Honey, I Shrunk The Parts

Welcome to the macro fab engineering podcast.

We are your host, Stephen Craig and Parker, Dolman.

This is episode 267.

So last week, we were talking about test procedures, test procedures. Yes. And one of the things I mentioned that we could have done like an entire episode on was stickers and labels. So we're gonna have an entire 50 minute podcasts about stickers and labels. Let's go. It won't be that long. But yes. So test procedures are good and great. Except you have to know what's been tested, what hasn't been tested? How do you do that?

I mean, there's a bazillion ways to do it, or accident, I shouldn't say there's a bazillion bad ways of doing it.

Yes, there's a lot of bad ways, there's very few good ways to do it. If you're doing small batches, like, you know, like 10 or 20 at a time, it's probably the best way is just like you have an inbox that are unprogrammed. And you have an outbox that are programmed. Easy to do that that way.

Yeah, that's like that's like the, how do I put this, that's like the bottom of the barrel way of doing that's the bottom

of the barrel. And so but once you start scaling up, you got to think of, kind of like inventory management, or PCB level tracking of what's tested, programmed, etc. What different stages of your product is in your build. And one of the easiest ways to do that is slapping a little sticker on it that serializes your board. There are a lot of questions I get a lot is, can I use silkscreen to serialize my board? Technically, yes, you can do that. You can have the PCB fabrication shop, like serialized each individual PCB with a silkscreen marking, but that becomes prohibitively expensive, because the board shop basically has to run a new silkscreen screen for every panel, because those numbers change. Some of them, depending on the fab shop, the lower volume fab shops, that actually tends to be cheaper, because they use like a machine to like laser like almost like jet print the silkscreen on so it's not as expensive because you don't have as much tooling cost. But that also

tends to be lower quality on the silkscreen itself. You get lower resolution, and it's just not as fantastic.

Yeah. So yeah, you could use silkscreen, it's expensive. And it would be if you're going with a shop that uses a printer to do it, and not printer, a inkjet printer. It'd be lower quality. So depends on what you're doing. But so the best what I always recommend to people is put stickers on the boards. And the the thing with stickers, though, is, like all the different ways of doing test management and program management, and inventory tracking. There's also a lot of a lot of different kinds of stickers. I would say most people, customers that come to us, they want serialization or tracking, except that they just that's what they give you. That's it. Yeah, we just want this. Yeah, we just want it which is fine. And I think most CRMs just have that kind of an internal kind of way of doing it. I'd like to know more about the product at that point, as a cm. Know, Like, what kind of label Do you want? Do you what kind of information do you want? Or is it just a number? Do you want date codes? Do you want a barcode?

Do you want a QR code?

QR code do you want? What kind of serial number? Is it something that it can be like a hex code not a hex code a, like a hash number, which is like a random characters that no one knows how to, like decode? Or do you want just 00001002 That's fine for you. Because some people don't want to tell how many units are out in the field. So being able to, like knowing that kind of level of information is very important. When you say I want stickers on my PCBs. So my first thing is, I guess you know, that's not the first thing this is the second thing. Second thing is talk to your CM about what kind of label printers they use. Because if you if you're developing on some other printer that they don't have, you're going to have to ship them the printer.

Well, above and beyond that. Not only you're going to have to do that, but you're gonna have to work with them to integrate. However it talks to a database or however it talks to a test procedure, or test script or whatnot. He got to make sure that you have all that squared away. You can't just be like, we want this size. And here's the little script. Good luck.

Yeah, yeah. So yeah, find out what label printers and how your CM integrates those is very important. At Mac fab, we use a lot of Zebra printers. There's also like Brady, there's T SC, a bunch of other different companies that do printers. And then the second thing is consumable items. And these are like the actual stickers themselves. And whether or not you need a ribbon. So what a ribbon is, is there's two different methods of transferring on technically history. But we'll talk about two two different ways of actually no data back, there's only two, one is just a duplicate the other one, anyways, you got thermal transfer, and direct thermal. So direct thermal is what everyone is usually most people are familiar with, which is like receipt paper, it's printing the papers got a it's, it's not it's it a chemical

dye, and honestly, I don't know is that one where if you run your finger across it really quick, you leave a streak on it.

Yeah, cuz it's basically you're heating up a certain section of the paper. And that's how you mark it. A lot of like shipping labels, that gone outside of boxes that comes up are our direct thermal printers, or direct oval labels. And the other kind is thermal transfer, which requires a ribbon that runs over the label, and it heats the ribbon up, and it prints onto the label. That's kind of like how a dot matrix printer would work with, if anyone remembers how those work, you have the ribbon in there. At macro fab, I typically use the ribbon more often, which is the thermal transfer, because it says they tend to resist like basically the PCB process better. They are more resistant to chemicals, washing heat. So if you have to do any kind of rework, if you have a direct thermal label on it, and you hit it with a hot air gun, it's going to turn black.

