does that make sense? 515 00:24:49,430 --> 00:24:51,717 AUDIENCE: Yeah, thanks. 516 00:24:51,717 --> 00:24:52,550 OLIVIER DE WECK: OK. 517 00:24:52,550 --> 00:24:55,880 And then another very popular frequency is 5.8 gigahertz. 518 00:24:55,880 --> 00:24:58,010 And then we can go to X band, which is closer 519 00:24:58,010 --> 00:25:00,020 to 8 gigahertz, et cetera. 520 00:25:00,020 --> 00:25:03,790 But anyway, so this is a kind of growing area 521 00:25:03,790 --> 00:25:06,560 is microwave power transfer. 522 00:25:06,560 --> 00:25:09,000 And then of course, the more old-fashioned, perhaps, 523 00:25:09,000 --> 00:25:11,660 but still very important is mechanical power, 524 00:25:11,660 --> 00:25:16,080 so transferring energy and power through a mechanical interface. 525 00:25:16,080 --> 00:25:19,010 So if it's linear, it's force times velocity. 526 00:25:19,010 --> 00:25:22,880 If it's rotary, it's torque times the angular rate 527 00:25:22,880 --> 00:25:24,950 of rotation. 528 00:25:24,950 --> 00:25:27,500 And then finally, energy-- 529 00:25:27,500 --> 00:25:29,810 typically, if you want to have energy flow, 530 00:25:29,810 --> 00:25:32,690 it does imply that there's a physical connection as well, 531 00:25:32,690 --> 00:25:36,810 like some kind of wires or a gearbox or something like this. 532 00:25:36,810 --> 00:25:43,810 However, a lot of interest now in wireless power transfer. 533 00:25:43,810 --> 00:25:46,050 But it's still kind of a-- 534 00:25:46,050 --> 00:25:50,020 for it to really work reliably at high power levels, 535 00:25:50,020 --> 00:25:54,020 this is pretty immature technology still. 536 00:25:54,020 --> 00:25:55,660 So here's some examples of-- 537 00:25:55,660 --> 00:25:57,572 again, this is from xerography. 538 00:25:57,572 --> 00:25:59,030 And you'll understand why I'm using 539 00:25:59,030 --> 00:26:01,700 these examples in a minute. 540 00:26:01,700 --> 00:26:03,380 So we have a paper. 541 00:26:03,380 --> 00:26:06,440 We have toner, unfused toner. 542 00:26:06,440 --> 00:26:08,400 And we want to fuse those two together. 543 00:26:08,400 --> 00:26:12,230 We do this through a heat roller and heat and a belt. 544 00:26:12,230 --> 00:26:13,730 And essentially, we're transferring 545 00:26:13,730 --> 00:26:16,880 energy from the heat roll to the paper, which 546 00:26:16,880 --> 00:26:18,920 allows it to fuse the toner. 547 00:26:18,920 --> 00:26:22,400 So we have transfer of energy from one to two, 548 00:26:22,400 --> 00:26:27,170 which is shown by this green mark here in the DSM. 549 00:26:27,170 --> 00:26:30,920 So heat energy is transferred from system one to system two. 550 00:26:33,710 --> 00:26:36,900 One question that often comes up is, well, 551 00:26:36,900 --> 00:26:39,890 what if we have like waste heat? 552 00:26:39,890 --> 00:26:44,720 There's heat losses, and there's parasitic energy flows. 553 00:26:44,720 --> 00:26:46,250 Do we care about those? 554 00:26:46,250 --> 00:26:48,060 Do we represent those? 555 00:26:48,060 --> 00:26:50,390 And the answer is yes, you should, 556 00:26:50,390 --> 00:26:52,400 especially if they affect your performance 557 00:26:52,400 --> 00:26:54,020 or your requirements. 558 00:26:54,020 --> 00:26:55,520 So the first thing you typically do, 559 00:26:55,520 --> 00:26:58,280 as you map out your interfaces, is you 560 00:26:58,280 --> 00:27:03,530 map the flows that you want to happen, that need to happen. 561 00:27:03,530 --> 00:27:05,900 And then on top of that, you map out 562 00:27:05,900 --> 00:27:07,730 the flows that happen even though you 563 00:27:07,730 --> 00:27:10,630 didn't design for that, but you need to account for them. 564 00:27:10,630 --> 00:27:13,400 So heat losses would be an example of that waste heat 565 00:27:13,400 --> 00:27:15,030 flux. 566 00:27:15,030 --> 00:27:17,310 Any questions about energy or power flows? 567 00:27:20,310 --> 00:27:31,010 MIT, any questions, energy or power flux? 568 00:27:31,010 --> 00:27:33,150 AUDIENCE: We're good, thanks. 569 00:27:33,150 --> 00:27:35,000 OLIVIER DE WECK: OK. 570 00:27:35,000 --> 00:27:36,640 Next one, mass flows. 571 00:27:36,640 --> 00:27:39,760 Mass flows is when matter, physical matter, 572 00:27:39,760 --> 00:27:42,940 is being exchanged between two elements or subsystems. 573 00:27:42,940 --> 00:27:46,960 So mass flows is dm dt, kilograms per second. 574 00:27:46,960 --> 00:27:53,320 And the form of the matter can be fluids, gases, solids. 575 00:27:53,320 --> 00:27:54,800 I guess plasma as well. 576 00:27:54,800 --> 00:27:56,725 There's a big plasma center here on campus. 577 00:27:59,710 --> 00:28:01,960 But the primary three are fluids-- so if it's fluids, 578 00:28:01,960 --> 00:28:04,630 it could be like a cooling liquid, a refrigerant, fuel, 579 00:28:04,630 --> 00:28:05,500 water. 580 00:28:05,500 --> 00:28:08,560 If it's gases, air, exhaust gas. 581 00:28:08,560 --> 00:28:13,120 And then for solids, examples would be toner, paper. 582 00:28:13,120 --> 00:28:16,900 If you're in the mining industry, it's iron ore. 583 00:28:16,900 --> 00:28:20,110 And typically, mass flow implies an underlying 584 00:28:20,110 --> 00:28:24,070 physical connection, like some kind of channel or conduit. 585 00:28:24,070 --> 00:28:27,760 So Rafael, your example of your sucking up and melting the snow 586 00:28:27,760 --> 00:28:33,070 and you said there's a hose, so the hose is itself a component, 587 00:28:33,070 --> 00:28:36,970 but it enables the mass flow of the melted snow 588 00:28:36,970 --> 00:28:39,910 into some holding tank or reservoir. 589 00:28:39,910 --> 00:28:42,589 And so the hose is the physical connection. 590 00:28:42,589 --> 00:28:44,380 And then the mass flow that happens through 591 00:28:44,380 --> 00:28:48,860 that hose is what's shown here. 592 00:28:48,860 --> 00:28:50,690 Mass flows can be open. 593 00:28:50,690 --> 00:28:54,120 So there's a source and there's a sink. 594 00:28:54,120 --> 00:28:57,100 The source and the sink could be inside or outside your system 595 00:28:57,100 --> 00:28:57,600 boundary. 596 00:28:57,600 --> 00:28:59,670 You have to think about it. 597 00:28:59,670 --> 00:29:03,270 But also, mass flows can form continuous loops, particularly 598 00:29:03,270 --> 00:29:05,220 if you recycle things. 599 00:29:05,220 --> 00:29:07,380 Like for example, in the refrigerator, 600 00:29:07,380 --> 00:29:13,680 your cooling liquid continuously cycles through your system. 601 00:29:13,680 --> 00:29:15,680 So here's some examples of mass flows. 602 00:29:15,680 --> 00:29:17,250 We have a heat exchanger. 603 00:29:17,250 --> 00:29:19,040 This is a U-type heat exchanger. 604 00:29:19,040 --> 00:29:23,255 So on the top, we have what's called here the shell side. 605 00:29:23,255 --> 00:29:27,710 A fluid is coming in and is running out on this end. 606 00:29:27,710 --> 00:29:30,290 And then we have horizontally, a fluid 607 00:29:30,290 --> 00:29:32,180 that's coming in on the lower side, 608 00:29:32,180 --> 00:29:35,960 cycling through, exchanging the heat, and moving out. 609 00:29:35,960 --> 00:29:37,700 So a heat exchanger is interesting 610 00:29:37,700 --> 00:29:40,350 because what kind of flows happen 611 00:29:40,350 --> 00:29:48,160 in a heat exchanger of the flows we've discussed so far? 612 00:29:48,160 --> 00:29:50,680 Energy and mass flow, right? 613 00:29:50,680 --> 00:29:52,450 So the two of them happen together. 614 00:29:52,450 --> 00:29:55,780 In order for the energy flow to happen or transfer to happen, 615 00:29:55,780 --> 00:29:57,980 the mass flow has to happen as well. 616 00:29:57,980 --> 00:30:01,160 So these two flows are tied to each other. 617 00:30:01,160 --> 00:30:02,810 Here's an example from xerography. 618 00:30:02,810 --> 00:30:05,510 So we have our photoreceptor belt. 619 00:30:05,510 --> 00:30:08,530 We have residual toner on the photoreceptor 620 00:30:08,530 --> 00:30:11,170 that needs to be removed from the photoreceptor. 621 00:30:11,170 --> 00:30:14,570 And we do this through a cleaning lamp, and then 622 00:30:14,570 --> 00:30:15,820 a blade. 623 00:30:15,820 --> 00:30:19,330 So in this case, you can show this as an OPM, 624 00:30:19,330 --> 00:30:23,860 and then you can show it here where that residual toner is 625 00:30:23,860 --> 00:30:28,030 transferred from photoreceptor one to the cleaning blade two. 626 00:30:28,030 --> 00:30:34,300 And this is shown in the DSM as a mass flow from one to two. 627 00:30:34,300 --> 00:30:38,570 And then we have one more, which is the information flows. 628 00:30:38,570 --> 00:30:41,020 So this is really where the big revolution 629 00:30:41,020 --> 00:30:43,600 has happened in the last two decades 630 00:30:43,600 --> 00:30:47,110 since a lot of functions that used 631 00:30:47,110 --> 00:30:50,500 to be done mechanically have been replaced with software. 632 00:30:50,500 --> 00:30:54,940 And here's some examples of these information flows. 633 00:30:54,940 --> 00:30:58,760 So if your system has a user interface, a GUI, 634 00:30:58,760 --> 00:31:03,130 a graphical user interface, or some kind of I/O, input/output, 635 00:31:03,130 --> 00:31:07,480 with the user, that's critical. 636 00:31:07,480 --> 00:31:09,970 That's an information interface. 637 00:31:09,970 --> 00:31:12,590 And so Rafael, again coming to your presentation-- 638 00:31:12,590 --> 00:31:14,780 so I'm going to pick on you a lot today. 639 00:31:14,780 --> 00:31:18,460 Not pick on you, but I'm going to cite you a lot. 640 00:31:18,460 --> 00:31:21,250 You said this rover is going to be controlled somehow, right? 641 00:31:21,250 --> 00:31:23,320 And the big question is always, how much autonomy 642 00:31:23,320 --> 00:31:26,620 do you put inside the thing, and how much do you 643 00:31:26,620 --> 00:31:28,990 rely on a user that's remote? 644 00:31:28,990 --> 00:31:33,490 And that's clearly an information interface. 645 00:31:33,490 --> 00:31:36,490 The nature of the information, is it analog? 646 00:31:36,490 --> 00:31:37,420 Is it digital? 647 00:31:37,420 --> 00:31:38,860 Is it wireless? 648 00:31:38,860 --> 00:31:40,810 So on the analog side, we're essentially 649 00:31:40,810 --> 00:31:43,690 talking about some signal, some voltage, 650 00:31:43,690 --> 00:31:47,680 and there's been a lot of work done on ADC, 651 00:31:47,680 --> 00:31:49,840 analog-to-digital conversion. 652 00:31:49,840 --> 00:31:54,410 And then the opposite is digital to analog. 653 00:31:54,410 --> 00:31:57,400 And depending on how many bits you have available to you, 654 00:31:57,400 --> 00:32:01,440 you're going to have some discretization errors. 655 00:32:01,440 --> 00:32:05,260 A purely digital interface, we refer to it as DIO, 656 00:32:05,260 --> 00:32:07,414 Digital Input/Output. 657 00:32:07,414 --> 00:32:08,830 And then of course, wireless, this 658 00:32:08,830 --> 00:32:11,470 is the standard, the 802.11 standard, 659 00:32:11,470 --> 00:32:14,380 which is really very important these days. 660 00:32:14,380 --> 00:32:17,920 If you think about why we need all this, 661 00:32:17,920 --> 00:32:20,290 mainly, the main reason is control, right? 662 00:32:20,290 --> 00:32:22,920 We have sensors, actuators, controllers. 663 00:32:22,920 --> 00:32:27,370 And because this information is never perfect-- 664 00:32:27,370 --> 00:32:31,450 there's gaps in the information or there's noise superimposed-- 665 00:32:31,450 --> 00:32:35,900 we have to also filter and amplify the information. 666 00:32:35,900 --> 00:32:38,620 And so if you think about the type of information flows-- 667 00:32:38,620 --> 00:32:42,111 for example, in spacecraft, we could have telemetry, sensor 668 00:32:42,111 --> 00:32:42,610 data. 669 00:32:42,610 --> 00:32:44,650 And telemetry fundamentally means, 670 00:32:44,650 --> 00:32:46,450 I want to know how is my system doing. 671 00:32:46,450 --> 00:32:47,530 Is it healthy? 672 00:32:47,530 --> 00:32:49,370 Are all the subsystems working? 673 00:32:49,370 --> 00:32:50,730 Are there some anomalies? 674 00:32:50,730 --> 00:32:53,630 Are components within their temperature limits? 675 00:32:53,630 --> 00:32:55,300 That's what we call telemetry. 