Beating the Heat

Hello, and welcome to the macro fab engineering podcast. We are your hosts, Steven Gregg and Parker Dolman. This is episode 273.

So I finally finished that fan controller hack. That might be the fastest turnaround for a project on the podcast in a long, long time.

would remind us all what, what that project was.

So the I'm working on my Jeep, and I have an electric fan. And the controller that I bought for that logic fan, from a company called delta Pag. The fan controller, the override signal was an active low signal line and seven active high. Most signals in a car are active high. I don't know why it's an active low, but for this controller, but it is. So I needed to invert the signal. And we had a podcast episode like four or five episodes ago, where we talked about different ways you could do that. Most people in automotive, they just use like a big chunky relay, because everyone's just got automotive relays lying around. When you're doing this mic, kind of stuff like working on your car, you have automotive relays. So you'd use that the invert signal and the 12 volt high to be a ground active signal. And then I was like, Okay, what about if we opened it up, open up the controller instead, and like putting our own circuitry to like invert the signal on the low level side. thought about doing that way. And then there was also like the idea of like, hey, let's open it up and like rewrite the firmware to be like active low or active high mean. But upon opening it up, we found that the microcontroller is a at 51 core like Chinese. Like like you Google it to Google that part number, it's going to be like that podcast episode. And like the Chinese datasheet it's like the only two hits or get and so I just had to go the in like the middle route, right, which is basically cut the trace inside the the unit and then I use a MOSFET style inverter, which is uses you know, and channel MOSFET, pulldown resistor and then a n series resistor to kind of like protect the gate of the MOSFET a bit. And then I threw in a clamping diode at like clamps at like 16 volts, just in case. There's a spike. And so I hacked it. And if you're on the stream, because we're live streaming, you can see this Oh no.

Look at that. It's actually it's actually a pretty, pretty attractive little heck there. Yeah. Wait, did you get a PCB made for the heck?

I made a PCB for

oh, yeah, that's great. Yeah, it looks good.

That looks Yeah, that looks really good. Now that's okay. So there's there's different levels of hacks, right. So like, no, no, no, seriously, like this goes. This goes back to like our different types of soldering. There's like Manhattan's style. There's, oh, Thunder ohm style, and then there's got style. And I think so. You're still involves wires. So it's still a hack, but it's like a really, it's like a high level hack because you went through all the trouble of making a PCB.

My hacks have monocles.

Yeah, did you did you like super glue that down to the board?

Yep. That's super glue down. Nice. Okay, so the the PCB the the extra PCB, super glue down and then all the wires I bonded or hot glue down.

Nice. Yeah. And that was okay. So that is a mod that's intended to be permanent. Or,

yeah, it's 100%. It's gonna be like, I'm gonna put this, this backing on the controller, and I'm never going to open it up ever again. Nice. So if anyone out there has a, like, I mentioned zero. People who listen to podcasts have a delta pagg motor fan controller that they want to invert their active low signal for the auxilary override. If you do though, you can go download my PCB and, and make that

little board. It's up on GitHub right now. Yeah, you know, okay there. But it's really nice that you did that because there might be that one guy 20 years down the road, and he's searching on new Google or whatever they call it. Google new and you Google. Yeah. And you Google. Yeah. And, and he's got this vintage car that has that little box in it, and finds just that podcast. And there you go. And then has the Wayback Machine, your GitHub. Yeah.

It's amazing that or someone has another controller and has an active load and they need a little signal. So let's cute. I like it. Yeah, I'm just glad it worked. Like the first try. Like i i Seriously, like soldered it up right before this podcast. You use the what a BSS. 171 38 138. Okay. Yeah, just a that's actually the house part number at macro fab for n channel MOSFETs. For a SOT 23 package.

It's about as Jellybean as you get for my as as

Jelly Bean as you can get. I think the only big problem with it is I think it's only got like an 18 volt gate, the source.

But what's the signal levels? You're sending it?

Oh, it's five and 12? Not a big deal I'm talking about sees a spike on the automotive line? Ah, yes, right? Yeah, that guy. That's why I put that little I put little 100 Ohm resistor in series just to be like, hey, if there's any spikes, at least that resistor is going to take all the energy out of it. So you know, that's something and there's a diode there to protect

it too. Right? Yeah. For any MOSFET. That is in any of my circuits. I just by default, like it's almost like, it's almost like I'm placing one component. But anytime I place a MOSFET there's always a resistor that just goes there.

And like I don't even consider it.

It really depends on like, if you have a MOSFET, because I'm researching MOSFET drivers. And you don't want a resistor between the driver and the MOSFET, of course. But you're at that point you're,

well, you have a low impedance driver that can actually hit the capacitance of the gate.

Yeah, and a lot of more constant current driven to instead of a voltage driven line, you wouldn't want to put a resistor there.

Yeah, well think about MOSFETs is they end up they ended up taking a particular amount of charge to actually turn on. So it effectively appears like somewhat of a capacitor, so you have to have enough oomph to actually drive it. And if you're trying to do something really fast, and switch something, your driver has to be able to handle that, that drive current fast enough, but in your application, you're just switching something and you don't want to ever switch when it shouldn't switch. So that was that and

I just want to protect that gate as a, from spikes and other EMF sources and having a little snubber reasons through there to kind of eat up the energy. Yeah, is fine. Because that's the thing is, like, the main reason why you wouldn't want to put a resistor there is if you're are, if you need a switch that, that that MOSFET really quickly, on and off, like, you know, PWM circuit,

or like an HR controller, power supply or something, anything like that.

