WEBVTT 00:00:00.000 --> 00:00:03.060 Welcome back to the Deep Dive. Before we even 00:00:03.060 --> 00:00:05.559 really get into the topic today, I actually want 00:00:05.559 --> 00:00:07.259 to ask you to do something. Oh, we are starting 00:00:07.259 --> 00:00:08.960 with an assignment. Yeah, just a quick one though. 00:00:09.000 --> 00:00:10.800 If you are listening to this right now, I want 00:00:10.800 --> 00:00:13.220 you to just look around the room you are in. 00:00:13.660 --> 00:00:16.120 Or your car if you are driving. Right, exactly. 00:00:16.300 --> 00:00:18.899 Look at the dashboard. Or if you are at home, 00:00:19.019 --> 00:00:22.199 look at your Wi -Fi router or your TV, maybe 00:00:22.199 --> 00:00:24.500 the charger block for your laptop. Pretty much 00:00:24.500 --> 00:00:26.660 anything plugged into a wall. Yeah, all of it. 00:00:26.739 --> 00:00:30.000 Because we live in this world that we generally 00:00:30.000 --> 00:00:32.780 define a solid state. Right. Like we think of 00:00:32.780 --> 00:00:36.000 our electronics as these completely rigid, frozen 00:00:36.000 --> 00:00:39.159 blocks of silicon and plastic. Just static things 00:00:39.159 --> 00:00:41.560 that process logic until the end of time. Exactly. 00:00:41.740 --> 00:00:44.100 But the thing is, there is a component inside 00:00:44.100 --> 00:00:46.320 almost every single device I just mentioned. 00:00:46.359 --> 00:00:48.740 And honestly, it is highly likely it's inside 00:00:48.740 --> 00:00:50.840 the device playing this audio right now that 00:00:50.840 --> 00:00:53.340 completely defies that solid state logic. It 00:00:53.340 --> 00:00:55.939 really does. It is the ubiquitous unsung hero 00:00:55.939 --> 00:01:00.420 of modern tech. And quite often it's the inevitable. 00:01:00.990 --> 00:01:04.109 point of failure. We are talking about the aluminum 00:01:04.109 --> 00:01:08.290 electrolytic capacitor or the E cap. And I have 00:01:08.290 --> 00:01:10.370 to say, the more I read through the source material 00:01:10.370 --> 00:01:13.510 for this deep dive, the more I realized that 00:01:13.510 --> 00:01:16.530 this is not just some boring standard electrical 00:01:16.530 --> 00:01:18.709 component. Oh, far from it. It's a contradiction. 00:01:18.890 --> 00:01:22.269 It is literally a piece of 19th century chemistry 00:01:22.269 --> 00:01:25.730 living inside our 21st century digital tech. 00:01:26.030 --> 00:01:29.689 It's a component that relies on. What the sources 00:01:29.689 --> 00:01:32.310 call useful rust. Useful rust, yeah. And essentially 00:01:32.310 --> 00:01:34.829 a wet sponge just to keep the entire internet 00:01:34.829 --> 00:01:37.209 running. That is a really evocative way to put 00:01:37.209 --> 00:01:39.549 it, but it is completely accurate. I tend to 00:01:39.549 --> 00:01:41.430 treat our electronics as if they are immortal. 00:01:41.709 --> 00:01:44.269 Right. We expect a microprocessor to work flawlessly 00:01:44.269 --> 00:01:46.390 forever unless you take a hammer and physically 00:01:46.390 --> 00:01:49.290 smash it. But to actually power that processor, 00:01:49.590 --> 00:01:52.069 we anchor it to a component that is fundamentally 00:01:52.069 --> 00:01:54.650 strictly mortal. It has a biological clock, basically. 00:01:54.650 --> 00:01:57.370 Exactly. It has a lifespan. It ages. It physically 00:01:57.370 --> 00:02:00.000 dries out. was the hook for me we are building 00:02:00.000 --> 00:02:02.260 these massive digital cathedrals but the foundation 00:02:02.260 --> 00:02:05.260 has a ticking clock built right into it so our 00:02:05.260 --> 00:02:07.500 mission today is to completely unpack the ecap 00:02:07.500 --> 00:02:09.840 we have a lot of ground to cover we really do 00:02:09.840 --> 00:02:12.439 we aren't just going to sit here and read a data 00:02:12.439 --> 00:02:14.860 sheet to you we are going to look at the anatomy 00:02:14.860 --> 00:02:17.319 which is way more complex than i ever thought 00:02:17.319 --> 00:02:22.159 and the chemistry and uh Surprisingly, the history. 00:02:22.319 --> 00:02:24.840 The history is wild. Right, because it involves 00:02:24.840 --> 00:02:27.860 this massive industrial espionage scandal back 00:02:27.860 --> 00:02:30.719 in the early 2000s that honestly nearly took 00:02:30.719 --> 00:02:33.340 down the entire computer industry. The Capacitor 00:02:33.340 --> 00:02:36.479 Plague. It is a classic cautionary tale of why 00:02:36.479 --> 00:02:39.120 you really shouldn't steal trade secrets if you 00:02:39.120 --> 00:02:41.259 don't actually understand the underlying chemistry. 00:02:41.840 --> 00:02:43.580 We're definitely going to get to the exploding 00:02:43.580 --> 00:02:45.620 computers later, but I think we should start 00:02:45.620 --> 00:02:48.300 with the basics. Good idea. So if I crack open 00:02:48.300 --> 00:02:50.900 an old radio or even just look through the vent 00:02:50.900 --> 00:02:53.879 holes of my desktop PC, I see these little cylinders. 00:02:54.080 --> 00:02:56.039 Right, the cans. Yeah, they look like tiny little 00:02:56.039 --> 00:02:58.360 soda cans just standing up on the green circuit 00:02:58.360 --> 00:03:00.860 board. Sometimes they have this plastic shrink 00:03:00.860 --> 00:03:02.780 wrap on them with numbers and sometimes they 00:03:02.780 --> 00:03:06.020 are just bare shiny metal. What exactly are we 00:03:06.020 --> 00:03:08.120 looking at there? So visually, yes, they just 00:03:08.120 --> 00:03:10.500 look like little cans. But that aluminum can. 00:03:11.210 --> 00:03:13.889 is really just the housing. It's the shell. The 00:03:13.889 --> 00:03:16.330 actual magic is all inside. Okay. If you were 00:03:16.330 --> 00:03:19.490 to take a pair of heavy pliers, which please 00:03:19.490 --> 00:03:21.449 do not actually do this at home because stuff 00:03:21.449 --> 00:03:24.629 inside is highly caustic and it will make a mess. 00:03:25.009 --> 00:03:28.150 But if you did peel that metal can open, you 00:03:28.150 --> 00:03:30.870 would not find a solid block of material in there. 00:03:30.969 --> 00:03:32.870 Right. The notes called it a jelly roll. Yes. 00:03:32.969 --> 00:03:35.449 A wound roll. You would find a roll of foils. 00:03:35.930 --> 00:03:38.330 But honestly, calling it a roll of foil really 00:03:38.330 --> 00:03:40.449 undersells the engineering involved. Because 00:03:40.449 --> 00:03:43.009 it's not just one piece of foil. It's a sandwich. 00:03:43.189 --> 00:03:45.310 It's a highly engineered sandwich that has been 00:03:45.310 --> 00:03:47.449 rolled up very tightly. You have an anode foil. 00:03:47.629 --> 00:03:50.409 You have a cathode foil. And you have a separator 00:03:50.409 --> 00:03:52.150 paper sandwiched right in between them. Okay, 00:03:52.189 --> 00:03:53.830 so let's pause right here because this is where 00:03:53.830 --> 00:03:56.409 my basic high school physics brain gets confused. 00:03:56.669 --> 00:03:59.550 Sure. Standard capacitor physics, right. You 00:03:59.550 --> 00:04:02.370 have two metal plates and you have an insulator 00:04:02.370 --> 00:04:05.110 in between them. You pump a charge into the plates 00:04:05.110 --> 00:04:07.069 and the energy is stored in the electric field 00:04:07.069 --> 00:04:09.050 between them. That is the textbook definition, 00:04:09.229 --> 00:04:11.889 yes. So in my mind, looking at this jelly roll, 00:04:12.050 --> 00:04:15.909 the anode foil is the positive plate. The cathode 00:04:15.909 --> 00:04:18.730 foil is the negative plate. And that paper is 00:04:18.730 --> 00:04:21.329 the insulator. And that right there is the most 00:04:21.329 --> 00:04:24.110 common misconception. Even a lot of first -year 00:04:24.110 --> 00:04:25.850 engineering students get that wrong on their 00:04:25.850 --> 00:04:28.769 first pass. Really? Yeah. Because if this were 00:04:28.769 --> 00:04:31.689 a film capacitor or a basic ceramic capacitor, 00:04:31.930 --> 00:04:34.560 you would be absolutely right. But in an aluminum 00:04:34.560 --> 00:04:38.160 electrolytic capacitor, The cathode foil is not 00:04:38.160 --> 00:04:40.800 actually the cathode. Wait, hold on. It is literally 00:04:40.800 --> 00:04:43.500 labeled cathode foil in all the diagrams in our 00:04:43.500 --> 00:04:45.680 sources. I know it is. It is incredibly confusing 00:04:45.680 --> 00:04:48.439 historical terminology. But you have to think 00:04:48.439 --> 00:04:50.860 about the physical structure of what we're trying 00:04:50.860 --> 00:04:54.139 to do here. Okay. To get high capacitance, which 00:04:54.139 --> 00:04:56.439 is just the ability to store a massive amount 00:04:56.439 --> 00:04:59.040 of electrical energy, you need a massive amount 00:04:59.040 --> 00:05:01.160 of surface area on your plates. Right. Bigger 00:05:01.160 --> 00:05:03.810 plates mean more storage. Exactly. Now, if you 00:05:03.810 --> 00:05:06.290 just take two strips of completely smooth aluminum 00:05:06.290 --> 00:05:08.410 foil from your kitchen and roll them up with 00:05:08.410 --> 00:05:10.970 some paper, you have very little actual surface 00:05:10.970 --> 00:05:13.480 area. Because it's just flat. Right. You would 00:05:13.480 --> 00:05:16.879 get maybe a few microfarads of capacitance, even 00:05:16.879 --> 00:05:19.180 if you made the thing the size of a giant beer 00:05:19.180 --> 00:05:22.060 can. But these e -caps are tiny. They're the 00:05:22.060 --> 00:05:24.480 size of a pencil eraser sometimes, and they have 00:05:24.480 --> 00:05:28.420 huge capacitance ratings. So where is all that 00:05:28.420 --> 00:05:31.379 surface area coming from if the foil is so small? 00:05:31.639 --> 00:05:34.680 It comes from a process called etching. Before 00:05:34.680 --> 00:05:37.279 they even wind that roll up at the factory, they 00:05:37.279 --> 00:05:40.060 take the anode foil. just the positive foil, 00:05:40.279 --> 00:05:42.600 and they run it through an electrochemical acid 00:05:42.600 --> 00:05:45.319 bath. So they are intentionally corroding it. 00:05:45.420 --> 00:05:47.839 Yes. They essentially eat away at the solid metal. 