It's just gone. Scott.

So speaking of that, the next thing is label material.

Oh, yeah. Well. So I know where you're going with this.

Let's, let's take one quick step back and consider at what step of the process are you wanting to put a label on your Exactly. So CMOS, think of things kind of forward, like, oh, I buy this board, I build all these things. A lot of customers think of things backwards, where they're like, I want a product, I sit there thinking about the the end result is what they want, and all the steps that are needed to get to that end result. So where does the sticker get placed on it that determines what sticker you choose?

Correct. And where it goes on the board. So you could put on the sticker at like, the board shows up at the CME and they get stickered. That's one method, you could do it. So that's like if you want to track throughout the entire process, like through stencil through, pick and place through oven through secondary assembly, all that good stuff. More commonly, though, is post PCBA assembly for tracking. So all the board is good eggs about to go into QA is when it gets stickered. So before it goes into QA, or QC, depending on background, we got QC and QA. And so depending on where it's going in that process is when that gets stickered or not. That's where most customers prefer to start their tracking is right before that, because that's the start of the you know, the testing and validation phase of PCB as the whole process starts there. So beforehand, if you're doing I guess, if you're doing some really high end stuff, you'd want more like if you're doing more one up style boards, like a big board that you want to do, okay? We want to know that this board was in this machine and the machine said it did this and you want that associated with the board, then you need it before you need it like you know, pre paste. But most customers don't typically need that.

You know, and I have an example actually from work where we have we have a board that we put some some LED switches on and these LED switches have a notorious failure rate from the factory. So A lot of times, not a lot of times on anytime we manufacture that product, after it leaves the pick and place machine, but before it goes to get through hole soldered, we do a full test phase right there. Because that's a great place where we can detect those switches, pop them off and rework them before going much further. Now we don't stick with the boards. But that could also be a place where, where it would make sense. And that's the you see the problem there is there's another stage where it has to go into a machine and into a machine that gets hot. So So if we were to stick or there, we'd have to take that into account.

Correct. And so that's where I'm going with the material is. There's two basically basic kind of materials for stickers that go on to printed circuit boards. Yeah, I have. I think it's polyester. Polyester, I believe, and then the polyimide Yeah. So probably mites polymerize.

The hot stuff.

That's the captain. Yeah. I think the other is polyester. I know, living looked it up right now.

Yeah, I'm not sure. Because I almost anytime I've used stickers, I've almost always used polyimide. Just because we've always needed them in that situation. Yeah,

it's polyester stickers. So those the polyester both if you use the ribbon style, so it's the thermal transfer. They're very resistant to solvents, and old and abrasion and all that good stuff. The difference is how hot they can get. And so if you are the polyester typically doesn't like temperatures over I think like 150 Celsius. Whereas the captain can handle like 500 see something ridiculous? Yeah, it does six minutes through a reflow. Oven. No problem. Yeah. And so if you have a if you're stickering, before the process before a hot process, then I would recommend polyester. If you're at the positive polyimide are kept on. And if it's afterwards, like right before QA QC, I would use polyester, because the price difference is like it's depending on the manufacturer, but five to 10x Difference per sticker.

Well, not only not only is it just a multiplier of price polyamide even though they're readily available, they are far more difficult to get you have to spend in different sizes. Yeah, yeah, yeah. So if you don't need the hot process sticker, then don't use it. As you say the readily available thing.

That's actually I think, I don't know, I haven't read an article about it. But I've noticed the stock on Mouser and DigiKey is like one usually they have like 10 or like their cases of these labels. I bought like the last roll of like this one kind. Everything. I can't find it anywhere else either. Yeah. So I'm like in delete backorders like eight weeks right now. So. So yeah, it's, uh, I wonder polyimide material is, is getting hard to get right now. Like the raw material make these labels? Maybe? Paul? Yeah. Paul yesterday, just tons of it right now. So doesn't seem to be a problem there. So yeah, it depends on that. And then what you also you said is, y'all have that rework step. And so if you if you find out that your product has more free work required, like a product like that, I would probably even if you didn't have a hot step, I would go with a polyimide. Because then you you don't have to worry about you know, shrinking that label.

You know, one other thing and this is macro fab, that has their own method of doing this. But but you could certainly implement this into your process, especially if you're doing like a large volume thing. If you have a PCB array of your, your parts, like say it goes into the pick and place machine, you could sticker the whole array on the rails of the of the of the whole array, such that you have tracking of an array, but as soon as they're actually populated, then they break apart and get their own individual stickers.