Exactly. Anything that because if because what you're doing is you're basically increasing the rise time of that MOSFET gate with that resistor, most time that that's if you're just going from zero or on off zero to one and your, your duty cycles, you know, 100%, right, because it's it's always on full zero full one. Yeah. And you you kind of want to do that to protect that back gate front end. But if you're actually PW I mean that you usually don't want to do that or use a really small value or you actually have a MOSFET driver. Interesting story about this is a story that Chris Kraft told me he was on the podcast a long long time ago. We should get him back on speaking of that, but um, he was he was the person who was really big into 3d printing and was kind of like did a lot of the hobby stuff back when he was first starting as well like the hobby 3d printing not when 3d printing first started but there was a problem with why I can't remember what controller there was a problem with but there it was like when they first started doing bed heating in hobby level 3d printers. So what for people who don't know what about 3d printing, you want your bed of basically the the material that you're printing on, you want that to be heated and so it prevents your your print from being warped, basically and helps adhesion And a bunch of other stuff. Anyways, the, the MOSFET kept burning up on the boards on these controller boards, and they could no one could figure out why. And, and people were using the MOSFET to drive another MOSFET board. That was like a bigger MOSFETs. It's kind of like that Arduino mentality of just like adding another or do we know. And basically, what ended up happening, though, was that MOSFET, the original MOSFET was plenty big enough to handle the current of the bed. But they were driving it at a frequency that that basically that MOSFET could never get into full saturation, it was always in that middle zone, where it was basically acting like an analog switch, right? That it's actually on. And so it's just superheating, because it's just it has resistance across have it across it. And they just dialed that frequency down the PWM and started working fine.

You know, so one other thing about MOSFETs, and that little resistor on there. So like the gate of a MOSFET is mega mega high impedance. I mean, it's, it's effectively glass right on the inside the gate of that. So you have a huge impedance. And so any parasitics that connect to that any any parasitic capacitance due to your PCB or wires, or inductance of the one that tends to aid in oscillation. And when a MOSFET oscillates, it doesn't oscillate at low frequency, it tends to oscillate really, really high. And in fact, a lot of times it oscillates well past what your circuit is wanting to operate it, and you don't even detect it, you don't even see it on your scopes, or anything like that. But that MOSFET sits there and just generates a bunch of heat. And I can actually damage itself, adding a little bit of resistance to the gait line. First of all, it basically cuts your trace width, because if you place that resistor very close to the pin, you're effectively swapping out the impedance. So no matter what with any MOSFET unless you're instructed not to buy like a driver IC or something like that, it's a good idea to just put some impedance in line with the gate of a MOSFET. Even if it's just a few ohms, then yeah, I think rule of thumb is just start with 100 to 1k. Somewhere in that range. And if if that's if that's making your circuit run too slow, then maybe you have the wrong MOSFET chosen, you know

are they getting that's what we did. That was the only thing we simulated on pin guitar. Was we is actually I went I took a week of vacation off. This is a couple years ago, and then stayed in Stephens basement and design the Penetang. Board.

Yeah, hey, that's a great, that's honestly, that's a great way to do design where you take a vacation, and that's like it, you're away from everything else. And like us, us focus, drank a lot of beer, worked on the Penetang board. And then I'm like, Huh, I

wonder. Like the only the only thing was I was using a completely new MOSFET basically, that had completely different gates, OR gate characteristics. And I was using a completely different kind of driver circuit. And I'm like, Hey, let's simulate this. And we simulated it. And it was like, Yeah, that looks fine.

Phil, actually found the model for it and imported it.

Yeah, yeah. Everything we were able to import and had proper models for Right, right. No,

although the more and more that I simulate things, the less and less I rely on simulation. And well, in other words, like I use simulation to simulate concepts and a lot less simulate, like, how is this going to happen on this edge case? Like the simulation is not going to work for that?

Yeah. But yeah, that's what I want to know is like, Okay, if I set it up with 100 Ohm resistor, or like, we actually we calculated what that resistor ideally should be. And it was actually really close to 100. I think it'd been like, it was like 80 Something ohms we're like, yeah, 100 is fine.

Well, you were trying to limit current output, because your driver chip had a maximum current output.

We weren't driving them from a driver chip. We're driving them from a sudden four hc 595 Shift Register, which has 20 milliamps max output. And so yeah, we put a basically calculate what the instantaneous, you know, turn on would current would have been, and basically then, because like we first we did zero, and it was like, it goes to infinity, right? Current goes to infinity, right? Well, habitable in real life. It's not infinity, right? But it's a lot and that could damage the drivers off that shift register over time. And so we just Kept basically increasing the resistance value until the peak was under 20 milliamps. And once we hit that, it was like, Okay, that was good. I think it was like 80 Something we got. And then we're like, okay, let's just use 100, because I'm using 100 elsewhere on the board,

right? It's, it's, you sacrifice a little bit of speed for some current limiting. And it's, and it's not like traditional current limiting its current limiting, because like, the gate has to be filled with CHARGE, basically. So you're just, you're making sure that it wouldn't ever go above 20 milliamps out of the pin. Because of a drive signal.

Yeah, because of the tribe. Yep. Yeah. And speaking of the Minotaur, it is now in production.

Finally, Oh,

congratulations that probably four years later, finally production?

And are you able to divulge what the production quantity is?

Over 250 units? Wow, nice. Yeah. So now now I got to the other side, which is like putting together a website with all the documentation and marketing. Oh, I'm not really. I'm marketing to OEM, so it's not really too big. It's not like I like marketing, like hobbyist and stuff. So it's a little bit easier because there's only like, I can count on my hand like how many pinball manufacturers there are still in America?

Probably in the world to

probably Yeah, probably.

So if a hobbyist did one, one.

Yeah, I will sell them. Yes, yes. I'm just not going to market to hobbyists, or anything like that

direct. The main goal isn't to sell them in singles. But you will. Somebody asked for one?

Yeah, yes. Oh, yes. Because because how do you design if you're going to go into production with a pinball machine? You need one just at least, you know, start testing stuff and putting together your first machine.