00:05:47.899 --> 00:05:51.000 They roughen it up on a microscopic level. The 00:05:51.000 --> 00:05:52.899 source material actually mentioned a magnification 00:05:52.899 --> 00:05:55.899 factor of up to 200. I was really trying to visualize 00:05:55.899 --> 00:05:57.660 that when I was reading it. It's hard to wrap 00:05:57.660 --> 00:05:59.660 your head around. Yeah. Like if I have a completely 00:05:59.660 --> 00:06:02.000 flat sheet of foil the size of my dining room 00:06:02.000 --> 00:06:05.360 table, how do I suddenly get 200 tables worth 00:06:05.360 --> 00:06:07.480 of surface area just by scratching it up with 00:06:07.480 --> 00:06:09.660 acid? You have to stop thinking in two dimensions. 00:06:09.939 --> 00:06:12.379 You have to think in 3D. Think about a city map. 00:06:12.579 --> 00:06:15.420 Let's use Manhattan as an analogy. If you look 00:06:15.420 --> 00:06:19.279 at a flat paper map of Manhattan, it covers a 00:06:19.279 --> 00:06:22.740 very specific number of square miles. That is 00:06:22.740 --> 00:06:25.660 what engineers call the projected area, the footprint. 00:06:25.920 --> 00:06:27.959 Right, just the outline of the island. But now 00:06:27.959 --> 00:06:30.980 imagine I give you a paintbrush and I tell you 00:06:30.980 --> 00:06:33.920 that you have to paint every single exterior 00:06:33.920 --> 00:06:37.019 surface in Manhattan, every single side of every 00:06:37.019 --> 00:06:40.490 skyscraper, every window frame. every alleyway 00:06:40.490 --> 00:06:43.689 wall down into the subway entrances. Oh, wow. 00:06:43.850 --> 00:06:45.790 Okay. How much paint do you need for that compared 00:06:45.790 --> 00:06:47.629 to just painting a flat parking lot the size 00:06:47.629 --> 00:06:50.069 of the island? A massive amount, probably hundreds 00:06:50.069 --> 00:06:52.870 of times more paint. Exactly. And that is exactly 00:06:52.870 --> 00:06:54.910 what the etching process does to the aluminum 00:06:54.910 --> 00:06:57.350 foil. It doesn't just put scratches on the surface. 00:06:57.569 --> 00:07:00.990 It digs these deep microscopic vertical tunnels 00:07:00.990 --> 00:07:03.009 straight down into the aluminum. Millions of 00:07:03.009 --> 00:07:05.709 them. Billions. Billions of microscopic tunnels 00:07:05.709 --> 00:07:08.680 per square centimeter. The solid metal foil essentially 00:07:08.680 --> 00:07:10.980 becomes a three -dimensional sponge of metal. 00:07:11.120 --> 00:07:12.819 Okay, I am totally with you on the Manhattan 00:07:12.819 --> 00:07:14.839 analogy. That makes perfect sense. So we have 00:07:14.839 --> 00:07:17.920 this incredibly complex rough metal sponge. Right. 00:07:17.980 --> 00:07:21.100 Now keep that in mind. If you just took a second 00:07:21.100 --> 00:07:24.579 piece of flat aluminum foil, the so -called cathode 00:07:24.579 --> 00:07:27.040 foil, and laid it directly on top of that etched 00:07:27.040 --> 00:07:30.040 sponge, where would it actually touch? It would 00:07:30.040 --> 00:07:32.300 only touch the very tops of the skyscrapers. 00:07:32.300 --> 00:07:34.420 It wouldn't get down into the alleyways or the 00:07:34.420 --> 00:07:36.759 tunnels. Precisely. You would instantly lose 00:07:36.759 --> 00:07:39.560 all of that extra surface area. You just spent 00:07:39.560 --> 00:07:41.620 all that time in asset creating. It would be 00:07:41.620 --> 00:07:44.139 completely useless. So you need a second plate 00:07:44.139 --> 00:07:46.139 that can somehow conform to all those tunnels. 00:07:46.339 --> 00:07:48.500 Yes. You need a second plate that can flow. You 00:07:48.500 --> 00:07:51.959 need a liquid. Ah, the electrolyte. The electrolyte. 00:07:51.959 --> 00:07:54.959 The liquid is the actual cathode. Wow. It literally 00:07:54.959 --> 00:07:57.980 floods the streets of Manhattan. It flows down 00:07:57.980 --> 00:08:00.740 into every single microscopic tunnel and coats 00:08:00.740 --> 00:08:03.500 the entire rough surface of the etched aluminum. 00:08:03.899 --> 00:08:06.560 That is why the liquid is the true second plate 00:08:06.560 --> 00:08:08.839 of the capacitor. That clicks so perfectly now. 00:08:08.920 --> 00:08:11.100 So the liquid is the functional negative side 00:08:11.100 --> 00:08:13.500 of the circuit. But then what is the physical 00:08:13.500 --> 00:08:15.680 cathode for actually doing in the roll? Why do 00:08:15.680 --> 00:08:17.639 we even need a second piece of metal in there 00:08:17.639 --> 00:08:19.579 at all? Well, think about how you connect this 00:08:19.579 --> 00:08:22.560 to a circuit board. You cannot solder a copper 00:08:22.560 --> 00:08:25.180 wire directly to a liquid. Right, that would 00:08:25.180 --> 00:08:27.939 be messy. You can't just stick a metal lead into 00:08:27.939 --> 00:08:31.000 a wet sponge and hope for a reliable high -speed 00:08:31.000 --> 00:08:34.139 electrical connection. You need a physical way 00:08:34.139 --> 00:08:37.000 to get the electricity from the outside world 00:08:37.000 --> 00:08:39.740 into the liquid itself. So the cathode foil is 00:08:39.740 --> 00:08:42.100 just the contact mechanism. Exactly. It is like 00:08:42.100 --> 00:08:44.820 dropping a giant copper cable into the ocean. 00:08:45.149 --> 00:08:47.570 The salt water of the ocean is the actual conductor, 00:08:47.789 --> 00:08:50.070 but the cable is just the interface to get the 00:08:50.070 --> 00:08:52.309 power in and out. So calling it the cathode foil 00:08:52.309 --> 00:08:54.669 is almost a lie. It's really just the current 00:08:54.669 --> 00:08:56.929 collector foil. Current collector foil would 00:08:56.929 --> 00:08:59.370 be a much, much more accurate name from a physics 00:08:59.370 --> 00:09:01.210 standpoint, yes. Okay, so we have this really 00:09:01.210 --> 00:09:04.649 rough metal sponge, which is the anode, and we 00:09:04.649 --> 00:09:07.029 have a liquid soaking deeply into it, which is 00:09:07.029 --> 00:09:10.210 the cathode. But wait. If the conductive liquid 00:09:10.210 --> 00:09:13.250 is physically touching, the conductive metal 00:09:13.250 --> 00:09:15.610 isn't that just a short circuit. Why doesn't 00:09:15.610 --> 00:09:17.450 the electricity just flow right through the whole 00:09:17.450 --> 00:09:19.909 thing? And this brings us to the actual magic 00:09:19.909 --> 00:09:22.870 of the construction. The useful rust we talked 00:09:22.870 --> 00:09:26.110 about. We need a dielectric, an insulator to 00:09:26.110 --> 00:09:28.590 separate the metal sponge from the liquid cathode. 00:09:28.690 --> 00:09:30.990 And in this specific case, it is aluminum oxide. 00:09:31.470 --> 00:09:34.370 Yes, Al2O3. Because aluminum inherently loves 00:09:34.370 --> 00:09:36.870 oxygen, right? Like if you leave a piece of bare 00:09:36.870 --> 00:09:39.889 aluminum outside in the yard, it instantly forms 00:09:39.889 --> 00:09:43.129 a thin oxide layer on its own. Exactly. That 00:09:43.129 --> 00:09:45.710 is why aluminum doesn't rust away into dust the 00:09:45.710 --> 00:09:48.309 way iron does. It naturally protects itself with 00:09:48.309 --> 00:09:50.769 a very thin barrier. But for a high voltage capacitor, 00:09:51.110 --> 00:09:54.330 I'm guessing that natural rust just isn't quite 00:09:54.330 --> 00:09:57.070 good enough. Not even close. It's way too thin 00:09:57.070 --> 00:09:59.750 and it's very patchy. We need to engineer a perfect 00:09:59.750 --> 00:10:02.690 layer. This is the anodization process from the 00:10:02.690 --> 00:10:04.830 outline. Right. So before the jelly roll is even 00:10:04.830 --> 00:10:07.549 assembled, they take that rough etched foil sponge 00:10:07.549 --> 00:10:09.529 and they run it through another chemical bath. 00:10:09.929 --> 00:10:12.129 But this time they apply a very specific voltage 00:10:12.129 --> 00:10:14.389 to it. They force it to rust. They force the 00:10:14.389 --> 00:10:16.590 oxygen in the bath to bond chemically with the 00:10:16.590 --> 00:10:19.149 aluminum surface. They grow a highly distinct, 00:10:19.289 --> 00:10:22.450 totally uniform layer of oxide over every single 00:10:22.450 --> 00:10:24.759 microscopic hill and valley. And the sources 00:10:24.759 --> 00:10:28.000 noted this concept of a valve metal. Yes, discovered 00:10:28.000 --> 00:10:32.159 way back in 1875 by a French physicist named 00:10:32.159 --> 00:10:35.720 Eugène Ducreté. He found that certain oxidized 00:10:35.720 --> 00:10:39.120 metals like aluminum or tantalum act like a one 00:10:39.120 --> 00:10:42.460 -way valve. They let current flow freely in one 00:10:42.460 --> 00:10:44.340 direction but totally block it in the other. 