WEBVTT 1 00:00:00.080 --> 00:00:03.919 Okay, so let's unpack this. It is Thursday, February twenty sixth, 2 00:00:04.040 --> 00:00:08.119 twenty twenty six, and today we are going to look 3 00:00:08.160 --> 00:00:10.880 at something that is literally all around. 4 00:00:10.599 --> 00:00:12.519 You right now, well absolutely everywhere. 5 00:00:12.640 --> 00:00:14.960 Yeah, Like it is probably on your mist it's definitely 6 00:00:15.000 --> 00:00:17.920 in your pocket, it's in your car, and it might 7 00:00:17.960 --> 00:00:20.120 even be in your toaster. But you never really. 8 00:00:19.879 --> 00:00:22.879 See it, right, which is I mean, that's the mark 9 00:00:22.920 --> 00:00:25.760 of a good design, right, If it's working perfectly, it's 10 00:00:25.800 --> 00:00:28.760 completely invisible. It's kind of the dark matter of the 11 00:00:28.800 --> 00:00:29.359 digital world. 12 00:00:29.480 --> 00:00:32.359 The dark matter. I love that exactly. So we are 13 00:00:32.399 --> 00:00:36.200 doing a deep dive today into the invisible digital infrastructure 14 00:00:36.280 --> 00:00:39.039 that basically runs the modern world. We are talking about 15 00:00:39.039 --> 00:00:43.560 embedded systems and to guide us through this absolute maze 16 00:00:43.719 --> 00:00:47.039 of you know, wires and logic gates, we're doing a 17 00:00:47.039 --> 00:00:50.039 deep dive into the source material here, which is the 18 00:00:50.079 --> 00:00:53.840 book Architecting High Performance Embedded Systems by Jim Leaden. 19 00:00:53.960 --> 00:00:56.039 It's a fantastic resource. 20 00:00:55.759 --> 00:00:57.479 It really is. Now. I have to admit, when I 21 00:00:57.520 --> 00:00:59.840 first saw the title, I thought, well, I thought, okay, 22 00:00:59.840 --> 00:01:02.280 this is going to be a super dry technical manual. 23 00:01:02.359 --> 00:01:03.759 Sure, yeah, it sounds like one. 24 00:01:03.799 --> 00:01:06.640 Right, but it's actually a beast of a book. I 25 00:01:06.640 --> 00:01:10.439 mean it covers everything from the literal physics of sensors 26 00:01:10.920 --> 00:01:13.239 all the way up to computer chips that can basically 27 00:01:13.280 --> 00:01:14.560 rewire themselves on the fly. 28 00:01:14.920 --> 00:01:17.400 It is a comprehensive guide for sure. And what I 29 00:01:17.439 --> 00:01:21.040 really appreciate about Lenins approaching this book is that he 30 00:01:21.079 --> 00:01:24.719 doesn't just treat these systems as well as like small computers. 31 00:01:25.280 --> 00:01:29.359 He really emphasizes the immense complexity and the precision that's 32 00:01:29.400 --> 00:01:33.200 required to actually make them work. Because you know, unlike 33 00:01:33.200 --> 00:01:35.560 your laptop, which just sits in a nice air conditioned 34 00:01:35.640 --> 00:01:39.519 room or on your desk hopefully right, hopefully, these embedded 35 00:01:39.519 --> 00:01:41.280 systems they live out in the real world there in 36 00:01:41.319 --> 00:01:44.079 the mud, the heat, the freezing rain, and you know, 37 00:01:44.560 --> 00:01:46.879 literally bolted inside exploding car engines. 38 00:01:47.000 --> 00:01:49.239 Yeah, the environment is brutal. So let's start with the 39 00:01:49.239 --> 00:01:52.439 basics for you listening, what actually counts as an embedded system? 40 00:01:52.519 --> 00:01:54.640 Because I feel like that term gets thrown around a lot, 41 00:01:54.719 --> 00:01:57.000 like I have a laptop, I have a smartphone, I 42 00:01:57.040 --> 00:02:00.280 have a car. Are they all embedded system? 43 00:02:00.480 --> 00:02:03.159 It's a really great question to start with. The core 44 00:02:03.200 --> 00:02:07.640 definition Leaden gives is pretty specific. An embedded system is 45 00:02:07.719 --> 00:02:11.439 digital processing that is integrated into a device that has 46 00:02:11.439 --> 00:02:13.800 a larger purpose beyond just computing. 47 00:02:14.000 --> 00:02:17.120 Well, okay, unpack that a bit, a larger purpose beyond computing. 48 00:02:17.240 --> 00:02:20.639 Yeah, so think about your laptop. Its entire purpose is 49 00:02:20.680 --> 00:02:23.400 to compute. It is a general purpose machine. Right. You 50 00:02:23.400 --> 00:02:25.400 can use it to write a novel, you can edit 51 00:02:25.439 --> 00:02:28.759 a video, you can calculate a giant spreadsheet. It's flexible. 