Okay, I got a question. That's, that's a total tangent, but it's still some somewhat related, I guess. Okay, what is the future of pinball? In other words, like, is pinball seeing like technological advances? Or is it all just the same stuff? rehash the different question? Yeah, like, is it just the same stuff with different mechs.

So wind back the time to the year 1999. Williams and valley. Pinball division is not doing so hot in the late 90s. It actually the pinball machines they were designing and making are some of the best that ever made like the themes, the gameplay, the quality of the pinball machines were like some of the best. But pinball was on the way out there just weren't making a lot of games. Just the whole coin op scene was changing. And so Williams Valley, went to the pinball division, and said, We need something different to turn this department around. And it came up with the pinball 2000 platform. And so I don't know if you've ever seen it, but the like revenge is revenge from the attack from Mars attack. Mars is like the wounds of like, regular people watching but I think it's like Revenge of the attack from Mars or something like that. But, and then they made Star Wars Episode One, because that had just come out. And then there was a unreleased game with like, wizards or something like that. But what they did was they did holograms on the pinball on the playfield. And how they did that was above the cabinets. So the cabinet instead of just having a typical, like, 90 degree head box on the back of the head box would back up, and then over the playfield. And there was a CRT monitor in there, that would display downwards. And then the glass was a special glass with a special coating on it, that would reflect what was on the CRT. And so with some, you know, math by stretching and skewing the image, they can make it look like there's a hologram on the playfield. And so you could do crazy stuff like you could change what targets look like and what the playfield toys look like. And the byway the Star Wars one looks really awesome. I don't think they play really well. They definitely look really cool. And they actually sold really well. And then Williams Valley just pulled the plug anyways, all of it. Basically, I think the I think the the brass of Williams Valley, we're going to pull the plug anyways. Unless it was like amazing, like so that was like kind of like a hurrah at the very end. Yeah, it's kind of like a Hurrah. So that was then no one pushed that platform really any further. Because we're still doing traditional demo machine still. In terms of tech, I don't know if there's been a couple attempts of like pushing like the tech and pinball. Like, we now have full color, HD, high, high HD graphics on on displays. But you still have companies that can also come out with like dot matrix stuff. But it's on like an LCD. Most of the advances are in lowering the cost. Not in like, the making the experience more high tech, because I don't think most pinball players are looking for that experience. They're not looking for the like, like, there's VR pinball. And there's like Virtual Pinball. And you can do technically anything in that space. And but people are still gravitating towards, they want something like physical pinball machine. So I don't know if there's going to be any crazy advances in terms of like, and that's space, because I don't think people are really looking for that.

Well, there's got to be machines out there. Like this shows my ignorance, they probably been around for a while. But there's got to be machines out there where like, it's just a cabinet with a big flat display in it. And they just loads up a game. Yeah, yeah. That Virtual Pinball. Oh, okay. Yeah, yep. Yeah. And it's got it looks like just a normal pinball cabinet. And but it's got a screen instead. And but it's not so cut chunk a chunk. Like I think there's there's a magic behind that for sure.

Yeah. And it's actually even though this is how crazy the Virtual Pinball stuff, it's actually really nice. Like, you can kind of trick your brain into thinking it's a real pinball machine, because they have head tracking now. And so it'll track your head, and then we'll move the playfield. And so it looks like there's an actual demo. It's like, hey, there has depth to it. Yeah, it's actually amazing that tech. But again, like, people, I know, there's some people that really like it. So I'm not going to. And I've actually played it quite a bit. It's just for me, it's not for me. I want physical playfield. And I want a steel ball. And yeah. So as far as for like tech advancement, most of that, at least in the physical pinball world is mostly in making stuff less expensive, and more high fidelity in terms of music and graphics on the screen.

Yeah. Because, because it seems like, okay, the tech behind what's actually physically happening on the playfield. I'm sure there's new ideas that can be spun up. But at the end of the day, it's like blinking light, it's like ramp, it's like door, it's like things that hit the ball. Like, those are all like, simple concepts that there's there's not a whole lot, specifically New there. There's just different flavors of it. Right?

There are, I mean, most games have different mechs in them, too. So that's where the innovation comes in is like, moving the ball around different way a different different kind of mechanism that's like custom to that machine. It's not a new moslem, it's not just a skin over it, either. It's an actual, it's a different, you know, mech altogether.

You know, maybe I've seen this, or maybe I've heard of it. But what comes to mind, in fact, I think maybe you told me about this. But regardless, the idea that the playfield could be on like a cylinder that turns such that, or think of think of maybe easier, maybe think of it on like a prism. Yes, it has three play fields. And you can play on one, and then it rotates, and you have a whole brand new play field on the inside. And then it could rotate again. And you'd have to like access different things, and it would spin and turn into a completely different field. That would be tight.

I don't know if anyone's actually prototype before that. It just only thing that comes to my mind is how freaking heavy that would be.

It would be heavy. And it would just be three. Like why would you have that if you could just have three pinball machines next to each other. Right. Oh, you're talking about? Oh, like inside the cabinet? The entire playfield it's a different game. Yeah. So

there's a couple No, I'm

not talking about a different game I'm talking about you have to play one and unlock the next one. Oh, go into it. So

there's a there is a pinball machine called. Man. This is amazing. But this podcast is this is not even our main topic.

Oh, not at all.

I thought it was a tangent. Yeah. But um, so going back to the tentage is there's a pinball machine called Bright a pin bots. As of course there's a there's a there's a machine called pin bot, which is like a robot. And then there's me she got bright pin bots. And on that, and so it's like a sequel I guess. In on that machine, it's got a a upper playfield that does exactly what you're talking about. Oh, it'll rotate. It has different faces of the of the bride. Maybe I was thinking that because you told me about that at

one point in time. Yeah, it could have been. What about a cylinder that actively rotates as you play? So if you shoot like diagonally, the whole field rotates with the speed of the ball.