00:10:44.639 --> 00:10:46.919 Which is exactly what an insulator needs to do 00:10:46.919 --> 00:10:50.539 here. But the key to this whole thing is the 00:10:50.539 --> 00:10:52.779 thickness of that rust layer, right? It is the 00:10:52.779 --> 00:10:54.799 most critical parameter. The thickness of this 00:10:54.799 --> 00:10:57.159 oxide completely determines the maximum voltage 00:10:57.159 --> 00:10:59.419 the capacitor can handle. The notes literally 00:10:59.419 --> 00:11:01.899 say the thickness is measured in nanometers per 00:11:01.899 --> 00:11:05.919 volt. It is unfathomably thin. And this is crucial 00:11:05.919 --> 00:11:08.480 for the performance because... capacitance the 00:11:08.480 --> 00:11:11.120 storage capacity is inversely proportional to 00:11:11.120 --> 00:11:13.159 the distance between your two plates so the closer 00:11:13.159 --> 00:11:15.059 you can get the metal plate and the liquid plate 00:11:15.059 --> 00:11:17.320 together without them actually touching the more 00:11:17.320 --> 00:11:19.940 energy you can store precisely you want the insulator 00:11:19.940 --> 00:11:22.200 to be as impossibly thin as you can make it without 00:11:22.200 --> 00:11:24.700 it breaking down i want to really get a solid 00:11:24.700 --> 00:11:27.539 sense of scale here for the listener if i have 00:11:27.539 --> 00:11:30.080 a standard everyday capacitor say one rated for 00:11:30.080 --> 00:11:32.200 16 volts which is super common in a lot of computer 00:11:32.200 --> 00:11:35.860 parts how thick is that insulator layer Well, 00:11:35.879 --> 00:11:38.480 the general rule of thumb in the industry is 00:11:38.480 --> 00:11:41.779 about 1 .4 nanometers of oxide thickness for 00:11:41.779 --> 00:11:44.879 every single volt of rating. So for a 16 -volt 00:11:44.879 --> 00:11:48.159 capacitor, you're looking at roughly 22 to maybe 00:11:48.159 --> 00:11:50.659 25 nanometers thick. Okay, that number obviously 00:11:50.659 --> 00:11:53.100 sounds very small, but let's contextualize it. 00:11:53.159 --> 00:11:56.470 I looked this up. A single strand of human DNA 00:11:56.470 --> 00:12:00.350 is about 2 .5 nanometers wide. Wow. Okay, so 00:12:00.350 --> 00:12:02.570 if you do the math there, the physical barrier 00:12:02.570 --> 00:12:04.590 that is keeping your computer's power supply 00:12:04.590 --> 00:12:07.549 from instantly blowing up right now is only about 00:12:07.549 --> 00:12:10.669 10 DNA strands thick. That is terrifyingly thin. 00:12:10.850 --> 00:12:12.769 It really is. It is walking an absolute molecular 00:12:12.769 --> 00:12:15.929 tightrope. And that is exactly why these components 00:12:15.929 --> 00:12:18.350 are so incredibly sensitive to voltage spikes. 00:12:18.690 --> 00:12:21.159 Because if you push the voltage too high... Say 00:12:21.159 --> 00:12:24.059 you accidentally put 30 volts across that 16 00:12:24.059 --> 00:12:27.019 -volt capacitor. The electrical field across 00:12:27.019 --> 00:12:30.820 that tiny gap becomes so intensely strong that 00:12:30.820 --> 00:12:33.500 it physically rips electrons right through that 00:12:33.500 --> 00:12:35.919 thin oxide wall. It just punches a hole right 00:12:35.919 --> 00:12:38.179 through it. Yes. That is called dielectric breakdown. 00:12:38.379 --> 00:12:40.299 And once you puncture that wall, you just have 00:12:40.299 --> 00:12:43.360 metal touching liquid. Which is a dead short 00:12:43.360 --> 00:12:47.779 circuit. Massive unregulated current flow. incredible 00:12:47.779 --> 00:12:51.320 heat, and then a very fast explosion. Yeah, we 00:12:51.320 --> 00:12:53.480 are definitely getting to the explosions. But 00:12:53.480 --> 00:12:55.500 before we leave the oxide layer, I want to touch 00:12:55.500 --> 00:12:57.460 on something the source material mentioned regarding 00:12:57.460 --> 00:12:59.820 the actual type of oxide they grow. Well, the 00:12:59.820 --> 00:13:01.899 structural types. Yeah, there was this big distinction 00:13:01.899 --> 00:13:04.980 between amorphous oxide and crystalline oxide. 00:13:05.220 --> 00:13:07.519 And reading it, it almost felt a bit like a personality 00:13:07.519 --> 00:13:09.840 test for the component. That is a great way to 00:13:09.840 --> 00:13:12.159 frame it. It is entirely about the physical structure 00:13:12.159 --> 00:13:14.519 of the molecules in that rust layer. So what's 00:13:14.519 --> 00:13:16.879 the difference? Amorphous oxide is kind of like 00:13:16.879 --> 00:13:19.200 a massive pile of bricks just dumped haphazardly 00:13:19.200 --> 00:13:22.440 on the ground. It is disordered. It is structurally 00:13:22.440 --> 00:13:25.799 messy. But because it is so messy, it is actually 00:13:25.799 --> 00:13:27.919 very flexible. It can shift around. Right. It 00:13:27.919 --> 00:13:30.740 covers the rough etched surface incredibly well. 00:13:30.820 --> 00:13:33.320 And it is very chemically stable over long periods 00:13:33.320 --> 00:13:36.360 of time. This is the exact personality you want 00:13:36.360 --> 00:13:38.519 for a general purpose capacitor that just needs 00:13:38.519 --> 00:13:41.379 to run quietly for 10 years. The reliable, steady 00:13:41.379 --> 00:13:44.019 workhorse. Exactly. But then on the other hand, 00:13:44.019 --> 00:13:46.480 you have crystalline oxide. Imagine taking those 00:13:46.480 --> 00:13:50.000 same bricks. and meticulously building a perfectly 00:13:50.000 --> 00:13:52.940 rigid, perfectly aligned wall. Highly organized. 00:13:53.320 --> 00:13:55.940 Everything is locked in place. Because it is 00:13:55.940 --> 00:13:58.299 so highly organized, it is physically denser, 00:13:58.320 --> 00:14:00.159 which means you can actually get away with a 00:14:00.159 --> 00:14:02.559 slightly thinner layer to get the exact same 00:14:02.559 --> 00:14:05.379 voltage protection. Which loops back to our rule 00:14:05.379 --> 00:14:08.720 from earlier. Thinner gap means even higher capacitance 00:14:08.720 --> 00:14:10.960 in the same physical footprint. Much, much higher 00:14:10.960 --> 00:14:14.000 efficiency, yes. But there is a huge trade -off. 00:14:14.350 --> 00:14:16.669 What happens to a perfectly rigid brick wall 00:14:16.669 --> 00:14:19.009 if the ground underneath it suddenly shakes? 00:14:19.210 --> 00:14:21.610 It cracks. It doesn't bend, it just shatters. 00:14:21.789 --> 00:14:25.429 It cracks. Crystalline oxide is incredibly brittle. 00:14:26.330 --> 00:14:29.330 If the machine bends the foil a tiny bit too 00:14:29.330 --> 00:14:31.450 much during the winding process at the factory, 00:14:31.629 --> 00:14:35.210 or even if the capacitor just heats up and thermally 00:14:35.210 --> 00:14:39.070 expands during use, that oxide can develop microscopic 00:14:39.070 --> 00:14:42.620 fractures. So where do they even use the crystalline 00:14:42.620 --> 00:14:45.639 type if it's so fragile? You typically only see 00:14:45.639 --> 00:14:48.100 it in very specialized high voltage applications 00:14:48.100 --> 00:14:50.600 where you desperately need that maximum efficiency. 00:14:50.879 --> 00:14:53.340 Things like photo flash capacitors for professional 00:14:53.340 --> 00:14:55.669 cameras. They need to sit there at 300 volts 00:14:55.669 --> 00:14:58.230 and then dump all that energy in a single millisecond. 00:14:58.289 --> 00:15:00.049 Right. They need the pure performance and they 00:15:00.049 --> 00:15:02.889 just accept the fragility. So for my simple laptop 00:15:02.889 --> 00:15:05.230 charger, I am definitely looking at the amorphous 00:15:05.230 --> 00:15:07.570 type because I prioritize it simply not dying 00:15:07.570 --> 00:15:10.149 on me over it being perfectly space efficient. 00:15:10.289 --> 00:15:11.929 Correct. You definitely want the amorphous stability 00:15:11.929 --> 00:15:14.289 for a daily driver. Okay, let's move on to the 00:15:14.289 --> 00:15:15.850 liquid itself. We've been calling it the wet 00:15:15.850 --> 00:15:18.929 sponge, but it is obviously not just plain tap 00:15:18.929 --> 00:15:21.259 water in there, right? The source notes list 00:15:21.259 --> 00:15:23.759 some fairly intense chemicals like ethylene glycol 00:15:23.759 --> 00:15:26.340 and boric acid. Right, the electrolyte. This 00:15:26.340 --> 00:15:28.259 is essentially the secret sauce of the entire 00:15:28.259 --> 00:15:31.440 industry. Every major manufacturer has their 00:15:31.440 --> 00:15:35.220 own highly specific proprietary recipe, and they 00:15:35.220 --> 00:15:37.399 guard those formulas like it's Coca -Cola. Because 00:15:37.399 --> 00:15:40.639 it's a huge competitive advantage. Massive. But 00:15:40.639 --> 00:15:43.120 historically, and honestly still very commonly 00:15:43.120 --> 00:15:47.179 today, the base liquid for that soup is ethylene 00:15:47.179 --> 00:15:51.019 glycol. Which is literally car antifreeze. Basically, 00:15:51.080 --> 00:15:54.019 yes. It is a solvent that doesn't boil off easily 00:15:54.019 --> 00:15:56.480 when it gets hot, and it doesn't freeze solid 00:15:56.480 --> 00:15:58.039 when it gets cold. Perfect for an electronic 00:15:58.039 --> 00:16:00.259 component that gets hot. Exactly. Then they take 00:16:00.259 --> 00:16:01.860 that base and they dissolve conductive salts 00:16:01.860 --> 00:16:04.580 into it. Very often things like boric acid or 00:16:04.580 --> 00:16:06.759 various ammonium salts, that is what actually 00:16:06.759 --> 00:16:08.860 makes the liquid conduct electricity. The notes 00:16:08.860 --> 00:16:11.379 called this the borax electrolyte. Yeah, it has 00:16:11.379 --> 00:16:14.139 been around since the 1930s. It's the gold standard 00:16:14.139 --> 00:16:16.570 for stability. But the liquid does more than 00:16:16.570 --> 00:16:19.009 just passively conduct electricity. And this 00:16:19.009 --> 00:16:20.690 is the part of the deep dive that I found totally 00:16:20.690 --> 00:16:23.809 counterintuitive. A self -healing. Yes. The liquid 00:16:23.809 --> 00:16:26.529 is actually acting as a medic. It is a microscopic 00:16:26.529 --> 00:16:30.120 repair crew living inside the can. This is arguably 00:16:30.120 --> 00:16:33.179 the single most important defining feature of 00:16:33.179 --> 00:16:35.919 the aluminum electrolytic capacitor. It is entirely 00:16:35.919 --> 00:16:38.259 why we still use them everywhere, despite all 00:16:38.259 --> 00:16:41.159 of their flaws. But how does a completely lifeless 00:16:41.159 --> 00:16:44.720 chemical soup heal a scratch on a piece of metal? 