52 00:02:29.240 --> 00:02:32.719 But a car, a car's purpose is to move people 53 00:02:32.719 --> 00:02:35.520 from point A to point B right to drive exactly. 54 00:02:35.919 --> 00:02:39.680 The computers inside the car are just there to facilitate 55 00:02:39.719 --> 00:02:41.639 that physical movement. They are specialists. 56 00:02:41.919 --> 00:02:44.240 Okay, So the laptop is the jack of all trades 57 00:02:44.400 --> 00:02:46.439 and the embedded system is the master of one. 58 00:02:46.560 --> 00:02:49.960 Precisely, And Lettin uses a really great simple example to 59 00:02:50.039 --> 00:02:52.560 draw the line here. He calls it the toothbrush test. 60 00:02:52.680 --> 00:02:55.080 The toothbrush test. Okay, I love this. Break it down. 61 00:02:55.199 --> 00:02:58.800 So take a standard electric toothbrush from say, maybe twenty 62 00:02:58.879 --> 00:03:01.919 years ago. Bat a motor and the little physical switch 63 00:03:02.240 --> 00:03:04.879 you turn it on, the circuit closes, the motor spins, 64 00:03:05.159 --> 00:03:08.000 and you brush your teeth. Is that an embedded system. 65 00:03:08.039 --> 00:03:12.280 I'm gonna guess. No, I mean it's just it's electric, right, 66 00:03:12.360 --> 00:03:13.960 It's just mechanics and a battery. 67 00:03:14.120 --> 00:03:17.759 Right. There is no digital processing happening. It's just dumb logic. 68 00:03:18.039 --> 00:03:18.280 Right. 69 00:03:18.439 --> 00:03:22.680 But take a modern smart toothbrush, the kind that lights 70 00:03:22.759 --> 00:03:25.639 up red if you press too hard against your gums. 71 00:03:25.680 --> 00:03:27.840 Oh yeah, I have one of those. It actively gets 72 00:03:27.840 --> 00:03:28.479 angry at. 73 00:03:28.400 --> 00:03:32.199 Me exactly, and that anger requires intelligence. It has a 74 00:03:32.199 --> 00:03:36.639 physical sensor to measure pressure. That sensor feeds analog data 75 00:03:36.680 --> 00:03:39.360 to a tiny micro controller okay, and then the micro 76 00:03:39.400 --> 00:03:43.479 controller processes that data and mathematically decides, hey, that is 77 00:03:43.520 --> 00:03:45.479 too much force. I need to turn on the red 78 00:03:45.599 --> 00:03:47.120 led and slow down the motor. 79 00:03:48.159 --> 00:03:51.080 That whole loop, that is an embedded system. It is 80 00:03:51.120 --> 00:03:54.400 a layer of digital thinking tucked inside an everyday object. 81 00:03:54.520 --> 00:03:57.240 Wow. And unlike my laptop, which, let's be honest, I 82 00:03:57.319 --> 00:03:58.599 have to reboot like once a week because it just 83 00:03:58.599 --> 00:04:02.080 gets confused. Yeah, things have to be incredibly robust. 84 00:04:02.560 --> 00:04:05.599 That is the other massive key distinction here. These systems 85 00:04:05.639 --> 00:04:08.400 don't usually have nice cooling fans. They have to survive 86 00:04:08.520 --> 00:04:12.319 intense vibration, dust, wild temperature swings. 87 00:04:12.000 --> 00:04:15.000 Because they're in the real world. Exactly. Think about it. 88 00:04:15.039 --> 00:04:19.279 If your laptop crashes, you lose an unsaved word document annoying, 89 00:04:19.439 --> 00:04:22.079 but fine. If the embedded system in your car is 90 00:04:22.160 --> 00:04:25.720 anti lock breaking, unit crashes, oh man. The consequences are 91 00:04:25.720 --> 00:04:27.639 physical and they are immediate. 92 00:04:27.759 --> 00:04:31.120 That makes total sense. And this naturally bleeds into I 93 00:04:31.199 --> 00:04:33.759 think the biggest buzzword of the last decade, which is 94 00:04:33.800 --> 00:04:36.720 the Internet of things or IoT because once you have 95 00:04:36.800 --> 00:04:40.680 that smart toothbrush, the next logical step for a company 96 00:04:40.759 --> 00:04:42.639 is to make it talk to your phone. Right. 97 00:04:42.759 --> 00:04:45.600 That is the natural evolution, and let draws a very 98 00:04:45.639 --> 00:04:48.759 clear line there in the source material. Embedded systems evolve 99 00:04:48.800 --> 00:04:51.160 into the Internet of things when they start communicating with 100 00:04:51.240 --> 00:04:54.800 central nodes. It's all about massive network communication. 