676 00:32:55,300 --> 00:32:58,120 And then commands are when you are sending a command 677 00:32:58,120 --> 00:33:02,800 to the system and say, you know, turn this way, turn that way. 678 00:33:02,800 --> 00:33:05,380 And fundamentally, telemetry and commands 679 00:33:05,380 --> 00:33:08,050 are treated quite differently. 680 00:33:08,050 --> 00:33:10,750 So where do you need better quality information? 681 00:33:10,750 --> 00:33:12,820 Where is the link more important, 682 00:33:12,820 --> 00:33:15,890 on the telemetry side or sending commands? 683 00:33:15,890 --> 00:33:16,640 What do you think? 684 00:33:20,490 --> 00:33:23,550 Actually, no. 685 00:33:23,550 --> 00:33:27,100 So if your telemetry is imperfect-- 686 00:33:27,100 --> 00:33:29,410 I mean, it depends a little bit what you do with it. 687 00:33:29,410 --> 00:33:31,710 But the commands are typically more critical. 688 00:33:31,710 --> 00:33:36,190 Like, if you have a bit error in a command, 689 00:33:36,190 --> 00:33:38,430 and it says fire the engine but you 690 00:33:38,430 --> 00:33:41,890 didn't want to fire the engine, you can lose your mission. 691 00:33:41,890 --> 00:33:47,100 So typically, what's known as the bit error 692 00:33:47,100 --> 00:33:49,650 rate has to be lower, meaning the link budget is 693 00:33:49,650 --> 00:33:51,300 much better for commands. 694 00:33:51,300 --> 00:33:54,500 So there's an asymmetry. 695 00:33:54,500 --> 00:33:57,070 OK, so here's some examples. 696 00:33:57,070 --> 00:33:58,800 So here's a typical control loop. 697 00:33:58,800 --> 00:34:02,670 You have your plants, your system, your actuator, sensors, 698 00:34:02,670 --> 00:34:05,820 a comparator that compares your reference signal-- 699 00:34:05,820 --> 00:34:09,060 this is how you want the system to behave-- 700 00:34:09,060 --> 00:34:13,080 and sends that to a controller, which then closes the loop. 701 00:34:13,080 --> 00:34:15,090 The example here, again from xerography, 702 00:34:15,090 --> 00:34:17,580 is you have some original document. 703 00:34:17,580 --> 00:34:20,469 You want to make a copy here. 704 00:34:20,469 --> 00:34:23,250 So there's a lamp that shines light onto your document. 705 00:34:23,250 --> 00:34:26,010 It goes through a lens, a laser diode, 706 00:34:26,010 --> 00:34:27,960 and then to your marking system. 707 00:34:27,960 --> 00:34:31,770 The OPM, you have your original, your optical system. 708 00:34:31,770 --> 00:34:34,469 You create a digital image file which then 709 00:34:34,469 --> 00:34:36,330 is sent to the marking system. 710 00:34:36,330 --> 00:34:37,980 So if we simplify all this, we can 711 00:34:37,980 --> 00:34:40,980 say there's information flow from the optical system 712 00:34:40,980 --> 00:34:42,449 to the marking system. 713 00:34:42,449 --> 00:34:47,110 From one to two, information is transferred. 714 00:34:47,110 --> 00:34:49,669 So any questions? 715 00:34:49,669 --> 00:34:52,020 So those are the four fundamental types 716 00:34:52,020 --> 00:34:54,659 of interfaces and examples with that. 717 00:34:54,659 --> 00:34:58,020 And so the theory, what the theory of design tells us 718 00:34:58,020 --> 00:34:59,880 is that this is universal. 719 00:34:59,880 --> 00:35:04,260 Any kind of interface falls into one of these types. 720 00:35:04,260 --> 00:35:07,680 And in fact, even in biological systems that's true. 721 00:35:07,680 --> 00:35:08,419 Yeah? 722 00:35:08,419 --> 00:35:10,375 AUDIENCE: When we have a system [INAUDIBLE]?? 723 00:35:15,827 --> 00:35:17,910 OLIVIER DE WECK: I'll show that to you in the DSM. 724 00:35:17,910 --> 00:35:19,650 You actually track them separately, 725 00:35:19,650 --> 00:35:21,360 but they're within the same-- 726 00:35:21,360 --> 00:35:24,690 it's one big interface, but it has these different subflows 727 00:35:24,690 --> 00:35:27,510 within it, which means that-- 728 00:35:27,510 --> 00:35:29,730 it's not just the fact that there is an interface, 729 00:35:29,730 --> 00:35:32,470 but the complexity of the interface is important. 730 00:35:32,470 --> 00:35:34,560 So the most complex interfaces obviously 731 00:35:34,560 --> 00:35:37,410 would be the ones where you have all four types, right? 732 00:35:37,410 --> 00:35:41,040 It's a physical connection, and there's an information flow, 733 00:35:41,040 --> 00:35:42,900 and there's an energy flow, and there's 734 00:35:42,900 --> 00:35:45,330 a mass flow across the interface. 735 00:35:45,330 --> 00:35:47,420 Usually, it's only a subset of those, 736 00:35:47,420 --> 00:35:51,570 but you can have interfaces where you have all these four 737 00:35:51,570 --> 00:35:53,690 present at once. 738 00:35:53,690 --> 00:35:57,075 Any questions about these types of interfaces at MIT? 739 00:36:00,529 --> 00:36:01,570 AUDIENCE: No, we're good. 740 00:36:01,570 --> 00:36:03,410 Thank you. 741 00:36:03,410 --> 00:36:04,840 OLIVIER DE WECK: OK. 742 00:36:04,840 --> 00:36:06,760 In general, can you hear better now? 743 00:36:06,760 --> 00:36:09,870 Is it pretty stable over there? 744 00:36:09,870 --> 00:36:11,388 AUDIENCE: Yeah. 745 00:36:11,388 --> 00:36:13,000 OLIVIER DE WECK: Good, all right. 746 00:36:13,000 --> 00:36:15,430 So let me talk about DSM. 747 00:36:15,430 --> 00:36:17,480 [INAUDIBLE] was asking about DSM. 748 00:36:17,480 --> 00:36:21,880 So DSM is a fairly recent method, the last couple 749 00:36:21,880 --> 00:36:25,450 of decades, for really making a map of your system 750 00:36:25,450 --> 00:36:28,910 that shows explicitly these interfaces. 751 00:36:28,910 --> 00:36:31,370 So here's a very simple example. 752 00:36:31,370 --> 00:36:33,920 This is a sample system with five components. 753 00:36:33,920 --> 00:36:38,110 It has a pump, a controller, a motor, a valve, a filter. 754 00:36:38,110 --> 00:36:40,330 And in this case, we have mass flow 755 00:36:40,330 --> 00:36:44,560 from the pump through the valve into the filter. 756 00:36:44,560 --> 00:36:48,700 And so this block diagram on the upper left, 757 00:36:48,700 --> 00:36:52,130 the exact same information is contained in this matrix. 758 00:36:52,130 --> 00:36:54,460 So a DSM is always a square matrix 759 00:36:54,460 --> 00:36:57,400 where your components are the rows and columns, equally 760 00:36:57,400 --> 00:36:59,070 labeled. 761 00:36:59,070 --> 00:37:01,420 And the diagonals here, there's no information 762 00:37:01,420 --> 00:37:03,120 on the diagonals. 763 00:37:03,120 --> 00:37:07,360 So the component connects with itself, obviously. 764 00:37:07,360 --> 00:37:11,800 So you keep the diagonals sort of blank. 765 00:37:11,800 --> 00:37:14,980 And then all of those interfaces and flows we mentioned 766 00:37:14,980 --> 00:37:17,080 are shown as off-diagonals. 767 00:37:17,080 --> 00:37:19,840 You have to be a little careful because there's 768 00:37:19,840 --> 00:37:23,260 two definitions of DSM depending on the inflows and then 769 00:37:23,260 --> 00:37:24,440 the outflows. 770 00:37:24,440 --> 00:37:27,190 So the definition that we mainly use in North America 771 00:37:27,190 --> 00:37:31,880 is where the inflows into a component are horizontal. 772 00:37:31,880 --> 00:37:35,860 So the controller sends information to the valve. 773 00:37:35,860 --> 00:37:38,980 This is that blue square right there. 774 00:37:38,980 --> 00:37:42,100 And you can see that's shown here on the block diagram. 775 00:37:42,100 --> 00:37:44,530 But the valve also sends information back 776 00:37:44,530 --> 00:37:45,906 to the controller. 777 00:37:45,906 --> 00:37:47,530 And that information would probably be, 778 00:37:47,530 --> 00:37:49,410 valve should open or close. 779 00:37:49,410 --> 00:37:51,400 And when the valve is actually opened, 780 00:37:51,400 --> 00:37:53,920 it would confirm that back to the controller 781 00:37:53,920 --> 00:37:58,910 that the status of the valve is open or closed. 782 00:37:58,910 --> 00:38:01,150 And then we have the energy flows. 783 00:38:01,150 --> 00:38:03,839 The numbers that are shown here, I 784 00:38:03,839 --> 00:38:05,380 don't want to get too much into this, 785 00:38:05,380 --> 00:38:07,870 but there's a really cool binary scheme 786 00:38:07,870 --> 00:38:10,390 for putting a number that you can then 787 00:38:10,390 --> 00:38:12,220 have one number in that cell. 788 00:38:12,220 --> 00:38:16,090 And based on that, you can, through the binary code, 789 00:38:16,090 --> 00:38:18,670 reverse engineer what kind of interface is it. 790 00:38:18,670 --> 00:38:19,720 Is it just physical? 791 00:38:19,720 --> 00:38:21,230 Is it just binary? 792 00:38:21,230 --> 00:38:25,810 It's a kind of trick for replacing these colors 793 00:38:25,810 --> 00:38:27,580 and cells with just a number that 794 00:38:27,580 --> 00:38:30,550 allows you to then reverse engineer what kind of interface 795 00:38:30,550 --> 00:38:32,290 is it. 796 00:38:32,290 --> 00:38:35,680 So if it's empty, it means there's no direct connection. 797 00:38:35,680 --> 00:38:39,050 And then you have mechanical, mass flows, information, 798 00:38:39,050 --> 00:38:41,320 and energy flows. 799 00:38:41,320 --> 00:38:47,390 So let me show you this on a real example in a minute. 800 00:38:47,390 --> 00:38:49,870 So this method, design structure matrix method, 801 00:38:49,870 --> 00:38:52,240 there's some synonyms for it-- 802 00:38:52,240 --> 00:38:55,270 design dependency matrix, n-squared matrix, 803 00:38:55,270 --> 00:38:57,460 n-squared diagram. 804 00:38:57,460 --> 00:39:01,600 In matrix or graph theory, it's called an adjacency matrix. 805 00:39:01,600 --> 00:39:03,950 Don't be fooled; it's the same thing. 806 00:39:03,950 --> 00:39:05,770 It's just a square matrix that shows 807 00:39:05,770 --> 00:39:07,540 you what connects to what. 808 00:39:07,540 --> 00:39:12,220 And then the simplest one is just a binary, 0's and 1's. 809 00:39:12,220 --> 00:39:15,640 But the richer DSMs, they contain more information 810 00:39:15,640 --> 00:39:18,040 about the type of interface. 811 00:39:18,040 --> 00:39:19,720 So a lot of work was done originally 812 00:39:19,720 --> 00:39:21,460 by Don Steward on this. 813 00:39:21,460 --> 00:39:23,920 This is an original reference from '81. 814 00:39:23,920 --> 00:39:27,310 And then at MIT, Professor Steve Eppinger at the Sloan School 815 00:39:27,310 --> 00:39:29,980 has done a lot of work on DSMs. 816 00:39:29,980 --> 00:39:33,160 So fundamentally, it's a matrix representation 817 00:39:33,160 --> 00:39:35,320 of your product architecture. 818 00:39:35,320 --> 00:39:37,270 Most of the literature you find just 819 00:39:37,270 --> 00:39:40,100 uses the simplest, this binary DSM. 820 00:39:40,100 --> 00:39:42,880 But I'm actually a fan of adding information 821 00:39:42,880 --> 00:39:45,190 into the DSM that tells you about the physics 822 00:39:45,190 --> 00:39:47,810 of the interface as well. 823 00:39:47,810 --> 00:39:50,330 So let's look at an example here. 824 00:39:50,330 --> 00:39:52,420 Do you remember the refrigerator case study 825 00:39:52,420 --> 00:39:54,430 we did a few sessions ago? 826 00:39:54,430 --> 00:40:00,650 So this is a liaison diagram, level one decomposition. 827 00:40:00,650 --> 00:40:04,390 So the refrigerator's been decomposed into 10 components-- 828 00:40:04,390 --> 00:40:06,610 the door, the condenser, power supply, 829 00:40:06,610 --> 00:40:08,620 all the way to the compressor. 830 00:40:08,620 --> 00:40:11,560 And all we're doing here is saying 831 00:40:11,560 --> 00:40:14,650 what is physically connected to what. 832 00:40:14,650 --> 00:40:17,380 And the reason it's called a liaison diagram 833 00:40:17,380 --> 00:40:20,200 is because it's going to help you think about the assembly 834 00:40:20,200 --> 00:40:22,670 sequence of how this should be put together. 835 00:40:22,670 --> 00:40:26,870 So another way to think about this is like the skeleton. 836 00:40:26,870 --> 00:40:30,590 This is the skeleton of your system. 837 00:40:30,590 --> 00:40:34,510 So if you do have a liaison diagram, shown again-- 838 00:40:34,510 --> 00:40:37,180 this is the same one on the left here-- 839 00:40:37,180 --> 00:40:40,330 you can just translate that to a matrix. 