Yeah, you could do that you can do that gives you kind of like batch level almost in more granular than a batch but not as granular as a PCB during the the SMT phase. Yeah.

So kind of actually what we were talking about last week as well. I feel like we're jumping all over the place here. But in your test procedure, or in any kind of assembly document where you're talking about stickers, make sure two things First of all, If you tell someone where their sticker goes on the board, and second, then it has space on the board. You're not putting a sticker over parts. Yeah,

well depends. Like one product I'm working on right now it goes on a component a big, big MOSFET. Actually, because that's actually the only spot on this board, do you put a sticker?

Well, okay, so that does happen, make sure it has a reasonable spot for it to go. Yes,

it is reasonable because it's, it's as a spot, this is actually a thing, it's a spot that's bigger than the sticker. Because if this these stickers in there, they have really, they usually have like a acrylic based adhesive. So it's really good at sticking to flat surfaces. But the moment it has the overhang, and you bump that overhang, it's kind of like, tempered glass. Tempered Glass is really strong when you hit it like on the flat surface, but you hit it on the edge and merely shatters. Maybe that's not the best analog. But it's, it's like that it's what they have when they kept on stickers, or the polymide stickers, like on the edge of your board, you can easily flick it off by just like hitting it on the edge. Whereas if it's like stuck down on, on a flat surface, it's really hard to peel it up. It's kind of weird. It's probably the difference between like a shear motion on the adhesive versus like a what would you call like if you were peeling it off? Like, like chewing gum? Anti compression? Anti oppression?

Yeah, I don't know. I don't know what that's got. I'm not a mechanical guy. Yeah.

Expansion,

you know, and kind of going on with with labeling individual components. Sometimes it's nice to like when you flash a component, have a printer printout saying like, this has been flashed to the Rev. And you stick it right on the on whatever poets programmed. I've also I've also seen, like on qf NS QR codes that are the size of the QFN gets stuck right to the part. Yeah.

So yeah, it just really depends on what you really want to track, and what kind of information you want on your board. But having more, when you go to your CME having a little bit more information, then I want stickers. And there's a lot of people that just say that too.

You know, okay, so we're, we're kind of broadcasting out that CMOS have this capability, and most CMOS do and most OEMs will work with you, even if they don't have that capability, they'll figure out a way to, to, ya know, one thing to keep in mind, though, is like, CMOs are not just, they're not just there to do anything you want them to do. And so be reasonable with what you're asking for, like, they're not going to want to print a sticker at every stage of the assembly and stick it on there, like and then also does your product actually need to the sticker for every stage in? Well, they will do that they

will just charge you for it. Yeah, but they'll

also complain about it. And here's another one, I've actually seen this before. And it's super annoying. Having a test procedure or an automated script that goes through and test your fancy widget and gives a pass fail, that's a great way for your to label your product because the computer can like oh, I detect a pass, print out a label, go ahead and stick it on. Great. I love that. But I've also seen the opposite where it prints a sticker anytime there's a failure, and you end up printing bazillions of failures. Because if you have the wrong cable connected or something gets off, yeah, or Wi Fi goes down, or blah, blah, blah, whatever, a bazillion things and you end up throwing away all of these tickets for failures. So maybe you need that. But just because you can doesn't mean you should. And that's a great situation where it's like consider Do you really need to print failed stickers? Because what's the point of printing a failed sticker? Well, you're going to stick it on the board and then hand it to some what like, I don't get the purpose.

Yeah, a better way. What I like to do when when someone comes to me and says I want stickers on my board for tracking for testing, is well I will typically do is recommend kind of like it's called pre loading. I call like pre loading the stickers. It's basically I have a database that has just unique IDs. So basically when when the start or the tracking needs to happen. All I have is a roll of QR codes that are how all unique IDs depending on how many we're building. And usually I just picked like a couple more digits up there up the hex code. So you have like, you know, possibility is like 6 billion combinations of numbers and letters right. So it doesn't even matter because it doesn't matter really as long as you don't want like incremental 01234? Because you don't really need that all you really need is a unique ID code. And then you look that unique ID code up. And you can see what the system you know, did it test that one first. Okay, good. That one's serial number one, then. Right. And so you go that way. Because that way at the beginning, you can just pre print a whole bunch of stickers and don't have to slow down your test procedure or your or your production, because like the printer messed up, or like you ran out of material, like if you actually ran out, like ribbon or something. So you just in that way, that person doesn't even have to wait on the printer, the person is just sticking on line, sticking the stickers down. And then and your test procedure, that's when you scan, it puts an entry into a database. And then you can say, Oh, this one failed. This one passed, this one, you know, V out was 5.832, vaults, stuff like that, you can you can put anything associated with that unique ID code in that database.

