Sep 18th, 2018
See all the show notes here.
Julie Ellis started her career as a representative for a semiconductor manufacturer after completing her Bachelor’s in Electrical Engineering. Now she is a Field Applications Engineer (FAE) at TTM Technologies, the third-largest circuit board manufacturer in the world. Listen to Julie and Judy discuss seamless global transfer and recommendations on working with offshore fabricators. Learn how to avoid excessive technical queries and how to migrate from prototype to production while optimizing global processes.
Bonus update on AltiumLive: Julie and Carl Schattke will be presenting at AltiumLive 2018, introducing new stackup and impedance tools in Altium Designer 19, so be sure not to miss them!
- Julie Ellis did a presentation about Documentation at AltiumLive 2017.
- What is Seamless Global Transfer? Transferring PCB manufacturing from onshore prototype level into production and offshore.
- Julie started her career at Hughes Aircraft, where she completed her Electrical Engineering Bachelor Degree - best decision of her life
- More women (not just circuit board barbie) need to get into STEM! #WomenInTech. Julie always encourages young women who are interested in STEM, to get a degree that will enable them to move into fascinating jobs with a variety of opportunities.
- Julie’s first job was as a semiconductor manufacturer’s representative; realized she liked the circuit board side of the business more than ICs and migrated over.
- On TTM: It’s like working at Google for circuit boards, I can always call someone for answers about manufacturing best practices.
- Seamless global transfer - the concept is that you aren’t just designing for the prototype but for global manufacturing i.e. avoid 100 technical queries
- What makes migrating over such a difficult process? Because the 6-Sigma 6Ms, are not the same when it transfers over to Asia.
- What are the 6Ms? Method, Mother Nature “Environmental”, (Man) People, Measurement, Machine, Materials.
- Equipment sets are different for mass production, production lines are longer, there is not as much human oversight, production lines must be scheduled and you cannot stop/start the process. The tolerances are different and they need to be accomodated in the designs.
- Throughput and drilling is always a bottleneck and to reduce this and reduce turn time, mass production sites have tweaked processes to get the highest yield.
- Internationally the general rule is 4 mil lines and spaces on half ounce copper; 10 mil is the most common size drill which results in an 8 mil finish hole size.
- As you go up in copper thickness you need to add a little bit to the pads.
- Blind vias are the ones that are on the outside but end up on an internal layer.
- Buried vias are buried completely inside the board.
- Working with offshore production house while still in prototype development phase.
- Recommendation - design for volume and technology. Qualify the design for the final production region and technology.
- HDI (High Density Interconnect) is anything 0.4 mm pitch and under that has a track running through the pads.
- Judy wants to throw everyone inside a fab house!
- There are at least 30 different processes required to manufacture one 4-layer board.
- Julie works directly with Carl Schattke and they will do a stackup presentation at AltiumLive 2018
- Materials are a significant cost in Asia, whereas here in the states the material is less of a cost (20% in USA, 50% in China).
- With production panels where you're trying to get as many cookies cut, you also need to consider and discuss with your manufacturer the tiny 2x2 inch pieces.
Links and Resources:
Hi everyone this is Judy with Altium's OnTrack Podcast thanks again for joining. We're happy to have you again.
I would like to continue to invite you to AltiumLive, and I also wanted to put a shout out that we have a call for presentations right now, so if you are an Altium Designer user, and you have some tips or tricks or some kind of breakthrough you've had on design please contact me at Judy.firstname.lastname@example.org and I'd love to hear from you ASAP.
Once again I have another talented and amazing guest with me; Julie Ellis from TTM technologies which as you know is one of the largest board manufacturing companies not only in North America but in the world today so Julie is an FAE at TTM and a very well respected technologist as well as a dear friend.
So Julie, welcome it's good to have you.
So Julie presented at AltiumLive last year on documentation. I've sat through many of her talks and learned much from her, so today we want to talk about what it takes to move jobs from onshore prototype level into production and offshore. She calls it seamless global transfer but before we get into that we'll hear a little bit about Julie's background. We both started in the printed circuit board industry in the 80s - which dates us a little bit I know - we're not going to say the year we're just gonna go with the round numbers the 80s but I always...