840 00:40:40,330 --> 00:40:45,150 The matrix, this DSM here, I'm using X's instead of 1's. 841 00:40:45,150 --> 00:40:49,630 But these two contain exactly the same information. 842 00:40:49,630 --> 00:40:53,860 One is a graph view, and the other is a matrix view, 843 00:40:53,860 --> 00:40:55,330 but they're identical-- 844 00:40:55,330 --> 00:40:57,220 the same thing, the same information, 845 00:40:57,220 --> 00:40:59,740 just presented in a different way. 846 00:40:59,740 --> 00:41:02,560 And of course, if you pay close attention to the matrix, 847 00:41:02,560 --> 00:41:04,450 you'll see it's actually symmetric. 848 00:41:08,240 --> 00:41:11,990 So then we can say, well, what kinds of interfaces? 849 00:41:11,990 --> 00:41:14,300 And here come the four types that we just 850 00:41:14,300 --> 00:41:17,480 discussed-- the physical connection, energy flow, 851 00:41:17,480 --> 00:41:19,950 mass flow, and information flow. 852 00:41:19,950 --> 00:41:23,740 So if we apply that idea to the refrigerator, 853 00:41:23,740 --> 00:41:29,220 here's our 10 by 10 physical connections. 854 00:41:29,220 --> 00:41:31,580 So we have a design scripture matrix 855 00:41:31,580 --> 00:41:33,080 with only the black marks. 856 00:41:33,080 --> 00:41:34,640 So it's symmetric. 857 00:41:34,640 --> 00:41:38,450 And this reflects the liaison diagram. 858 00:41:38,450 --> 00:41:41,900 And then we can add, on top of that, the mass flows, which 859 00:41:41,900 --> 00:41:44,010 are in red. 860 00:41:44,010 --> 00:41:45,710 So the mass flows, the way you read 861 00:41:45,710 --> 00:41:49,040 this is we have a mass flow from the compressor 862 00:41:49,040 --> 00:41:54,110 to the condenser, and then from the condenser 863 00:41:54,110 --> 00:41:58,940 to the evaporator, and then from the evaporator back 864 00:41:58,940 --> 00:42:00,160 to the compressor. 865 00:42:00,160 --> 00:42:01,550 Do you see that? 866 00:42:01,550 --> 00:42:05,060 And we could rearrange these to be closer to each other. 867 00:42:05,060 --> 00:42:08,290 We could regroup the DSM. 868 00:42:08,290 --> 00:42:15,500 But essentially, these red marks represent the refrigerant cycle 869 00:42:15,500 --> 00:42:19,850 within the refrigerator, within the machine. 870 00:42:19,850 --> 00:42:23,990 And so you can check by tracing these from row to column. 871 00:42:23,990 --> 00:42:28,370 You can actually check whether the loop is properly closed. 872 00:42:28,370 --> 00:42:33,440 So then energy flows, since the refrigerant loop that I just 873 00:42:33,440 --> 00:42:37,250 talked to you about is important, 874 00:42:37,250 --> 00:42:40,370 you can see that those connections 875 00:42:40,370 --> 00:42:45,050 we just talked about always also carry a energy flow with them. 876 00:42:45,050 --> 00:42:48,740 And then finally, information flow is very little. 877 00:42:48,740 --> 00:42:52,140 So in this case, there's only one that I've represented, 878 00:42:52,140 --> 00:42:56,480 which is the flow from the thermostat to the power supply. 879 00:42:56,480 --> 00:42:59,540 The thermostat defines the temperature at which you 880 00:42:59,540 --> 00:43:01,340 want to hold the refrigerator. 881 00:43:01,340 --> 00:43:04,580 And if temperature goes up, the power supply 882 00:43:04,580 --> 00:43:09,710 kicks in and drives the compressor. 883 00:43:09,710 --> 00:43:14,510 And as soon as the temperature falls below some threshold, 884 00:43:14,510 --> 00:43:18,230 then it reverses, and the compressor turns off. 885 00:43:18,230 --> 00:43:20,300 So when your refrigerator kicks in, 886 00:43:20,300 --> 00:43:22,820 that's an information flow from the thermostat 887 00:43:22,820 --> 00:43:26,960 to the compressor to the power supply. 888 00:43:26,960 --> 00:43:31,520 If you own one of those fancy, new Samsung refrigerators where 889 00:43:31,520 --> 00:43:37,460 the door has a flat screen TV in it and you can push buttons, 890 00:43:37,460 --> 00:43:38,830 there's a lot more blue here. 891 00:43:38,830 --> 00:43:40,580 There would be a lot more blue in a really 892 00:43:40,580 --> 00:43:43,760 high-end refrigerator these days. 893 00:43:43,760 --> 00:43:46,363 OK, any-- yes, please. 894 00:43:46,363 --> 00:43:51,454 AUDIENCE: [INAUDIBLE] 895 00:43:51,454 --> 00:43:53,120 OLIVIER DE WECK: Yeah, so the question-- 896 00:43:53,120 --> 00:43:54,911 I'm just going to repeat the question-- is, 897 00:43:54,911 --> 00:43:57,140 can you also map the external interfaces? 898 00:43:57,140 --> 00:43:59,690 And the answer is, absolutely. 899 00:43:59,690 --> 00:44:01,640 So what you have to do, though, then 900 00:44:01,640 --> 00:44:05,360 is to kind of draw a bigger box around this 901 00:44:05,360 --> 00:44:08,210 and add here the external elements-- 902 00:44:08,210 --> 00:44:16,090 so the human user, the wall outlet, the grocery store, 903 00:44:16,090 --> 00:44:18,620 wherever you're including in your external world that's 904 00:44:18,620 --> 00:44:20,030 interfacing with it. 905 00:44:20,030 --> 00:44:23,180 You would have to represent at least those elements that 906 00:44:23,180 --> 00:44:25,520 have a direct connection with any of it 907 00:44:25,520 --> 00:44:27,380 across the system boundary. 908 00:44:27,380 --> 00:44:29,530 You would add them here, and then show-- 909 00:44:29,530 --> 00:44:32,660 but you would have to then put a big box around it 910 00:44:32,660 --> 00:44:35,918 to make it clear what's inside the system and what's outside. 911 00:44:35,918 --> 00:44:45,480 AUDIENCE: [INAUDIBLE] 912 00:44:45,480 --> 00:44:46,815 OLIVIER DE WECK: That's correct. 913 00:44:46,815 --> 00:44:47,440 That's correct. 914 00:44:47,440 --> 00:44:50,460 And we can argue-- of course for the refrigerator, 915 00:44:50,460 --> 00:44:52,290 the main interface that you typically 916 00:44:52,290 --> 00:44:55,590 have-- well, there are two. 917 00:44:55,590 --> 00:44:57,660 Let's say we go for what you're saying. 918 00:44:57,660 --> 00:45:01,620 We're going to add some external interfaces. 919 00:45:01,620 --> 00:45:08,610 For a refrigerator, what are the two or three most important? 920 00:45:08,610 --> 00:45:12,729 AUDIENCE: [INAUDIBLE] 921 00:45:12,729 --> 00:45:13,770 OLIVIER DE WECK: Correct. 922 00:45:13,770 --> 00:45:15,570 So where would that enter here? 923 00:45:20,500 --> 00:45:31,360 AUDIENCE: [INAUDIBLE] 924 00:45:31,360 --> 00:45:33,940 OLIVIER DE WECK: The power supply, right? 925 00:45:33,940 --> 00:45:38,030 So we would add, like, wall AC power. 926 00:45:38,030 --> 00:45:42,850 If we added that, then in this cell corresponding to that, 927 00:45:42,850 --> 00:45:45,620 you would have an inflow of power. 928 00:45:45,620 --> 00:45:46,120 And then-- 929 00:45:46,120 --> 00:45:50,597 AUDIENCE: [INAUDIBLE] 930 00:45:50,597 --> 00:45:52,430 OLIVIER DE WECK: So this is the power supply 931 00:45:52,430 --> 00:45:55,760 inside of the refrigerator, inside the system boundary. 932 00:45:55,760 --> 00:45:59,807 But if you had the wall outlet, then that would be an external. 933 00:45:59,807 --> 00:46:01,640 And then you would have an interface across, 934 00:46:01,640 --> 00:46:04,280 but into the power supply. 935 00:46:04,280 --> 00:46:05,060 So that's one. 936 00:46:05,060 --> 00:46:06,410 What's the other one? 937 00:46:06,410 --> 00:46:08,600 Do you remember we talked about the Carnot cycle? 938 00:46:12,558 --> 00:46:15,346 Well, yes, what do you have to do with the heat? 939 00:46:15,346 --> 00:46:16,120 AUDIENCE: Burn? 940 00:46:16,120 --> 00:46:18,620 OLIVIER DE WECK: Yeah, so the heat has to be removed, right, 941 00:46:18,620 --> 00:46:20,180 radiated out. 942 00:46:20,180 --> 00:46:21,740 And where does that happen? 943 00:46:21,740 --> 00:46:33,070 Which element here is where the heat gets removed, 944 00:46:33,070 --> 00:46:36,034 which one of these 10? 945 00:46:36,034 --> 00:46:38,964 AUDIENCE: [INAUDIBLE] 946 00:46:38,964 --> 00:46:40,130 OLIVIER DE WECK: Not really. 947 00:46:40,130 --> 00:46:42,940 The refrigerant does the internal heat transfer, 948 00:46:42,940 --> 00:46:46,810 but the heat transfer to the outside, to the air. 949 00:46:46,810 --> 00:46:50,200 Refrigerators don't work well in vacuum chambers. 950 00:46:50,200 --> 00:46:52,720 Remember that. 951 00:46:52,720 --> 00:46:55,370 It would be the evaporator. 952 00:46:55,370 --> 00:46:57,430 It's at the evaporator that you're 953 00:46:57,430 --> 00:46:59,470 radiating the heat to the surrounding 954 00:46:59,470 --> 00:47:00,490 air, the environment. 955 00:47:00,490 --> 00:47:04,330 So you'd have the surrounding air as an external system, 956 00:47:04,330 --> 00:47:08,770 and then and energy transfer from the evaporator 957 00:47:08,770 --> 00:47:11,060 to the outside. 958 00:47:11,060 --> 00:47:12,830 But it's a good question. 959 00:47:12,830 --> 00:47:16,000 So you have to always define your system boundary. 960 00:47:16,000 --> 00:47:20,450 So the question then is, when do you generate these DSMs? 961 00:47:20,450 --> 00:47:23,900 And fundamentally, there's two ways, top-down or bottom-up. 962 00:47:23,900 --> 00:47:27,930 So top-down, you start with a model of your system. 963 00:47:27,930 --> 00:47:30,110 And we talked about those system modeling languages, 964 00:47:30,110 --> 00:47:34,190 like SysML, or in this case, OPM. 965 00:47:34,190 --> 00:47:35,600 And then you hide some things. 966 00:47:35,600 --> 00:47:37,190 You hide attributes and states. 967 00:47:37,190 --> 00:47:39,590 You collapse all your processes into 968 00:47:39,590 --> 00:47:44,030 these tagged structural links, and then you generate your DSM. 969 00:47:44,030 --> 00:47:47,000 And then the other way to do it is bottom-up where you already 970 00:47:47,000 --> 00:47:48,950 have a design or a system. 971 00:47:48,950 --> 00:47:50,270 You decompose it. 972 00:47:50,270 --> 00:47:52,610 For example, you do a product dissection. 973 00:47:52,610 --> 00:47:55,880 You do your bill of materials or your parts list. 974 00:47:55,880 --> 00:47:59,660 Then you do your liaison diagram. 975 00:47:59,660 --> 00:48:01,850 You show the physical connections. 976 00:48:01,850 --> 00:48:03,980 And then on top of those physical connections, 977 00:48:03,980 --> 00:48:07,190 you map your mass, energy and information flows. 978 00:48:07,190 --> 00:48:11,240 And once you have that, then you can start manipulating the DSM. 979 00:48:11,240 --> 00:48:13,040 You can group components differently. 980 00:48:13,040 --> 00:48:15,930 You can show modules and so forth. 981 00:48:15,930 --> 00:48:18,680 So I want to show you a real world-- this is not just 982 00:48:18,680 --> 00:48:21,860 sort of an academic example, but this is a real-world example 983 00:48:21,860 --> 00:48:26,590 of a DSM that we built for a research project with Xerox. 984 00:48:30,020 --> 00:48:33,440 The machine that's represented here is in the upper right. 985 00:48:33,440 --> 00:48:37,760 This is the iGen3 digital printing press. 986 00:48:37,760 --> 00:48:42,890 And represented here is its architecture, its interfaces, 987 00:48:42,890 --> 00:48:45,890 as an 84 by 84 DSM. 988 00:48:45,890 --> 00:48:50,310 So the machine has been decomposed into 84, really, 989 00:48:50,310 --> 00:48:53,910 subsystems, because you have a lot more than 84 parts. 990 00:48:53,910 --> 00:48:58,020 But at this level of abstraction of 84 by 84, 991 00:48:58,020 --> 00:49:04,470 we have a total of 572 physical connections, 45 mass flows, 992 00:49:04,470 --> 00:49:10,399 167 energy flows, and 165 information flows. 993 00:49:10,399 --> 00:49:11,940 And there's a little more detail here 994 00:49:11,940 --> 00:49:13,670 in terms of the types of flows. 995 00:49:13,670 --> 00:49:18,360 So in red, we have mass flows-- paper, toner, air, 996 00:49:18,360 --> 00:49:23,790 ozone and dirt, because you've got to keep dirt out of the-- 997 00:49:23,790 --> 00:49:26,540 the printing engine is shown here in the middle. 998 00:49:26,540 --> 00:49:28,600 Let me just show this. 999 00:49:28,600 --> 00:49:31,350 So you're probably scratching your head now. 1000 00:49:31,350 --> 00:49:34,320 Ozone, really? 