Yeah, for sure. And on top of that, if only only I would suggest only caring about serial numbers to a certain degree, as in, if, if you need tracking of serial numbers, then build a system that works with that work with your CM to build a system that allows you to track those things. But like Parker was saying the, the discrete individual serial number doesn't matter shouldn't matter as much. So in other words, like if you have packing instructions, or it's like play serial number x through y in this box, that would be a huge ask for a CRM, because they would have to go and find those serial numbers or they'd have to be extremely organized and good at putting the correct serial numbers in. And then it gets into like a giant mess if something gets fails, or we can't find them or whatnot. So use serial numbers and stickers and tracking where it makes sense. But don't just try to micromanage every single unit through your see.

So we do have a customer that's very similar to that. But we solved that problem with because they wanted the serial numbers on the outside of case packs. And which is so how I sold that is I made a little little web app. This is actually way back when when I was talking about building web apps to print Zebra printers, zebra labels, and that was actually for this. And so you would take so you'd have all the boxes that would have the serial number on the outside. And you would just scan all eight boxes that go into case pack and then hit print. And it would print the case pack label with all the serial numbers on it. So then you don't have to go. You don't have to go find consecutive boxes and then put them in the case pack. You can just go boop, boop, boop, boop, boop boop boop and put them all in.

And that's so that's a really viable option. Because that's not dependent on specific serial numbers.

Correct. Like, it is slightly different than what you were saying. But yeah, that's the way if you wanted serial numbers on the outside of your case back, because there's a lot of use cases for that. Sure. That would allow you to do that easier. So on the database side, the tool I've been playing around with a lot to do that is this application called Air table. And I like you can do this with like Microsoft Access, or you can do Google Drive spreadsheets. I've done that before. Google Drive is not the easiest thing to make a your own self rolled application talk to, I just did that, unlike I all I wanted to do was like upload a file to Google Drive. That took like half a day to figure out. But, um, so I've been using this application called Air table. It's like a spreadsheet, except it has. It's like spreadsheets plus, I guess it allows you to do like, length lookups. And the cool thing about it is the, it has history for each cell. And you can make like reports and a bunch of other crazy stuff I really liked because it actually just gives you a API endpoint for like, a spreadsheet. And so you can go, Hey, make an entry in this spreadsheet with an API call and put this information in it. So you can do stuff like scan a label, and then the test report sends an API link up. So instead of having to do like these hack around workarounds, on like, filling out like a CSV form, like table and like uploading that to some like MySQL database, this is just so much easier, at least for me. In my capabilities, this worked out for me. And I like it because I can hand the I can hand like the login information. Well, not my login feature, but I can have the customer can have their own account that is also can look at this information.

Yeah, you can you can build custom views and dashboards for them. Correct.

Yeah. And it's also great for, like the rest of the engineering team, because they can just, they can easily look at this information as well. It's just been pretty nice tool, not, we're not sponsored at all, by this tool. If I find something better, I'll probably immediately switched to it. By like it a lot. So far, it's been really easy because the airtable, the API is pretty easy to use.

You know, it's funny, so we use air table at WMD. For okay, it's, it's interesting, we have, we have an MRP system at WD. So that handles all of our builds that handles our bill of materials. And it handles, you know, material allocation and consumption and things of that sort. But it's, it's just MRP, it's pretty bad at at saying, hey, Steven is doing this right now in Billy is doing that and say rose doing this

task management.

It's awful at this? Well, it's MRP. It's not supposed to do that. So we have tied together our MRP system and air table, because the owner of the company has literally not been in the shop for a year. And he doesn't want to call every single person to say what are you doing? So we created an air table system where everyone can report what's going on. And we can move cards through our manufacturing process. And he can just add a glance, see wherever there is going on, and he doesn't have to ask any questions of anyone it and what's what's great is we've been able to eliminate these onerous meetings where we're all discussing whether what's happened with like stand ups? Yeah, we've just completely got rid of them. Because everybody just updates their information. And then the owner if he rarely has to ask questions anymore, because it's just right there in front of them. So yeah,

that's a very interesting use case for air table, because we haven't we use JIRA for task management, which has its downfalls and greatnesses. Well, we're

not a software company, and we don't have software people. So we don't use JIRA.

Yeah.