Yeah, there was a, there was a, I can't remember the pinball machine. But it's got like, it's similar concept you think about but like inverted. So instead of like a flat playfield. It's like, like a molded injection molded, clear plastic, like, Valley kind of thing. It's not fun to play. It just so it just feels so unpredictable when you hit the ball.

Okay, how about this? This is this is another good one. You can't

remember one is that it? DJ? Was it? I think that's what it is.

Okay, how about this, it's a conveyor belt, such that as you shoot the ball and it goes towards the top that conveyor moves down. So you can like scroll towards like higher levels. But then if the ball falls, it scrolls back down conveyor belt style.

And there was another one. Now we're just going down the list of really weird, like non conventional playfields. Pinball circus, I think it's what's called and it's like a vertical pinball machine.

Oh, like Plinko.

Kind of, but you but Plinko goes from the top down, and you can't really interact with it. This is like you're shooting upwards and like going through like mechs going upwards. But like, there's a traditional playfield like normally to it's really weird. I've actually never played rivaling seen pictures.

Super tangent, we're going on another tangent here. There's some really great YouTube videos about the math behind Plinko. And the fact that every space on Plinko is a if you if you drop enough pucks it becomes a normal distribution Gaussian curve. It Yeah, Gaussian curve. It's a normal distribution. So if you ever go play Plinko, anywhere, and you want to hit a particular number, the best possible place to drop your puck is directly over that number, because it has the highest. What's it called? Probability of dropping in that?

So because that's because when it hits a pin, in theory, it's 5050. Whatever way it drops, Caribbean, the next pins 5050 Whether way? Yeah, I see. So if

you look at the very, like, if you put it, if you look at the very edges of distribution, it would have to be 5050 5050, like all the way down to reach the edges, but it has the highest probability of constantly being basically zigzagging back and forth to the one you want. So if there's like 1000 bucks right in the middle, put it right in the middle and has a higher likelihood you there.

So DJ 02 7x and chat says, Yeah, orbiter one has the concaved playfield with spinny pop, like it's got like an injection molded like plastic basin almost, and widely regarded as the worst pin ever, is not a lot of fun to play, because it just feels so random. And that's like the exact opposite of what you want. You want your player to feel like they are in control of, of the ball.

You know? Okay, I think that's, I'm looking at it right now. It looks like the playfield is melted. Like it looks like because I mean, it just has like, I don't know. Yeah, so it's wavy, wavy, almost like a kind of like a minefield, you know, that's been exploded. I don't know.

That's how I view it as like a Martian landscape. I think that's actually what's supposed to look like

got it. You know, and so the whole thing I was going out there for a second was that me being just like the most pinball I've ever played is Parker's to games.

Got a very, that's a very interesting selection there too. Because,

one that's super hard and one that's like at least and has. What? It's nice to newbies.

Yeah, but like both of them are regarded as some of the best shooting games of all time. Okay, the space shuttle in Congo, right? Yeah, so space shuttle, which came out at four. I think it's like the number one sold pinball machine as well. Like they made more special rules than like any other place machine ever. Don't quote me on that, but I'm pretty sure that's correct. And then Congo, which is a be game because no one really wants it because what sells pinball machines is theme theme one and

the movie was terrible. Yeah, and

the movie what everyone thought that was going to be the big New game because who wrote it? The people who are the person who's addressed the park, I can't remember his name. But Congo was going to be the next big thing. And that didn't happen. No. So that played that game, most of them, most Congos got turned into other games. But it's a really, and that was from the era of like, Medieval Madness and attack from Mars. From Williams, and so it's actually a really well shooting game, it's put together well, the ruleset is a little basic, because basically what happened, I think is they found that the movie wasn't going to be good. And so they kind of just stopped working on the code. There's not a lot of depth to the code, but it shoots really well.

And that's the same part. And that's fun. Right? Yeah,

it's a lot of fun. Because the shots are, there's some challenging shots, some easy shots, but when you hit the shots, it feels good. And it makes you feel like you're in control of the game.

So So, okay, again, that is that shoots? Well, in my opinion, seems like it would be a great party game where you get a few people together, you get some beers, and if you play against each other, but a game that's really deep, and has a lot of extension to the code, seems like it would be really great for one person to play and like trying to surmount everything.

Correct cracker? Great. Yeah. Yeah. So you see, the big thing nowadays, because especially since bar, like bars don't really have pinball machines or anything. I mean, there's some bars that do. Yes, but most bars don't have pinball machines, because no one cares anymore. And so most games now are really deep in code. Because the people who are buying them are putting them in their man cave. And they typically don't have, you know, 40 people showing up to a party. It's just them in their mancave. Drinking. Global for a bit. Yeah, playing a game.

Well, okay, and so sorry. Back to the reason I even brought that up. The whole orbiter one game feeling random. Pinball, to me, in a lot of ways feels random, because I don't have much experience with it. So like, my, whenever I play it, the whole goal is to just stay alive.

So good. Static, though, but I think that's everyone's tactic in it. First, go to a machine I've never played before. That's my tactic is just stay alive. Yeah. So yeah, space shuttle is a really fun game. It's just brutal, because that's from that era. It's tough. It's fun, that era of pinball, where, like, people were pumping coins into into machines. And so the less the ball time was, the more quarters you get. So that game eats your would eat your quarters, whereas Congo is from the era of like, they want people to feel good when you're playing it. And so ball times a little bit longer is easier shots. It doesn't drain as harsh. Yeah. Cago is a very good introductory game, where a space shuttle is like, you know, playing on on ultra nightmare.

Is there a doom pinball machine?

I don't think

so. Oh, man, that would be legit.