101 00:04:55.120 --> 00:04:57.519 The book mentions some examples that really show the sheer 102 00:04:57.519 --> 00:05:00.879 scale of this. On one hand, you have everyday stuff 103 00:05:00.959 --> 00:05:04.639 like smart speakers, your Amazon Echo, Google Nest sitting in 104 00:05:04.680 --> 00:05:06.759 your living room just waiting for a wake word. 105 00:05:06.720 --> 00:05:10.879 Right, an admitted system that's listening, processing audio locally, and 106 00:05:10.920 --> 00:05:12.199 then communicating with the cloud. 107 00:05:12.439 --> 00:05:15.879 But then you connect that same concept to the industrial side. 108 00:05:16.079 --> 00:05:19.800 Think about an oil pipeline. You have embedded systems monitoring 109 00:05:19.879 --> 00:05:24.959 flow rates, pressure and temperature over thousands of miles of 110 00:05:25.040 --> 00:05:26.040 pipe exactly. 111 00:05:26.120 --> 00:05:28.079 So instead of a guy in a truck having to 112 00:05:28.160 --> 00:05:29.920 drive out into the middle of nowhere to check a 113 00:05:29.959 --> 00:05:30.800 physical gauge. 114 00:05:30.879 --> 00:05:33.759 The pope itself basically sends a text message saying, Hey, 115 00:05:33.839 --> 00:05:34.399 I'm leaking. 116 00:05:34.720 --> 00:05:39.079 Yes. Yeah, it's incredibly efficient, but this is a big butt. 117 00:05:39.800 --> 00:05:43.319 The book issues a pretty stark warning here regarding security. 118 00:05:43.399 --> 00:05:45.680 Oh absolutely, because when you take a device that was 119 00:05:45.720 --> 00:05:49.360 originally designed to be a simple, isolated specialist, like a 120 00:05:49.360 --> 00:05:52.560 home thermostat or a pipeline valve, and you suddenly give 121 00:05:52.560 --> 00:05:55.279 it a live Internet connection, you are opening a door. 122 00:05:55.560 --> 00:05:58.519 And historically these things aren't exactly built with Fort Knox 123 00:05:58.600 --> 00:05:59.839 level security, are they No. 124 00:06:00.199 --> 00:06:03.399 Historically they are not. And that's exactly the danger. If 125 00:06:03.439 --> 00:06:07.120 a malicious actor hacks your personal laptop, they might steal 126 00:06:07.240 --> 00:06:11.519 your credit card info. That is bad, obviously, but it's manageable. 127 00:06:11.680 --> 00:06:15.000 But if they hack an IoT system like a medical 128 00:06:15.000 --> 00:06:18.199 pacemaker or that oil pipeline we mentioned, or the steering 129 00:06:18.240 --> 00:06:21.240 control of a connected car, we aren't talking about data 130 00:06:21.279 --> 00:06:25.879 loss anymore. We are talking about severe risks to physical safety. 131 00:06:25.920 --> 00:06:28.959 It's the difference between a privacy nightmare and a literal 132 00:06:29.000 --> 00:06:29.800 safety crisis. 133 00:06:29.879 --> 00:06:34.120 Yeah, exactly. Leaden emphasizes that in this new IoT world. 134 00:06:34.439 --> 00:06:37.560 Security cannot just be a software pass you add on later. 135 00:06:38.120 --> 00:06:40.680 It has be baked into the very architecture of the 136 00:06:40.720 --> 00:06:41.879 hardware from day one. 137 00:06:42.160 --> 00:06:45.519 It's terrifying but also fascinating. Okay, let's step back a 138 00:06:45.519 --> 00:06:47.879 bit and look at how these systems actually interact with 139 00:06:47.920 --> 00:06:50.079 the physical world. Because a computer chip, at the end 140 00:06:50.079 --> 00:06:52.839 of the day only knows ones and zeros. It doesn't 141 00:06:52.839 --> 00:06:55.439 know what hot or fast or loud actually feels like. 142 00:06:55.600 --> 00:06:57.879 It lives in this dark digital bubble. 143 00:06:58.160 --> 00:07:00.439 And this is the realm of sensors. This is how 144 00:07:00.480 --> 00:07:05.399 the system literally senses the outside world. Letting categorizes them 145 00:07:05.399 --> 00:07:07.519 into two main buckets in the book that are super 146 00:07:07.560 --> 00:07:10.360 helpful to understand, passive and active sensors. 147 00:07:10.480 --> 00:07:13.199 Okay, passive sounds a bit safer. What is a passive sensor? 