That's interesting. I wonder if we have to take a look at that. See how well that works. In our table. We also have, we use air table now for all of our RMAs, which is great, because we have multiple people doing repairs, including some in house and we have some people who work from home who get our amazing go home and take them and diagnose them. So it's really it's really, really easy to work out who gets what, what are what are like attempted fixes and things. It just Yeah, turns out nice.

Alright, let's move on from the macro fab airtable, gush over. That you use that tool, too, I have been starting to like it a lot.

Yeah, it's convenient. Alright. So I got I got a question real quick for everyone. Because I have an application that I'm looking for a solution for. And I would have thought that there would be a simple solution, but I'm kind of running into a brick wall right now. So I'm reaching out to the hive mind to see if anyone has a solution. I want some test equipment that basically acts as a low frequency network analyzer. So in other words, I want to basically plot frequency responses of my circuits, so that I want to inject a signal, read the output, and just plot my my frequency response. And, and I'm just coming up blank on good methods to do this. And I say good as in like, I would prefer to have something do it for me, because I could do it manually, right? I could put a function generator into a circuit, and I could step it by known amounts, and then measure my amplitudes and plot it myself. But having something do it for me would be would be preferable. And so network analyzers are readily available, and they're cheap. But 99% of the time, they they're network analyzers for high frequency that, you know, they they start at, like 100 kilohertz, I would like a network analyzer that starts at like, one hurt, and goes all the way up to like 100k. And it doesn't have to be blinding, ripping fast. Like, if I want to do a sweep from one hurt to 100 kilohertz and it takes 15 minutes, that's fine with me. And it would probably have to take a long time because it's acquisition time to get low frequencies would have to be through the roof. I understand that. But it's I'm surprised that those don't really exist, as in readily available, and perhaps I'm searching for something, I'm using the wrong search terms. Because but Well, I'm just not coming up with with anything. So if anyone is aware of a tool that will do this, and I would like it to be moderately accurate. So in other words, like, I don't want to use a microphone output, or a speaker output on a computer and a microphone input, like, those are cute, and those are great and all but I'd like to have some. I'd like to have something more accurate than that. Let's just put it that way. It's something that that actually reads, voltage. And months and months and months ago, I had Parker send me the arachnid labs tsunami, which was a Kickstarter.

Yeah, it was a Kickstarter is one of the it was in the top first 10 products we built at the fab.

Yeah, and this is actually Parker's personal unit here that I have. And okay, so this device that I'm holding, it's super cool. And it's, it's effectively exactly what I'm looking for. But it just doesn't work great. And it doesn't like, okay, so effectively what this is, is a network analyzer that does zero to two megahertz. wide sweep range, which that's perfect. And it runs on an Arduino core, which, okay, fine, whatever I can deal with that no problem. It has a DDS on the output, such that it can produce pretty not pretty but pretty damn accurate frequency output, and it, you're able to control the output amplitude. The thing about it is it it's kind of noisy, it doesn't do acquisition super well, I had to write some monstrous averaging filters to get anything reasonable out of it. And, and it doesn't really like like other solutions, it seems to work a lot better at higher frequencies past the frequencies of that I'm interested in. So if I want to plot the frequency response of one of my circuits, it seems like right now the best way to do it is to just sit there and do it. Do it the college lab way where you get a frequency generator, and you just step it 50 hertz, and then write down what you do and 50 but it just seems like there's got to be something out there.

This sounds like something a, like a, like an NI LabVIEW. Yeah. Acquisition module would be good for,

or we've talked about this before. Did you ever do with the NI ELVIS or whatever.

I there was a lad that used that at my school. But I was not in that lab. I just saw it. It looks fancy. It's basically

just a breadboard that connects to a computer via USB. And I say just like, I mean, we used it in one of my electronics classes to plot frequency responses. In fact, the NIA Elvis did exactly what I'm looking for. But the NFL was was kind of garbage because it was beat up by college kids who didn't know what they were doing. And and like, so our our filter responses were just terrible. So I don't know, maybe like I said, I'm just confused. I may perhaps I'm putting the wrong search terms in. But hive mind if you know anything that suits what I'm looking for, please let me know. And it would be awesome. If you know, it's it's output capabilities where say, I don't know, if you could output five volts peak to peak. And it's input capability was 20 volts peak? I don't know. I'm just throwing numbers out there. But

you can get beat up and I Elvis's on eBay for two for $300.