I don't think there is one but like original Doom. Oh, like a bit. Yeah. So anyways, back onto penetrator. So, you know, we've been talking about these supply chain issues. I have one component that has supply chain issues, and it's not what you would think. Okay, it is the ASC locking connector I'm using for the RGB lights. No, really? Yeah, yeah. Everyone's buying jst Oh, I bought the world supplier this part right now. Did you? Wow. So that Yeah. So I have like the world supply I still don't have I have enough for like the bill dependent tours. Don't have enough to build all the RGB lights yet. I have enough to get started. But, you know,

what are you gonna do? I'm working with some some clients right now that are a little bit new to contract manufacturing. They've done their first production run. And it went successful and everything was good. We had a post mortem. And we discussed

not talk boards, modem, whatever. We after the after the first run, we discussed everything. And now they're ready to go for round number two. And they're running into parts shortages. And they're like, what the first run is, what well, it was like, well, things are different now. Welcome to production. Welcome to Manufacturing, electronics, manufacturing anything really well. Yeah. Yeah, it's it's always fun. And it's always easy. That's what I tell them.

So I'm not too worried about it, mainly because I have enough to get going. And the lead times don't look that bad. It's only a couple months. But yeah, I basically bought everything I could find of this part.

Just out of curiosity, are we talking about four digits? five digits? Of what? Quantity of those connectors?

I bought four digits worth. Okay, cool. So, I need I need a six digits worth.

So you got some?

No, no. Is that right? I need five digits worth. Okay. I got about I got about a little over a third of what I need. Okay,

so it's not, I mean, you're, the demand for them isn't particularly high at the moment.

I think what it is, is, it's the locking style, which actually not a lot of people use, most people just use the kind of like, friction fit style. I actually don't like the friction fit, because one, it takes a lot of force to remove those. And we're using them is on all the RGB lights, which are only, you know, attached with actually have one right here. They're only attached with one screw. And so if you're yanking on that cable, you don't want to dive in, you're just gonna flex this, this knot at this thing. And so you had the locking. And so when you move the tab, it just freely pops out. Yeah. So there's it's a, there's low insertion force, basically. So I don't think a lot of them because I can just go and buy the non locking right now. Like, there's like, there's six digits worth those in stock? Yeah. I don't think they just don't make a lot of these. And so and then I stroll up and like, I need, you know, 25,000 of these. And everyone's just like, oh,

yeah, yeah.

They're probably gonna have to, they see a bot. Oh, what they had. And they're spooling up a lot, not a line, but spooling up a run of them eventually.

Well, yeah, you'll buy out whatever that Ron is.

Probably. Yeah.

So we're 39 minutes into this episode. We're getting into, I guess, what you could potentially call main topic. I don't know if we really have

those. Now, this is we talked about pinball was the main topic. Yeah,

effectively by this point? Well, and I. So I want to talk about thermal management a little bit, mainly because I've run into that this week. And I was like, oh, that's kind of cool. That would be kind of fun. And I don't think we've talked about thermal management a whole lot on the podcast,

not a ton. I mean,

to some degree, actually, we've probably talked about it in terms of the SSPs. A bit, right, because that's a monster thermal management.

Well, we did a little bit and we've talked about using, like, through hole vias and solder to kind of channel heat away from your components. Yeah, but yeah, not not, we haven't had a in depth discussion, I

guess, is good way. Well, and this isn't necessarily going to be an in depth discussion. But this will be some some things that I'm doing right now that are related to it effectively. So if you're if you're part of our Slack channel, and you look in the the MEP channel in there, you can see I've been posting a bunch of pictures of a project I'm working on, you know, which is funny, because let me let me tangent just for a quick second, Parker said earlier, you know, like, that little box that you worked on, was one of the fastest projects you've ever cranked out did on the podcast. I'm literally working on the longest project that I've ever had this project. People are gonna laugh at this. You know, the rule of thumb in general with like, electrolytic capacitors is if they're over a decade old, then you should start to consider replacing them. I have a project that is so old that before I even finished it, I've decided to recap it. Except for over a decade, swear to God, I started this project and you haven't heard them up yet? Yeah, yeah. Haven't even fired them up. And I just pulled them off and replaced all of them.

This is also why you don't buy electrolytic capacitors though. From like surplus stores. Oh, yeah. You have no clue. They do have date codes on them.

I think last time I went to one, like the date code was like from like the 80s. Yeah. So I actually in parroted one time 80,000 electrolytic capacitors, and I kept maybe like a few 100 of them. And I started contacting places around town. Like, would you buy these from me? And everywhere I contacted was like, No. And I was like, what if I give them to you? And they were like, No, I don't want that. Because because they're electrolytic capacitors, they can't do anything. So yeah, it was a waste. Yeah, it was a bad situation. So I had 80,000 capacitors that I had to do something with, I don't even remember what I did with them. Okay, regardless, multiple tangents back. So I've got this project I'm working on that I didn't design. Well, I've designed parts of the project. But I didn't design the whole thing. I purchased PCBs for this project in 2010. And I have some like ancillary schematics and things like that from way back then from forum post back then. And this is a vacuum tube compressor, audio compressor. And one of the one of the or two of the PCBs that go inside of this thing are DC heater circuits. So they basically take a tap from my transformer, I rectify it, I put it into these wreck these regulators, and they regulate it down to the correct voltage for the vacuum tube to run off.

Yeah, back into heaters. So I got a question before you jump to a bunch of others. What condition were the circuit boards in?