-we were child savants though so we say we were 12
-we were 12
-okay five, yeah we were five.
So Julie just came in and noticed my super cool Career Barbie of 2018, which is a Robotics Engineer Barbie. She's got circuit board patterns on her shirt and a laptop, she kind of looks like us, so we're just gonna call her Circuit Board Barbie and you know blondes... smart ladies you know. Finally, there's a Barbie we can really relate to, and we want women to get into science and STEM and everything - so go for it and aspire to be this Barbie here.
Right on, yeah! Girl Power. We want to get more women in here, and it's just about exposure and motivating others, so we hope that throughout this podcast we inspire maybe somebody to give a girl a little nudge out there. We've enjoyed long enjoyable careers. So, okay Julie before we get started, why don't you kick off and tell our audience a little bit about who you are and your background - how you got into this wonky industry?
I am Julie Ellis; I started as a Design Engineer at Hughes Aircraft Company. I was awarded a student engineering scholarship there, which paid for most of my schooling - the rest of my schooling after I moved out here from Iowa - so I always tell everybody that getting a Bachelor of Science in Electrical Engineering was the best decision I've ever done in my life, so I really do encourage people. If you're interested in Math, Science, Biology - anything - get a good STEM degree so that you can always move forward into interesting, fascinating jobs with a lot of variety of opportunities.
It's a really, really good way to go and I encourage youth and people to get into this kind of field. So I started as - once I graduated from Cal State Fullerton - I stayed on it until the 1990s when things got really tight in the military market, and I was on loan to one department, but I couldn't hire in, so I took the first job that was offered to me as a semiconductor manufacturer’s rep and I had circuit board industry or circuit board experience at Hughes and as a rep I also had a couple of circuit board lines and I really, really liked the printed circuit board side even compared to the ICS and memory sales and everything. So I ended up migrating toward the printed circuit boards. Fast forward eight years, landed a great job at TTM as a Field Applications Engineer just a little over four years ago and it's been a fantastic opportunity.
TTM is the world's third-largest printed circuit board fabricator, and we would probably be number two if it didn't include Flex because the top two manufacturers have a lot more flex and rigid-flex than we do, so I'm surrounded by experts in this field. It's like living in Google for printed circuit boards because whenever I really want to know something I can go call somebody within my company and find the answer, so it's it's really good working here.
Really it's impressive, and you're right - like you really can go to anyone to get the latest and greatest information on manufacturing best practices which are really, really fun.
So we wanted to talk today about Seamless Global Transfer, and I know that we've talked a lot on this podcast about there's no such thing as the quick and dirty prototype so why don't we just jump off from there? Like what does it mean? So you design a board it's gonna go into production, but you've got deadlines, you need to crank it out really quick, you crank it out really quick and then it's like: hey it works let's migrate offshore!
That’s right - that way exactly.
That would be like the worst case scenario, like you heading for Niagara Falls and not knowing it. So why don't you talk about the myth of the quick and dirty prototype and why you really need to think about global manufacturing up front while you're just developing the circuit board and designing it?
Yes, so Seamless Global Transfer is the concept that you're not just designing for your prototype to get it through a quick turn shop here in the United States in five days. Because I worked - one of the numerous positions - was as a Circuit Board Commodity Manager and a contract manufacturer and a lot of the projects we got had already been tested and proven and developed here in the United States. They sent it to us for mid-level production, we’d try to send the parts overseas, and everybody would come back with a hundred technical queries and say: we can't build it because we don't have this capability over in China. Oh, you need to change this on the design - it's not going to work, and by the time you've given your job to a contract manufacturer your engineers do not want to make changes to the design that they've already tested.
So global seamless transfer plans ahead and thinks about: what is our migration path from quick turn development prototype and taking it over until long-term production and so there's a lot of background that goes into that, and that's what Judy and I wanted to talk about here.
So what is it that you think that makes that migrating over makes that a difficult process?