00:16:44.940 --> 00:16:47.220 Well, think back to that unimaginably thin oxide 00:16:47.220 --> 00:16:51.279 layer, the 10 DNA strands. It is practically 00:16:51.279 --> 00:16:53.320 inevitable that there will be microscopic defects 00:16:53.320 --> 00:16:56.139 in that wall. Right. Impurities in the original 00:16:56.139 --> 00:16:58.620 aluminum or tiny micro cracks from when they 00:16:58.620 --> 00:17:00.799 rolled it up. Or just thermal stress over time. 00:17:01.019 --> 00:17:03.259 Now, in a solid state capacitor like a ceramic 00:17:03.259 --> 00:17:05.900 one, a crack in the insulator is just a permanent 00:17:05.900 --> 00:17:09.240 path for a short circuit. It is game over for 00:17:09.240 --> 00:17:11.960 the component. It's just dead. But here in the 00:17:11.960 --> 00:17:14.200 ECAP, you have a conductive liquid physically 00:17:14.200 --> 00:17:16.519 pressing against that crack. And that liquid 00:17:16.519 --> 00:17:19.200 contains dissolved oxygen. So when the electrical 00:17:19.200 --> 00:17:21.319 current tries to rush through that crack in the 00:17:21.319 --> 00:17:22.960 wall. The current actually drives a chemical 00:17:22.960 --> 00:17:25.460 reaction. It's basically the anodization process 00:17:25.460 --> 00:17:27.900 happening all over again, but highly locally 00:17:27.900 --> 00:17:31.579 in real time. The leakage current literally forces 00:17:31.579 --> 00:17:34.980 the oxygen out of the electrolyte and binds it 00:17:34.980 --> 00:17:37.200 with the raw exposed aluminum at the very bottom 00:17:37.200 --> 00:17:40.180 of the crack. It dynamically grows fresh aluminum 00:17:40.180 --> 00:17:42.980 oxide plugging the hole. It literally scabs over. 00:17:43.180 --> 00:17:46.160 Yes. It patches the breach, the leakage current 00:17:46.160 --> 00:17:49.140 stops flowing, and the capacitor just fixes itself 00:17:49.140 --> 00:17:51.380 while your computer is running. That sounds absolutely 00:17:51.380 --> 00:17:54.440 perfect, almost too perfect. So what is the downside 00:17:54.440 --> 00:17:56.900 here? There has to be a physical cost to doing 00:17:56.900 --> 00:17:59.960 that. The cost is chemistry. You do not get that 00:17:59.960 --> 00:18:02.200 fresh oxide for free. Right. Conservation of 00:18:02.200 --> 00:18:05.099 mass. To make that AL2O3 molecule, you have to 00:18:05.099 --> 00:18:07.640 physically steer oxygen atoms from the electrolyte 00:18:07.640 --> 00:18:10.420 solution itself, which means every single time 00:18:10.420 --> 00:18:13.180 it heals. You are physically permanently consuming 00:18:13.180 --> 00:18:15.700 a tiny bit of the liquid. You are using up the 00:18:15.700 --> 00:18:17.920 lifeblood of the capacitor just to fix the wounds. 00:18:18.160 --> 00:18:20.579 Exactly. And there is a chemical byproduct to 00:18:20.579 --> 00:18:22.660 this process, too. When you strip the oxygen 00:18:22.660 --> 00:18:25.779 out of water or out of these solvents, what chemical 00:18:25.779 --> 00:18:28.339 element is left over? Oh, hydrogen. Hydrogen 00:18:28.339 --> 00:18:31.640 gas. So every time the capacitor patches a tiny 00:18:31.640 --> 00:18:35.079 microscopic hole, it generates a tiny microscopic 00:18:35.079 --> 00:18:38.819 bubble of hydrogen gas inside the sealed can. 00:18:39.400 --> 00:18:41.819 Yes. And since the can is hermetically sealed, 00:18:42.000 --> 00:18:45.119 that gas has absolutely nowhere to go. Correct. 00:18:45.200 --> 00:18:47.960 Now, the healing process happens very slowly 00:18:47.960 --> 00:18:51.599 over years. That gas might just reabsorb into 00:18:51.599 --> 00:18:54.839 the liquid or diffuse out incredibly slowly through 00:18:54.839 --> 00:18:56.799 the rubber seal at the bottom. But if it's happening 00:18:56.799 --> 00:18:59.599 constantly. If you have a ton of defects or if 00:18:59.599 --> 00:19:01.579 you are running the capacitor really hot and 00:19:01.579 --> 00:19:03.680 accelerating the chemistry, the hydrogen gas 00:19:03.680 --> 00:19:06.259 builds up pressure fast. It essentially turns 00:19:06.259 --> 00:19:09.059 the tiny aluminum can into a pressure cooker. 00:19:09.400 --> 00:19:11.819 Okay. This feels like the absolute perfect transition 00:19:11.819 --> 00:19:14.319 to the history segment of our deep dive because 00:19:14.319 --> 00:19:16.400 the phrase pressure cooker literally describes 00:19:16.400 --> 00:19:19.180 the entire computer industry in the early 2000s 00:19:19.180 --> 00:19:21.240 perfectly. I want to talk about the capacitor 00:19:21.240 --> 00:19:23.339 blight. This is a story that I think really highlights 00:19:23.339 --> 00:19:26.140 exactly how incredibly fragile our entire global 00:19:26.140 --> 00:19:28.519 technology supply chain actually is. So set the 00:19:28.519 --> 00:19:31.119 scene for us here. The year is roughly 2001. 00:19:31.519 --> 00:19:33.880 Right. The tech world is absolutely booming. 00:19:34.000 --> 00:19:36.779 The dot -com crash happened, but hardware is 00:19:36.779 --> 00:19:39.039 pushing forward fast. processors are getting 00:19:39.039 --> 00:19:41.839 much, much faster. The Pentium 4 is just coming 00:19:41.839 --> 00:19:44.900 out. Huge power draws. Yes. These new, extremely 00:19:44.900 --> 00:19:47.759 dense processors needed a ton of power, and they 00:19:47.759 --> 00:19:50.980 needed it delivered in microsecond bursts. To 00:19:50.980 --> 00:19:53.039 do that, the power supplies on the motherboards 00:19:53.039 --> 00:19:56.019 desperately needed capacitors with incredibly 00:19:56.019 --> 00:19:58.720 low electrical resistance. Which the industry 00:19:58.720 --> 00:20:01.680 calls ESR. We'll define the physics of that in 00:20:01.680 --> 00:20:04.220 a bit. But basically, they needed high -performance 00:20:04.220 --> 00:20:06.759 sports car capacitors. Right. Now, if you are 00:20:06.759 --> 00:20:08.660 a chemist and you want to lower the electrical 00:20:08.660 --> 00:20:11.079 resistance of your electrolyte, the absolute 00:20:11.079 --> 00:20:13.579 easiest knob you can turn is to just add more 00:20:13.579 --> 00:20:15.759 water to the recipe. Because water conducts really 00:20:15.759 --> 00:20:17.819 well. Much better than the thick glycol, yes. 00:20:18.240 --> 00:20:21.259 But water is incredibly dangerous inside an e 00:20:21.259 --> 00:20:24.309 -cab because pure water... aggressively attacks 00:20:24.309 --> 00:20:27.269 bare aluminum it just rusts it continuously it 00:20:27.269 --> 00:20:30.109 corrodes it constantly creating hydrogen gas 00:20:30.109 --> 00:20:32.630 even when the device is completely turned off 00:20:32.630 --> 00:20:35.630 so adding water was always seen as a huge no 00:20:35.630 --> 00:20:38.720 -go But then a highly respected Japanese component 00:20:38.720 --> 00:20:42.019 company called Rubicon managed to develop a revolutionary 00:20:42.019 --> 00:20:44.400 new formula. They cracked the code. They did. 00:20:44.500 --> 00:20:47.039 It was a very high water content electrolyte. 00:20:47.039 --> 00:20:49.240 But they had discovered this complex proprietary 00:20:49.240 --> 00:20:53.579 mix of chemical additives, inhibitors, that successfully 00:20:53.579 --> 00:20:56.119 neutralized the water. It completely stopped 00:20:56.119 --> 00:20:58.019 it from eating the aluminum. So they got all 00:20:58.019 --> 00:20:59.940 the performance of the water with none of the 00:20:59.940 --> 00:21:02.940 explosive downsides. The holy grail. It absolutely 00:21:02.940 --> 00:21:05.880 was. And naturally, every single competitor in 00:21:05.880 --> 00:21:08.339 the market wanted that formula desperately. And 00:21:08.339 --> 00:21:10.839 this is exactly where the spy enters our story. 00:21:10.980 --> 00:21:13.859 A rogue material scientist. Allegedly, yes. A 00:21:13.859 --> 00:21:16.299 researcher from a competitor. The exact details 00:21:16.299 --> 00:21:18.660 get murky and stories vary. Some sources say 00:21:18.660 --> 00:21:21.220 they defected to a Chinese manufacturer. Others 00:21:21.220 --> 00:21:23.480 point to a Taiwanese outfit often cited in the 00:21:23.480 --> 00:21:26.059 industry as Luminous Town Electric. But the core 00:21:26.059 --> 00:21:28.259 gist is they walked out the door with the golden 00:21:28.259 --> 00:21:30.319 recipe. They stole the formula, yes. But they 00:21:30.319 --> 00:21:32.740 messed up the heist. They really did. Yeah. Because 00:21:32.740 --> 00:21:35.440 they didn't steer the complete formula. They 00:21:35.440 --> 00:21:38.019 got the base recipe perfectly, the exact ratios 00:21:38.019 --> 00:21:42.519 of water, the solvents, the primary salts. But 00:21:42.519 --> 00:21:45.299 they seemingly missed the supplemental page. 00:21:45.950 --> 00:21:48.410 with the proprietary stabilizing additives. Or 00:21:48.410 --> 00:21:50.650 they just completely misunderstood how the additives 00:21:50.650 --> 00:21:52.650 interacted. Right. It's like breaking into a 00:21:52.650 --> 00:21:54.569 vault to steal a famous recipe for a loaf of 00:21:54.569 --> 00:21:56.190 bread, but you completely forget to write down 00:21:56.190 --> 00:21:58.799 the yeast. Or honestly, in this context, it's 00:21:58.799 --> 00:22:00.799 more like stealing the blueprints for a bomb 00:22:00.799 --> 00:22:02.900 and completely forgetting to build the timer 00:22:02.900 --> 00:22:05.339 mechanism. That is unfortunately a much better 00:22:05.339 --> 00:22:08.440 analogy because this competitor took this flawed, 00:22:08.640 --> 00:22:11.079 deeply unstable recipe and they just started 00:22:11.079 --> 00:22:13.859 manufacturing these defective capacitors by the 00:22:13.859 --> 00:22:16.039 absolute millions. And because they were stealing 00:22:16.039 --> 00:22:18.660 the tech, they undercut everyone on price. They 00:22:18.660 --> 00:22:21.380 sold them to major assembly houses at a massive 00:22:21.380 --> 00:22:23.940 discount. And those assembly houses were building 00:22:23.940 --> 00:22:26.019 boards for everybody. They put these ticking 00:22:26.019 --> 00:22:28.779 time. bombs into everything. Bell Optiplex desktops. 00:22:28.880 --> 00:22:33.559 Apple iMac G5s. HP workstations. networking gear, 00:22:33.720 --> 00:22:36.700 everything. And the truly insidious part of this 00:22:36.700 --> 00:22:38.920 whole plague is that the capacitors actually 00:22:38.920 --> 00:22:40.900 work perfectly at first. They passed all the 00:22:40.900 --> 00:22:42.859 factory testing. They passed the quality control 00:22:42.859 --> 00:22:45.779 checks with flying colors because the pure water 00:22:45.779 --> 00:22:48.579 corrosion reaction is relatively slow. It takes 00:22:48.579 --> 00:22:50.660 months of heat and chemistry to build up the 00:22:50.660 --> 00:22:53.099 pressure. So the computers all ship out to customers. 