Okay. A few of them we're in we're in great condition. They were just dusty. So I just washed them with alcohol and they look great. These DC heater boards that I'm talking about here, were actually they were oxidized. Really bad. So I when I went to solder them, like solder wasn't sticking to the pads, I had a flux, I had to stop and clean the living bejesus out of them. And then I just got out the RA flux. Because when you want to actually solder, something you pull out are a flux, you know, that stuff will eat through it. So yeah, surprisingly, I was actually concerned about the boards because they were in

what surface finish with a Enid or hassle. Oh,

they look like hassle. I don't know what it is. But given the timeframe, and given that this was just like a forum project, it's probably hassle. That's probably why they're okay. But but like I said they oxidized really bad. Which isn't necessarily an indicator of hassle. That's true. They were they were in, they just weren't in great shape. The solder didn't want to stick to the pads. Let's just put it that way. So but you know, a toothbrush with some ethanol and some ra flux will fix anything. Yeah. So okay, so So here's the thing. There's two PCBs inside this circuit that have four tubes, each two of those tubes are have a have a load of 400 milliamps, and two of the tubes have a load of 900 milliamps. So I have two separate circuits that are 2.6 amps each. So five point, whatever, 5.2 amps total load on this, but I have two separate DC heater boards. So you can just consider it just think about one. So it's a 2.6 amp heater load on it at 6.3 volts. So it's a pretty hefty amount of juice going through there just to heat these things up. So I've got my transformer that's cranking out some DC voltage. And then I'm regulating down to that 6.3 volts. But at 2.6 amps going through this regulator, you could, you know, maybe the gears are turning in your head, like, that's a lot of heat. That's a lot of power to dissipate. So I actually took some measurements of this. So we'll start out I actually have some calculations here because I wanted to I wanted to kind of back calculate some stuff. And well, let me just throw out some numbers and then we'll we'll move forward. So right off my rectifier, I'm getting about 10 10.8 volts. So that's my like input DC voltage 10 point, and I want 6.3 volts out. So my regulator has to drop that 10.8 down to 6.3. So that's four and a half full differential there. So four and a half volt differential, and my expected load is 2.6 amps. So the amount of heat that the regulator itself has to adjust to is 4.5 volts times 2.6 amps. That's 11.7 watts in my regulator, which is not an insignificant amount of heat, right? That's, that's a ton of power for regulator. This regulator that I'm using is an LM 338, which is if you're in a 202 20 package, which is a, you know, it's a pretty classic little regulator there. And it's one of those adjustable type where it has like a 1.25 volt reference. And then, depending on the resistors on the output of it, you can adjust its output voltage, which, on my boards, I have a little trim pot that I can do. Now, mind you, I didn't design these boards. And I'm not saying that because I would have done it differently. But I didn't design these boards, I just populated them. And I knew that there was going to be some heat there. But I didn't go through the exercise of calculating it. I was like, I know these things are gonna get hot. So I'm, I just bolted these transistors directly to my chassis. I mean, not directly I put a an insulating spacer on there, but like a mica barrier. Yeah, like a Michael barrier and an insulating shoulder washer in the in the tab. So they're electrically isolated, but I could I thermally coupled them. I also put thermal paste on it, because I knew that these things were gonna get hot. But like, is the chassis itself enough of a heatsink to get rid of the power? That's sort of what I wanted to back calculate here? Because I can tell you right now, the answer is no, it's not. Not enough. And because it's not enough, I thought I would go through the exercise of actually calculating, first of all, how much is it not enough? And then how much do I need for it to be enough? And I did. So these calculations are, I'm going to preface this right now by saying these calculations are a little bit rough. So deal with me on this. Or bear with me, I should say on this. But they at least give like their ballpark numbers. Let's just put it that way.

So the thing is, I know I have about 11 and a half or 11.7 watts that eventually needs to just go somewhere. It's super inefficient system. And a to Well, I'm sorry, let me let me back up real quick and just talk about how I tested this. So

I guess you could just say I plugged it in and saw what happened, right? So so here's the thing, because I have vacuum tubes that I can pull out of their sockets, I can adjust the load, right I can, I can adjust the load by just plugging it in or pulling it out. And I started adding tubes one at a time and checking my output voltage and adjusting my little trimmer to make sure that the regulator was was regulating. And you know, at 400 milliamps, it was fine at 800 milliamps fine, I can I can get the voltage I want. As soon as I put an additional 900 milliamps on it, the output voltage just drops. And then I put an additional 900 milliamp load on the output it dropped yet again, and dropped as in like, I'm turning my trim pot to adjust the regulator and it's just like, Ah, I'm not doing this. It's not going you're not supposed to do that, though. Well, no, right. That's the whole point of a regulator, right? Yeah, it's supposed to adjust automatically for the load. So in looking at the circuit, like there's nothing wrong with the circuit, it's functioning properly. Here's what's happened. I started dissipating too much heat, and I ran up against the thermal barrier inside the regulator, the regulator automatically adjusts its output such that it maintains a maximum heat inside of the actual regulator, which that's not a good place to operate your regulators that that's not their happy zone or anything

like that. Did you only realize this after like the powder coating on your chassis starts bubbling?

Hey, this is anodized aluminum, this isn't 19 Okay,

the anodizing starts as coloring.