Because the Six Ms: man, machine, materials, environment - which is another M that I can't remember. Everything is not the same when it transfers over to Asia. The equipment sets are different for mass production, the production lines are so much larger and often much more automated, so they can't get the human element, you know. Watch this, watch that, we don't get the babysitting of our project over in China like we can here. In Asia or China, we have to schedule the production lines, and you can't just interrupt a line there to quickly throw this job in front of everybody else. The schedules are a lot different, the process tolerances are different, and because the process tolerances are different, we have to accommodate those in our designs.
Okay, so there seems to be a perception anyways that once we have a pretty robust design here that we can just kind of throw it over the pond. Why is that, I mean you just talked about some reasons but what are some like tangible snags you're gonna run into if you try to do that?
A lot of it has to do with the drilling. Like over in China most mass production shops, except for the really advanced HDI shops which would all go laser micro vias all the way through, as a rule don't drill using six mil drill bits because they're expensive, they break and they can't be re-sharpened and they break more easily and they have to be changed out twice as often as bigger drill bits. And bigger drill bits can be stacked, or you know, panels can be stacked. So if you can drill two or three panels at one time you've just got your throughput and drilling which is one of the largest bottlenecks in fabrication. You reduce your turn time significantly and time is money. What we're paying for in printed circuit boards besides materials, is the time it takes to get through the processes. So Asia and mass production sites have all tweaked their processes to achieve the highest yields, in the least amount of time, at the lowest cost. But there is a sacrifice to that and sometimes at the sacrifice of we need a better, bigger pad around a drill hole. We're going to stack two or three panels high instead of drilling a six mil drill and our plating processes are a little bit different so we may have to have more edge compensation. Which means that, that will drive a little bit larger requirements for line, width, and space.
So on those, is there a recommended - that's kind of a broad question - but are there recommended kind of hole sizes and pad sizes and/or trace and space sizes to help on the throughput? If you have it.
Yeah kind of the general rule of thumb internationally, is 4 mil lines and spaces, on half ounce copper is a good start. Anything under that on half ounce copper is going to be a premium. And ten mil is the most commonly sized drill which would drop you down to an eight mill finish hole size. And we'd like to see the hole size plus ten mil for the pad. So if you've got an 18 or an 8 mil finished hole size, we would drill it probably at 10 or 12. We'd like to see at least an 8+, 10 and 18 mil pad on that hole. That's just for a single lamination through-hole in multi-layer printed circuit boards. As we go up in copper thickness, we need to start adding a little bit to the pads.
Okay, and how does that change when you start adding buried and blind vias in?
It depends on the construction. If we're talking like a real traditional blind via board; blind vias are the ones that are on the outside, and they end up on an internal layer. Buried are vias that are buried completely inside the board, and those are different technologies. But so if we're talking standard blind vias where we might have 1 to 6 and then 7 to 12, both being blind via stack-ups, we would actually stack up the material layers 1 to 6, drill and plate, and then we would stack up the materials layer 7 to 12 - drill and plate. And then we would laminate all those together, and then we would drill and plate and etch the outer layers.
So those definitely have different rules because the two outer layers already have plating - additional plating - on the outer layers which means that we have to etch through thicker copper because of the foil plus the plating, and we're going to require slightly bigger line widths and spaces on that particular design.
So one thing we were chatting about as we were preparing for the podcast, that I thought was obvious, but also fascinating, is the idea of working with your - you know, I kind of want to move into now, sort of takeaways for our audience. So you were talking about working with your offshore production house while you're in your prototype development stage which I think is kind of counterintuitive. I don't know, is it?
Actually, if we are in the prototype development stage, it's the best way to do it because if - I always recommend that my clients design for volume. Whatever their final volume is you know, we all know the term DFM, but we really have to take it to heart to figure out, qualify the design for the final production region. Final production technology, whether it's a single lamination or a multi-lamination that's not HDI board like I just brought up, or whether it's an HDI board that has blind and buried vias, but with laser micro vias and advanced HDI board which I categorize as anything 0.4 millimeter pitch and under, that has a track running through the pads. So if you start at your before-prototype stage, qualified the design for the volumes and the technology so that you can pick your final production sites, get the design guidelines for those sites, get the stack up for those sites, and have the stack up and the design guidelines identified before you even route the board.
