Silicon Condensors

Hello, and welcome to the macro fab engineering podcast.

We got to save this take. Yeah, just restart. Hello and welcome to the mag fed engineering podcast. We're your host, even Craig and Parker.

Sorry about that Josh just pulled a really funny joke and kind of rolled over into the intro. Perfect way to start the podcast. Yeah, and this is episode 12. Episode 12. Steve, Steven, you're working on a really cool mic actually for Josh a power supply I think for Mike. Yeah, yeah. So Okay. All right, Mike.

So Josh and I've been been friends for a while now and and I do some some repair work for him. I've been actually doing audio repair work for a handful years now. And so Josh actually gave me a, an old Sony C 37. A microphone, old tube microphone that he had that was in need of a power supply. Because these these, these mics actually run on a couple 100 volts and with with heater voltage and all that jazz, and actually the kit that he had didn't include the correct power supply. It had a power supply for a completely separate tube mic. So So Josh asked if I could build up a power supply for that. So I've been I've been working on that for the past. Oh, good while now and and finally got it finished.

And for our listeners, Josh is the person who lets lets us record here and does all our editing Well, minimal editing basically makes it so that we sound decent on Mike's just looking listen, I think we talked about Josh once so

Okay, yeah, yeah. Josh, the audio guy. Yeah.

And he actually just got done drying out his recording studio from the Texas flood of 2016. The torrential downpour. Yeah. One of a 500 year flood, supposedly,

yeah. Monday was a lot of fun here in Houston.

So yes. What, um, I know you had some problems building this thing. You're because I saw you building it on Saturday, this last past Saturday? Yeah. And you were basically it looked like you were using like a tuning rod because you're holding them transformer, like rotated in space, and there's going rain or air around.

So okay, for the listeners, these old tube mics have a really interesting design feature. In the fact that feature, right? Well, yeah, feature, it's a great word. But so they include the entire power supply, so that that's two rails one at about 200 to 300 volts somewhere in that range. Another one that's semi regulated at six and a half ish volts. And all the audio wiring in a tiny box. And that audio wiring includes an output transformer. And they try to shove all of this in, in a tiny little box. So the the actual configuration of the mic itself is basically if you think of it in transistor terms, it's kind of like a, a common common cathode, a common cathode, so it's basically a buffer, you have the the sensing capsule, and then it just goes into a buffer. And that's sent down to the power supply, where it's actually converted into a balanced signal using the output transformer. So you actually have to send a couple 100 volts up to the mic, and then send your tiny little audio signal down and convert it to a balanced signal inside the power supply. And that comes with a lot of issues with the fact that you have a big output transformer right next to your power transformer. So grounding is key. And the orientation of the two iron components in relation to each other is absolute key.

And he's talking about the iron in the Transformers.

Well, yeah, right i on the transformer, just because the magnetic field from your power transformer gets coupled very easily onto your audio lines. Because the the thing about this mic that's crazy is it technically doesn't have any gain. In terms of positive gain, it's, since it's a since it's a common cathode, it actually technically has slightly less than unity gain coming off of it. So whatever comes out of that capsule, that's the voltage you get effectively. So your output transformer picks up that magnetic field very easily and you were talking about kind of me moving a transformer around. I had to physically align and move the transformer to find the lowest angle at which it picked up the least amount of noise and then bolt it to the chassis in that in that order. And so that was a little bit of a chore. And on top of that, having a really, really strict star grounding system, I found that I probably tried six or seven different ground schemes on it. And every single one produced noise until I just said, You know what, screw it, I'm going to take every ground wire and connect it to chassis at one point and do a true star ground. And I found that that's the only way that I could get the noise to a reasonable limit.

Yeah. So on that reasonable limit, what did you consider reasonable?

I actually measured it a goodwill measure in quotes, because I was looking at it in my audio workstation. So it's not, I'm not looking at it in terms of test equipment. Yeah. But it was negative between negative 15 Negative 60. Decibel. That's, that's really when I applied a bunch of game to it. Yeah, after applying game, so it wasn't, it's not terrible, I would love for it to be in the negative 80 range. But 5060 can be acceptable, especially after putting a bunch of gain on it. Cool. But once again, I don't know how calibrated those numbers are, because I'm looking at it through my audio workstation. Not an actual test gear. But yeah, just not not terrible.

So what kind of applications you would use this kind of tube mic for?