Well, okay, so here's the thing, once again, this is a project from 2020 10. That's just like a forum thing where people are like, I buy these boards that work for you. But like, there's no instructions on like, Oh, you're going to need a monster heatsink for these things or anything like that. And like I said, I just kind of like, plugged them in to just try to give it a go kind of thing. So I'm kind of laughing at him like a lot of these people that I see who were also building this wouldn't have even expected that you would have needed a heatsink at all, they would have just started the transistor to the board and been like, Oh, it's good, because the circuit says it's good, you know? They would have just bought it and soldered it. So

yeah, I mean, whatever. But at least Hey, at least I get to have some fun with it. And here's one of the one of the cool things about this. I have no idea in general, what is the thermal dissipation of a plate of aluminum that has a particular thickness and it's anodized on one side, like, would you even like what would you even guess is the thermal dissipation of that, and I don't want to go and simulate that and like I have no clue how to calculate that I didn't take that class in college. So this is a fun way to at least back calculate that and Get a gut feel for what it is right? So okay, so check this out a to 220 package component, in general, doesn't matter what it is has a thermal dissipation of about 70 degrees Celsius per watt. That is, like, if you're just starting from scratch in any design, that's a good number to use 70 C per watt. So what that means is, if the device itself is dissipating one watt, if you're asking it to dissipate one, one, you can expect its temperature to be 70 degrees Celsius above ambient, ambient. Yep. So I think a lot of people get that wrong, especially when you like, back to the Arduino crowd, like you were joking about earlier. Like, they're like, Oh, 7805, I'll just put that in and just start cranking a bunch of voltage from it, or a bunch of current from it, and it gets really damn hot really fast. Because they put 18 volts on the input, and they're asking an amp out of it, you know, to regulate down to five volts, like, you got to consider the, that differential gets chewed up as heat.

In fact, I, it's funny, the day I kind of connected these dots, maybe it was a lot later than I should have. But I'd also never really been asked this question before. I remember a mechanical engineer, came up to me at my first job. And he was asking about, like, power dissipation in in our circuits is like, I've taken a little bit of circuit classes and stuff, but I kind of confused about something he's like, you put voltage and current into a device. And that device consumes power. And he goes, where does that power go? What is that power? And like, I thought about it was like, some of it is useful, but like 99% of it just becomes heat. Right? It turns entropy? Well, yeah, effectively, yeah, you go to make we as electrical engineers make lots and lots and lots of little space heaters. That's like the majority of our stuff, right? But like, I guess, I guess I knew that, but I never really connected the dots until he asked me there's like, oh, yeah, almost all the power we put into our circuits is just heat on a small scale, but it's just just heat, right?

Unless you make the most efficient thing ever, where 100% of your power goes to the thing you want it to? Which we're not there yet. So okay, yeah. So a to 220 dissipates about 70 degrees Celsius per watt. So if I had just a to 220, that that was just on this board. And I tried to have that to 220 by itself, dissipate 11.7 Watts, that's 11.7 times 70, above ambient. Effectively, you know, if we were in a simulation world, that would be 844 degrees Celsius of is what I would be asking this little regulator to do. That ain't happening, right. So so what's happening in my situation is, as I'm adding load to it, I'm increasing the heat on this regulator, and the regulator is saying, I'm not cool with that. So it's, it's dropping the dropping its voltage until it's at its maximum thermal capability that the internal circuitry is allowing for. So with a 1.6 amp load, I pulled two of the tubes, and I still have 1.6 amps on this, I saw that my output dropped from 6.3 volts down to five volts. So I could kind of back calculate based off of that differential that I know there. So 10.8 Minus five gives me 5.8 volts across the regulator. So that that meant a power of about 9.28 Watts, a dissipating in the, in the regulator. And I also know that the regulator, according to the datasheet has a maximum junction temperature of 125 degrees Celsius. So basically, given those numbers, what I could do is say that my ambient temperature is 25 volts, it raised 100 volts to get to 125. So 100, sorry, 100 volts 100 cells Celsius, 100 degrees Celsius divided by 9.28 gives me a thermal resistance of about 10.78 degrees Celsius per watt. That is what my aluminum was doing. So for every watt of juice that I was asking this regulator to do, the the temperature of the regulator itself would rise by 10, almost 11 degrees Celsius, that's a hell of a lot better than 70 degrees Celsius if regulators just sitting there alone. So the thing that's cool about that is it at least gives me a gut feel for what aluminum would do. You know, like if I ever had to do this again, I could say like, oh, I can use 11 C per watt as just like a overall value. And in fact, I have a PCB that's arriving next week that I'm going to build up that also gets bolted to the chassis and it has much lower power needs. But I'm kind of curious, I want to, I want to test its power and see if it also gets a value somewhere around 11 C per watt. So I've got a thermal camera or I borrowed a thermal camera the other day, I want to give it a shot, turn this thing on, heat it up, see if that works. So here's the thing. 10, or we'll just go with 1111 degrees Celsius per watt is not enough for this circuit to run, because I'm running into the thermal overload of this regulator. So back calculating around, what kind of heatsink what I need at bare minimum in order for this circuit to work. So I need to dissipate 11.7 watts. So if I just back calculate that 100 degrees C rise divided by 11.7, I would need 8.54 Celsius per watt heatsink. So in other words, my we've already found my chassis is not enough to dissipate that much heat. But I went on to Amazon and I bought these big old chunky aluminum heatsinks. I'm showing the Slack channel here. So I'm going to drill and tap these and basically thermal paste and just stick these on the back of the of the chassis and see if I can get that down to the eight and a half Celsius per watt.

But those you can't go and go I need the eight C per watt. How do you spec that out? Well, that just went on Amazon bought

something I went on Amazon and bought something I did look through the comments and somebody actually had the thermal resistance of this. I don't remember exactly what it was, I think it was like five or six Celsius per okay. So so this is pretty good. However, that would be if I bolted my transistors directly to this thing. I'm bolting them through a layer of aluminum and a spacer and thermal grease. So like it's worse than that, right?

It was it what's gonna be interesting is I did a lot of research for this. You remember the reload Pro? Oh talk that we did like those like product like two that we did at macro fab? Yep. anodized aluminum, that anodizing layer is very thermally non conductive? Yep. Like thermal. The heat does not like to go through aluminum oxide. Okay,

but he loves to go through aluminum. Not the oxide.