00:22:53.220 --> 00:22:55.440 They get set up on office desks around the world. 00:22:55.819 --> 00:22:58.599 And inside every single one of those little cans, 00:22:58.859 --> 00:23:02.599 the highly unstable water is just slowly... relentlessly 00:23:02.599 --> 00:23:05.420 eating away at the aluminum foil. We can actually 00:23:05.420 --> 00:23:07.220 do a quick chemical equation check here from 00:23:07.220 --> 00:23:12.160 the sources. It's 2Al plus 6H2O yields 2AlOH3 00:23:12.160 --> 00:23:15.119 plus 3H2. Right. So aluminum metal plus the water 00:23:15.119 --> 00:23:17.119 equals aluminum hydroxide plus that dangerous 00:23:17.119 --> 00:23:19.220 hydrogen gas we talked about earlier. Exactly. 00:23:19.220 --> 00:23:21.980 The pressure slowly built up and up and up. And 00:23:21.980 --> 00:23:24.440 then right around 12 to 18 months after these 00:23:24.440 --> 00:23:27.039 machines were sold pop. Yeah, I distinctly remember 00:23:27.039 --> 00:23:29.319 working in an IT support job around that exact 00:23:29.319 --> 00:23:31.740 time. You would open up a desktop computer case 00:23:31.740 --> 00:23:33.859 that had just stopped booting, and it genuinely 00:23:33.859 --> 00:23:36.240 looked like a microscopic war zone in there. 00:23:36.400 --> 00:23:39.500 It was a mess. The shiny silver tops of the capacitors 00:23:39.500 --> 00:23:42.779 were all domed and bulged outward, or they had 00:23:42.779 --> 00:23:46.039 literally split wide open and spewed this completely 00:23:46.039 --> 00:23:49.200 dried up brown, crusty chemical stuff everywhere. 00:23:49.599 --> 00:23:52.380 And that brown crust was the conductive electrolyte 00:23:52.380 --> 00:23:55.660 itself. It physically sprayed out onto the sensitive 00:23:55.660 --> 00:23:58.740 motherboard traces, totally shorting out data 00:23:58.740 --> 00:24:01.420 lines, killing the CPU, killing the memory. It 00:24:01.420 --> 00:24:04.579 was devastating. It cost the PC industry billions 00:24:04.579 --> 00:24:07.259 of dollars. Dell alone had to take a massive 00:24:07.259 --> 00:24:09.720 financial gutter. Hit some estimates, put it 00:24:09.720 --> 00:24:12.019 in the hundreds of millions, just replacing thousands 00:24:12.019 --> 00:24:14.339 and thousands of ruined motherboards under warranty. 00:24:14.640 --> 00:24:17.039 All because one person stole Incomical Formula 00:24:17.039 --> 00:24:19.900 they simply didn't understand. It is the ultimate 00:24:19.900 --> 00:24:21.940 lesson in the difference between know -how and 00:24:21.940 --> 00:24:24.319 know -what. They knew what physical chemicals 00:24:24.319 --> 00:24:27.000 were in the jar, but they had absolutely zero 00:24:27.000 --> 00:24:29.299 idea how they worked together dynamically. That 00:24:29.299 --> 00:24:31.180 really drives home the reality that these little 00:24:31.180 --> 00:24:33.240 cylinders are not just static buckets holding 00:24:33.240 --> 00:24:35.460 electricity. They are highly active little chemical 00:24:35.460 --> 00:24:37.799 reactors. They're entirely alive in a chemical 00:24:37.799 --> 00:24:40.130 sense. So let's pivot and look at the actual 00:24:40.130 --> 00:24:42.750 parameters, the personality traits of these capacitors, 00:24:42.849 --> 00:24:45.029 because really knowing these traits helps us 00:24:45.029 --> 00:24:48.230 fully understand why they eventually fail, even 00:24:48.230 --> 00:24:50.450 when there isn't some botched stolen formula 00:24:50.450 --> 00:24:52.970 involved. Let's do it. Let's start with polarity. 00:24:53.029 --> 00:24:55.269 We mentioned the valve metal concept earlier, 00:24:55.470 --> 00:24:58.890 the one -way street. Yes. Aluminum electrolytic 00:24:58.890 --> 00:25:01.630 capacitors are strictly polarized. They have 00:25:01.630 --> 00:25:04.809 a very distinct physical positive and a distinct 00:25:04.809 --> 00:25:08.230 negative lead. You absolutely cannot mix them 00:25:08.230 --> 00:25:10.410 up when you install them. What actually happens 00:25:10.410 --> 00:25:12.869 physically if I just grab a soldering iron and 00:25:12.869 --> 00:25:15.009 put one in backward on a circuit board? It is 00:25:15.009 --> 00:25:17.609 catastrophic almost immediately. Remember that 00:25:17.609 --> 00:25:20.710 thin oxide layer, the insulator wall, is originally 00:25:20.710 --> 00:25:23.650 formed at the factory by applying a strong positive 00:25:23.650 --> 00:25:27.559 voltage to the anode foil. It is an active electrochemical 00:25:27.559 --> 00:25:30.160 growth process. Right. If you apply a massive 00:25:30.160 --> 00:25:32.420 reverse voltage, meaning you push a heavy negative 00:25:32.420 --> 00:25:35.279 charge into that positive anode, you actively 00:25:35.279 --> 00:25:38.019 reverse that entire chemical process. You start 00:25:38.019 --> 00:25:39.980 aggressively, actively dissolving that oxide 00:25:39.980 --> 00:25:42.420 layer right back into the liquid. You are literally 00:25:42.420 --> 00:25:44.900 stripping the insulation right off the wires 00:25:44.900 --> 00:25:47.660 while the power is on. Exactly. You are stripping 00:25:47.660 --> 00:25:50.559 the insulation. And at the exact same time, the 00:25:50.559 --> 00:25:52.900 chemistry starts desperately trying to form a 00:25:52.900 --> 00:25:56.059 brand new oxide layer on the... actual cathode 00:25:56.059 --> 00:25:58.859 foil which was never designed or etched to handle 00:25:58.859 --> 00:26:01.819 it. So the electrical resistance just plummets 00:26:01.819 --> 00:26:04.720 to near zero. It drops instantly. The current 00:26:04.720 --> 00:26:08.740 wildly spikes and all that energy dumps straight 00:26:08.740 --> 00:26:10.839 into the internal liquid. It hits the boiling 00:26:10.839 --> 00:26:12.779 point almost instantly. And the gas pressure. 00:26:13.000 --> 00:26:14.819 It goes completely off the charts in seconds. 00:26:15.369 --> 00:26:17.569 The aluminum can will physically violently explode. 00:26:17.750 --> 00:26:20.250 Which brings up a very noticeable visual feature. 00:26:20.410 --> 00:26:22.609 This is exactly why they have those score marks 00:26:22.609 --> 00:26:25.049 stamped into the top of the can, right? Like 00:26:25.049 --> 00:26:26.869 if you look closely at the top, there is always 00:26:26.869 --> 00:26:30.069 a little K shape or a Y shape physically dented 00:26:30.069 --> 00:26:32.849 into the shiny metal. The safety vent, yes. It 00:26:32.849 --> 00:26:35.150 is a highly engineered, controlled structural 00:26:35.150 --> 00:26:38.190 weakness. It's a blowout panel, basically. Exactly. 00:26:38.670 --> 00:26:41.690 The engineers designing these know perfectly 00:26:41.690 --> 00:26:44.609 well that if the internal pressure gets too high, 00:26:45.079 --> 00:26:47.619 The can will absolutely fail. They can't stop 00:26:47.619 --> 00:26:50.000 it. So they want to deliberately decide how it 00:26:50.000 --> 00:26:52.339 fails. Because the alternative is terrifying. 00:26:52.720 --> 00:26:54.880 Right. Because without that weakened vent score, 00:26:55.119 --> 00:26:58.140 the entire aluminum shell might just rip off 00:26:58.140 --> 00:27:00.059 the bottom seal and shoot straight up into the 00:27:00.059 --> 00:27:03.019 air like a bullet. It literally becomes a metal 00:27:03.019 --> 00:27:05.920 projectile inside your TV. Wow. But with the 00:27:05.920 --> 00:27:07.900 vent. With the vent. As the pressure builds, 00:27:08.099 --> 00:27:11.420 the metal just... bulges and then neatly tears 00:27:11.420 --> 00:27:14.279 open right along those stamped lines, it's safe. 00:27:14.339 --> 00:27:17.019 It is the absolute best way to visualize it. 00:27:17.279 --> 00:27:20.339 Think of an ideal, purely theoretical capacitor. 00:27:20.779 --> 00:27:24.180 It would store energy and release energy with 00:27:24.180 --> 00:27:26.779 absolutely zero loss. It would be a completely 00:27:26.779 --> 00:27:29.779 frictionless bucket of electricity. But the real 00:27:29.779 --> 00:27:32.799 world has friction. Right. In a real eCAP, the 00:27:32.799 --> 00:27:35.160 liquid electrolyte inherently has some electrical 00:27:35.160 --> 00:27:38.150 resistance. The paper spacer. blocking the path 00:27:38.150 --> 00:27:41.230 has resistance. Even the physical crimped connection 00:27:41.230 --> 00:27:43.630 between the internal foil and the external metal 00:27:43.630 --> 00:27:46.369 leg has resistance. And ESR is just the grand 00:27:46.369 --> 00:27:48.650 total of all those little obstacles added up. 00:27:48.730 --> 00:27:51.269 Yes. It is the sum total of friction. And basic 00:27:51.269 --> 00:27:53.569 physics tells us that whenever you push an electrical 00:27:53.569 --> 00:27:55.789 current through a resistance, you generate heat. 00:27:56.049 --> 00:27:58.710 Correct. And this directly leads us to the next 00:27:58.710 --> 00:28:01.839 massive parameter, the ripple current. This is 00:28:01.839 --> 00:28:04.519 the parameter that usually silently kills capacitors 00:28:04.519 --> 00:28:06.859 in modern power supplies over time. Because in 00:28:06.859 --> 00:28:09.240 a computer power supply, the capacitor isn't 00:28:09.240 --> 00:28:11.980 just sitting there full. It is constantly rapidly 00:28:11.980 --> 00:28:14.660 charging and discharging hundreds of thousands 00:28:14.660 --> 00:28:17.579 of times every single second just to smooth out 00:28:17.579 --> 00:28:19.700 the choppy voltage coming from the wall. It's 00:28:19.700 --> 00:28:22.059 working out constantly. It is a high -intensity 00:28:22.059 --> 00:28:25.000 interval workout that never stops. And all that 00:28:25.000 --> 00:28:27.599 moving charge is what we call ripple current. 