1001 00:49:34,320 --> 00:49:35,270 Ozone? 1002 00:49:35,270 --> 00:49:38,140 Why do we keep track of ozone? 1003 00:49:38,140 --> 00:49:40,270 What do you think? 1004 00:49:40,270 --> 00:49:43,211 Why do we care about ozone? 1005 00:49:43,211 --> 00:49:46,620 AUDIENCE: [INAUDIBLE] at ground level. 1006 00:49:46,620 --> 00:49:49,080 OLIVIER DE WECK: It's a pollutant at ground level. 1007 00:49:49,080 --> 00:49:51,430 And where does the pollutant come from? 1008 00:49:54,136 --> 00:49:55,490 AUDIENCE: [INAUDIBLE] 1009 00:49:55,490 --> 00:49:58,480 OLIVIER DE WECK: So it's generated inside the machine. 1010 00:49:58,480 --> 00:50:02,545 The reason is these machines use very high voltages. 1011 00:50:02,545 --> 00:50:06,130 Xerographic processes use pretty high voltages. 1012 00:50:06,130 --> 00:50:08,280 So if you apply a very high voltage, 1013 00:50:08,280 --> 00:50:11,350 you actually generate ozone through a-- 1014 00:50:11,350 --> 00:50:13,200 it's an electrochemical reaction that 1015 00:50:13,200 --> 00:50:15,670 happens inside the machine. 1016 00:50:15,670 --> 00:50:19,590 Nobody used to care about this until people working 1017 00:50:19,590 --> 00:50:23,010 in these print shops started to feel symptoms 1018 00:50:23,010 --> 00:50:24,940 and started to complain. 1019 00:50:24,940 --> 00:50:27,660 So in the US, it's called OSHA. 1020 00:50:27,660 --> 00:50:31,414 That's the Occupation Safety and Health Administration. 1021 00:50:31,414 --> 00:50:33,330 They came in, and they took some measurements. 1022 00:50:33,330 --> 00:50:37,420 And wow, the ozone level is through the roof. 1023 00:50:37,420 --> 00:50:42,870 It's multiple times above the allowable maximum limit. 1024 00:50:42,870 --> 00:50:44,760 And the reason is because of this kind 1025 00:50:44,760 --> 00:50:47,080 of xerographic process. 1026 00:50:47,080 --> 00:50:50,800 So now, these machines have to control the ozone. 1027 00:50:50,800 --> 00:50:56,100 And essentially, this box here, the printing engine, is sealed. 1028 00:50:56,100 --> 00:50:58,050 It has its own air conditioning system 1029 00:50:58,050 --> 00:51:00,480 to prevent the ozone from leaking out. 1030 00:51:00,480 --> 00:51:06,960 It's a great example of how occupational safety drove 1031 00:51:06,960 --> 00:51:09,090 new requirements which then increased 1032 00:51:09,090 --> 00:51:10,690 the complexity of the machine. 1033 00:51:10,690 --> 00:51:13,140 So we keep track of the ozone flows. 1034 00:51:13,140 --> 00:51:16,110 And then in green, we have the usual suspects-- 1035 00:51:16,110 --> 00:51:19,890 high voltage, low voltage, DC, mechanical energy, and then 1036 00:51:19,890 --> 00:51:20,710 heat energy. 1037 00:51:20,710 --> 00:51:23,700 So there's some coding that goes on here. 1038 00:51:23,700 --> 00:51:31,510 So in total, we have 1,033 non-empty cells here, 1039 00:51:31,510 --> 00:51:35,280 which is about 3.7%, so about 4%. 1040 00:51:35,280 --> 00:51:39,310 This DSM has a density of about 4%. 1041 00:51:39,310 --> 00:51:41,210 And that doesn't sound very high. 1042 00:51:41,210 --> 00:51:46,360 But believe me, this is a very complex machine. 1043 00:51:46,360 --> 00:51:51,030 This DSM that you see here is version 31. 1044 00:51:51,030 --> 00:51:55,400 So it took 31 versions to create this DSM. 1045 00:51:55,400 --> 00:51:58,230 And so why did it take that long? 1046 00:51:58,230 --> 00:52:03,390 Because this was published as an official document 1047 00:52:03,390 --> 00:52:06,820 capturing the architecture of this machine. 1048 00:52:06,820 --> 00:52:10,140 So what that means in practice is of these 84 components 1049 00:52:10,140 --> 00:52:13,980 or subsystems, every one of those components or subsystems 1050 00:52:13,980 --> 00:52:17,670 is owned by a different person or a different team. 1051 00:52:17,670 --> 00:52:25,100 And we went through a process where all of these 84 teams, 1052 00:52:25,100 --> 00:52:29,780 for their component, had to sign off and say, yep, 1053 00:52:29,780 --> 00:52:32,270 you've accurately captured my interfaces. 1054 00:52:32,270 --> 00:52:35,630 My physical, mass, information and energy interfaces 1055 00:52:35,630 --> 00:52:39,170 have been correctly captured in this matrix. 1056 00:52:39,170 --> 00:52:44,930 And oh, by the way, the empty cells have also been certified. 1057 00:52:44,930 --> 00:52:46,940 So if there's an empty cell here, 1058 00:52:46,940 --> 00:52:49,035 there's no question whether, well, maybe it's 1059 00:52:49,035 --> 00:52:51,560 just somebody forgot to put the interface there. 1060 00:52:51,560 --> 00:52:52,760 No. 1061 00:52:52,760 --> 00:52:56,750 The owner of that row and column certified 1062 00:52:56,750 --> 00:52:59,240 that there is no direct interface 1063 00:52:59,240 --> 00:53:01,430 between this component and the other component. 1064 00:53:01,430 --> 00:53:03,640 That's why it took 31 versions. 1065 00:53:03,640 --> 00:53:07,670 But now that you have this map, you can do a lot of things 1066 00:53:07,670 --> 00:53:08,220 with it. 1067 00:53:08,220 --> 00:53:11,090 You can do architectural analysis. 1068 00:53:11,090 --> 00:53:13,640 You can benchmark against other products. 1069 00:53:13,640 --> 00:53:16,160 You can do what we call technology infusion 1070 00:53:16,160 --> 00:53:17,780 analysis, which was done. 1071 00:53:17,780 --> 00:53:19,370 You can say, well, what if we add 1072 00:53:19,370 --> 00:53:21,780 this feature in the machine? 1073 00:53:21,780 --> 00:53:26,060 OK, well, where is that feature going to go? 1074 00:53:26,060 --> 00:53:28,310 What are the interfaces of that new feature? 1075 00:53:28,310 --> 00:53:29,930 Do we have to change the software? 1076 00:53:29,930 --> 00:53:33,530 Do we have to physically integrate it? 1077 00:53:33,530 --> 00:53:36,980 This is a fantastic, very powerful tool 1078 00:53:36,980 --> 00:53:42,700 for showing the interfaces and managing them. 1079 00:53:42,700 --> 00:53:47,460 Any questions about this DSM approach? 1080 00:53:47,460 --> 00:53:51,360 The real way to do this is to actually build one yourself. 1081 00:53:51,360 --> 00:53:54,600 And you can see it's a lot of work, but once you have it, 1082 00:53:54,600 --> 00:53:55,835 it's very powerful. 1083 00:53:55,835 --> 00:53:56,543 Yes, [INAUDIBLE]? 1084 00:53:56,543 --> 00:54:04,590 AUDIENCE: [INAUDIBLE] 1085 00:54:04,590 --> 00:54:06,930 OLIVIER DE WECK: So there's templates, 1086 00:54:06,930 --> 00:54:08,430 like Excel templates. 1087 00:54:08,430 --> 00:54:10,080 But there's also software. 1088 00:54:10,080 --> 00:54:12,120 There's a website called DSMweb.org. 1089 00:54:14,760 --> 00:54:17,570 And I'm happy to send that out for you as a resource. 1090 00:54:17,570 --> 00:54:20,360 There's even yearly conferences. 1091 00:54:20,360 --> 00:54:24,250 There are DSM conferences where people get together 1092 00:54:24,250 --> 00:54:27,930 from academia, from different companies-- 1093 00:54:27,930 --> 00:54:29,830 like BMW, for example. 1094 00:54:29,830 --> 00:54:31,320 I'm a big fan of BMW. 1095 00:54:31,320 --> 00:54:34,920 I think they do a great job. 1096 00:54:34,920 --> 00:54:37,680 Their system engineering is very, very strong. 1097 00:54:37,680 --> 00:54:39,550 They use this on all their vehicles, 1098 00:54:39,550 --> 00:54:41,490 on all their subsystems. 1099 00:54:41,490 --> 00:54:44,190 And so this DSMweb.org, you'll see 1100 00:54:44,190 --> 00:54:48,360 sort of the software that's available and so forth. 1101 00:54:48,360 --> 00:54:51,130 There's also some commercial software available. 1102 00:54:51,130 --> 00:54:52,470 So yeah, please, go ahead. 1103 00:54:52,470 --> 00:54:56,634 AUDIENCE: [INAUDIBLE] 1104 00:54:56,634 --> 00:54:58,440 OLIVIER DE WECK: Ah. 1105 00:54:58,440 --> 00:55:01,540 So let's think about this. 1106 00:55:01,540 --> 00:55:05,430 The 7 plus or minus 2, how does the 7 plus or minus 1107 00:55:05,430 --> 00:55:07,650 2 relate to the 84? 1108 00:55:07,650 --> 00:55:10,290 This is an 84 by 84 DSM. 1109 00:55:10,290 --> 00:55:12,848 So how do the two relate to each other? 1110 00:55:12,848 --> 00:55:15,032 AUDIENCE: [INAUDIBLE] 1111 00:55:15,032 --> 00:55:15,990 OLIVIER DE WECK: Right. 1112 00:55:15,990 --> 00:55:21,406 But what's the relationship of that rule with the 84 by 84? 1113 00:55:21,406 --> 00:55:25,110 AUDIENCE: [INAUDIBLE] 1114 00:55:25,110 --> 00:55:27,630 OLIVIER DE WECK: Yeah, so this machine might have-- 1115 00:55:27,630 --> 00:55:29,310 if you really want every little detail, 1116 00:55:29,310 --> 00:55:33,500 this is probably a level 4, level 5, right? 1117 00:55:33,500 --> 00:55:36,310 But what's shown here, this is essentially the equivalent 1118 00:55:36,310 --> 00:55:39,250 of a level two representation. 1119 00:55:39,250 --> 00:55:41,140 So do you remember 7 plus or minus 1120 00:55:41,140 --> 00:55:49,600 2 squared is somewhere between 25 at the low end and 81. 1121 00:55:49,600 --> 00:55:52,200 So we're just slightly above this. 1122 00:55:52,200 --> 00:55:55,800 So my argument here is this is essentially a level two 1123 00:55:55,800 --> 00:56:00,860 DSM where you decompose the system down two levels, 1124 00:56:00,860 --> 00:56:03,530 and then you show the DSM the interfaces 1125 00:56:03,530 --> 00:56:05,360 at this level of abstraction. 1126 00:56:05,360 --> 00:56:07,170 You could go level three, but now you're 1127 00:56:07,170 --> 00:56:11,030 going to have a 500 by 500 DSM, which 1128 00:56:11,030 --> 00:56:15,560 is almost impossible for a human to really understand and get 1129 00:56:15,560 --> 00:56:17,370 any use out of it. 1130 00:56:17,370 --> 00:56:20,480 There's also been some work in hierarchical DSMs. 1131 00:56:20,480 --> 00:56:26,000 So for example, if you look at the feeder system-- 1132 00:56:26,000 --> 00:56:28,586 you see in the upper left here that says "feeder"? 1133 00:56:28,586 --> 00:56:30,020 Well, what is the feeder? 1134 00:56:30,020 --> 00:56:34,070 It's all this upfront stuff that feeds your paper and your media 1135 00:56:34,070 --> 00:56:35,630 into the print engine. 1136 00:56:35,630 --> 00:56:36,830 So you could take-- 1137 00:56:36,830 --> 00:56:38,660 this is just two rows here. 1138 00:56:38,660 --> 00:56:42,060 You could take those and break those into much greater detail. 1139 00:56:42,060 --> 00:56:45,380 So you'd have-- kind of like you guys did the requirements tree, 1140 00:56:45,380 --> 00:56:46,910 the requirements hierarchy. 1141 00:56:46,910 --> 00:56:49,180 You can have a hierarchy of DSMs. 1142 00:56:49,180 --> 00:56:51,650 I'm not showing that here, but that's the relationship. 1143 00:56:51,650 --> 00:56:55,490 This is essentially a level two decomposition DSM. 1144 00:56:55,490 --> 00:56:57,650 Good question. 1145 00:56:57,650 --> 00:56:59,000 MIT, any questions? 1146 00:57:04,153 --> 00:57:06,820 AUDIENCE: Yeah, I just have one question about the liaison 1147 00:57:06,820 --> 00:57:08,440 diagram. 1148 00:57:08,440 --> 00:57:09,400 OLIVIER DE WECK: Yeah? 1149 00:57:09,400 --> 00:57:12,070 AUDIENCE: Is there, I guess, any meaning or information 1150 00:57:12,070 --> 00:57:15,970 that comes from the distances between-- like when you draw 1151 00:57:15,970 --> 00:57:19,330 it, the way that you draw it? 1152 00:57:19,330 --> 00:57:21,830 OLIVIER DE WECK: No, there really isn't. 1153 00:57:21,830 --> 00:57:25,900 So the tricky thing is if it's a very complex system, 1154 00:57:25,900 --> 00:57:29,390 like if we did the liaison diagram for the 84 by 84-- 1155 00:57:29,390 --> 00:57:31,760 let me just go back here-- 1156 00:57:31,760 --> 00:57:33,770 it would be pretty messy. 1157 00:57:33,770 --> 00:57:38,160 And so you want to sort of do the layout of the liaison 1158 00:57:38,160 --> 00:57:41,080 diagram so you minimize the number of crossing lines, 1159 00:57:41,080 --> 00:57:42,520 for example. 1160 00:57:42,520 --> 00:57:47,470 There's actually some graph visualization algorithms 1161 00:57:47,470 --> 00:57:52,960 that try to create as clean as possible of a graph. 1162 00:57:52,960 --> 00:57:53,920 One of those is-- 1163 00:57:53,920 --> 00:57:56,920 it's called a spring energy algorithm. 