Yeah, not the oxide. Yeah. And so I hate to scrape off that aluminum. That oxide layer? Man,

I might have to. I'm gonna try it without. And just I would try it without see if you can get there. Yeah. Here's the beautiful thing about heat sinks. So when you the kind of calculations I just went through to find what I need. First of all, that's what I need at a bare minimum to make sure that the regulator doesn't go into thermal shutdown. I need you to be well better than that to make sure that it doesn't go into thermal shutdown, right? This is purely the bare minimum. And that's also assuming that this thing is always in ambient temperature of 25 degrees Celsius, which it won't be right. So I need probably 2x Better than that, if not more than that. So that was just a fun, like academic exercise to kind of like back out

sounds like you should change that power supply out.

I probably should. Yeah. Or redesign one or just go with something else. I'm just I'm having fun with this though. The beautiful thing about heatsinks is if you go to something like Mouser or DigiKey, they have a search field that is their thermal resistance basically, or their disguise. Yeah. So you can just go and be like, oh, I need an 8.5 Celsius per watt or better. And you just type that in, and it'll show you everything of that.

I wonder how that's measured, because it's also dependent on contact area.

There are tons of standards around them. In fact, that's something else I wanted to bring up. Because there is there is a big pitfall in datasheet for new people. If you go to look at the datasheet for this lm 338 regulator or pick any other part that's in a to 220 Most of the time, and I don't mean some of them, I mean most of the time, the values that you see for their thermal dissipation are not what is reality. You have to dig deeper and you have to go to different documents to find it. Because if you look at the one that that sort of dictates everything, which is theta J A, which is the thermal resistance between the junction to ambient, that tells you that's that 70 value that I gave her where for every watt it rises by 70. If you go look at most data sheets for parts that are in 220s, it doesn't say 70 It says something like 23 or 24 or something like that. You have to look deeper and look in the notes because the They test that to some JEDEC standard. And if you go read that standard, they're allowed to bolt it to a heatsink with a certain value and then measure it at that point. So they give you this nice like, ooh, that value is great. That means that this part can can dissipate three watts of power, when in reality, it barely does one, right? So be really careful with all the thermal stuff because I've it's been my personal experience that they're bloated, those numbers and they're nowhere near real. So yeah, that's just something to watch out for. I chuckle every time I see that, because like, it's on every damn datasheet for it to 220 part. I'm like that is not real in any way. That is not real. In fact, just for fun, I took this exact same power supply that it was handling 800 milliamp load fine, I got the voltage, I want everything. So I was like, Okay, what happens if I just unbolt the the regulator from the, from the chassis and just run it in free air? Yeah, there. It got, like, the voltage was like 200 millivolts on the output. Like, it was trying so hard to do anything. And I was I was afraid of letting out the smoke. Because like, I mean, I'm asking 800 millivolts across, I don't know, 10 volts or something like that. Or sorry, 800 milliamps. And it just, I mean, it just immediately went into thermal shutdown, like from the second I turned that on, so but if it but if there was like a nice 20 degrees Celsius per watt value, like there was in the datasheet it would have been fine. It would have been Yeah, it would have been a lot better. So just watch out for that. So yeah, I'm gonna I'm gonna try to in fact, later tonight, I'm hoping to just drill and tap these, one of these heat sinks, plug it in and just see if it's good enough. I don't know. It's kind of fun. Actually, it's, it's interesting. I went to a, an interview once at an engineering firm. And they did some really impressive system level design some really big stuff. And I remember asking them in there, I was like, Oh, this is great. How do you guys handle your thermal stuff? Do you guys go and calculate or are you doing like 3d models? And like, you know, everything? And they're like, oh, yeah, no. And their solution was like, we turn it on and see if it gets hot. And I'm like, okay, cool. Like, it's interesting. They add, they'll, they'll do things like add a fan or whatnot, if it just gets hot.

They design it. So they could do that.

Maybe not even, maybe, like, maybe they it's all afterthought in that sense. Thermal stuff is weird. Because like, do you think about it? At the beginning stages? I guess it depends on your industry. It depends on what you're doing.

On pinnacle. I was thinking about it when I was designing it. Yeah,

sure. I mean, I think I think

there's not a lot of thermal stuff on the printed car. So

I have a product that I'm designing at work, that's just, it gets ridiculously hot, I've got too much analog, let's just put it that way, running on high voltage rails that don't need to be high voltage rails. And that's this is sort of the magic of switch mode power supplies because for a little bit of heat, currency that you pay in, you can shift voltages without just dumping that as heat. That's sort of one of the magic, magical things about switch mode power supplies is you don't need a transformer to switch things over.

So Casey ate a PF. That's a that's a username on the Twitch chat says he's or she is currently working with a heatsink vendor to design a solution for a 400 watt ASIC. So they're doing a lot of simulations.

That's cool. God, I'd love to see some of those simulations. That'd be fun.

That'd be pretty cool to see. So so people do do the simulations. It's just that one company you're interviewing with was not

well, okay, so yeah, but But in that particular situation, if you've got one individual part that's cranking out 400 Watts, then then I think you're just you pretty much have to write you don't have a choice. You can't ignore it. This other company had think they told me 13 individual PCBs inside of a system a large system enclosure, and they they had different teams designed each board and then they brought them all together. And in testing, they found out if they needed heat or needed thermal management or not. Ah, Parker, you turned into a robot man it might just be on my side. I do okay, so yeah, it must be it must be for everyone seems like it seems like this the chat see here in it to do you need me to? Oh, you want me to close out the podcast? Okay, I see how it goes. Well okay, thanks for joining us everyone in the slack chat we appreciate you being here each week. So with that that was the macro fab engineering podcast we were your host and Steven Craig take it easy Thank you. Yes, you our listener for downloading our podcast. If you had a cool idea, project or topic let Stephen and I know Tweet us at macro Feb at Longhorn engineer or at analog EMG or email us at podcast at macro fab.com Also check out our Slack channel. You can find us at macro fab.com/slack