00:28:28.619 --> 00:28:31.539 Every single time that current sloshes back and 00:28:31.539 --> 00:28:35.319 forth, it is physically pushing against the ESR 00:28:35.319 --> 00:28:38.279 friction, which generates internal heat. If the 00:28:38.279 --> 00:28:41.079 inherent ESR of the part is too high, or if the 00:28:41.079 --> 00:28:43.240 ripple current demand from the processor is too 00:28:43.240 --> 00:28:46.200 punishing, the capacitor simply heats up from 00:28:46.200 --> 00:28:48.599 the dead center inside out. And as we keep saying, 00:28:48.799 --> 00:28:51.220 heat is the absolute enemy of this component. 00:28:51.400 --> 00:28:53.440 Heat is the ultimate killer. We will look at 00:28:53.440 --> 00:28:55.440 the exact math of the Arrhenius equation in a 00:28:55.440 --> 00:28:58.190 minute, but the short version is. Heat physically 00:28:58.190 --> 00:29:00.690 makes the wet electrolyte evaporate faster. And 00:29:00.690 --> 00:29:03.170 as it turns to vapor and slowly escapes... There 00:29:03.170 --> 00:29:05.490 is physically less conductive liquid left inside 00:29:05.490 --> 00:29:07.630 to carry the electricity. So what happens to 00:29:07.630 --> 00:29:09.910 the friction? The resistance, the ESR, goes even 00:29:09.910 --> 00:29:12.809 higher. Exactly. The ESR goes up, which generates 00:29:12.809 --> 00:29:15.450 even more heat for the exact same amount of ripple 00:29:15.450 --> 00:29:17.769 current. Which immediately causes even more rapid 00:29:17.769 --> 00:29:20.609 evaporation. It is a totally inescapable thermal 00:29:20.609 --> 00:29:23.599 death spiral once it starts. Okay, there is one 00:29:23.599 --> 00:29:25.519 more parameter we need to cover before we move 00:29:25.519 --> 00:29:28.400 on to predicting failures. And honestly, this 00:29:28.400 --> 00:29:30.299 one sounds like a ghost story from the source 00:29:30.299 --> 00:29:34.319 material. Soakage. Or dielectric absorption. 00:29:34.759 --> 00:29:37.859 Oh yeah. It is a physical phenomenon that genuinely 00:29:37.859 --> 00:29:41.619 terrifies anyone who works on old high voltage 00:29:41.619 --> 00:29:45.279 tube amplifiers or massive industrial power grids. 00:29:45.359 --> 00:29:47.799 Let's lay out the scenario. Imagine you are repairing 00:29:47.799 --> 00:29:50.480 a massive amplifier. You have this big capacitor 00:29:50.480 --> 00:29:52.759 sitting there fully charged up to 400 volts. 00:29:52.940 --> 00:29:55.359 You do the safe thing. You turn off the wall 00:29:55.359 --> 00:29:58.180 power. You take a heavy screwdriver with a thick 00:29:58.180 --> 00:30:00.880 insulated handle, obviously, and you carefully 00:30:00.880 --> 00:30:03.079 short the two metal terminals together to completely 00:30:03.079 --> 00:30:05.099 discharge it. Right. You hear the big crack, 00:30:05.240 --> 00:30:07.819 the spark jumps, and the voltage drops instantly 00:30:07.819 --> 00:30:10.359 to zero. You think the bucket is completely empty 00:30:10.359 --> 00:30:12.789 and safe. It is empty. So you walk away to go 00:30:12.789 --> 00:30:14.369 have a sandwich. You come back an hour later. 00:30:14.730 --> 00:30:16.630 You reach down to unscrew the terminals. And 00:30:16.630 --> 00:30:18.289 the second you touch the bare metal, you get 00:30:18.289 --> 00:30:21.549 a massive, dangerous electric shock. Where did 00:30:21.549 --> 00:30:23.650 the voltage come from, the zombie voltage? Where 00:30:23.650 --> 00:30:25.910 was it physically hiding for an hour? It was 00:30:25.910 --> 00:30:29.109 hiding literally inside the molecular structure 00:30:29.109 --> 00:30:32.420 of the oxide dielectric. Really? Yes. When you 00:30:32.420 --> 00:30:35.000 pump 400 volts into a capacitor for a long time, 00:30:35.180 --> 00:30:37.640 you aren't just piling loose electrons onto a 00:30:37.640 --> 00:30:40.799 metal plate. That intense electric field physically 00:30:40.799 --> 00:30:43.859 stretches and aligns the actual molecules inside 00:30:43.859 --> 00:30:47.119 the rigid aluminum oxide layer. It polarizes 00:30:47.119 --> 00:30:49.680 them physically. Like bending a spring. Exactly. 00:30:50.019 --> 00:30:51.980 When you short the terminals with a screwdriver, 00:30:52.319 --> 00:30:54.359 all the loose surface electrons leave instantly. 00:30:54.440 --> 00:30:58.059 The voltage reads zero. But those stretched oxide 00:30:58.059 --> 00:31:00.819 molecules take a long time to relax back to their 00:31:00.819 --> 00:31:03.390 normal neutral state. And as they slowly relax, 00:31:03.529 --> 00:31:05.869 as they slowly untwist over the next hour, they 00:31:05.869 --> 00:31:07.990 physically release their stored mechanical energy 00:31:07.990 --> 00:31:10.309 back onto the metal plates as an electrical charge. 00:31:10.630 --> 00:31:13.230 The voltage literally magically creeps back up 00:31:13.230 --> 00:31:15.549 out of nowhere. That is just wild. It is like 00:31:15.549 --> 00:31:17.470 pushing down hard on a memory foam mattress. 00:31:17.730 --> 00:31:19.470 You take your hand away and it leaves an imprint, 00:31:19.609 --> 00:31:22.109 but then it very slowly, agonizingly, slowly 00:31:22.109 --> 00:31:25.329 pops back up to full size. That memory foam analogy 00:31:25.329 --> 00:31:28.390 is absolutely spot on. Engineers actually call 00:31:28.390 --> 00:31:31.619 it dielectric memory. And it is entirely why, 00:31:31.779 --> 00:31:35.400 if you ever look at incredibly large industrial 00:31:35.400 --> 00:31:38.480 capacitors shipping from the factory, they almost 00:31:38.480 --> 00:31:41.400 always arrive with a thick piece of copper wire 00:31:41.400 --> 00:31:44.380 firmly twisted across the two terminals, shorting 00:31:44.380 --> 00:31:47.380 them out. Ah. Because if they ship them in an 00:31:47.380 --> 00:31:49.880 open circuit, the tiny bits of static electricity 00:31:49.880 --> 00:31:52.200 from the factory or atmospheric changes could 00:31:52.200 --> 00:31:55.130 slowly build up via that soakage effect. Exactly. 00:31:55.130 --> 00:31:57.630 The capacitor could literally slowly charge itself 00:31:57.630 --> 00:32:00.390 up in the cardboard box and potentially lethally 00:32:00.390 --> 00:32:03.170 shock the guy unpacking it a month later. Okay. 00:32:03.529 --> 00:32:05.930 That is properly spooky. Let's move into the 00:32:05.930 --> 00:32:09.150 final major phase of this deep dive, failure 00:32:09.150 --> 00:32:12.069 and reliability. The inevitable end. Right. We 00:32:12.069 --> 00:32:14.130 know perfectly well that these ECAPs fail. We've 00:32:14.130 --> 00:32:15.829 talked about how they dry out and pop. But from 00:32:15.829 --> 00:32:18.269 an engineering standpoint, how do we actually 00:32:18.269 --> 00:32:20.430 predict when that is going to happen? The source 00:32:20.430 --> 00:32:22.869 material leans heavily on this concept of the 00:32:22.869 --> 00:32:25.410 bathtub curve. The bathtub curve is really the 00:32:25.410 --> 00:32:28.029 fundamental standard reliability model for almost 00:32:28.029 --> 00:32:31.109 all electronics. Imagine the physical cross -section 00:32:31.109 --> 00:32:34.549 shape of a bathtub graph. Okay, so high on the 00:32:34.549 --> 00:32:37.470 left dips down low and flat in the middle and 00:32:37.470 --> 00:32:39.890 then spikes back up high on the right side. Right, 00:32:39.950 --> 00:32:42.769 so that steep high wall on the far left side 00:32:42.769 --> 00:32:45.950 is what engineers call infant mortality. That 00:32:45.950 --> 00:32:49.559 is a grim industry term. It really is. But it 00:32:49.559 --> 00:32:51.839 describes failures that happen almost immediately 00:32:51.839 --> 00:32:54.900 right out of the factory box. These are sheer 00:32:54.900 --> 00:32:58.240 manufacturing defects. A completely bad rubber 00:32:58.240 --> 00:33:02.319 seal, a tiny scratch on the foil, a short circuit 00:33:02.319 --> 00:33:04.440 in the internal winding. Stuff that was just 00:33:04.440 --> 00:33:07.640 built wrong on a Tuesday. Exactly. And honestly, 00:33:07.759 --> 00:33:09.339 usually these are caught by the manufacturer 00:33:09.339 --> 00:33:11.900 during a process called burn -in before the part 00:33:11.900 --> 00:33:14.920 even ships. But a few always sneak through. Then 00:33:14.920 --> 00:33:16.859 you move past that and you hit the long, flat 00:33:16.859 --> 00:33:19.420 bottom of the tub. Right. the random failure 00:33:19.420 --> 00:33:21.940 phase. During this period the component is working 00:33:21.940 --> 00:33:24.559 exactly as intended. The statistical failure 00:33:24.559 --> 00:33:27.079 rate is incredibly extremely low. If something 00:33:27.079 --> 00:33:28.859 dies here it's usually just a totally random 00:33:28.859 --> 00:33:31.460 lightning strike or a power surge kind of event. 00:33:31.819 --> 00:33:34.319 And for a lot of purely solid state parts, like 00:33:34.319 --> 00:33:37.960 a simple carbon resistor or a basic silicon transistor, 00:33:38.240 --> 00:33:39.799 they just stay on that flat bottom practically 00:33:39.799 --> 00:33:42.299 forever, right? Pretty much, yes. They don't 00:33:42.299 --> 00:33:44.319 really have a steep wear -out wall. They just 00:33:44.319 --> 00:33:47.160 keep going until some random external event finally 00:33:47.160 --> 00:33:49.599 kills them 50 years later. But the electrolytic 00:33:49.599 --> 00:33:51.619 capacitor is entirely different. It hits the 00:33:51.619 --> 00:33:53.900 right side of the bathtub, the wear -out phase. 00:33:54.680 --> 00:33:57.500 Because of the wet liquid electrolyte, the ECAP 00:33:57.500 --> 00:34:01.849 actually has a rigidly defined... physical end 00:34:01.849 --> 00:34:05.349 of life. The great drying out. Yes. The rubber 00:34:05.349 --> 00:34:07.769 stopper sealing the bottom of the can is a very 00:34:07.769 --> 00:34:10.630 good polymer, but on a molecular level, it is 00:34:10.630 --> 00:34:13.929 not completely perfect. Over 10, 15, 20 years, 00:34:14.130 --> 00:34:17.230 the microscopic vapor in the electrolyte simply 00:34:17.230 --> 00:34:19.769 diffuses right through the solid rubber molecules. 