148 00:07:13.439 --> 00:07:16.480 A passive sensor just sits there and measures the environment 149 00:07:16.560 --> 00:07:19.800 exactly as it is. It's observant. A classic example he 150 00:07:19.839 --> 00:07:22.399 gives as a thermister. A thermister, Yeah, it's a special 151 00:07:22.439 --> 00:07:25.120 type of resistor that changes its electrical resistance based on 152 00:07:25.120 --> 00:07:27.720 the ambient temperature. It's not doing anything to the air 153 00:07:27.759 --> 00:07:30.079 around it. It's just reacting to the heat. 154 00:07:30.279 --> 00:07:33.000 Okay, so kind of like a sponge soaking up water. 155 00:07:33.399 --> 00:07:36.399 It changes its state based on the environment, but it 156 00:07:36.480 --> 00:07:38.000 doesn't spray water back out. 157 00:07:38.160 --> 00:07:41.680 That's a great analogy. Now contrast that with an active sensor. 158 00:07:41.680 --> 00:07:44.360 An active sensor actually goes out and pokes the world 159 00:07:44.639 --> 00:07:45.519 to see what happens. 160 00:07:45.639 --> 00:07:47.560 It pokes the world. 161 00:07:47.199 --> 00:07:49.959 Well in a way. Yes, think about an ultrasonic sensor. 162 00:07:50.120 --> 00:07:51.920 These are used a lot in robotics and in the 163 00:07:51.959 --> 00:07:53.879 parking sensors on your car bumber. 164 00:07:53.600 --> 00:07:55.360 Oh, the ones that beep when you get too close 165 00:07:55.360 --> 00:07:55.879 to a wall. 166 00:07:56.079 --> 00:08:00.759 Exactly, that sensor actively sends out a ping, a high 167 00:08:00.839 --> 00:08:03.680 frequency sound wave. Yeah, and then it immediately switches into 168 00:08:03.759 --> 00:08:06.959 listening mode and waits to hear the echo bouncing back like. 169 00:08:06.920 --> 00:08:10.360 A bat using echolocation or submarine. 170 00:08:09.879 --> 00:08:13.040 Exactly like that. It uses a concept called time of flight. 171 00:08:13.639 --> 00:08:15.839 The system knows the exact speed of sound, so it 172 00:08:15.879 --> 00:08:19.000 sends the ping out, starts a tiny internal stopwatch. Here's 173 00:08:19.040 --> 00:08:21.040 the echo return, and stops the watch. 174 00:08:21.160 --> 00:08:22.800 Oh, and then it just does the math. 175 00:08:22.800 --> 00:08:25.439 Right, the time it took tells you exactly how far 176 00:08:25.439 --> 00:08:28.480 away the physical wall is. That is active sensing. You 177 00:08:28.560 --> 00:08:31.639 generate a stimulus and you measure the real world response, and. 178 00:08:31.560 --> 00:08:35.159 The source material mentions lightar and radar working on similar principles. 179 00:08:35.240 --> 00:08:39.159 Right, Yes, lightar is completely fascinating. Instead of sound waves, 180 00:08:39.240 --> 00:08:43.120 it uses pulses of laser light. By spinning that laser 181 00:08:43.159 --> 00:08:46.159 around and measuring millions of those time of flight returns 182 00:08:46.200 --> 00:08:49.519 every single second, an autonomous vehicle can build what's called 183 00:08:49.519 --> 00:08:52.440 a point clide a point cloud. Yeah, it's basically a 184 00:08:52.480 --> 00:08:55.519 real time three D map of the entire world around 185 00:08:55.519 --> 00:08:58.440 the car. The computer knows that this specific cluster of 186 00:08:58.519 --> 00:09:01.840 data points is a pedestrian walking, that cluster over there 187 00:09:01.960 --> 00:09:04.759 is a stationary tree, and that fast moving cluster is 188 00:09:04.799 --> 00:09:09.000 another car. It's literally seeing the world in three D geometry. 189 00:09:09.200 --> 00:09:11.840 That is just incredible. But here is where it gets 190 00:09:11.879 --> 00:09:15.600 really technical and honestly, I think really interesting. The real 191 00:09:15.639 --> 00:09:20.080 world is analog like. Temperature doesn't just jump from seventy 192 00:09:20.120 --> 00:09:23.080 degrees to seventy one degrees in a single instant step. 193 00:09:23.600 --> 00:09:28.039 It flows smoothly, it's continuous, right, But computers are digital. 194 00:09:28.320 --> 00:09:31.960 They only understand distinct steps. They're choppy. So how do 195 00:09:32.039 --> 00:09:35.240 we bridge that gap between a smooth world and a 196 00:09:35.320 --> 00:09:36.039 choppy computer. 197 00:09:36.399 --> 00:09:39.519 That crucial job falls to the ADC. The analog to 198 00:09:39.600 --> 00:09:43.759 digital converter. It acts as the ultimate translator. Imagine you 199 00:09:43.799 --> 00:09:46.799 are drawing a smooth, sweeping curve on a piece of 200 00:09:46.799 --> 00:09:47.399 graph paper. 