This is this is purely aesthetics, I would, I would pretty much say. So this is this is of what I know, of the research I had to do to do this, get this power supply for the mic work. And this is a highly sought after mic. And in the recording industry. This is this is a go to, in a way in terms of pristine sound. Okay. So interesting thing is, there's not a lot of information in terms of schematics and things like that. So

interesting. Now like most like amplifier, like guitar amps, schematics everywhere form, yeah,

you can usually find those pretty easily. Yeah.

And then I've been working on more test fixtures at macro fab for customers, and mainly for the pin, heck, not in more things at tests on the board. And I actually just finished about 10 minutes before I started to drive over here. Got all the pick three, two side stuff tested code, the pic 32. And actually the parallax propeller 100%. Test every single IO pin on that pin. Heck, so all 160 some odd pogo pins are tested now. Well, the last thing to implement is the pic 32, i square C tests where it will communicate to the I squared C devices that are on the pin tech board and do a self test basically say hey, is the Real Time Clock talking? Is the EEPROM talking? That kind of stuff. That's it. Awesome. Yeah. And that whole test only takes I think now takes two minutes and 10 seconds. Wow. Just rips through it. Yeah. And all you do is just like us talking the serial terminal? Yes, yes, yes, yes. And then at the end, it spits out errors, if there's any. The only interesting issue I found was on the solenoid tests, it basically the solenoid banks have moss spots on it. And channel MOSFET, because they're low side switches, is if you put a NPN transistor, and it's placed like a tip one or two, which we had tip one and twos on that board. So if one accidentally gets placed in the solenoid spot, it will actually test fine. Because it drives that NPN like a switch. Okay, but the moment you probably put 50 volts and try to put like five, six amps through it, it'll probably make nice smoke. Yeah. And the problem is, you can't actually test that on the test fixture. There's no way to do it. And the only way I could think you could do it is if you actually read the analog signal. Mm hmm. And actually measure the voltage of all those pins. Yeah. But the problem is, you would only see a point two volt ish difference, which might be under your tolerance for all those FETs I didn't run the numbers, but it might be and so it might be you just can't tell if you have a NPN versa. Versa MOSFET Well,

what would be the reason why you would need to check that just a so operator mess up or something? Yeah, so

if the operator or the The person whose stuff in the boards Yeah, puts a tip one or two in a IRL 540. Spot. Yeah. And then the board went on to chuck up and spooky pinball and they plugged it in. Yeah. And that solenoid fired, it would probably blow out that tip one or two, instantly. Make a lot you might last a while. Depends on what what which one it is. If it's one of the ones that fires really quickly, yeah, it probably will be okay. Yeah, it was one of the ones that like the the magnet that stays on for, you know, 10 to 15 seconds, that might do it, it would just overheat and overheat and toast that tip. The good thing is tip one or twos have a different color grounding tab, or it's not ground time. I think they're attached to the emitter. I think the tabs connected to JMeter. Really?

Or is it? It's usually typically usually connected to collectors.

Okay, it has to be collector. Yeah. So it's connected to the collector. But that tab is a completely different shape and color than an IRL 540 MOSFET.

So when you look I mean, the actual metal. Yeah, the metals. toe is easy. Kind of like a like a more copper ish color. No, the

the moss bits are. Okay, moss bits are copper color. Yeah. Whereas the tip one or twos are are actually tendons. So they're very matte silver colored. Yeah, like a kind of brushed aluminum ish. Yeah, yeah. And so it's actually very easy to visually tell. But you can't tell through the test. Gotcha. Oh, that's the only thing I've found that would be possibly a problem. Yeah. But I haven't run into that issue yet. I just thought of that. Just like what if this happened? So that's what we've been doing this week. Oh, yeah. And I've been working on a fan controller for my jeep. No, yeah. So I've been riding a display driver, for a ginormous Oh, yeah, you brought that VFD in. Yeah, it's a ginormous four by 40. Character VFD. That runs off that HD four for what five, zero, whatever, that basically a normal character display right at a really cheap, and basically runs out that same protocol, except it has to eat lines, which basically, they use the lines. It's kind of like clocks. And so basically, the first two lines use E one. And the next two lines use E two, which is kind of weird. Some least a couple, four line character displays I've looked at, they use one line, and they just expanded the memory and said, I'm just making two separate banks.

Yeah, that's that's kind of strange.