1164 00:57:56,920 --> 00:58:00,100 It basically treats these links as though they were springs. 1165 00:58:00,100 --> 00:58:02,400 And then depending on how far things are, 1166 00:58:02,400 --> 00:58:04,360 there's spring energy. 1167 00:58:04,360 --> 00:58:07,160 And you want to minimize the total energy of the system. 1168 00:58:07,160 --> 00:58:12,250 And by doing that, usually, the graph looks pretty clean. 1169 00:58:12,250 --> 00:58:16,240 But there's no inherent meaning in how you lay it out. 1170 00:58:16,240 --> 00:58:18,620 Does that makes sense, Sam? 1171 00:58:18,620 --> 00:58:20,370 AUDIENCE: Yeah, thank you. 1172 00:58:20,370 --> 00:58:22,296 AUDIENCE: I have a question too-- 1173 00:58:22,296 --> 00:58:23,520 OLIVIER DE WECK: Go ahead. 1174 00:58:23,520 --> 00:58:26,070 AUDIENCE: --about the liaison diagrams. 1175 00:58:26,070 --> 00:58:28,170 I was doing something similar, at least where 1176 00:58:28,170 --> 00:58:31,560 I had each block as a functional aspect of the system. 1177 00:58:31,560 --> 00:58:33,780 It was for an avionics system. 1178 00:58:33,780 --> 00:58:36,390 And I found that I needed to, like, make power 1179 00:58:36,390 --> 00:58:38,900 a separate thing because it was more of like a tree. 1180 00:58:38,900 --> 00:58:40,980 We had one power distribution system. 1181 00:58:40,980 --> 00:58:44,970 But the thing I wanted to bring up was the information flow. 1182 00:58:44,970 --> 00:58:47,190 Originally I just had one color for that, 1183 00:58:47,190 --> 00:58:49,250 but I didn't find that too useful. 1184 00:58:49,250 --> 00:58:51,930 What I plan on doing is splitting it 1185 00:58:51,930 --> 00:58:54,400 up into command and telemetry. 1186 00:58:54,400 --> 00:59:00,990 So is drawing information just as one entity typically done? 1187 00:59:00,990 --> 00:59:04,320 Because I think it's more useful to have command and telemetry, 1188 00:59:04,320 --> 00:59:09,210 or like data and commands. 1189 00:59:09,210 --> 00:59:11,430 OLIVIER DE WECK: So you know, that's fine. 1190 00:59:11,430 --> 00:59:13,680 And depending on what you're going to use it for, 1191 00:59:13,680 --> 00:59:16,770 you can redefine the colors. 1192 00:59:16,770 --> 00:59:20,220 There's no global standard for how to properly do a DSM. 1193 00:59:20,220 --> 00:59:22,020 That's still kind of evolving. 1194 00:59:22,020 --> 00:59:25,080 So if you have different types-- 1195 00:59:25,080 --> 00:59:27,420 and you saw in the Xerox example, 1196 00:59:27,420 --> 00:59:30,750 we used letters as well for the different types of mass flows-- 1197 00:59:30,750 --> 00:59:33,170 P for paper, T for toner. 1198 00:59:33,170 --> 00:59:37,140 So the way you're going to code it is really up to you. 1199 00:59:37,140 --> 00:59:41,640 There's no global, right one way to do it yet. 1200 00:59:41,640 --> 00:59:42,730 So that sounds good. 1201 00:59:42,730 --> 00:59:45,852 Actually, you did the right thing there. 1202 00:59:45,852 --> 00:59:46,810 AUDIENCE: Cool, thanks. 1203 00:59:46,810 --> 00:59:49,478 OLIVIER DE WECK: How are we doing time-wise? 1204 00:59:49,478 --> 00:59:50,450 AUDIENCE: [INAUDIBLE] 1205 00:59:50,450 --> 00:59:52,670 OLIVIER DE WECK: OK. 1206 00:59:52,670 --> 00:59:54,210 Any other questions? 1207 00:59:54,210 --> 00:59:55,126 Yes? 1208 00:59:55,126 --> 00:59:58,271 AUDIENCE: What kind of [INAUDIBLE]?? 1209 00:59:58,271 --> 01:00:00,020 OLIVIER DE WECK: So why are we doing this? 1210 01:00:00,020 --> 01:00:01,796 AUDIENCE: Yes, exactly. 1211 01:00:01,796 --> 01:00:03,420 OLIVIER DE WECK: Well, pretty pictures. 1212 01:00:03,420 --> 01:00:06,890 No, I'm sort of half-joking. 1213 01:00:06,890 --> 01:00:12,110 Part of it is definitely a graphical, visual, powerful way 1214 01:00:12,110 --> 01:00:15,080 to convey the whole system architecture. 1215 01:00:15,080 --> 01:00:17,180 If you think about it, there aren't too many ways 1216 01:00:17,180 --> 01:00:20,390 of actually visualizing the whole system. 1217 01:00:20,390 --> 01:00:22,220 This is a very abstract view. 1218 01:00:22,220 --> 01:00:26,200 But there's a-- and then this DSM here at Xerox, 1219 01:00:26,200 --> 01:00:29,856 it printed out as a huge, wall-sized poster. 1220 01:00:29,856 --> 01:00:30,980 And you put it on the wall. 1221 01:00:30,980 --> 01:00:33,380 And people actually go up there and say, ah, 1222 01:00:33,380 --> 01:00:34,580 that's my subsystem. 1223 01:00:34,580 --> 01:00:36,660 And then they argue and so forth. 1224 01:00:36,660 --> 01:00:39,255 So part of it is definitely a visualization. 1225 01:00:39,255 --> 01:00:40,880 But the other thing you can do with it, 1226 01:00:40,880 --> 01:00:43,400 particularly if you start putting the numbers in, 1227 01:00:43,400 --> 01:00:47,280 you can use this for complexity quantification. 1228 01:00:47,280 --> 01:00:50,420 So you can quantify the complexity of your system 1229 01:00:50,420 --> 01:00:53,960 and then compare it against a competing system 1230 01:00:53,960 --> 01:00:55,790 or a prior generation system. 1231 01:00:55,790 --> 01:00:58,339 AUDIENCE: [INAUDIBLE] 1232 01:00:58,339 --> 01:00:59,380 OLIVIER DE WECK: Correct. 1233 01:00:59,380 --> 01:01:01,972 AUDIENCE: You could not have a macro view on it. 1234 01:01:01,972 --> 01:01:02,930 OLIVIER DE WECK: Right. 1235 01:01:02,930 --> 01:01:05,570 So this is a macro view, but if you really-- then you 1236 01:01:05,570 --> 01:01:08,300 need to put numbers for the component 1237 01:01:08,300 --> 01:01:11,580 complexities, the interface complexities and so forth. 1238 01:01:11,580 --> 01:01:14,030 And there's quite a bit of research in this. 1239 01:01:14,030 --> 01:01:17,000 So you can then move this from a purely visual tool 1240 01:01:17,000 --> 01:01:19,160 to becoming an analytic tool. 1241 01:01:19,160 --> 01:01:19,887 Yes? 1242 01:01:19,887 --> 01:01:33,304 AUDIENCE: [INAUDIBLE] 1243 01:01:33,304 --> 01:01:34,220 OLIVIER DE WECK: Yeah. 1244 01:01:34,220 --> 01:01:36,020 And one thing we'll talk about next 1245 01:01:36,020 --> 01:01:41,360 is interface control documents as an important-- so in theory, 1246 01:01:41,360 --> 01:01:43,820 in theory, every one of these interfaces, 1247 01:01:43,820 --> 01:01:48,400 you can have a document or a digital file 1248 01:01:48,400 --> 01:01:50,450 if you click on this. 1249 01:01:50,450 --> 01:01:53,330 And this is where the tools are not fully mature yet. 1250 01:01:53,330 --> 01:01:55,670 But you'd like to be able to click on this, 1251 01:01:55,670 --> 01:01:57,980 and then it links you to another document. 1252 01:01:57,980 --> 01:01:59,750 And what is that other document? 1253 01:01:59,750 --> 01:02:02,300 It's the ICD, the Interface Control Document 1254 01:02:02,300 --> 01:02:03,992 that defines that interface. 1255 01:02:03,992 --> 01:02:05,450 And then you have a lot of details. 1256 01:02:05,450 --> 01:02:09,290 Like if it's an electrical connector, how many pins, 1257 01:02:09,290 --> 01:02:11,930 what's the function of each pin? 1258 01:02:11,930 --> 01:02:16,220 And that's really where things become useful, OK? 1259 01:02:16,220 --> 01:02:18,160 All right, let me move on. 1260 01:02:18,160 --> 01:02:19,920 But I did upload for you guys. 1261 01:02:19,920 --> 01:02:26,780 So it's really up to you if you want to do a DSM at your PDR. 1262 01:02:26,780 --> 01:02:30,480 It would be nice, but I don't absolutely-- 1263 01:02:30,480 --> 01:02:32,900 if you like this, if you think it's useful, 1264 01:02:32,900 --> 01:02:34,400 try it out for the PDR. 1265 01:02:34,400 --> 01:02:36,950 But it's not a must-have requirement. 1266 01:02:36,950 --> 01:02:42,200 But in the full-year SDM class that we teach, 1267 01:02:42,200 --> 01:02:44,010 we're just doing that right now. 1268 01:02:44,010 --> 01:02:46,670 And so you can use it for reverse 1269 01:02:46,670 --> 01:02:49,260 engineering the architecture of a software package. 1270 01:02:49,260 --> 01:02:51,290 You like open source software. 1271 01:02:51,290 --> 01:02:53,000 You just download it. 1272 01:02:53,000 --> 01:02:55,310 You generate a DSM, and you can start 1273 01:02:55,310 --> 01:02:58,130 seeing what are the different modules in the code, what 1274 01:02:58,130 --> 01:03:00,320 do they do, how do they talk to each other. 1275 01:03:00,320 --> 01:03:01,825 It's incredibly powerful. 1276 01:03:01,825 --> 01:03:06,740 But what I did do is I uploaded for you guys-- 1277 01:03:06,740 --> 01:03:08,870 I said there's a top-down, and then there's 1278 01:03:08,870 --> 01:03:11,930 a reverse engineering or bottom-up way of doing it. 1279 01:03:11,930 --> 01:03:16,000 This is a 17-step procedure, pretty detailed, 1280 01:03:16,000 --> 01:03:20,120 a document that I uploaded under the reading section. 1281 01:03:20,120 --> 01:03:23,410 So if you go on the website, under the reading section, 1282 01:03:23,410 --> 01:03:24,320 you will find-- 1283 01:03:24,320 --> 01:03:27,440 this is like a four or five-page document that says step 1284 01:03:27,440 --> 01:03:29,110 by step how do you do a DSM. 1285 01:03:31,640 --> 01:03:35,000 And again, so at a high level, then, what you get out of it 1286 01:03:35,000 --> 01:03:36,350 is-- 1287 01:03:36,350 --> 01:03:38,300 here, I've sort of hidden the details 1288 01:03:38,300 --> 01:03:39,530 from what you saw before-- 1289 01:03:39,530 --> 01:03:42,320 is the high-level architecture you can see. 1290 01:03:42,320 --> 01:03:44,540 So in the case of the Xerox example, 1291 01:03:44,540 --> 01:03:47,480 we have our frontend system, the feeder system, 1292 01:03:47,480 --> 01:03:48,770 the main marking engine. 1293 01:03:48,770 --> 01:03:51,020 You can see the modules. 1294 01:03:51,020 --> 01:03:53,680 And then you have the finishing system down here. 1295 01:03:53,680 --> 01:03:59,610 And so many systems and products have similar architectures. 1296 01:03:59,610 --> 01:04:04,690 OK, any questions before I move on. 1297 01:04:04,690 --> 01:04:08,950 Let's talk about ICDs, Interface Control Documents. 1298 01:04:08,950 --> 01:04:12,820 And in the system engineering engine from the handbook, 1299 01:04:12,820 --> 01:04:14,560 this is right here in the middle. 1300 01:04:14,560 --> 01:04:17,140 This is the technical management processes. 1301 01:04:17,140 --> 01:04:20,020 Number 12, interface management. 1302 01:04:20,020 --> 01:04:21,610 It's a huge deal. 1303 01:04:21,610 --> 01:04:24,550 If you don't do it well, if you don't 1304 01:04:24,550 --> 01:04:26,170 do interface management well-- 1305 01:04:26,170 --> 01:04:28,480 and we heard some examples-- 1306 01:04:28,480 --> 01:04:30,870 it's really trouble. 1307 01:04:30,870 --> 01:04:34,510 So the purpose is for controlling your development 1308 01:04:34,510 --> 01:04:35,280 effort. 1309 01:04:35,280 --> 01:04:37,900 It's really important-- I just highlighted a few words here-- 1310 01:04:37,900 --> 01:04:41,050 if you have geographically diverse technical teams. 1311 01:04:41,050 --> 01:04:44,230 And maintaining interface definition, 1312 01:04:44,230 --> 01:04:49,430 once you've decomposed your system, is absolutely critical. 1313 01:04:49,430 --> 01:04:51,820 So let me talk briefly about the different-- 1314 01:04:51,820 --> 01:04:54,400 so this is a little bit more like traditional system 1315 01:04:54,400 --> 01:04:57,910 engineering where you define the interface with a document. 1316 01:04:57,910 --> 01:05:00,430 We've said that system engineering is moving 1317 01:05:00,430 --> 01:05:02,050 to become more model-based. 1318 01:05:02,050 --> 01:05:05,230 We don't want big piles of paper anymore. 1319 01:05:05,230 --> 01:05:08,860 But the reality is that many projects and programs still 1320 01:05:08,860 --> 01:05:10,900 have a lot of these documents. 1321 01:05:10,900 --> 01:05:13,150 Ideally, you'd just have a model, a digital model 1322 01:05:13,150 --> 01:05:14,020 of the interface. 1323 01:05:14,020 --> 01:05:17,170 Then you don't need a separate document. 1324 01:05:17,170 --> 01:05:20,830 But if you are working with suppliers or contractors that 1325 01:05:20,830 --> 01:05:24,010 don't have the same modeling capability as you have, 1326 01:05:24,010 --> 01:05:25,660 these documents are critical. 