201 00:09:47.440 --> 00:09:48.360 Okay, I'm picturing it. 202 00:09:48.519 --> 00:09:52.200 The real physical world is that smooth curve. The ADC 203 00:09:52.440 --> 00:09:55.240 is looking at that curve at very specific tiny points 204 00:09:55.240 --> 00:09:57.480 in time. This is called sampling it, and it's turning 205 00:09:57.519 --> 00:10:00.279 each specific point into a precise digital number. 206 00:10:00.399 --> 00:10:02.240 So if you don't sample it fast enough, you miss 207 00:10:02.279 --> 00:10:04.000 all the details of the current exactly. 208 00:10:04.039 --> 00:10:06.840 That's a problem called aliasing. If you blink too slowly, 209 00:10:06.879 --> 00:10:09.080 you might miss the baseball flying right past your face. 210 00:10:09.320 --> 00:10:12.000 The source material talks about the XCDC hardware on the 211 00:10:12.080 --> 00:10:13.360 RDA seven board. 212 00:10:13.240 --> 00:10:16.960 Right, the RDA seven, which is a specific FPGA development 213 00:10:17.000 --> 00:10:18.200 board they use as an example. 214 00:10:18.360 --> 00:10:22.399 Yes, that specific hardware can sample the analog world up 215 00:10:22.480 --> 00:10:24.480 to one million times per second. 216 00:10:24.879 --> 00:10:28.360 One million times a second. That is, I mean, I 217 00:10:28.360 --> 00:10:29.759 can't even wrap my head around that. 218 00:10:29.840 --> 00:10:32.600 And honestly, that's just considered standard for high performance control. 219 00:10:33.000 --> 00:10:36.200 Some advanced radio frequency systems sampled into the billions of 220 00:10:36.200 --> 00:10:39.600 times per second. It creates this absolutely massive fire hose 221 00:10:39.600 --> 00:10:42.039 of numbers that the processor then has to crunch. 222 00:10:41.759 --> 00:10:43.679 Okay, so we have the data. Now, the sensor has 223 00:10:43.679 --> 00:10:46.919 picked up the physical world, the ADC has translated it 224 00:10:46.960 --> 00:10:50.159 into digital numbers. Now, how does that data actually get 225 00:10:50.200 --> 00:10:52.200 to the main brain of the system. We need a 226 00:10:52.240 --> 00:10:53.000 nervous system. 227 00:10:53.240 --> 00:10:56.440 We absolutely do. And these are called the communication protocols. 228 00:10:56.840 --> 00:10:59.240 Depending on exactly what you're trying to do, you pick 229 00:10:59.240 --> 00:11:00.600 a different digital language. 230 00:11:00.720 --> 00:11:03.200 Let's run through the big ones mentioned in Letin's book. 231 00:11:03.360 --> 00:11:06.360 Start with the absolute simplest one GPIO. 232 00:11:06.679 --> 00:11:11.159 Okay, GPIO General purpose input output. This is basically the 233 00:11:11.159 --> 00:11:14.039 brute force method. It's usually just a single wire that 234 00:11:14.120 --> 00:11:17.919 is either on or off high voltage or low voltage. 235 00:11:17.600 --> 00:11:19.879 One or zero, kind of like a physical light switch. 236 00:11:20.039 --> 00:11:23.679 Perfect analogy is the machine's safety cover open, yes or no? 237 00:11:24.559 --> 00:11:28.200 Is the emergency button currently pressed? Yes or no? It's 238 00:11:28.240 --> 00:11:32.519 incredibly fast, it's instant, but it obviously doesn't carry complex data. 239 00:11:32.600 --> 00:11:35.279 You can't send a JPEG image over a light switch. 240 00:11:35.120 --> 00:11:37.519 Right, Okay, So what if we do need to send 241 00:11:37.639 --> 00:11:40.279 actual complex data. The book talks about a protocol called 242 00:11:40.360 --> 00:11:40.879 I two C. 243 00:11:41.840 --> 00:11:45.080 I two C or inter integrated Circuit. I always like 244 00:11:45.159 --> 00:11:46.639 to call this one the polite conversation. 245 00:11:46.799 --> 00:11:47.879 Polite Why polite? 