Yeah, it's a little weird. But it works now. Yeah, it looks great. Yeah. And I all I've got to implement now is the brightness control. Since it's the VFD. The brightness controls actually through the controller, not through an external basically led that you PWM.

How do you write a value to? Yeah,

there's just a different command that you send. Okay. It follows one of the setup instructions. And so what you can do is you basically send the setup again, and send a brightness command.

Is it just a value that you send it? Or is there like high and low?

Well, you said I, you set the RS value, RS pin the high and then you send the command? And the commands got a certain certain kind of like, look to it? Yeah. And then it knows, okay, this is the brightness command, and then the last two bits in that, that, that, uh, that bite or the brightness, so you have four levels.

Okay. Okay. Yeah. So you it's not like you have, you know, 1000 different levels? No, no, you

only get four different levels. Okay, it's only two bits that you set. Sure. So a default 00 is a highest brightness, and then it goes all the way down to very dim. But basically, I want to do that do brightness control. And basically what the, the fan controller is going to do is actually read the a light bulb in my dash, because when you turn the lights on, the dash actually gets dimmer. Okay? Because at night, you don't want really blinding lights in your eyes. Sure. And so it dims the dash. And so the, the fan controller is going to have an ADC on it that will read that and say, hey, the dash is supposed to be dim, dim this display so it's not blinding the driver.

Yeah, yeah, I could see that being a big trouble because that VFDs it's really bright.

Yeah. And I guess we'll roll right into the RFO. Yeah. There's a really cool video that I can't remember the person who put out the video. It might have been in video. It could have been Yeah, I'm not sure. But basically Nvidia allowed someone to come in and record They're this guy named Howard. That's at their silicon failure analysis lab. Mm hmm. And they do the r&d and of testing, right? How do they test their chips and all that stuff? Well, along with the failure analysis when something goes bad, yes. And so the crazy thing about this is they are able to find so like, the chipsets on these Nvidia chips, they have billions of transistors, he was saying 9 billion and 9 billion transistors, and they can find one transistor out of that entire 9 billion that failed.

That's insane. Yeah, if you're into silicon dye, and semiconductor physics and all that lovely jazz, you will just absolutely drool over this video. It

is awesome. And the best thing is, whenever he introduces a new machine, he also introduces the price tag.

Oh, every single machine. Yeah, this is two and a half million dollars.

All those machines are worth more than I make in a year, like 10 times over.

Yeah. Yeah, absolutely. No. Like, yeah, electron microscopes. T VMs. All kinds of crazy thermal testers. 3d X rays. Yeah. All kinds of cool stuff.

Yeah. It's

that's all to make sure that you can gain faster.

Yeah, make sure you get higher frames per second. Yep. Oh, and compute computational because they also make workstation cards.

Yeah. And it was funny. I was actually having a chat with Parker earlier. It makes total sense that they would have all of this stuff, but you don't really realize it just because, you know, they make graphics cards. And it's not. I mean, they're they're expensive piece of machinery. There's a lot of technology that goes behind it, in some cases, a lot more than what you actually have as your main computing. Hardware. Yeah. But you would you just wouldn't think that they have this, you know, 30 $40 million fab just sitting there making sure you can get that 60 frames a second, you know, yeah, it's

one of those things where you would think they would test the chip. And that chip failed, like, oh, okay, whatever, you know, they have a certain yield. Yeah. But they actually test all those failures. Oh, yeah. That's it's kind of interesting that they actually did test all that stuff, I guess, I guess that's part of improving their yield is fixed these issues that are causing these chips to fail?

Well, when you when you release a new GPU, or anything that has a significant amount of transistors in it, the the industry standards, you can kind of expect somewhere around 30% yield upon release of that chip. And by maturity, you expect closer to 80 to 90% yield. And so these guys are probably the guys who are tasked with getting it from the 30 up to the 80 to 90.

Yeah, that makes a lot of sense. Yeah. I wonder on those 30% If that's closer to the center of the wafers, you know, I don't know. Because, yeah, it's been a long time since I've taken solid state design, by recall that that's usually the case. Then I have new technology that's fixed that though.

Yeah. And if you think I mean, most fabs are running 300 millimeter wafers. Nowadays, so what constitutes the center? Yeah. Is it like is there an inch around the border that is just gets messed up and all the rest is considered center? I don't know.

I don't know. Interesting. Yeah, very interesting stuff. Go we'll have the link in the blog post. Go give a look to it. Yeah, super cool. Super cool stuff. And then microchip came out with the pic dim lab two, which is basically a little development board that you basically plug any kind of pic or dip pic. Yeah, into it. Well, it's

meant for all of their eight bit side.