1327 01:05:25,660 --> 01:05:29,290 So the first one, it's called IRD, Interface Requirements 1328 01:05:29,290 --> 01:05:30,370 Document. 1329 01:05:30,370 --> 01:05:34,300 And this is-- do you remember when you did the requirements 1330 01:05:34,300 --> 01:05:36,670 analysis in assignment two? 1331 01:05:36,670 --> 01:05:38,710 This is one of the six classes of requirements 1332 01:05:38,710 --> 01:05:41,030 is interface requirements. 1333 01:05:41,030 --> 01:05:43,570 So that's the idea, is all the interface requirements 1334 01:05:43,570 --> 01:05:46,220 are pulled together into an IRD. 1335 01:05:46,220 --> 01:05:48,070 It defines the functional, performance, 1336 01:05:48,070 --> 01:05:51,820 electrical, environmental, human, physical requirements 1337 01:05:51,820 --> 01:05:55,480 and constraints that exist at the common boundary between two 1338 01:05:55,480 --> 01:05:58,060 or more functions or system elements. 1339 01:05:58,060 --> 01:06:02,350 So you haven't designed the interface yet, physically, 1340 01:06:02,350 --> 01:06:05,090 but you've defined what the interface should be able to do, 1341 01:06:05,090 --> 01:06:08,140 its functionality. 1342 01:06:08,140 --> 01:06:10,570 Then we have-- this is the most common. 1343 01:06:10,570 --> 01:06:12,720 You've probably heard this acronym before-- 1344 01:06:12,720 --> 01:06:16,790 is ICDs, Interface Control Document or Interface Control 1345 01:06:16,790 --> 01:06:17,890 Drawing. 1346 01:06:17,890 --> 01:06:21,280 And so that is essentially the specification of the interface. 1347 01:06:21,280 --> 01:06:23,360 What does the interface look like? 1348 01:06:23,360 --> 01:06:26,740 So if you look at a ICD of an electrical connector, 1349 01:06:26,740 --> 01:06:30,190 you'll see the pin-out, you see the voltage levels, 1350 01:06:30,190 --> 01:06:33,640 you'll see the purpose of each signal for each pin. 1351 01:06:33,640 --> 01:06:35,630 Maybe there's some pins that are unused that 1352 01:06:35,630 --> 01:06:37,600 are reserved for future use. 1353 01:06:37,600 --> 01:06:39,940 That's also very clearly defined. 1354 01:06:39,940 --> 01:06:45,160 And the key about the ICD is it's a two-sided document. 1355 01:06:45,160 --> 01:06:48,010 So there's a subsystem A and a subsystem B, 1356 01:06:48,010 --> 01:06:49,990 and they share a common interface. 1357 01:06:49,990 --> 01:06:52,390 And the ICD, the Interface Control Document, 1358 01:06:52,390 --> 01:06:55,990 will define both sides of the interface as opposed 1359 01:06:55,990 --> 01:07:01,390 to IDD, Interface Definition Document, 1360 01:07:01,390 --> 01:07:05,240 which is only a one-sided or unilateral document. 1361 01:07:05,240 --> 01:07:08,290 So for example, let's say your company makes 1362 01:07:08,290 --> 01:07:13,920 a very fancy camera or sensor to be used on a spacecraft, 1363 01:07:13,920 --> 01:07:17,020 but you don't know who is going to use that camera, 1364 01:07:17,020 --> 01:07:18,370 how are they going to use it. 1365 01:07:18,370 --> 01:07:21,490 So what you're defining is essentially 1366 01:07:21,490 --> 01:07:23,320 your side of the interface-- 1367 01:07:23,320 --> 01:07:25,630 physical, functional. 1368 01:07:25,630 --> 01:07:31,480 And then it's up to whoever your customer is to take the IDD 1369 01:07:31,480 --> 01:07:33,730 and turn it into an ICD. 1370 01:07:33,730 --> 01:07:37,300 But the IDD is only one-sided, unilateral. 1371 01:07:37,300 --> 01:07:40,570 And the ICD is bilateral, interface definition. 1372 01:07:40,570 --> 01:07:42,370 Does that make sense? 1373 01:07:42,370 --> 01:07:43,780 OK. 1374 01:07:43,780 --> 01:07:47,080 So here's-- again, this is from the handbook, 1375 01:07:47,080 --> 01:07:49,690 the system engineering handbook-- 1376 01:07:49,690 --> 01:07:50,980 interface management. 1377 01:07:50,980 --> 01:07:52,570 On the input side, you start with 1378 01:07:52,570 --> 01:07:55,150 your interface requirements, any changes 1379 01:07:55,150 --> 01:07:57,790 to the interface that come up during the lifecycle 1380 01:07:57,790 --> 01:07:59,050 of the project. 1381 01:07:59,050 --> 01:08:02,080 You go through the steps and out come your interface control 1382 01:08:02,080 --> 01:08:06,370 documents, any requirements changes, and then interface 1383 01:08:06,370 --> 01:08:08,080 management work products, which might 1384 01:08:08,080 --> 01:08:10,310 be other supporting documents. 1385 01:08:10,310 --> 01:08:14,020 For example, a user manual or an operator manual 1386 01:08:14,020 --> 01:08:17,979 that will actually tell you how to properly use the interface 1387 01:08:17,979 --> 01:08:22,149 would be an example of a supporting document. 1388 01:08:22,149 --> 01:08:25,470 OK, any questions about that before we 1389 01:08:25,470 --> 01:08:28,800 get to system integration? 1390 01:08:28,800 --> 01:08:31,399 So who's actually read an ICD or had 1391 01:08:31,399 --> 01:08:34,680 an actual ICD in their hands? 1392 01:08:34,680 --> 01:08:36,440 You have, yeah? 1393 01:08:36,440 --> 01:08:39,000 [INAUDIBLE],, I'm sure you've had that many times. 1394 01:08:39,000 --> 01:08:41,316 Can you tell us what it was? 1395 01:08:41,316 --> 01:08:43,361 AUDIENCE: [INAUDIBLE] 1396 01:08:43,361 --> 01:08:45,069 OLIVIER DE WECK: Let me give you the mic. 1397 01:08:45,069 --> 01:08:49,560 So-- [INAUDIBLE] 1398 01:08:49,560 --> 01:08:53,460 AUDIENCE: So we were working on a project 1399 01:08:53,460 --> 01:08:58,770 where a Chinese asked us to contribute a camera payload 1400 01:08:58,770 --> 01:09:00,689 for their satellites. 1401 01:09:00,689 --> 01:09:08,760 And so we had to define all the electrical and data interfaces. 1402 01:09:08,760 --> 01:09:11,279 I was in charge of the data interfaces. 1403 01:09:11,279 --> 01:09:18,120 And there were design documents saying how will the interaction 1404 01:09:18,120 --> 01:09:19,380 work between the two systems. 1405 01:09:23,051 --> 01:09:23,550 Let's see. 1406 01:09:23,550 --> 01:09:26,069 Because [INAUDIBLE] protocol, I had 1407 01:09:26,069 --> 01:09:32,460 to tell the satellites what to do if the camera fails and how 1408 01:09:32,460 --> 01:09:35,010 should the camera fail if there is 1409 01:09:35,010 --> 01:09:39,702 a problem on the computer, so all the failure cases 1410 01:09:39,702 --> 01:09:40,660 and this kind of thing. 1411 01:09:40,660 --> 01:09:44,380 OLIVIER DE WECK: So it sounds like [INAUDIBLE].. 1412 01:09:44,380 --> 01:09:46,860 AUDIENCE: Yeah, it's an ICD. 1413 01:09:46,860 --> 01:09:48,210 OLIVIER DE WECK: Good. 1414 01:09:48,210 --> 01:09:50,853 And was it used then, or-- 1415 01:09:50,853 --> 01:09:54,830 AUDIENCE: [INAUDIBLE] 1416 01:09:54,830 --> 01:09:56,500 OLIVIER DE WECK: Great, excellent. 1417 01:09:56,500 --> 01:10:02,320 Any examples of an ICD or IDD or IRD on the MIT side? 1418 01:10:02,320 --> 01:10:04,051 Anybody have experience with this? 1419 01:10:11,049 --> 01:10:11,590 AUDIENCE: No. 1420 01:10:14,350 --> 01:10:16,610 OLIVIER DE WECK: Was that you, your honor? 1421 01:10:16,610 --> 01:10:19,850 AUDIENCE: Yes, I personally have experience with ICDs, 1422 01:10:19,850 --> 01:10:21,839 but no one in the room does. 1423 01:10:21,839 --> 01:10:24,380 OLIVIER DE WECK: OK, well, why don't you tell us about yours? 1424 01:10:24,380 --> 01:10:27,482 Because I know you worked on this gravity gradient sensor. 1425 01:10:27,482 --> 01:10:28,440 AUDIENCE: Yes, exactly. 1426 01:10:28,440 --> 01:10:30,065 OLIVIER DE WECK: So maybe that example. 1427 01:10:30,065 --> 01:10:31,800 Yeah, why don't you tell us about that? 1428 01:10:31,800 --> 01:10:33,850 AUDIENCE: So each system had its own ICD. 1429 01:10:33,850 --> 01:10:37,340 And that was primarily because the rest of the team 1430 01:10:37,340 --> 01:10:39,410 needed to understand the inputs and outputs 1431 01:10:39,410 --> 01:10:41,837 and how the specific system works. 1432 01:10:41,837 --> 01:10:43,670 So it not only works for system integration. 1433 01:10:43,670 --> 01:10:46,190 It just works for the team as a whole 1434 01:10:46,190 --> 01:10:47,750 in its understanding of the project 1435 01:10:47,750 --> 01:10:49,095 and the particular systems. 1436 01:10:51,730 --> 01:10:53,410 OLIVIER DE WECK: OK, good. 1437 01:10:53,410 --> 01:10:57,820 All right, so this is a really, really big deal in practice. 1438 01:10:57,820 --> 01:11:01,350 If you do something that's-- 1439 01:11:01,350 --> 01:11:03,790 so for example, if you make a design, 1440 01:11:03,790 --> 01:11:06,085 you modify the interface because you think, ah, 1441 01:11:06,085 --> 01:11:06,910 it's better for me. 1442 01:11:06,910 --> 01:11:09,710 I'm going to change this. 1443 01:11:09,710 --> 01:11:11,830 And it's different from the ICD. 1444 01:11:11,830 --> 01:11:14,410 And then there's a failure in your system. 1445 01:11:14,410 --> 01:11:17,110 And then you say you're going to blame the supplier and say, 1446 01:11:17,110 --> 01:11:18,760 your box failed. 1447 01:11:18,760 --> 01:11:21,730 My mission got screwed up because of your box. 1448 01:11:21,730 --> 01:11:24,640 And then it turns out, ah, you actually 1449 01:11:24,640 --> 01:11:30,430 implemented an interface that's in conflict with the ICD. 1450 01:11:30,430 --> 01:11:35,410 The supplier will step back and say, that's not my problem. 1451 01:11:35,410 --> 01:11:37,780 You didn't follow the interface definition, 1452 01:11:37,780 --> 01:11:40,810 and just changing one pin or one polarity 1453 01:11:40,810 --> 01:11:43,330 can lead to a system failure. 1454 01:11:43,330 --> 01:11:46,120 So really, you have to pay a lot of attention 1455 01:11:46,120 --> 01:11:48,170 to the details here. 1456 01:11:48,170 --> 01:11:50,780 OK, let's talk about system integration. 1457 01:11:50,780 --> 01:11:51,950 So what is it? 1458 01:11:51,950 --> 01:11:54,850 It's essentially the process of deliberately-- 1459 01:11:54,850 --> 01:11:56,920 so let's say you've designed your system. 1460 01:11:56,920 --> 01:11:59,430 The interfaces are clear. 1461 01:11:59,430 --> 01:12:01,860 When you physically put together your system, 1462 01:12:01,860 --> 01:12:04,979 you have to integrate the system. 1463 01:12:04,979 --> 01:12:06,270 So that means different things. 1464 01:12:06,270 --> 01:12:08,730 Physical assembly of the parts, that's 1465 01:12:08,730 --> 01:12:11,790 the liaison diagram, what fits into what. 1466 01:12:11,790 --> 01:12:14,310 And we'll talk about sequencing in a minute. 1467 01:12:14,310 --> 01:12:18,510 You need to connect your different conduits and hoses. 1468 01:12:18,510 --> 01:12:21,180 You need to fill in various consumables-- 1469 01:12:21,180 --> 01:12:24,900 fuel, lubricants. 1470 01:12:24,900 --> 01:12:27,870 An example in space flight would be 1471 01:12:27,870 --> 01:12:31,140 if you're going to use a cryocooler, 1472 01:12:31,140 --> 01:12:33,540 you might have to use some gases or liquids 1473 01:12:33,540 --> 01:12:36,570 for the cyrocooler that are consumables, 1474 01:12:36,570 --> 01:12:38,100 finite consumables. 1475 01:12:38,100 --> 01:12:41,100 Connecting all your electronics to the power sources, 1476 01:12:41,100 --> 01:12:42,480 the avionics. 1477 01:12:42,480 --> 01:12:45,060 Wire harnesses are still used a lot. 1478 01:12:45,060 --> 01:12:47,430 We would love to get rid of wire harnesses, right? 1479 01:12:47,430 --> 01:12:48,040 They're heavy. 1480 01:12:48,040 --> 01:12:48,831 They're cumbersome. 1481 01:12:48,831 --> 01:12:50,570 They're expensive. 1482 01:12:50,570 --> 01:12:54,900 And we've been trying to replace wire harnesses with wireless, 1483 01:12:54,900 --> 01:12:57,630 but it's really hard to make it work reliably. 1484 01:12:57,630 --> 01:13:01,000 So we still use wire harnesses quite a bit. 1485 01:13:01,000 --> 01:13:02,730 And then you need to upload and test 1486 01:13:02,730 --> 01:13:07,690 your operational software that sits inside your system. 1487 01:13:07,690 --> 01:13:10,950 So that's a lot of work, and it has to be done very carefully. 