00:34:20.300 --> 00:34:21.980 It just goes, walks right through the seal. It 00:34:21.980 --> 00:34:24.360 does. It leaves the can forever. The capacitance 00:34:24.360 --> 00:34:27.400 value drops off a cliff. The ESR skyrockets to 00:34:27.400 --> 00:34:29.519 massive levels and the circuit completely stops 00:34:29.519 --> 00:34:31.539 filtering electrical noise. And we can actually 00:34:31.539 --> 00:34:34.500 accurately predict exactly how fast that evaporation 00:34:34.500 --> 00:34:36.719 happens using the Arrhenius equation. Now, I'm 00:34:36.719 --> 00:34:38.760 not going to ask you to do complex calculus on 00:34:38.760 --> 00:34:40.559 audio, but I really want to understand the core 00:34:40.559 --> 00:34:42.659 concept here because this isn't just some loose 00:34:42.659 --> 00:34:45.760 rule of thumb. This is hard physics. It is fundamental 00:34:45.760 --> 00:34:49.099 baseline chemistry. A Swedish chemist named Svante 00:34:49.099 --> 00:34:52.119 Arrhenius figured this math out way back in the 00:34:52.119 --> 00:34:55.659 late 1800s. The equation simply describes exactly 00:34:55.659 --> 00:34:58.460 how chemical reaction rates change radically 00:34:58.460 --> 00:35:00.980 with temperature. Okay, so how does that apply 00:35:00.980 --> 00:35:03.900 to our wet sponge inside the can? Think about 00:35:03.900 --> 00:35:06.420 the actual individual molecules swimming around 00:35:06.420 --> 00:35:09.280 in the liquid electrolyte. Scientifically speaking, 00:35:09.579 --> 00:35:12.280 physical temperature is nothing more than a direct 00:35:12.280 --> 00:35:15.260 measure of kinetic energy. It measures exactly 00:35:15.260 --> 00:35:17.519 how fast those little molecules are vibrating 00:35:17.519 --> 00:35:20.039 and violently banging into each other. Right. 00:35:20.139 --> 00:35:22.579 Hotter molecules just dance much faster. And 00:35:22.579 --> 00:35:25.480 crucially, when they dance faster, they collide 00:35:25.480 --> 00:35:28.800 with much, much more force. Most chemical reactions, 00:35:28.900 --> 00:35:31.500 like the slow corrosive breakdown of the electrolyte 00:35:31.500 --> 00:35:34.000 or the evaporation process itself, require a 00:35:34.000 --> 00:35:36.260 very specific threshold amount of collision energy 00:35:36.260 --> 00:35:38.710 to actually happen. Chemists call that the activation 00:35:38.710 --> 00:35:40.929 energy. Like a toll booth you have to pass through. 00:35:41.050 --> 00:35:42.909 Exactly. When you raise the overall temperature 00:35:42.909 --> 00:35:45.150 of the computer case, you are giving billions 00:35:45.150 --> 00:35:47.949 of more molecules enough kinetic energy to smash 00:35:47.949 --> 00:35:50.469 through that toll booth and react. So even a 00:35:50.469 --> 00:35:53.150 fairly small bump in temperature can suddenly 00:35:53.150 --> 00:35:55.750 lead to a completely massive spike in the speed 00:35:55.750 --> 00:35:57.730 of the chemical reaction. It changes exponentially, 00:35:58.030 --> 00:36:00.750 yes. And this elegant bit of physical chemistry 00:36:00.750 --> 00:36:03.849 translates directly into what the engineering 00:36:03.849 --> 00:36:06.269 world calls the 10 -degree rule. This is the 00:36:06.269 --> 00:36:08.449 Golan rule of electronics right here. It is. 00:36:08.530 --> 00:36:11.670 The rule states very simply, for every single 00:36:11.670 --> 00:36:14.210 10 degrees Celsius, you drop the operating temperature 00:36:14.210 --> 00:36:17.309 of an electrolytic capacitor, you exactly double 00:36:17.309 --> 00:36:19.980 its expected physical lifespan. That is just 00:36:19.980 --> 00:36:23.219 a massive, massive lever to pull. It is the biggest 00:36:23.219 --> 00:36:25.539 lever you have. Let's actually do the rough math 00:36:25.539 --> 00:36:27.800 here to show how crazy it gets. Let's say you 00:36:27.800 --> 00:36:30.699 buy a very standard e -cap. The data sheet says 00:36:30.699 --> 00:36:33.820 it is rated for 2 ,000 hours of life at its maximum 00:36:33.820 --> 00:36:37.059 temperature of, say, 105 degrees Celsius. 2 ,000 00:36:37.059 --> 00:36:38.719 hours really does not sound like very much time. 00:36:38.800 --> 00:36:41.079 That is, what, like maybe 83 continuous days? 00:36:41.300 --> 00:36:43.059 Right, just under three months. If I bought a 00:36:43.059 --> 00:36:45.340 brand new router and it died completely in 83 00:36:45.340 --> 00:36:48.409 days, I would be absolutely furious. Well, of 00:36:48.409 --> 00:36:51.369 course. But the trick is that inside your router, 00:36:51.489 --> 00:36:54.610 sitting in your cool living room, it is nowhere 00:36:54.610 --> 00:36:58.570 near 105 degrees. Hopefully it is operating at 00:36:58.570 --> 00:37:01.489 a nice mild 45 degrees Celsius. Okay, so how 00:37:01.489 --> 00:37:03.590 many 10 degree steps is that? The difference 00:37:03.590 --> 00:37:07.809 between 105 and 45 is exactly 60 degrees. Yes, 00:37:07.809 --> 00:37:11.550 60 degrees. Yeah. Which is exactly six 10 degree 00:37:11.550 --> 00:37:13.849 intervals. Yes. So according to Arrhenius, we 00:37:13.849 --> 00:37:17.389 get to double our baseline. 2 ,000 hours, six 00:37:17.389 --> 00:37:19.809 separate times. Okay, let's double it. 2 ,000 00:37:19.809 --> 00:37:24.130 becomes 4 ,000. 4 becomes 8 ,000. 16, 32, 64. 00:37:24.329 --> 00:37:27.929 All the way up to 128 ,000 continuous hours. 00:37:28.190 --> 00:37:30.909 And if you divide 128 ,000 hours by the number 00:37:30.909 --> 00:37:33.869 of hours in a year? You get just about 14 .6. 00:37:34.030 --> 00:37:37.010 Let's round to 15 years. Exactly. About 15 years 00:37:37.010 --> 00:37:40.070 of continuous 24 -7 operation. That is truly 00:37:40.070 --> 00:37:42.250 incredible. So the literal difference between 00:37:42.250 --> 00:37:45.010 your expensive device dying in three measly months 00:37:45.010 --> 00:37:48.170 versus it happily running for 15 straight years 00:37:48.170 --> 00:37:51.050 is literally just keeping the room cool and ensuring 00:37:51.050 --> 00:37:53.849 airflow. That is exactly it. This is precisely 00:37:53.849 --> 00:37:57.769 why dust accumulation is such a massive silent 00:37:57.769 --> 00:37:59.849 killer of electronics. It acts like a sweater. 00:38:00.320 --> 00:38:03.059 It is a perfect insulating blanket. If your PC 00:38:03.059 --> 00:38:05.900 exhaust fans get heavily clogged with dust and 00:38:05.900 --> 00:38:09.659 dog hair, the internal ambient temperature easily 00:38:09.659 --> 00:38:13.139 slowly rises by maybe 20 degrees Celsius. Two 00:38:13.139 --> 00:38:16.059 10 -degree steps. Yes. You just quartered the 00:38:16.059 --> 00:38:18.460 total lifespan of your entire power supply. You 00:38:18.460 --> 00:38:20.960 dropped it from 15 years down to less than four 00:38:20.960 --> 00:38:23.360 years just because of some dust. So, honestly, 00:38:23.440 --> 00:38:25.619 the most practical takeaway here is... Clean 00:38:25.619 --> 00:38:27.980 your fans, people. Just buy a can of compressed 00:38:27.980 --> 00:38:30.679 air. Seriously, it saves hardware. So there's 00:38:30.679 --> 00:38:33.739 one more truly bizarre failure mode I want to 00:38:33.739 --> 00:38:36.940 touch on from the outline. storage rot. Ah, the 00:38:36.940 --> 00:38:39.239 shelf life problem. Right. This is definitely 00:38:39.239 --> 00:38:41.920 one for the antique radio restorers or the retro 00:38:41.920 --> 00:38:44.739 computing nerds listening. If I go into my grandparents' 00:38:45.000 --> 00:38:47.340 hot attic and I find a totally gorgeous, pristine 00:38:47.340 --> 00:38:50.579 vacuum tube radio from 1950, why shouldn't I 00:38:50.579 --> 00:38:52.320 just take it downstairs and immediately plug 00:38:52.320 --> 00:38:54.480 it into the wall socket to see if it works? Because 00:38:54.480 --> 00:38:56.840 the capacitors inside it have essentially been 00:38:56.840 --> 00:39:00.579 asleep for way too long. Yes. When a wet aluminum 00:39:00.579 --> 00:39:03.949 capacitor just sits on a shelf, Completely unpowered, 00:39:04.010 --> 00:39:06.989 with zero electrical volts pushing through it 00:39:06.989 --> 00:39:09.889 for several decades, the liquid electrolyte, 00:39:09.889 --> 00:39:12.409 which, remember, is a mildly acidic soup, just 00:39:12.409 --> 00:39:16.469 sits there and very, very slowly eats away at 00:39:16.469 --> 00:39:18.750 our precious aluminum oxide layer. The useful 00:39:18.750 --> 00:39:21.989 rust just dissolves away. It does. With no leakage 00:39:21.989 --> 00:39:24.190 current to trigger the self -healing medic process, 00:39:24.570 --> 00:39:26.849 the thin insulation wall just gets thinner and 00:39:26.849 --> 00:39:30.019 thinner and thinner until... It's basically bare 00:39:30.019 --> 00:39:32.079 metal again. The protective wall just crumbles 00:39:32.079 --> 00:39:35.099 into the moat. Exactly. So if you just enthusiastically 00:39:35.099 --> 00:39:38.079 plug that antique radio straight into modern 00:39:38.079 --> 00:39:40.880 wall power, you are instantly violently hitting 00:39:40.880 --> 00:39:43.500 that completely weakened microscopic wall with 00:39:43.500 --> 00:39:45.800 hundreds of volts instantly. It has zero chance 00:39:45.800 --> 00:39:48.400 of holding that pressure back. None. The internal 00:39:48.400 --> 00:39:50.739 leakage current will be massive. The electrolyte 00:39:50.739 --> 00:39:52.500 will hit a rolling boil in a matter of seconds, 00:39:52.639 --> 00:39:54.820 and those old metal cans will just dramatically 00:39:54.820 --> 00:39:57.519 explode right there on your workbench. Wow. Okay, 00:39:57.579 --> 00:40:00.079 so if I can't just plug... How do we safely wake 00:40:00.079 --> 00:40:02.360 the machine up? You have to perform a process 00:40:02.360 --> 00:40:06.139 called reforming. Honestly, it is very much like 00:40:06.139 --> 00:40:09.619 incredibly slow physical therapy for a totally 00:40:09.619 --> 00:40:12.000 atrophied muscle. How does that work? You hook 00:40:12.000 --> 00:40:14.000 the old equipment up to a specialized variable 00:40:14.000 --> 00:40:17.179 power supply, usually called a variac. You start 00:40:17.179 --> 00:40:20.300 at absolute zero volts. And then very slowly 00:40:20.300 --> 00:40:23.179 over the course of many, many hours, you dial 00:40:23.179 --> 00:40:25.679 the voltage up manually. Give it 10 volts. Then 00:40:25.679 --> 00:40:28.699 you wait. Then 20 volts. So you were just gently 00:40:28.699 --> 00:40:32.539 nudging the chemistry. Yes. This incredibly slow. 00:40:33.179 --> 00:40:36.280 Gentle, creeping voltage gives the dormant self 00:40:36.280 --> 00:40:38.539 -healing mechanism the actual time it needs to 00:40:38.539 --> 00:40:41.280 kick in and work. It slowly drives the oxygen 00:40:41.280 --> 00:40:44.039 out of the old liquid and forces it to painstakingly 00:40:44.039 --> 00:40:46.920 rebuild the oxide wall atom by atom before the 00:40:46.920 --> 00:40:48.760 internal heat can ever get out of control. You 00:40:48.760 --> 00:40:50.980 are literally nursing the chemistry back to health. 