Okay. What say what Josh

you can plug any of those in.

Yes, any dip pic, you can plug into. Anyways, the cool thing about this is we actually used a pic dem lab, one that Steven brought up to prototype, the

macro watch. Yeah. It's really convenient because it just has a whole bunch of dip sockets. So you can just plug in whatever chip you have. And it has a bunch of headers where you plug in the picket. You're often programming.

Yeah. So I mean, sure. That's cool. And all this new lab to thing. I think it just supports different versions or something like that. But the interesting thing is the recommended videos after. And this is actually something Steven found he recommends a video by Dave Jones from the Eevee blog. Yeah, where he completely trashes the picket three,

and okay, so I stumbled upon this because I was I was floating around he, he Web. Yeah, that's where I was. And on the web, there is a microchip has their own like sub kind of domain on there where they broadcast all this microchip stuff. And it's all like, Hey, we're microchip. We're awesome. And then all the videos that popped up it's like, you might want to watch these and is everyone is Dave just trashing my coach.

Yeah, they've doesn't seem to really like microchip too much. He doesn't like the pic kit. Yeah, the picket three, like the picket too. Yeah, love the

kid, too. And, you know, I the thing is, I never used the picket two. I started on the three. So I don't see what's I don't see why it's bad. But I don't know either. Maybe if I pick up a two, I'll be like, Oh, my gosh,

this is amazing thing that what it is, is the three can't provide as much power power devices as much. I think that's one of the issues.

And there's something about memory. It doesn't have either any or enough memory. I can't remember what the deal was.

It doesn't matter. Pick it three works. At least works for us.

I've I've always been able to program chips.

Yep. And then there was a really cool Hackaday project of a desktop siege cannon.

Oh, yeah, this thing was awesome. So this cannon actually uses a flash paper, or what was it nitrocellulose? Yeah. It's kind of

gotten cotton almost too.

Yeah. Yeah, effectively. And it basically the guy who designed it use two six volt batteries, and just effectively electrocuted this gun cotton. And it just, it makes an amazing fireball. Yeah. I wish it was something I could have on my desk. Yeah.

So we're talking about design to make this work for macro fab. And I was thinking is you put an electrolytic down the down the barrel barrel? Yeah. Backwards. And then so the leads are sticking out? No, no, no, you put it down normal. But you reverse bias the cap. Oh, so you'd have like a capacitor bank is charged up with a couple 100 volts, and then hit a electro lytic backward like a 10 volt electrolytic? Backwards? Yeah, but hit it with like 1000 volts or, like 1000 Nothing was shoot out of the barrel. Does it constitute a biological weapon? Or with all the electrolytic flying

on the inside? Yeah.

Oh, well, every really cool looking.

It would be awesome.

I think we had to build one at least in as a personal project. Yeah, yeah, I'm totally down for that. Yep. And then, big Clive had a really cool video. This is a guy that does really cool videos on YouTube. Yeah. Mainly breakdowns of really bad and dangerous products that will probably electrocute you. Okay, yeah. So this big Clyde basically took apart a USB soldering iron. And I mean, USB can only provide what is it? 2.5 watts. Is that really all it is? That's five volts.

Oh, yeah, half an amp. You're right. Yeah, that makes sense. But

this thing's actually pulling eight watts. So you can't actually plug it into a normal USB 2.0 You have to plug it into a basically a power pack or a charger USB charger. Alright. Something that just can deliver lots of power. Yeah, a bit more. And the cool thing about it is actually used as a 555 doesn't have a microcontroller in it uses a 5.5 timer. It just PWM is the heater element. Yeah. Nice. And it has a timeout. So it has a little low vibration spring. And if it doesn't sense any movement, the 5.5 timer times out and turns off the iron. That's nifty. It's a really neat design. And he actually shows them soldering, like MOSFET connectors. So it actually does work fairly well. Yeah. And then also, of course, me being an electrical engineer. I'm like, How could you make that better? You need more power? Sure. Well, the USB 3.1 spec over type C connector can do three amps over five volts. Okay, so watts. Yeah. But then you can also do USB power delivery over the same connector. Yeah. Which is 20 volts at five amps, which is 100 100 watts. 100 Watts is plenty to do a soldering iron. Oh, I mean, this guy was soldering with 800 Watt, you'd be able to solder. You can bring one gauge wire. Yeah, you can raise some stuff together. Yeah. And so I actually did a little research on how you would make this work. So I was looking at the the TI TPS 65986. Okay, yeah, a number. Yeah, that part is basically a USB C, or USB 3.0 and 3.1, type C style. Chip. And using that chip is you can talk to USB 3.1 and configure your device. So you can set up all this power stuff.