1488 01:13:10,950 --> 01:13:13,590 A lot of this is done in facilities 1489 01:13:13,590 --> 01:13:16,140 that are very clean, clean rooms, 1490 01:13:16,140 --> 01:13:19,000 because you really need to do every step properly. 1491 01:13:19,000 --> 01:13:20,650 You need to document every step. 1492 01:13:20,650 --> 01:13:21,735 There's quality control. 1493 01:13:24,250 --> 01:13:28,080 And that's why it's expensive and slow. 1494 01:13:28,080 --> 01:13:30,240 So the next point here is the sequence 1495 01:13:30,240 --> 01:13:32,910 in which the integration occurs may be important. 1496 01:13:32,910 --> 01:13:34,470 And I want to-- 1497 01:13:34,470 --> 01:13:37,410 so the reading, the post-reading associated with this lecture, 1498 01:13:37,410 --> 01:13:43,440 is a paper by Ben-Asher, who's a professor at Technion in Israel 1499 01:13:43,440 --> 01:13:46,380 who I think has done some of the best work, theoretical work, 1500 01:13:46,380 --> 01:13:48,270 on system integration. 1501 01:13:48,270 --> 01:13:52,080 And unfortunately, in complex systems, a lot of problems 1502 01:13:52,080 --> 01:13:53,220 are errors. 1503 01:13:53,220 --> 01:13:56,220 We often discover them during system integration and testing. 1504 01:13:56,220 --> 01:14:00,180 So that's also a key point here. 1505 01:14:00,180 --> 01:14:02,320 OK, so let me-- 1506 01:14:02,320 --> 01:14:05,387 I hope you have a chance to read this Ben-Asher paper, 1507 01:14:05,387 --> 01:14:06,720 because it's pretty interesting. 1508 01:14:06,720 --> 01:14:10,780 But I'll give you sort of the summary here on one slide. 1509 01:14:10,780 --> 01:14:13,800 So what they looked at was, does it 1510 01:14:13,800 --> 01:14:17,610 matter whether you do a single-stage integration, 1511 01:14:17,610 --> 01:14:19,620 everything at once, or you do what's 1512 01:14:19,620 --> 01:14:21,930 called incremental integration? 1513 01:14:21,930 --> 01:14:23,820 So the idea of incremental integration 1514 01:14:23,820 --> 01:14:27,630 is you just integrate two components or subsystems 1515 01:14:27,630 --> 01:14:29,550 in isolation at a time. 1516 01:14:29,550 --> 01:14:33,840 You check them, and then you gradually integrate as opposed 1517 01:14:33,840 --> 01:14:35,640 to doing it all at once. 1518 01:14:35,640 --> 01:14:40,830 And the example in the paper is UAV, Unmanned Aerial Vehicle. 1519 01:14:40,830 --> 01:14:44,130 And there's not a lot of detail, but there's 1520 01:14:44,130 --> 01:14:46,650 four major subsystems-- the engine, 1521 01:14:46,650 --> 01:14:50,430 the flight control system, the optronic payload, 1522 01:14:50,430 --> 01:14:53,010 and then the payload control system. 1523 01:14:53,010 --> 01:14:58,250 And so in one case, you would do a single-stage integration. 1524 01:14:58,250 --> 01:15:00,930 So here's your four subsystems, and you just 1525 01:15:00,930 --> 01:15:02,871 integrate them all at once. 1526 01:15:02,871 --> 01:15:04,620 And that's shown here in this upper graph. 1527 01:15:04,620 --> 01:15:08,920 So this is not the graph of components. 1528 01:15:08,920 --> 01:15:11,820 This is the graph of integration steps. 1529 01:15:11,820 --> 01:15:13,770 And then you have to put the system, 1530 01:15:13,770 --> 01:15:16,770 once it's integrated, these step seven, eight, nine, 1531 01:15:16,770 --> 01:15:18,420 through different tests. 1532 01:15:18,420 --> 01:15:20,910 And then the idea is that the amount 1533 01:15:20,910 --> 01:15:24,450 of time that it takes to do each of these integration steps 1534 01:15:24,450 --> 01:15:26,190 is somewhat stochastic. 1535 01:15:26,190 --> 01:15:28,890 It's not a deterministic number. 1536 01:15:28,890 --> 01:15:31,140 And so this upper graph here that 1537 01:15:31,140 --> 01:15:37,410 looks kind of like a ski jump, this is time, number of days, 1538 01:15:37,410 --> 01:15:40,320 versus the probability. 1539 01:15:40,320 --> 01:15:44,310 So what you see here is that it's right-skewed-- 1540 01:15:44,310 --> 01:15:47,190 or left-skewed, depending on how you define it. 1541 01:15:47,190 --> 01:15:50,280 But the peak here is to the right. 1542 01:15:50,280 --> 01:15:55,410 And so the idea is that if everything goes super smoothly, 1543 01:15:55,410 --> 01:15:57,720 actually, single-stage integration 1544 01:15:57,720 --> 01:15:59,430 could work very well. 1545 01:15:59,430 --> 01:16:02,760 But the likelihood of that happening is not high. 1546 01:16:02,760 --> 01:16:04,950 And if you think about it, if you do all at once, 1547 01:16:04,950 --> 01:16:07,290 single-stage integration, troubleshooting 1548 01:16:07,290 --> 01:16:09,760 becomes more complex as well. 1549 01:16:09,760 --> 01:16:14,700 And so that's why this curve is actually shifted to the right. 1550 01:16:14,700 --> 01:16:17,290 Incremental integration, in this case, 1551 01:16:17,290 --> 01:16:21,540 you're just integrating two and two, two and two, and then 1552 01:16:21,540 --> 01:16:24,461 you integrate your pre-integrated subsystems 1553 01:16:24,461 --> 01:16:24,960 together. 1554 01:16:24,960 --> 01:16:28,170 So it looks more gradual. 1555 01:16:28,170 --> 01:16:29,610 And then here is-- 1556 01:16:29,610 --> 01:16:31,890 again, using their model. 1557 01:16:31,890 --> 01:16:34,680 They have made certain assumptions about this-- 1558 01:16:34,680 --> 01:16:39,310 you get a somewhat broader distribution of outcomes. 1559 01:16:39,310 --> 01:16:42,930 But the mean, the expected amount of time for integration, 1560 01:16:42,930 --> 01:16:45,570 is actually less. 1561 01:16:45,570 --> 01:16:48,540 So you have a faster integration time 1562 01:16:48,540 --> 01:16:52,090 on average with incremental approach. 1563 01:16:52,090 --> 01:16:57,540 But in the worst case, you may be a little bit longer 1564 01:16:57,540 --> 01:16:58,970 than the single-stage integration. 1565 01:16:58,970 --> 01:17:02,460 But what really matters is, what's the average? 1566 01:17:02,460 --> 01:17:05,690 So the conclusion in the paper is that the integration 1567 01:17:05,690 --> 01:17:09,230 sequence is important, that the integration sequence can 1568 01:17:09,230 --> 01:17:12,140 be optimized, that there's advantages 1569 01:17:12,140 --> 01:17:14,940 of doing incremental integration, 1570 01:17:14,940 --> 01:17:17,850 primarily from a risk perspective. 1571 01:17:17,850 --> 01:17:23,900 So again, Octanus 1 project, maybe you guys 1572 01:17:23,900 --> 01:17:25,580 haven't thought about this too much yet. 1573 01:17:25,580 --> 01:17:27,390 Maybe you have. 1574 01:17:27,390 --> 01:17:31,130 But when you do your final vehicle, 1575 01:17:31,130 --> 01:17:34,250 the sequence in which everything gets integrated and tested 1576 01:17:34,250 --> 01:17:35,970 is important. 1577 01:17:35,970 --> 01:17:37,550 That's really what this says. 1578 01:17:37,550 --> 01:17:43,690 And an incremental approach can be very helpful. 1579 01:17:43,690 --> 01:17:45,530 The more components you have, the more 1580 01:17:45,530 --> 01:17:47,790 the combinatorial space of integration 1581 01:17:47,790 --> 01:17:49,720 sequences gets bigger and bigger. 1582 01:17:49,720 --> 01:17:53,730 So you have to do this at some level of abstraction. 1583 01:17:53,730 --> 01:17:54,860 Is that clear? 1584 01:17:54,860 --> 01:18:03,434 AUDIENCE: [INAUDIBLE] 1585 01:18:03,434 --> 01:18:04,350 OLIVIER DE WECK: Yeah. 1586 01:18:04,350 --> 01:18:05,766 And so the way you describe-- this 1587 01:18:05,766 --> 01:18:07,290 is driven by your supply chain. 1588 01:18:07,290 --> 01:18:09,000 When are things becoming available? 1589 01:18:09,000 --> 01:18:13,110 But you have to be careful because it could 1590 01:18:13,110 --> 01:18:15,460 be sort of risk-driven as well. 1591 01:18:15,460 --> 01:18:18,540 So OK, good. 1592 01:18:18,540 --> 01:18:20,520 Any questions at MIT? 1593 01:18:20,520 --> 01:18:21,770 Any comments about this? 1594 01:18:26,980 --> 01:18:28,675 AUDIENCE: I have a question. 1595 01:18:28,675 --> 01:18:29,550 OLIVIER DE WECK: Yup. 1596 01:18:29,550 --> 01:18:30,925 AUDIENCE: So I guess the takeaway 1597 01:18:30,925 --> 01:18:34,740 is that incremental integration is usually preferred 1598 01:18:34,740 --> 01:18:39,150 for most things, on average? 1599 01:18:39,150 --> 01:18:40,650 OLIVIER DE WECK: Yeah, I'm not sure. 1600 01:18:40,650 --> 01:18:43,140 Read the paper. 1601 01:18:43,140 --> 01:18:46,950 I think the conclusion is that for new systems 1602 01:18:46,950 --> 01:18:49,100 that you've never integrated before, 1603 01:18:49,100 --> 01:18:55,050 where your level of knowledge and the TRL levels-- or you've 1604 01:18:55,050 --> 01:18:58,380 never done this before, first of a kind of integration, 1605 01:18:58,380 --> 01:19:01,530 that incremental is better because if there's a problem, 1606 01:19:01,530 --> 01:19:04,230 then troubleshooting is easier. 1607 01:19:04,230 --> 01:19:07,540 Now, if you're integrating a very well known system-- 1608 01:19:07,540 --> 01:19:12,510 so it's sort of the 10th system, but you've sort of figured out 1609 01:19:12,510 --> 01:19:14,700 where the bugs are-- 1610 01:19:14,700 --> 01:19:18,410 so the risk level, the uncertainty is low. 1611 01:19:18,410 --> 01:19:21,030 Maybe then, single-stage integration is OK. 1612 01:19:21,030 --> 01:19:25,000 So it really depends on the novelty and risk. 1613 01:19:25,000 --> 01:19:27,250 So I don't think the conclusion of the paper 1614 01:19:27,250 --> 01:19:31,960 is single-stage integration is bad, always. 1615 01:19:31,960 --> 01:19:36,550 I think it's that you should think carefully 1616 01:19:36,550 --> 01:19:39,510 about the sequencing, and it's driven by the amount of risk. 1617 01:19:42,520 --> 01:19:44,480 AUDIENCE: OK, thanks. 1618 01:19:44,480 --> 01:19:48,040 OLIVIER DE WECK: All right, good. 1619 01:19:48,040 --> 01:19:50,090 So we're almost at the end here. 1620 01:19:50,090 --> 01:19:52,580 I want to ask you a quick question. 1621 01:19:52,580 --> 01:19:55,340 This is our concept question for today. 1622 01:19:55,340 --> 01:19:58,850 And this is about interface standards. 1623 01:19:58,850 --> 01:20:04,170 So over time in different industries, 1624 01:20:04,170 --> 01:20:06,350 rather than reinventing the wheel 1625 01:20:06,350 --> 01:20:09,350 and coming up with custom interfaces for every project 1626 01:20:09,350 --> 01:20:11,930 you do, you say, do the interface 1627 01:20:11,930 --> 01:20:15,090 according to this standard or that standard. 1628 01:20:15,090 --> 01:20:17,720 So I'm just curious to know from you guys-- 1629 01:20:17,720 --> 01:20:21,380 and I asked this question last week from another class 1630 01:20:21,380 --> 01:20:22,460 I teach at MIT. 1631 01:20:22,460 --> 01:20:26,810 So please answer this, and then I'm going to show your answers 1632 01:20:26,810 --> 01:20:29,900 and the answers of the other class combined. 1633 01:20:29,900 --> 01:20:34,270 So the interface definitions here are 802.11, 1634 01:20:34,270 --> 01:20:42,230 [? MIL-STD-1553, ?] RS-232, BB_aJ23100. 1635 01:20:42,230 --> 01:20:45,290 And then if you know of any other interface standards 1636 01:20:45,290 --> 01:20:47,690 that you think are important, just click Other. 1637 01:20:47,690 --> 01:20:49,100 This is a multi-- 1638 01:20:49,100 --> 01:20:51,440 there's not just one right answer here. 1639 01:20:51,440 --> 01:20:55,510 This is a multi-checkbox question. 1640 01:20:55,510 --> 01:21:00,890 All right, so here's the responses. 1641 01:21:00,890 --> 01:21:04,250 So 56%-- so this is your results, 1642 01:21:04,250 --> 01:21:08,420 and then my class from last week, kind of mixed together. 1643 01:21:08,420 --> 01:21:11,750 So 56% know the 802.11. 1644 01:21:11,750 --> 01:21:12,770 So what is-- 1645 01:21:12,770 --> 01:21:17,510 I thought it might be more than 56%, but what is 802.11? 1646 01:21:17,510 --> 01:21:18,390 AUDIENCE: Wi-Fi. 1647 01:21:18,390 --> 01:21:20,389 OLIVIER DE WECK: It's the Wi-Fi standard, right? 1648 01:21:20,389 --> 01:21:23,630 And now, G I think is the latest, I think. 1649 01:21:23,630 --> 01:21:26,830 It's sort of ABC, 802.11 ABC. 1650 01:21:26,830 --> 01:21:30,850 So it's kind of evolving standard-- 1651 01:21:30,850 --> 01:21:33,030 very important, obviously. 