246 00:11:47.960 --> 00:11:51.000 Because it uses just two wires. One wire is a 247 00:11:51.039 --> 00:11:53.879 clock to keep the timing synchronized and the other is 248 00:11:53.879 --> 00:11:56.840 for the actual data. But it can connect multiple different 249 00:11:56.879 --> 00:12:00.320 devices using a clever address system. It uses a master 250 00:12:00.360 --> 00:12:04.039 servant architecture. Okay, so the master controller says, hey, Sensor A, 251 00:12:04.240 --> 00:12:07.120 it's your turn to speak, and Sensor A talks. Then 252 00:12:07.159 --> 00:12:09.720 the master says, okay, thank you, Sensor B, now it's 253 00:12:09.720 --> 00:12:10.159 your turn. 254 00:12:10.559 --> 00:12:12.519 Ah, So they actually take turns on the wire. 255 00:12:12.639 --> 00:12:15.120 Yes, it's what we call half duplex. You can only 256 00:12:15.159 --> 00:12:18.120 send data or receive data at one specific time, not both. 257 00:12:18.679 --> 00:12:21.600 It's great for simple things like reading a temperature sensor 258 00:12:21.720 --> 00:12:23.600 once a second, but it's a bit slow. 259 00:12:23.840 --> 00:12:26.080 So what if I have a real need for speed, 260 00:12:26.200 --> 00:12:29.559 like if I'm streaming that crazy one million sample per 261 00:12:29.600 --> 00:12:31.559 second data from the ADC, then you. 262 00:12:31.559 --> 00:12:36.120 Need to upgrade to SPI Serial Peripheral Interface. This is 263 00:12:36.159 --> 00:12:39.200 the speedy connection. It uses four wires instead of two, 264 00:12:39.919 --> 00:12:43.639 and because it has completely separate dedicated wires for sending 265 00:12:43.720 --> 00:12:45.679 and for receiving, it is full. 266 00:12:45.559 --> 00:12:48.399 Duplex, meaning it can talk and listen at the exact 267 00:12:48.480 --> 00:12:51.679 same time exactly. Sounds like a chaotic argument where everyone 268 00:12:51.720 --> 00:12:52.919 is just yelling over each other. 269 00:12:53.080 --> 00:12:56.159 Actually, it's more like a highly efficient high speed telephone 270 00:12:56.159 --> 00:12:59.399 call where both people are talking and perfectly understanding each 271 00:12:59.399 --> 00:13:02.360 other simultananeously. It can run much much faster than it 272 00:13:02.480 --> 00:13:05.639 two see easily up to fifty mega herds for more wow. 273 00:13:05.720 --> 00:13:09.039 Okay, but what about inside cars? Because I know modern 274 00:13:09.080 --> 00:13:12.360 cars are an absolute nightmare of electrical noise. You have 275 00:13:12.480 --> 00:13:15.799 spark plugs firing thousands of times a minute, giant alternator 276 00:13:15.879 --> 00:13:20.159 spinning that creates massive electromagnetic interference. 277 00:13:19.840 --> 00:13:21.559 Very hostile environment for electronics. 278 00:13:21.639 --> 00:13:24.600 So how do the tiny computer chips communicate without their 279 00:13:24.679 --> 00:13:27.440 data just getting completely scrambled by all that engine noise? 280 00:13:27.799 --> 00:13:30.840 That is exactly where the CANbus comes in the controller 281 00:13:30.879 --> 00:13:35.399 area network. It is the absolute gold standard for automotive systems, 282 00:13:36.000 --> 00:13:39.120 and the way it physically handles that noise is just 283 00:13:39.360 --> 00:13:40.279 brilliant physics. 284 00:13:40.360 --> 00:13:41.519 Okay, how does it work. 285 00:13:41.960 --> 00:13:45.720 It uses something called differential signaling. Imagine you have two 286 00:13:45.759 --> 00:13:48.480 separate wires running right next to each other through the car. 287 00:13:49.240 --> 00:13:51.840 When the system wants to send a signal, it sends 288 00:13:51.879 --> 00:13:55.159 a positive voltage down one wire and a negative voltage 289 00:13:55.200 --> 00:13:55.960 down the other wire. 290 00:13:56.080 --> 00:13:58.159 Okay, so like a plus one on the first wire 291 00:13:58.240 --> 00:13:59.799 and a minus one on the second wire. 292 00:14:00.120 --> 00:14:02.759 Right now, the receiving chip at the other end doesn't 293 00:14:02.799 --> 00:14:05.559 look at the absolute voltage on either wire. It strictly 294 00:14:05.600 --> 00:14:08.440 looks at the difference between the two wires. So positive 295 00:14:08.440 --> 00:14:11.639 one minus negative one equals two. The intended signal is two. 296 00:14:11.840 --> 00:14:12.919 Okay, I follow the math. 297 00:14:13.200 --> 00:14:16.679 Now, imagine a huge spike of electrical noise from the 298 00:14:16.759 --> 00:14:20.399 engine suddenly hits those wires. Because the wires are physically 299 00:14:20.399 --> 00:14:24.159 twisted together, the noise hits both wires equally. Right, Let's 300 00:14:24.159 --> 00:14:26.200 say it adds ten volts of pure noise, So the 301 00:14:26.240 --> 00:14:28.799 plus one wire becomes plus eleven and the minus one 302 00:14:28.799 --> 00:14:29.879 wire becomes plus nine. 303 00:14:30.039 --> 00:14:32.600 Ah, so both numbers instantly jump up. 304 00:14:32.919 --> 00:14:35.759 They do, but the receiver only cares about the mathematical 305 00:14:35.759 --> 00:14:39.759 difference between them, and eleven minus nine is still two. 306 00:14:39.799 --> 00:14:42.600 The signal is perfectly preserved exactly. 