Okay? Is that how you knew what? No, what's it called? The power? Not power on the go? Or on the go? No, like the 20 volt five?

Oh, power delivery, power

delivery? Do you have to enable that?

Yeah, you basically have to talk to it and say, Hey, I'm a device that can that needs this. Can you give it to me? Hmm. Because it's only going to do that on certain occasions?

I know, are they planning on having this as an available outlet on your desktop laptops?

Laptops are gonna charge this way. So instead of having a barrel Jack like we do now? Yeah, you're gonna have a USB C power delivery plug. Oh, that's cool. So you plug that your your your laptop, power supply will plug into that. Yeah. And then you can also plug in your cell phone into it. And so your cell phone can super juice off of it. And they actually say you could do device The device to so one left one laptops charged up. The other one's dead. You can just charge one laptop to one laptop using the same cable and connections. That's cool. That's pretty cool stuff.

I wonder what kind of safeguards they're gonna have to put in place with there are

lots. Yeah, I was looking at the schematics and stuff for this stuff. It's very different from how USB USB 2.0 is handled. Yeah, just be 2.0 was handled is like you have a MOSFET that base up a p channel high side MOSFET. That turns off and on for numeration. Yeah. And then you have a usually you just have an inductor to prevent inrush. Mm hmm. That's about it. This is way more complicated, because also that connectors can be flipped over. Yep. And so actually, these chips have what's way the cables plugged in? It can detect which way Oh, so

it knows what one is. Which, yes, choose your own adventure, basically. So no, no, this is actually an interesting question. So to get that 20 volt five amp, it's gonna have to run on a dc dc switcher. To get it up to that level? Well, it

depends on what it is like your laptop, power brick would probably just out we'll just output 12 volt 20 volts, because they most of them output 19 already.

Right? Right. Okay, but so what I'm just curious about is if you're pulling the full 100 watts, watts, and you're talking digital communication over that line.

I'm doing I don't know if you can. Okay, that's I haven't looked too much inspect. I want to make a noise on that. It's got to be terrible. I'm going to assume you probably can know. Wow.

Good. I mean, yeah, of course it would. It's DC but pulling that much. You'd have to have a pretty clean line to not interfere with all the rest of the

asiata 10 gigabits a second. Yeah, that's,

I'm gonna go do some research on this. I want to learn some more.

Yeah, I want to do some more research. We'll be back next week with probably I'll probably try to have a rough schematic, or at least a block diagram of how to do simple enumeration and get power delivery working Yeah, like a feasibility to see if this would actually work because I kind of want to make it so that you can power a soldering iron off a cell phone type C connector,

what I love about this little thing is like this is sort of the new cutting edge technology and you want to make a soldering iron. I love it. And then

I was looking at a if you go into Mouser and just type in like type C connectors. Some really cool stuff pops up and we want a good bit available. Yeah, from like Molex. Of course, and like MC eyes got some ti not not ti What is it to the Tyco? No, I think it's ti ti

interconnect. interconnect. Yeah,

yeah, that's them. And then J E electronics. That's actually the problem I want to use. It's a really big long part number that's on my sheet. I'll just post in the blog. Yeah. It's like 20 characters long, but that one looks really cool because it actually had mounting tabs in the front with SMT pads in the back. Oh with the rest of the net all surface mount. Yeah. Business in the front party in the back.

Yeah. I you know, I the the through hole legs. call them a pain. There's so much more reliable.

Well, yeah, especially with a if you're going to have this on the back of a soldering iron, it's going to be flexed a lot. So I'm thinking as much mechanical stability as I can. You should

you should back that sleeve

that we had no one give us ideas on Twitter.

Well, yeah, listeners, unfortunately.

I might have been the best. Yeah. I think that's uh, yep. We're at the Balmer sheet here.

You have anything else to add? No, I think that's it. Alright, cool. Cool. Sign us out.

Yeah, well, this was the macro FEHB engineering podcast where your host, Stephen Craig and Parker don't have a good day

later, guys.