OnTrack: The PCB Design Podcast

From Tesla to consumer devices, Mihir Shah has been a PCB designer among the best. Now, as Director of Special Projects at Royal Circuit Solutions, he is pioneering ways to make manufacturing easy for printed circuit board designers. Meet Mihir and Jon Lass, co-founder and VP of Engineering at Royal Circuits, and listen in as they discuss manufacturing best practices and share various insights on symmetrical stackups, solder mask, and copper weights. Get a wide range of PCB design tips and learn how same-day turn times on printed circuit boards is possible at high-quality board shops like Royal Circuits.  

Watch the video and read get all show notes. 

 

Hey everyone this is Judy with Altium's OnTrack Podcast. We are glad to have you back again, the podcast continues to grow and we thank you for listening and I know that you are tuning in because I have amazing guests like I have today so I would like to introduce you to my guest but before I do I would like to  invite you to connect with me on LinkedIn. I'd love to connect with you and share a lot of information relative to PCB design and engineering and also on Twitter I'm @AltiumJudy and Altium is on Facebook, Twitter, and LinkedIn so we like to have conversations with you, not just monologues; so please connect and make sure you subscribe to our podcast too so we can keep making these.

Today we are with a couple of great people that are involved in US manufacturing of printed circuit boards. I'd like to introduce you to Mihir Shah, who is Director of Special Projects at Royal Circuits. Mihir actually was an EE and has lots of experience - hands-on experience - being in the trenches and doing design work. His father right, it's your dad that owns Royal Circuits Mihir?

[Mihir nods]

And so his dad somehow sucked him into the manufacturing industry. So we're glad to have him there actually, we need more young blood and also, we have Jon Lass who is the VP of Engineering. He's also one of the original founders of Royal Circuits, so gentlemen welcome thanks for joining today.

Thanks this is great.

So Mihir, I'm going to start with you. Why don't you tell us a little bit about your background and then give us a little blurb on Royal Circuits?

Sure, so as you kind of alluded to I'm an Electrical Engineer. I started my career at Tesla Motors where I was working on a lot of the cool things with Model X, and Model X, back in the early beta days of that vehicle. I just got crazy hands-on experience learning how to design your own boards hands-on for everything, just do things quick and get a design approach to rapid prototyping, which was great. And then I went to Taser - now known as Axon - where we worked on consumer devices that are - I mean literally - the Taser device.

So I was more on the power electronics side of the Taser weapon and some things on the body camera so a really, really great experience albeit limited, but really great in the short time that I was a fulltime design engineer. And then you know, somehow, some way, my dad convinced me to join the manufacturing side of things.

Yay Dad!

I'm kind of - one of the roles is always like: look, you were buying boards and designed them for a while, now come here on the other side and try to make it as easy, clean, and simple as possible for people to order them now that you've seen often times, what a pain it is, or all the mistakes that you kind of made, or things that delayed the time, cost etc can help on the manufacturing side and now, Royal Circuits, just to give a brief overview of who we are, and kind of our main value proposition. We're a big US manufacturer of purely quick-turn, prototype, printed circuit boards. The whole idea is one, two, three-day turns in the Bay Area, same-day turns and weekend turns - totally acceptable, and all owned and operated by us, all here in the United States.

So we have two factories: one right here in Hollister where I am now, right in the Bay Area, and then we have a factory down in Los Angeles that's purely flex and rigid flex. We really focus on that technology down there, so we've been doing this for over 20 years. You know, we have our customers range anywhere from large tech companies like ones you've heard of, like Google etc. Students, Stanford, Caltech, UPN, MIT and military.We're ITAR certified, and then thousands and thousands of other customers in that group and more. So we do everything from simple two layer boards all the way to 30 layer, HDI  , High Density Interconnect, PCBs, fab and assembly - no minimum order quantity.

So really, really focus on the low volume, super quick turn, with an incredible focus on customer service and making sure that people get their boards when they need them and at the price that they want them at, right here.

Wow, Jon can you tell us a little bit about your background in the industry and your history at Royal and what you do there?

Sure, so my background has been Engineering for about 30 years. I started out in the CAM/CAD industry and was involved in the very early days of photo plotting when we used to build boards with films that have,   actual direct imaging as we're doing today.

My dad and I founded the company 20 years ago here in Hollister and we - like Mihir pointed out - it's always been about quick turn, 1 to 3 dating, prototypes, all the way from 2 to 30 layers and, very exotic type of materials and boards. So we've been around for 20 years servicing our customers and we still have some of the same customers 20 years later.

That's a good report card! That's an excellent report card.

That's how we look at it too.

Well my favorite stat about the company - just to interject - is that we really do have a 1 percent turnover in 20 years. I mean, I really encourage you to find another company in the United States that has such a low employee turnover. Everywhere I'm looking, Jon is a testament to that, people don't leave, we just keep growing here and in LA. And in some of the other kind of businesses that we run, same deal. Customers first.