Yeah, no. I remember I remember plotting, just first order RC filters on those things, and then having to write lab reports on them. And I remember my tea I have a TI my TA even telling me he was like, yeah, if you see anything weird and the graphs just right in there, literally this is a anomaly due to the NIA Elvis. And so like our Labra Yeah, our lab reports were completely filled with like, well, this is what it should look like. But you know, and I Elvis or TIA Oh yeah, it was

terrible. Or is that on the back of it? And I Elvis,

I don't remember. It's not a

it's better than a parallel port. What do you hook that into? Oh, is

it a GPI B port or whatever? Something like that. Yeah, I don't remember. Anyways, I think it's those were those are a very cool concept. And of course, like, of course, my mind is gone. Like, I could just design what I'm looking for. It wouldn't be that hard to do. But I'm like, no, no, you can't do that. Like, I'm not I'm not allowing myself to even think about that. Because it wouldn't ever have

Yeah, exactly. I. So I found this article earlier this week, all about circuits, and a size does matter, world's smallest components hit the market. And it's an article that's covering, like new chips that are really tiny, integrated circuits that are really tiny. And it kind of touches on this one subject that actually want a deep dive on its world smallest claims can come with trade offs. And so what I really want to go dive into is that statement, and what things as a designer and a manufacturer, you have to worry about when components get tiny, as they keep getting tinier, as it yeah, as I keep getting tinier. And so right off the bat, the first thing I came to my mind is, design rule checks for your PCB. As you get components smaller, you're going to have to tighten the tolerances of your traces. And of your of your drill hits your registration hits your solder mask registration, because your your your pins are smaller. If that your pin is like the width of your solder mask registration, you might not even have solder mask, you might have solder mask over your pads. So going to different PCB design rule checks is like the first step when you start getting into smaller components, which directly drives PCB pricing. Oh, go figure. Yes. It's like the number one thing that drives your Well, number two thing that drives your PCB pricing. Number one's like how many layers you have.

Actually, you know, isn't that funny? How many layers you have? Not physical size?

Correct? Yeah. Yeah. So yeah, different PCB design rule checks, directly effects in your PC Parts. So that's like, Okay, if you got a product that you're using smoke, if you can get a basic you can get away with using slightly larger components. Do it. If you can fit in? Oh 603 component on your board. Use that. Don't go out and buy the Oh 105 component and put it on the board. If you can fit in? Oh, 603 use it right now. Oh, six oh, threes are like the cheapest component right now. Like package size. Oh, eight. Oh fives typically are more expensive per value. And then oh, four twos are slightly more expensive per value. Right?

Down at Oh, six. Oh, yeah. dips down. right in that sweet spot. Yeah. And then And then beyond that, it goes

way up? Way, way up.

Yeah. Well also, like unless you just unless you just love soldering through hole components with it, I'm sorry, surface mount components without a a scope. Don't like there's no point in just like peppering your board with like 1210 size components, when you could just put Oh 603 Yes.

So that's the first thing is is don't go smaller if you don't have to, because of these reasons, which are increase PCB pricing, because you probably have to go with more layers to fit those components on there. And increase tolerances for your for your design rule checks. And then next thing is okay, you start increasing the density of your pins on your components, you start getting to the point where you can't get better PCB tolerances. And so you have to start using other PCB technologies like via and pad which is it's evil evil in quotes, which is basically a via on your board a hole in your pad, where the lead of your component is at ServiceNow Park which doesn't sound like a bad idea, but what happens is those holes will suck heat away from your lead and also suck the suck is the wrong word draw is a better term it draws it wick caliper was it capillary action capillary yes capillary action wicking action basically ill wick the the solder away from the joint of your of your lead. So, what vn pad does is you basically fill the via with with a non conductive or conductive epoxy. The difference between that is one can conduct electricity one cannot reason why you have differences is non conductive is typically cheaper and so usually go non conductive unless you need current cable current capability flow Go through the news conductive. There might be other reasons why, but that's the only thing I know of. And so you fill it with epoxy, and then you plate over it. And so the via is like, can't be seen. It's like plated over so no solder can go in there. There's other reasons you would use the impact and leave them. Excuse me, leave them on, filled, like big ground pads underneath components, but that's for a different time. So you'd have to start using V and pad to escape components, like Denta, fan out the part, you have to use V pad, and then that's a price added, basically, you pay per via the philam.

It gets expensive very quickly. Yes.

So that's the big thing there. And a long, long time ago, Stephen wrote an article on our blog, that is like, still our number one article for people hitting. It's about escaping BGA. So we'll put that link. And that kind of covers this. It's like it covers tolerances and calculate how to do it. How to do it. Yeah, that when you need to increase your,

yeah, increase your layer count, right? Yeah. So for, for every image for every two layers of balls on a BGA, you add two layers to your board. So if you if your BGA only has four balls on it, you only need a two layer board. But as soon as you start to get more and more, for every two layers, add to two rows of balls, you add two layers to your board. And so when you start getting into those bigger BGA, you can count and estimate right away how many layers your board is going to be to that that's if you need to have access to most if not every ball on your BGA. So at least helps you kind of get a guesstimate as you're looking for parts.