00:40:51.119 --> 00:40:53.340 You are. And honestly, if you do it patiently 00:40:53.340 --> 00:40:55.460 enough and the rubber seals haven't totally failed, 00:40:55.619 --> 00:40:58.460 a 50 -year -old capacitor can often wake up and 00:40:58.460 --> 00:41:00.719 work just absolutely fine. That is genuinely 00:41:00.719 --> 00:41:02.559 surprisingly touching for a hardware discussion. 00:41:02.760 --> 00:41:04.980 It really strongly treats the little electronic 00:41:04.980 --> 00:41:07.239 component like a living biological thing that 00:41:07.239 --> 00:41:09.420 just needs some tender care. It's an electrochemical 00:41:09.420 --> 00:41:12.119 ecosystem in a tiny jar. Okay, so bringing this 00:41:12.119 --> 00:41:15.369 all fully together. We have explored this completely 00:41:15.369 --> 00:41:19.469 humble, mundane component. It is a piece of metal 00:41:19.469 --> 00:41:22.510 etched like the streets of Manhattan. It is coated 00:41:22.510 --> 00:41:25.090 in a nanometer thin layer of engineered rust. 00:41:25.369 --> 00:41:28.989 It is filled with heavily modified car antifreeze 00:41:28.989 --> 00:41:31.630 and it continuously heals its own wounds until 00:41:31.630 --> 00:41:34.289 it finally completely dries up and dies. That 00:41:34.289 --> 00:41:36.929 is the perfect summary. So the obvious closing 00:41:36.929 --> 00:41:39.969 question is. Why are we still using millions 00:41:39.969 --> 00:41:41.730 of these things? Why haven't we just entirely 00:41:41.730 --> 00:41:43.849 moved on to something definitively solid state 00:41:43.849 --> 00:41:46.409 by now? Like the sources mentioned solid polymer 00:41:46.409 --> 00:41:49.170 capacitors or pure tantalum ones. Why stick with 00:41:49.170 --> 00:41:51.449 the wetsuit? As with almost everything in engineering, 00:41:51.570 --> 00:41:54.170 it always aggressively comes down to two inescapable 00:41:54.170 --> 00:41:57.869 things, energy density and sheer cost, specifically 00:41:57.869 --> 00:42:00.070 when dealing with very high voltages. So those 00:42:00.070 --> 00:42:02.789 newer techs just can't compete there. Not perfectly, 00:42:02.849 --> 00:42:05.880 no. Solid polymer capacitors are genuinely amazing. 00:42:06.019 --> 00:42:07.860 They are completely solid. They have ultra -low 00:42:07.860 --> 00:42:10.199 ESR, and they flat out do not dry out over time. 00:42:10.619 --> 00:42:13.119 But physically manufacturing them to reliably 00:42:13.119 --> 00:42:15.780 handle high voltages, anything significantly 00:42:15.780 --> 00:42:19.219 above, say, 35 volts, gets incredibly difficult, 00:42:19.420 --> 00:42:22.079 bulky, and insanely expensive. And what about 00:42:22.079 --> 00:42:24.500 tantalum? Tantalum capacitors are incredibly 00:42:24.500 --> 00:42:26.840 dense. You can store a massive amount of energy 00:42:26.840 --> 00:42:30.139 in a tiny, tiny grain. But tantalum is a rare, 00:42:30.159 --> 00:42:34.090 expensive conflict mineral. And crucially, well, 00:42:34.190 --> 00:42:36.630 if a solid tampon capacitor fails from a voltage 00:42:36.630 --> 00:42:38.869 spike, it doesn't just quietly vent some fishy 00:42:38.869 --> 00:42:41.369 steam. It tends to violently, aggressively catch 00:42:41.369 --> 00:42:43.050 fire and burn a hole through your motherboard. 00:42:43.369 --> 00:42:45.610 Okay. Yeah. Actively catching fire is definitely 00:42:45.610 --> 00:42:47.989 not ideal for a laptop charger. No, it's not. 00:42:48.110 --> 00:42:50.730 So despite all of its weird liquid quirks, the 00:42:50.730 --> 00:42:53.110 classic wet aluminum electrolytic capacitor is 00:42:53.110 --> 00:42:55.610 genuinely still the absolute undisputed king 00:42:55.610 --> 00:42:58.420 of the power grid. It is still the only available 00:42:58.420 --> 00:43:01.139 technology that can reliably hand you hundreds 00:43:01.139 --> 00:43:04.619 of microfarads of storage at 400 volts for just 00:43:04.619 --> 00:43:06.860 a few pennies per unit. It really is the ultimate 00:43:06.860 --> 00:43:09.099 triumph of perfectly good enough. Good enough 00:43:09.099 --> 00:43:12.340 is the true core driver of all industrial engineering. 00:43:12.539 --> 00:43:15.320 It really is. So as we fully wrap up this deep 00:43:15.320 --> 00:43:18.440 dive today, what is the biggest ultimate takeaway 00:43:18.440 --> 00:43:21.460 for you personally from all these sources? We 00:43:21.460 --> 00:43:23.719 have gone so deep into the microscopic chemistry 00:43:23.719 --> 00:43:27.480 and the Arrhenius physics. What stands out? For 00:43:27.480 --> 00:43:30.579 me, honestly, it is just about firmly breaking 00:43:30.579 --> 00:43:32.599 that psychological illusion of the perfectly 00:43:32.599 --> 00:43:35.840 clean black box. We as consumers tend to treat 00:43:35.840 --> 00:43:38.800 our sleek modern electronics as these sterile, 00:43:38.800 --> 00:43:41.980 perfect logic magic boxes. But when you really 00:43:41.980 --> 00:43:44.920 fundamentally understand the e -cap, you realize 00:43:44.920 --> 00:43:46.820 that sitting inside that beautifully machined, 00:43:46.820 --> 00:43:49.360 sleek aluminum laptop charger, that ultra -thin 00:43:49.360 --> 00:43:51.880 flat -screen TV, there is an incredibly messy, 00:43:52.079 --> 00:43:55.559 highly active, wet chemical reality. There are 00:43:55.559 --> 00:43:57.519 physical ions frantically swimming around in 00:43:57.519 --> 00:43:59.179 there. Chemical gases are occasionally being 00:43:59.179 --> 00:44:01.420 vented. Molecules are violently vibrating with 00:44:01.420 --> 00:44:03.679 heat. We haven't actually escaped the messy physical 00:44:03.679 --> 00:44:06.179 world of chemistry at all. We've just brilliantly 00:44:06.179 --> 00:44:09.420 engineered a way to cleanly package it into tiny 00:44:09.420 --> 00:44:11.599 little sealed metal cans. I really love that 00:44:11.599 --> 00:44:13.360 perspective. It honestly makes the high -tech 00:44:13.360 --> 00:44:15.659 stuff feel so much more incredibly grounded in 00:44:15.659 --> 00:44:18.059 reality. It does. And I actually want to leave 00:44:18.059 --> 00:44:20.539 the listener today with this final slightly provocative 00:44:20.539 --> 00:44:23.019 thought, a bit of a philosophical hardware paradox 00:44:23.019 --> 00:44:26.489 to just chew on. We spend so much of our daily 00:44:26.489 --> 00:44:29.550 modern lives completely striving for absolute 00:44:29.550 --> 00:44:33.150 digital perfection. We demand completely lossless 00:44:33.150 --> 00:44:36.789 high res audio. We want absolutely flawless 4K 00:44:36.789 --> 00:44:39.690 streaming video. We want mathematically perfect 00:44:39.690 --> 00:44:42.250 error corrected code. We fundamentally want our 00:44:42.250 --> 00:44:44.170 digital data and our memories to be immortal. 00:44:44.429 --> 00:44:46.690 We really do. We demand permanence. Right. And 00:44:46.690 --> 00:44:48.730 yet the very fundamental thing that actively 00:44:48.730 --> 00:44:51.429 physically powers this entire modern quest for 00:44:51.429 --> 00:44:54.150 digital immortality is a cheap little component. 00:44:54.280 --> 00:44:56.179 that fundamentally operates on the basic principle 00:44:56.179 --> 00:44:58.599 of inevitable decay. Entropy always eventually 00:44:58.599 --> 00:45:02.039 wins the long game. Your entire pristine digital 00:45:02.039 --> 00:45:06.760 life is quietly silently sustained by a very 00:45:06.760 --> 00:45:10.360 slowly evaporating pool of liquid. It is entirely 00:45:10.360 --> 00:45:13.139 governed by the cold math of the Arrhenius equation 00:45:13.139 --> 00:45:16.019 and the sheer physical integrity of a 20 -cent 00:45:16.019 --> 00:45:18.579 piece of stamped factory rubber. In a very weird 00:45:18.579 --> 00:45:21.559 way, every single massive multibillion -dollar 00:45:21.559 --> 00:45:24.539 server farm and every single supercomputer on 00:45:24.539 --> 00:45:27.800 Earth has a very real biological heart hidden 00:45:27.800 --> 00:45:30.400 inside it. It beats, it ages, it slowly degrades, 00:45:30.420 --> 00:45:32.719 and eventually it just stops. That is a deeply 00:45:32.719 --> 00:45:35.739 profound and admittedly slightly anxiety. inducing 00:45:35.739 --> 00:45:37.639 perspective to end on. I like to keep people 00:45:37.639 --> 00:45:39.679 thinking. So seriously, go grab a flashlight 00:45:39.679 --> 00:45:41.400 and check your really old networking electronics 00:45:41.400 --> 00:45:43.739 for bulging tops. Give a quick little nod of 00:45:43.739 --> 00:45:46.280 genuine respect to the chaotic 19th century chemistry 00:45:46.280 --> 00:45:48.340 silently, fiercely bubbling away right inside 00:45:48.340 --> 00:45:50.380 your power brick right now. And please, for the 00:45:50.380 --> 00:45:52.079 love of your hardware, keep those exhaust fans 00:45:52.079 --> 00:45:54.539 clean. Thanks for listening to this deep dive. 00:45:54.719 --> 00:45:55.780 See you on the next one.