And if you do that then you're not going to route a whole board, send it over to China, and China is going to say: oh sorry those line widths and spaces, there's not enough space for us to compensate the etch and artwork during etch, we can't build it this way. Go increase your spaces, and if you don't have room on a tightly designed board, or if your pads aren't big enough to achieve the annular ring that you're asking for, your design is no good for manufacturing. So my term is ‘design for volume,’ but it means whatever your volume is. And the reason I'm doing that, or I'm saying 'your volume' is because we have customers that do 200 printed circuit boards a month, and we have customers that do a million circuit boards a month. And the shop that does the million circuit boards a month is not going to take the 200 circuit boards per month order, but they have a much higher level technology - so I can't design for that technology knowing that I could never run it in that particular site.
Right, so it's both volume and technology.
I feel like it's such a good service, in many ways on the prototype end, that we can kind of do push-button ordering now, but I also feel like what's has been lost is how complex the fabrication process is and I just wish - I want to throw everyone inside a fab shop. Because it's like when you - think you can just push a button and then a package shows up on your door; you know what I'm saying?
That every shop is a little unique is for a variety of reasons. It's not - for reasons that enable different types of technologies - they do it with high intention and lots of precision and all of that, and so you have to design for that shop. It's not just push-a-button and out it comes. Especially the more complex the board gets, so, on the one hand, I'm a fan to get the prototypes out fast, onshore when you can, have maybe available that kind of service. But on the other side, if you're going into volume, I don't know - I think it gives people sort of a false perception of what it's like on the other end.
Talk about - I think you mentioned this stack up; getting this stack up right at the... I really like that DFM right, design for volume, that was kind of a new concept to me that you introduced me to. So you're saying that the stack up should be kind of vetted and worked out with the volume as well as, what kind of board, what kind of technology buried/blind vias, you have the space levels to also work out the stack of details.
Yeah we need all that information to be able to create the stack up because most of those multi-layer boards with VGAs also require controlled impedance like for the high-speed digital that we're doing all the autopilot, industrial controls, medical controllers, everything seems to be working off some sort of USB and PCI, and we need to manage the controlled impedance. Controlled impedances based on line width, space, and how thick the dielectric is and to a little teeny effect, how thick the copper is. So we have to play all these together while creating a stack up and also keeping track of, if we're doing stacked or offset micro vias. We build those from the inside out and just keep adding layers, drill the outer layer down to the next layer, then on both sides then we add two more layers drill the outer layer down to the next layer.
But each time we do that, we have to figure out how we're going to plate those and how thick the plating is going to be and those are process variances are you know. There are process capabilities and variations from site to site, and there can be unintended consequences along the way, like putting additional copper on that outer layers - it's the more complex it gets you have these: if you do this, then this you know, what I'm saying there's so many!
Anybody who has seen my presentations knows that I always say that I'm always splitting hairs. Because a human hair is about 2.5 to 3 mils in diameter, and I'm always worrying about unintended consequences because if a customer comes in and they say: I want thick plating inside my hole walls you know, give me 2 mils of plating inside my hole walls. Well for one I can't think of one fabricator in China that would do that. The IPC standard for class three is 1 mil average plating in the hole walls. But the other thing is, remember whenever we plate inside the hole walls we're also plating the surface, the outer surfaces, the thicker those outer surfaces get, the harder they are for us to etch fine lines and spaces.
Well, why don't you just put it through the machine that just spits out the board Julie?
We need a magic machine!
If I could do that I wouldn't have to be here... I'd be somewhere on my own Island in Bora Bora...
Barbie we need a magic machine to spit out - maybe Barbie will get you to know either a Barbie plane and maybe she'll have a Barbie magic PCB?
That'd be great.