And then the other thing with that is you also get to a point where vn pad won't be enough and you have to switch Yes, you have to switch V and pad and also micro vias and are also known as like,

basically blind vias. Micro vias and blind vias are almost the same thing.

So that what that means is it's a hole that doesn't go all the way through the board at that point, and so that you could stop at a certain depth. And that way, you can have traces on different layers, like to basically start running to annular ring tolerances, tolerancing problems of these vias, and you need to stack them closer and closer together. So you need to have annular rings on different layers and not be on other layers. Yeah, it starts being like this whole 3d Tetris puzzle of annular rings and traces underneath these packages.

It gets messy really fast.

Yes. So that's another thing you got thinking about when you're starting to get small with the components. The other thing that people don't really think about is silkscreen. When you're starting to do these, I call high density designs, you get to the point where you can't put silkscreen on the board that's meaningful of like we're see one or you 12 Is that. So so it becomes your documentation package that goes with your Gerber's or your ODB plus plus file becomes even more important, because you don't have designators on your board that tell you where the parts now you have to know you. One is here because of a text file that tells you now and that's usually like a part placement file. Most EDA tools generate that now. That's actually like the Microsoft platform. Required requires that those kind of files, so it's not too big of a deal. And depending on the board's like assembly documents that go that like a PDF that will show where those parts are at or at as well. So it's more of a visual guide than a text file. And then kind of more on the CME side is TruLine starts to get more expensive. So at macro fab, we've got a machine called paste jetter micronic paste jetter and so what that allows us to do is for small orders to not have to pay tooling fees for a stencil. And because it basically think about like a a laser inkjet printer that instead of depositing ink on a piece of paper, it actually dive bombs, solder paste at the locations on the board, and it's actually really cool you watch the videos of these like the how it works. It like releases it ahead of time and has like purchase Next where the solder paste is going to land on the board and a slight parabolic arc. It's pretty cool stuff. But the problem with that is that it's only there's a minimal dot size, and it's 12 mils in diameter. And so we on our end is we have a DRC tool that will look we're looking at the pace, we're analyzing the pace. And if your dot size is smaller than that, then we require a stencil. Because we are paced gender can't pay shut that small, the size. And so you have to pay for your tooling for a stencil. And so it's not that expensive and high volume, like you know, you're playing a couple 100 bucks for a stencil, so it's not too big of a deal. Now, let's say you get an even smaller of components, smaller leads. Now you're talking about more expensive stencils because you need special coatings that the pace can actually release because when you get a really small opening, I mean that pace is like peanut butter it wants to stick around on the stencil, it doesn't really want to release off the off the stencil and so you need like there's like nano slick, gold slick does a lot of different like name brand. Yeah, the

electro washer electro polish on all the apertures. On top of that if your board incorporates items that have really, really tiny apertures, but then some that have very large apertures, you end up having to get step stencils, which have different levels and thicknesses and heights, such that you can properly apply the right amount to the big thick pads and just the right amount to the small pads. And anytime you have like specialty custom stuff just like you know dollar signs appear in your eyes.

But truly, it's not it's a it's a cry once kind of thing. For this kind of stuff. You typically buy it once and then it's a cry every prototype. That's what Yeah, crying for every new prototype face. Yeah. And the thing about actually, that's one thing that has popped in my mind about step stencils that I haven't actually thought of or looked up yet is wouldn't you if you had a step stencil, so you had an area of the stencil that's thicker. So let's say it was a good one area that is got the smaller apertures. So you have a thinner foil, let's say we were doing like two mil or three mil foil, and then most foils or four mil thickness stencils. But let's say you had a big MOSFET that needed a lot of solder pad on its on its on its a, it would be its drain

depends on the situation

depends on the package. And I think most of them are drained the big tabs that drain on like a to a to 220. So you need to put like a six mil like stencil there, which is that's ginormous, but you know, can be what you need. If you tried to put those components right next to each other, you're going from a six mil thickness to a two mil thickness stencil. And so you're going to have this like that, because how the stencil machines work is it squeegees it across and well now there's a gap in that squeegee. How do you handle that? I don't know. Ah, do you just not worry about it? Like you don't have to worry about like the shadow of like the squeegee step.