307 00:14:42.960 --> 00:14:46.039 The noise literally cancels itself out. It is an incredibly 308 00:14:46.120 --> 00:14:49.320 robust design. That is exactly why your anti lock breaks 309 00:14:49.320 --> 00:14:52.080 still work perfectly even when your engine is screaming at 310 00:14:52.120 --> 00:14:52.600 the red line. 311 00:14:52.639 --> 00:14:56.120 That is so smart. It's just elegant engineering. 312 00:14:55.679 --> 00:14:57.799 It really is. And there's one more very cool thing 313 00:14:57.799 --> 00:15:01.480 about cambus. Unlike it two C, there is no master controller. 314 00:15:01.879 --> 00:15:03.279 It's entirely peer to peer. 315 00:15:03.720 --> 00:15:06.480 Wait, so who decides who gets to talk? If there's 316 00:15:06.480 --> 00:15:08.720 no master it sounds like it would be chaos. 317 00:15:08.840 --> 00:15:11.679 The priority is actually built directly into the message ID 318 00:15:11.799 --> 00:15:15.159 number itself. If it deploy breaks message and the turn 319 00:15:15.240 --> 00:15:17.879 up radio volume message both try to talk on the 320 00:15:17.919 --> 00:15:20.759 network at the exact same millisecond. The system is physically 321 00:15:20.799 --> 00:15:23.080 designed so that the breaks have a lower ID number, 322 00:15:23.279 --> 00:15:26.279 and on a canvas, a lower number mathematically wins out 323 00:15:26.480 --> 00:15:27.679 and gets higher priority. 324 00:15:27.919 --> 00:15:30.600 That is amazing. The radio automatically just shuts up and 325 00:15:30.600 --> 00:15:31.360 waits yep. 326 00:15:31.679 --> 00:15:34.159 The physics of the wire forced the lower priority message 327 00:15:34.159 --> 00:15:34.720 to back off. 328 00:15:34.919 --> 00:15:38.679 I honestly wish human conversations worked like that, Like important 329 00:15:38.720 --> 00:15:41.759 safety info just mathematically overrides small talk. 330 00:15:42.320 --> 00:15:45.320 It would definitely make office meetings a lot more efficient. 331 00:15:45.759 --> 00:15:48.000 No kidding, Okay, So we had our sensors and we 332 00:15:48.039 --> 00:15:50.879 have the nervous system protocols. Now, let's talk about the brain. 333 00:15:51.440 --> 00:15:55.159 And this is where the book really focuses heavily. We 334 00:15:55.200 --> 00:15:58.519 aren't just talking about normal CPUs here, we are talking 335 00:15:58.519 --> 00:15:59.879 about FPG. 336 00:16:00.519 --> 00:16:03.960 Yes, this is where the engineering gets really interesting. FPGA 337 00:16:04.159 --> 00:16:06.960 stands for field programmable gate array. 338 00:16:07.200 --> 00:16:10.000 Field programmable makes it sound like I can literally fix 339 00:16:10.039 --> 00:16:11.879 it while I'm standing out in a cornfield. 340 00:16:12.000 --> 00:16:14.799 I mean, basically, yes, it means you can completely change 341 00:16:14.840 --> 00:16:17.200 how the physical chip works after it has already been 342 00:16:17.240 --> 00:16:18.840 manufactured and deployed in the field. 343 00:16:19.039 --> 00:16:22.480 How is that fundamentally different from a normal computer processor 344 00:16:22.720 --> 00:16:23.799 like the one in my laptop. 345 00:16:23.919 --> 00:16:26.720 Well, a normal CPU like an Intel or AMD chip 346 00:16:27.279 --> 00:16:31.399 has a totally fixed hardware structure. It is etched in silicon. 347 00:16:31.759 --> 00:16:35.000 It reads software instructions one by one from memory. Add 348 00:16:35.000 --> 00:16:37.519 this number, then move this data, then check this condition. 349 00:16:38.000 --> 00:16:41.440 It operates sequentially right, one step at a time, exactly. 