Then you know, in Barbie's world I think we'll just spit it out, I know - it's very complex and by the way. Let me stop right here and say that Julie helps every top brand that you could probably think of in Silicon Valley and beyond; helps them to do their stack-ups and come up with these you know, calculations to help work out all this hair-splitting and she's very skilled and capable. And that's why she will be presenting at AltiumLive with a senior PCB designer who she works with directly which is Carl Schattke, I cannot tell you what brand he works for, or I would get in trouble, but suffice it to say he's in Silicon Valley and works for the top electric car manufacturer and I am delighted that Julie and Carl will present on stack up on this very subject, and you couldn't get two more qualified people - I’m so excited that you're doing that.
Thanks, we are too - I think it will be fun.
It'll be really fun, and so they're so used to being deep in the weeds they'll be such a resource.
So before you move on though, it's not just the stack up, it's also the pad stack line, widths, and spaces that need to be provided to the customer with the stack up. Because we want to make sure that they know all of those design requirements before the board guy starts routing everything.
You talked about DFM and DRC's for final site and prep for the prototype. Is that - I just wrote myself a note here - have we covered most of that here?
Yeah, we have for the stack up and the design rules. But one thing I'd like to bring up is everybody's trying to stay competitive and because of the processes and the way that production panels are laid out in Asia. Materials are a significant cost over in Asia compared to here in the prototype shops. Here we pay for the quick turns, for the setups and things like that which are insignificant compared to those. So the material here is only about 20% of the average cost compared to 50% of the cost in Asia.
So if you can also plan your size to fit well up on a production panel so that like, imagine an 18 by 24 inch production panel, and you're trying to get as many cookies cut on that production panel, but you also want to think if you've got really small pieces your assembler is not going to be able to load those tiny little 2x2 inch pieces. Their conveyor equipment can't hold them, and it would take them forever to go through those linearly, so another really cost-saving exercise is to work with both your fabricator and your assembler to come up with a multiple up-array for smaller boards and also make sure that you know whether you've got enough clearance on the two long sides of your array, or your printed circuit board for the parts to be conveyed through assembly.
There's sometimes parts hanging off the edge which really makes things fun.
Yeah and that needs to be planned for in advance, whether: do you need an extra rail on a leading edge, because there's a big connector hanging there, or is the assembler going to put that on after the fact? But if you also take into account design for assembly - put all your test points on the board because once the board is designed and you can access test points, nobody's going to be able to go back in and design an in-circuit test fixture or functional test fixture and unpick those plates.
So don't just design for volume. Like I said really, truly design for DFX, design for fabrication, assembly, test, and long-term reliability.
Good, good, good, good advice. So can you give some real-life examples from your real life career? We won't name names of brands but suffice it to say there; you work with major consumer brands that if we could say names everyone would recognize them and tell us some of the, you know challenges that they had by actually not thinking about some of these ideas ahead. And these are the brightest of the brightest - I think what we want to share here is, everybody is challenged in this area, right? It's a challenging area, so we're not saying, oh we're so smart, and you know the audience what do they know? No, the top designers, the top printed circuit board designers almost in the world, are challenged by some of these issues. So just talk about some real-life examples and how it went wrong or how it went right?
Okay one real-life example in the last quarter was a major commercial customer like you said, they had worked with a - probably a Silicon Valley shop - they built their boards, tested them out, proved them, and they wanted to go into mass production. Their start date is like August, to start delivering mass production so that they can you know, start shipping their product. Well it turns out they had a design that had a six mil drill - mechanical drill through a standard thickness board with a ten mil pad and when I said, remember I said like, do your finish hole size plus ten for the pad, this only gave the hole size plus four, and it wasn't enough to make sure that people wouldn't totally drill you know, have too much because of misregistration material movement.
Every time you add a process, you add misregistration. Nobody in Asia would take this business, and we actually had to help the customer convert the whole design to another via structure type to be able to pull it off. And the way this happens is one of two things: if you're a major customer and you go to a, you know like a smaller shop, they are going to be so hungry for your business they're not going to say, no, no, no - we can't do that. They are going to babysit every single panel and put them in the drill machine by hand and make darn sure that they're going to get that for you.
Or there are probably a few select super, super advanced shops that are just doing onesie-twosie jobs and they can meet these kind of requirements, and these tight process tolerances, using direct imaging everywhere you know, using single headed drills for the production panel rather than five or six spindles that we use. And so it's not even saying that that particular circuit board fabricator was a bad designer - it's just that they're only designing for their site capabilities and probably pushing technology to make a big customer happy.