Yeah, the the amount. So we deal with that, like we use step stencils quite a bit now. Because we have a lot of surface mount headers that we use for board to board connections. And so you have those sticker we do build up on almost every PCB SMT header

that we have, what thickness Do you normally run there then think we do

four or five or five, six, somewhere in that range?

On the headers, yeah, then a four everywhere else. Yeah. Okay.

Yeah, but we also do the polished stencils. And we just high dollar. Well, you know, I mean, we do enough volume that they're not

too big. Again, it's one of the it's a cry once. And it's like okay, we just

purchased a brand new printer in December. Got it online last week in December. And we've been running that and it I put I put these these boards under the fresh pasted boards under the scope. They're flawless. I mean, like if you have a corner, a 90 degree corner on an aperture and you look at the paste, that paste has a 90 degree corner on it like this stuff comes out. Absolutely perfect.

The solder balls are 90 degrees as well. Yeah, yeah, like they Yeah, if you pressed if you want the heck if you want like, I don't know, some random shape aperture and you pasted that it would come out perfect. I'm talking about that in our Slack channel a couple weeks ago and Different we're talking about like, different aperture shapes for solder paste. And because like there's no normal squares and there's like a home plate design. And there's a C shape and these are old designs of paste to kind of like help eliminate tombstone mean. And I came up with the with the star shaped

starship What was that for? I can't remember.

Well, someone it was, it was, I think was Tom Anderson. He mentioned one company was doing moon shaped like crescents. Oh, that's right. Yeah. So I'm like, Moon star. So we're about to start five pointed stars. And that would be like the best shape? Oh,

yeah, go figure. Yeah, cuz it's

random tangent? Yeah. So yes, I guess I guess you don't worry about then. I don't worry about what you don't have to worry about, like, the shadow of like the step,

you know. And the funny thing is, okay, so the answer is no. But maybe we should I haven't had to yet. I haven't had a problem with it. And we do have, you know, as small as Oh, four Oh, twos close to these surface mount headers. And I haven't run into an issue with the

yet. Yeah, I guess it would be. This might be asking a little much, but like, next time you run the squeegee over it and see if there's like any pace shadowing in that step region? Or is that pace jetter? Not patient or up pace? Printer, apply enough pressure where it can squeegee that out of that? That step?

Well, that would defeat the purpose, wouldn't it like because then it would just depress the the build up area?

It could be the blade is flexing?

It could be it could be I doubt it, though, you know, I need to look at it the next time. The next time we run one of those, which we do. I'm sure we run one of those multiple times a week. So yeah, I'll I'll see if I can if I can take a look at it. Because now I'm thinking about it, I don't know the answer to that. We do get to, we do get to do we have control over how much force the squeegee presses down on the board. And we actually have we, we got this really cool

new feature on this printer, where it has a bed underneath that has, it's like a bed of nails, but they're all but they're all on little plungers. And what it does is it applies like your board comes in and gets clipped in by the conveyor. And then compressed air pushes all of these plungers up. But it's it's very light. So it just makes contact with everything. And then these pins come out and lock all the pins in place. So it gets this 3d plate that your board can rest on such that when you do put pressure with the squeegee, it doesn't deflect the board at all. And so you know, the first time it goes through whatever the bottom of your board is flat, but if you're doing double sided assembly, now all of those plungers conform to all the parts of the board. Yeah, it's

super cool to cool. That's cool. So next time I'm in Denver, I'm gonna have to go look at that machine. Yeah, it's it's really cool. And I've, you know, the

the, the micronic that Mac fab has I was always astounded at how fast it can paste a board. But then I then I see one of these machines, it's like eight times faster. Yeah,

well, that's that's the trade off is the micronic pace shutter is rolling at all, but it just takes a little longer. And there's a limitation on how small of an aperture right but the moment that you're like okay, we're going to print like 100 panels. Yeah, the it's a no brainer to go to a stencil machine.

I mean, yeah, absolutely. Like in our assembly process. The stencil printer is always waiting on the pick and place machine. There's never like a pause the other direction.

I really want to see that. The fingers. I can't remember

the name of it. But yeah, it's super cool.

Well, we only talked about 30 minutes with stickers. Not not the full hour. Not full episode.

Cool. Well, with that that was the macro of engineering podcast. We were your host Stephen Craig and Parker Dolman Take it easy.

Later everyone thank you. Yes, you our listener for downloading our podcast. Do you have a cool idea project or topic? Let Stephen and I know Tweet us at Mac fab at Longhorn engineer or at analog EEG, or emails that podcast at Mac fab.com Let all your friends know about the podcast and also check out ours Slack channel you can find it at mcwrap.com/slack I think we just reached over 555 a 500 timer people actually went 556 now so we're dual 555 So see you there