1652 01:21:33,030 --> 01:21:39,940 [? MIL-STD-1553, ?] 24% know about that one. 1653 01:21:39,940 --> 01:21:44,770 Anybody here, any [INAUDIBLE],, [? MIL-STD-1553? ?] So 1654 01:21:44,770 --> 01:21:47,270 you've done rural engineering, I can tell. 1655 01:21:47,270 --> 01:21:49,310 So what is the 1553? 1656 01:21:49,310 --> 01:21:50,720 AUDIENCE: It's [INAUDIBLE]. 1657 01:21:56,046 --> 01:21:57,420 OLIVIER DE WECK: Yes, it's a DoD. 1658 01:21:57,420 --> 01:21:59,700 So that's what's a military standard. 1659 01:21:59,700 --> 01:22:02,250 It's essentially a data bus. 1660 01:22:02,250 --> 01:22:07,590 So for example, on military planes, on many satellites, 1661 01:22:07,590 --> 01:22:10,290 it's essentially the architecture of the avionics. 1662 01:22:10,290 --> 01:22:11,790 It's a bus architecture. 1663 01:22:11,790 --> 01:22:15,990 So you typically have like a mission computer that's 1664 01:22:15,990 --> 01:22:19,335 the brain, and then it directs how the messages 1665 01:22:19,335 --> 01:22:21,660 are sent from the different-- 1666 01:22:21,660 --> 01:22:25,440 like the flight control system or the payload. 1667 01:22:25,440 --> 01:22:28,890 It's kind of an old standard, but it's a very robust standard 1668 01:22:28,890 --> 01:22:31,290 for avionics architectures. 1669 01:22:33,950 --> 01:22:35,580 OK, at MIT, RS-232? 1670 01:22:38,180 --> 01:22:40,110 This is also kind of an older standard. 1671 01:22:40,110 --> 01:22:41,460 Anybody chose that? 1672 01:22:46,160 --> 01:22:46,910 AUDIENCE: It's a-- 1673 01:22:46,910 --> 01:22:48,360 OLIVIER DE WECK: Did anybody-- 1674 01:22:48,360 --> 01:22:49,200 go ahead. 1675 01:22:49,200 --> 01:22:52,820 AUDIENCE: --communication protocol for electronics. 1676 01:22:52,820 --> 01:22:55,760 OLIVIER DE WECK: Right, but it's a very specific one. 1677 01:22:55,760 --> 01:22:57,950 There's a keyword I'm looking for. 1678 01:22:57,950 --> 01:22:59,519 AUDIENCE: Serial. 1679 01:22:59,519 --> 01:23:00,810 OLIVIER DE WECK: Serial, right. 1680 01:23:00,810 --> 01:23:02,720 It's a serial interface, as opposed 1681 01:23:02,720 --> 01:23:04,340 to a parallel interface. 1682 01:23:04,340 --> 01:23:08,030 So it's a serial interface. 1683 01:23:08,030 --> 01:23:09,920 And still, it's kind of old-fashioned, 1684 01:23:09,920 --> 01:23:11,800 but it's still used. 1685 01:23:11,800 --> 01:23:17,360 All right, now comes BB_aJ23100. 1686 01:23:17,360 --> 01:23:22,940 And nobody, out of 119, knew what that is. 1687 01:23:22,940 --> 01:23:25,730 So-- and this is kind of a trick question-- 1688 01:23:25,730 --> 01:23:29,760 this is a BioBricks interface. 1689 01:23:29,760 --> 01:23:33,870 So are you familiar with synthetic biology? 1690 01:23:33,870 --> 01:23:36,320 So synthetic biology, the basic idea 1691 01:23:36,320 --> 01:23:41,390 is that you can do what we do here for biological systems. 1692 01:23:41,390 --> 01:23:45,900 So there's libraries of parts, standard library 1693 01:23:45,900 --> 01:23:50,060 of biological parts, like different proteins, 1694 01:23:50,060 --> 01:23:53,580 that you can actually assemble organisms. 1695 01:23:53,580 --> 01:23:55,325 There's an annual competition. 1696 01:23:55,325 --> 01:23:57,800 It's been going on for almost a decade now. 1697 01:23:57,800 --> 01:23:59,210 It's kind of scary, actually. 1698 01:23:59,210 --> 01:24:03,650 But so far, nothing really bad has happened, I guess. 1699 01:24:03,650 --> 01:24:08,930 But the idea is that in biology, it's going this way. 1700 01:24:08,930 --> 01:24:11,990 So you can look this up. 1701 01:24:11,990 --> 01:24:14,930 You can google for this particular reference, 1702 01:24:14,930 --> 01:24:19,280 and it'll guide you to essentially a protein 1703 01:24:19,280 --> 01:24:20,060 interface. 1704 01:24:20,060 --> 01:24:23,900 It's essentially a molecular, biological interface. 1705 01:24:23,900 --> 01:24:30,800 So the prediction here is that 10, 20, 30 years from now, 1706 01:24:30,800 --> 01:24:33,240 people will know what this is. 1707 01:24:33,240 --> 01:24:36,710 And there will be a very internationally recognized set 1708 01:24:36,710 --> 01:24:43,100 of biological interfaces that people actually use and know. 1709 01:24:43,100 --> 01:24:45,950 And then anybody chose Other here? 1710 01:24:48,470 --> 01:24:49,510 Anybody choose other? 1711 01:24:49,510 --> 01:24:50,450 Yeah, you did? 1712 01:24:53,086 --> 01:24:56,690 AUDIENCE: Like standard USB and HDMI? 1713 01:24:56,690 --> 01:24:58,900 OLIVIER DE WECK: And USB has evolved as well, right? 1714 01:24:58,900 --> 01:25:01,480 We had USB 1.0, and then USB 2.0. 1715 01:25:01,480 --> 01:25:03,600 Isn't that the latest still? 1716 01:25:03,600 --> 01:25:04,712 AUDIENCE: USB-C. 1717 01:25:04,712 --> 01:25:05,670 OLIVIER DE WECK: C, OK. 1718 01:25:05,670 --> 01:25:08,060 AUDIENCE: It's reversible. 1719 01:25:08,060 --> 01:25:10,070 OLIVIER DE WECK: Yeah, so standards 1720 01:25:10,070 --> 01:25:11,917 evolve as well, right, as the technology 1721 01:25:11,917 --> 01:25:13,250 behind the standards-- you know. 1722 01:25:13,250 --> 01:25:15,740 So this is a huge deal in practice. 1723 01:25:15,740 --> 01:25:20,768 Anybody at MIT chose Other, other examples? 1724 01:25:20,768 --> 01:25:23,120 AUDIENCE: Yeah, I chose Other. 1725 01:25:23,120 --> 01:25:28,880 Things that come to mind are like SPI and I2C, 1726 01:25:28,880 --> 01:25:32,330 more like serial electronic communication protocols 1727 01:25:32,330 --> 01:25:35,580 that are useful for talking to chips. 1728 01:25:35,580 --> 01:25:38,660 OLIVIER DE WECK: OK, good. 1729 01:25:38,660 --> 01:25:41,750 So those are sort of inside the IC world, right? 1730 01:25:46,970 --> 01:25:48,980 OK, excellent. 1731 01:25:48,980 --> 01:25:52,400 So all right, well, we are actually 1732 01:25:52,400 --> 01:25:54,390 at the end of today's lecture. 1733 01:25:54,390 --> 01:25:57,320 So I just want to summarize the key points. 1734 01:25:57,320 --> 01:25:59,880 Why is interface management important? 1735 01:25:59,880 --> 01:26:03,650 There are many reasons, but I think the two top reasons are-- 1736 01:26:03,650 --> 01:26:07,320 think of interfaces as a potential Achilles heel. 1737 01:26:07,320 --> 01:26:10,490 They're potentially seeds for failures. 1738 01:26:10,490 --> 01:26:12,170 We saw bottlenecks. 1739 01:26:12,170 --> 01:26:14,150 We saw structural failures. 1740 01:26:14,150 --> 01:26:17,360 We saw the wrong kind of information 1741 01:26:17,360 --> 01:26:19,730 being passed through software. 1742 01:26:19,730 --> 01:26:22,560 And so be careful. 1743 01:26:22,560 --> 01:26:26,240 And if there's nobody in charge in your project 1744 01:26:26,240 --> 01:26:30,150 to really look after the interfaces, that's a problem. 1745 01:26:30,150 --> 01:26:34,890 So you want to really think about this 1746 01:26:34,890 --> 01:26:38,760 because almost any project today, you're decomposing it, 1747 01:26:38,760 --> 01:26:41,640 and you're not doing everything in-house yourself. 1748 01:26:41,640 --> 01:26:43,320 You're going to deal with suppliers. 1749 01:26:43,320 --> 01:26:45,210 You're going to deal with partners. 1750 01:26:45,210 --> 01:26:49,200 So of course you have an organizational, contractual 1751 01:26:49,200 --> 01:26:52,650 interface with them, but there's a physical interface as well. 1752 01:26:52,650 --> 01:26:55,530 The longer you wait to define that interface 1753 01:26:55,530 --> 01:26:59,190 with your partners and suppliers, the fuzzier it is, 1754 01:26:59,190 --> 01:27:01,350 the more problems you will have. 1755 01:27:01,350 --> 01:27:04,800 So critical to really define the interface, 1756 01:27:04,800 --> 01:27:10,200 and then stick with the interface very sharply. 1757 01:27:10,200 --> 01:27:13,570 Interface management has different aspects to it. 1758 01:27:13,570 --> 01:27:15,510 So first of all, types of interfaces, 1759 01:27:15,510 --> 01:27:18,090 there are many, many dozens or hundreds 1760 01:27:18,090 --> 01:27:21,000 of types of interfaces, but they can all be boiled down 1761 01:27:21,000 --> 01:27:23,190 to the four canonical types-- 1762 01:27:23,190 --> 01:27:27,660 physical connection, mass, energy, and information flows. 1763 01:27:27,660 --> 01:27:31,500 I've introduced you to the DSM, the Design Structure Matrix 1764 01:27:31,500 --> 01:27:32,790 method. 1765 01:27:32,790 --> 01:27:37,110 And that's where you basically decompose your system, 1766 01:27:37,110 --> 01:27:40,380 and you show it as an n by n matrix. 1767 01:27:40,380 --> 01:27:42,600 And then you put all the interfaces 1768 01:27:42,600 --> 01:27:45,690 into that matrix that you know about, 1769 01:27:45,690 --> 01:27:48,840 not just the interfaces that you want to have, 1770 01:27:48,840 --> 01:27:51,070 but the interfaces that are actually there. 1771 01:27:51,070 --> 01:27:56,580 So if there's noise, or vibration, or EMI, 1772 01:27:56,580 --> 01:28:00,660 Electromagnetic Interference, or waste heat, 1773 01:28:00,660 --> 01:28:02,760 those are also important because you're 1774 01:28:02,760 --> 01:28:06,720 going to have to deal with them in some way. 1775 01:28:06,720 --> 01:28:09,030 Finally, ICDs. 1776 01:28:09,030 --> 01:28:12,480 So ICDS-- and we said there's IRDs, Interface Requirements 1777 01:28:12,480 --> 01:28:16,580 Documents, which are the requirements for the interface; 1778 01:28:16,580 --> 01:28:20,440 IDDs, one-sided; and then ICDs are the most important. 1779 01:28:20,440 --> 01:28:22,980 It's the two-sided interface definition. 1780 01:28:22,980 --> 01:28:27,570 That's the NASA approach, but [? ISA ?] uses that. 1781 01:28:27,570 --> 01:28:30,540 Many, many industries now use these interface control 1782 01:28:30,540 --> 01:28:31,890 documents. 1783 01:28:31,890 --> 01:28:34,710 Ideally, in the long-term, we want to get rid of them. 1784 01:28:34,710 --> 01:28:36,780 We don't want documents in system engineering. 1785 01:28:36,780 --> 01:28:39,390 We just want to have models that we share. 1786 01:28:39,390 --> 01:28:43,010 But it's still moving in that direction. 1787 01:28:43,010 --> 01:28:45,600 OK, and then finally, the point about system 1788 01:28:45,600 --> 01:28:48,690 integration-- do not just do it randomly. 1789 01:28:48,690 --> 01:28:52,740 Think about, how will you assemble the system physically? 1790 01:28:52,740 --> 01:28:55,320 How are you going to connect all your electronics, 1791 01:28:55,320 --> 01:28:57,000 the consumables? 1792 01:28:57,000 --> 01:28:59,460 How will you load your software? 1793 01:28:59,460 --> 01:29:00,600 How will you test it? 1794 01:29:00,600 --> 01:29:03,300 We'll talk about testing, verification and validation 1795 01:29:03,300 --> 01:29:07,050 next week, but the way in which you integrate, 1796 01:29:07,050 --> 01:29:10,750 the sequence of integration could be very important. 1797 01:29:10,750 --> 01:29:12,810 And it's a pretty active area of research, 1798 01:29:12,810 --> 01:29:14,760 so in sort of system engineering research, 1799 01:29:14,760 --> 01:29:20,040 is really understanding and modeling system integration 1800 01:29:20,040 --> 01:29:22,470 as something you can optimize. 1801 01:29:22,470 --> 01:29:24,750 And then the last point was industry standards. 1802 01:29:24,750 --> 01:29:27,180 Really understand what are the key standards 1803 01:29:27,180 --> 01:29:30,460 in your industry, what are the key interface standards. 1804 01:29:30,460 --> 01:29:33,600 And if you're going to deviate, if you're 1805 01:29:33,600 --> 01:29:37,410 going to choose something else than an industry standard, 1806 01:29:37,410 --> 01:29:39,330 you better know why you're doing it. 1807 01:29:39,330 --> 01:29:42,450 Because the minute you decide not to use an industry 1808 01:29:42,450 --> 01:29:45,750 standard, you've probably just inflated the cost 1809 01:29:45,750 --> 01:29:49,170 by a factor of 10 because now there's no longer 1810 01:29:49,170 --> 01:29:51,780 any off-the-shelf components. 1811 01:29:51,780 --> 01:29:53,800 You have to basically do everything yourself, 1812 01:29:53,800 --> 01:29:58,700 but it's possible that there's good reasons for doing that.