350 00:16:41.919 --> 00:16:45.000 And FPGA, on the other hand, is basically a blank 351 00:16:45.080 --> 00:16:49.840 box of digital lego bricks. Raw logic gates, memory blocks 352 00:16:49.840 --> 00:16:52.960 flip flops. When you program it. You don't write software 353 00:16:52.960 --> 00:16:55.759 that runs on the chip. You write code that literally 354 00:16:55.799 --> 00:16:59.000 rewires the electrical connections between those internal lego bricks. 355 00:16:59.480 --> 00:17:03.080 Wait, I'm physically well electronically changing the actual circuit. 356 00:17:03.159 --> 00:17:06.880 Yes, you are literally creating custom hardware on the fly. 357 00:17:07.200 --> 00:17:09.279 Let's break down those lego bricks you mentioned. The book 358 00:17:09.279 --> 00:17:10.599 talks a lot about logic gates. 359 00:17:10.759 --> 00:17:13.640 Right, These are the fundamental decision makers of the digital world. 360 00:17:13.640 --> 00:17:16.599 You have eighty gates which say if input A andy 361 00:17:16.680 --> 00:17:20.200 input B are true, then do this. You haver gates, 362 00:17:20.680 --> 00:17:23.839 exoor gates, and then you have flip flops, which act 363 00:17:23.880 --> 00:17:26.359 as high speed memory units. They hold a single bit 364 00:17:26.400 --> 00:17:28.559 of data for a split second based on the ticking 365 00:17:28.559 --> 00:17:29.359 of the system clock. 366 00:17:29.440 --> 00:17:31.799 And a single FPGA is just made of millions of 367 00:17:31.839 --> 00:17:33.359 these little gates, millions. 368 00:17:33.079 --> 00:17:35.039 Of them, and because you are wiring them all up 369 00:17:35.079 --> 00:17:38.519 yourself in whatever pattern you need, you unlock the FPGA's 370 00:17:38.839 --> 00:17:43.279 true superpower parallelism. Parallelism that is definitely the key word here. 371 00:17:43.519 --> 00:17:46.799 Think of a standard CPU as a genius level math professor. 372 00:17:47.160 --> 00:17:49.720 He can solve absolutely any problem you give him, but 373 00:17:49.799 --> 00:17:52.640 he can only solve one problem at a time. He 374 00:17:52.720 --> 00:17:57.039 is very fast, but he's working entirely alone. Okay, And 375 00:17:57.200 --> 00:18:00.400 FPGA is more like a giant room flow of ten 376 00:18:00.440 --> 00:18:03.480 thousand middle school students. They might not be as smart 377 00:18:03.480 --> 00:18:07.000 individually as the professor, but they can all solve a specific, 378 00:18:07.319 --> 00:18:09.759 simple problem at the exact same time. 379 00:18:09.960 --> 00:18:12.720 So if I'm processing an image that has a million pixels, 380 00:18:12.799 --> 00:18:15.119 the normal CPU has to look at pixel one, then 381 00:18:15.160 --> 00:18:17.039 pixel two, then pixel three exactly. 382 00:18:17.279 --> 00:18:20.720 Even if it does it fast, it's still sequential. The FPGA, however, 383 00:18:20.799 --> 00:18:22.960 can be wired to look at all one million pixels, 384 00:18:22.960 --> 00:18:26.240 simultaneously process every single one of them, and be completely 385 00:18:26.279 --> 00:18:28.359 done in a single tick of the clock. That is 386 00:18:28.440 --> 00:18:31.400 true hardware acceleration. It's not just about running faster, it's 387 00:18:31.400 --> 00:18:32.480 about doing everything at once. 388 00:18:32.599 --> 00:18:35.519 That honestly blows my mind. But you don't use regular 389 00:18:35.559 --> 00:18:39.039 programming languages like Python or C plus plus to program 390 00:18:39.079 --> 00:18:39.759 these things, do you. 391 00:18:40.000 --> 00:18:42.960 Well? You can use C plus plus with some newer 392 00:18:43.160 --> 00:18:46.720 high level synthesis tools now, which definitely helps. But underneath 393 00:18:46.759 --> 00:18:50.440 it all the native tongue of the FPGA is VHDL 394 00:18:50.599 --> 00:18:53.920 or verolog. These are hardware description languages. 395 00:18:53.960 --> 00:18:54.799 Hardware description. 396 00:18:55.039 --> 00:18:58.240 Right when you type in violog, you aren't writing a 397 00:18:58.279 --> 00:19:01.000 sequential to do list for the computer. You are describing 398 00:19:01.039 --> 00:19:04.079 a physical structure. You are telling the software I need 399 00:19:04.119 --> 00:19:06.640 you to connect this wire to that specific gait. 400 00:19:07.000 --> 00:19:09.759 It's drawing architectural pootprints, not writing a recipe. 401 00:19:09.799 --> 00:19:13.000