Right, and that may be their niche, that may be their niche market - but again they're not thinking particularly ahead, they're trying to help their customer be - - so it's kind of myopia. They're just designing for that, and they're great shops, they're great shops very, very capable, but not unless you tell them up front or you start this conversation up front it can go bad like that. On a consumer product that, okay it's August let's go into production and then wait, stop. Stop everything and the cost, the headache to that customer, they have to respin the board, run the protos over again and do all the testing over again. And now, schedules are lost, time to market is lost, you know so that it can become really painful very quickly and very costly.
Yeah very costly.
And I had another similar design that my customer had a design with 5 mil mechanical drills and 9 mil pads and most shops I know don't really drill mechanical 5 mils. So that was a tough one for him to go into production. So that's a real common example. The wrong size drill with the wrong size pad, or one that I just saw recently, was a really thick dielectric that still needed a blind hole and it was planned on being a laser hole because they wanted some big RF circuits on the outer layer. Which means they needed a thick dielectric and normally if you're using laser micro vias you have very thin dielectrics. And I was able to confirm that we can do it over in China but it's - it wouldn't have been my first choice for a design you know, and it kind of set me back but - but we were capable on that one.
Yeah so, you have a saying that I like which is: pick your experts wisely. So tell us what that means? What you mean when you say that; pick your experts wisely?
Well if you're going to listen to an expert, they're going to segue you to the path that they know, and if you pick the wrong expert and they take you down a garden path that nobody else can fabricate. I know that there are shops that they'll say: let's do this and let's have the customer design it this way because then they can't go anywhere else. It's a way to guarantee their business.
I can confirm that you know, I've seen entirely that. It locks you into that job.
It locks you into that job, and you know, I can see both sides. I'm like this ambidextrous Gemini so I can see both sides of the story. I can see an internal engineer wanting to secure future business for their location. But on the other hand, it may not be good long-term for the customer. And I'm in it for the long haul you know, I've been both sales and technical support, and a lot of times I have to work with customers to make slight modifications and design engineers; these are your babies. You don't want to have somebody coming in from the outside and saying, you know what, I really can't quite achieve that. Can we tweak your design a little bit? Who wants to hear that?
But if I have credibility, as somebody who's thinking for the customer, for the fabricator, and working towards the best solution long-term. I've - you develop trust, and you can get better work done. So, I choose to do the good path even though it probably means that I tell everybody I'm a conservative designer and so that means that if you design a stack up - if I design your stack up, give you the design rules, if you can meet them chances are one of my competitors can also do the work. Yeah, but on the other hand you know, the relationship most of the time means a lot.
Right it does, and not everybody has both the technical depth that you have, the integrity you have, and you have reached to top, top fabrication experts in the world. So that gives you a really broad perspective which I appreciate.
So Julie thank you so much. This has been so great, and I feel like we've just scraped the surface but I would like to invite our listeners, if you are available, to come to AltiumLive and Julie will dig into - she and Carl Schattke have an hour-long presentation plus QA and, will be introducing some new stack up and impedance tools in Altium Designer 19, and so they will be giving a really rich treatment of the subject of stack up. So if you want to hear more from Julie, come on out to AltiumLive, and we would love to have you. Thanks again Julie, it's always - I always learn from you every-
- - thank you.
Every time we talk.
So there is one other thing that we should talk about.
What should we talk about?
Okay the other background of seamless global transfer is that if you're working with a company that has multiple sites like DTM - we can take that - we can take the lessons learned from the prototypes, and transfer the data, and transfer the lessons learned over to the final fab site, so that it's not a new learning curve once it goes overseas. And that's a real advantage about really paying attention to this.
Right, which is a good point.
Do you transfer the learning curve along with just the data files right?
That's right yeah.
So good point. Okay, thanks for inserting that again. This has been Judy Warner with Altium's OnTrack podcast, and Julie Ellis of TTM.
We look forward to you joining us again next time. Until then, remember to always stay OnTrack.