Ok so I'm going to become a board industry geek for a moment but I want to point out something about that, that may or may not be obvious to our audience; but something that I've noted when you do work with a board house that has low turnover, your quality remains consistent because there aren't people coming in muddying the waters all the time, are on a learning curve, or trying to insert something and so your processes stay a lot tighter and cleaner. And that may be something obvious but it's just something that I observed over the years working for multiple board shops and assembly shops.

It was a statistic, before I chose to represent one of those places is, what is your turnover? Because I knew that would create a lot of chaos not only for me, but for my customers, because of the fluctuation, customers, designers, will say to me: I was doing business with XYZ company and all of a sudden - they were great for eight years - and all of a sudden they lost the recipe and I go: uh-oh, they have had employees change. I know exactly what happened. So I know, we've all seen it, maybe an obvious point but something I thought worth pointing out to our listeners.

We appreciate that point, and I will also say that our Production Manager's been here for 18 of the 20 years, so again, it does make a difference.

It does make a huge difference. And Jon, your tenure there and being - yeah that's just wonderful. That's again, a great report on you guys.  So this morning what I thought we'd talk about is stack up and impedance, but from a manufacturing point - what you guys can teach designers and engineers that are laying out boards. How you can help them sort of avoid some pitfalls relative to stack up and impedance from a manufacturing standpoint?

So Jon, maybe I'll start out with you, or maybe you both want to kind of ping-pong this one  for the uninitiated, let's just talk about what kind of implications there are specific to stack ups with materials?

I mean that's where you start.

Do we have enough time?

[laughter]

We could do a whole thing on materials maybe we need to do that? Because you just said you had a lot of exotic materials I'm like: oh they're one of those. Okay so all right, let's talk about materials, sorry.

You know, I'm just starting - again just the very basics. Our main is FR4 high-temperature FR4 materials but we do get into a lot of Rogers materials for the RF type designs   a lot of hybrid combinations a little bit of Teflon so just there's different variances on what you can use.

But diving into stack ups, what a lot of people don't think about from the impedance standpoint is, what are we doing with the the outer layers as far as the copper weights and the plating? And I'm touching on that real quick, because when you start out with a half ounce copper foil and then you plate up another additional ounce - sometimes when they're doing the modeling in the software - they're putting in half ounce and they model it and they get a certain number. But in reality when you manufacture it, you're plating on the surface, so a lot of times I'll get, from design engineers: well my model shows that it should be, 50 ohms and you guys are coming out at 55 ohms?

It's like: well, you're not taking into consideration all the plating on the surface and that makes a big difference. And so we get a lot of that where there's a lot of model software out there on the internet people can go to. We use a software called 'Archeo' it's a very very deep system, as far as it actually takes into consideration all the dielectric constants of the materials you're using in your stack up.

For example, different cores are built with different prepregs, and so they have different dielectric constants. Some of the modeling software on the internet gives you one setting so you can put in 4.1 or 4.2 for your DK value but in reality, depending on how the materials build you have different DK values that can range all the way from maybe 3.8 to 4.2 on a certain materials like these; I sold a 370 HR for example.

So when creating the stackup, we have  all of that in there, we have all the pretty products the laminates that are being used, even the LPI and all the dielectric constants and when we're modeling that, impedance becomes very very accurate compared to the models are on the internet.

So - I wanted you to pause right there, you said LPI, so that's Liquid Photo Image of a solder mask. So do you typically put the solder mask in when you're doing your models?

Absolutely you do, because that's a big, big critical part. Another example - I'm glad you brought that up because again, they go out and model on the internet; they're not putting on the solder mask, they're not putting on the copper plating. As I mentioned, they get a completely different value, and then when we come to model it we're going back and telling them we need to change their stackup they may have defined on their fab drawing, because it doesn't meet the impedance requirements. And also you do get a lot of designers that understand that and they'll put notes on the fab drawing saying the manufacturer can adjust the dielectric, spacing, or the trace width within, plus or minus 10% to obtain that value.

Yeah, and like you said, I think that's a good point, say in the case of Isola, or any laminate, they might put a datasheet that’s about 3.8, but it's not exactly 3.8, it can vary in a minor way, from lot to lot, is that correct?

Not so much lot to lot as it is from material to material so they -  if you build all the way from 3 core to 47 core, they're using 106 and 108s and these are all prepreg styles that I'm mentioning, and each one has a different dielectric constant. So if you get a combination of them, you end up with a different value - and that again - depending on how your stackup is generated, one discussion Mihir and I had earlier today, is about designers that specify in their fab drawing the stackup they want you to follow, and they can send it to board house X, Y and Z but if, for example, let's say they have a four-layer and they want to specify they want 8 mil dielectric spacing between 1 & 2 and 4 & 3.

Well, we may use a different series of prepregs to obtain that than another fab house and again, DK values - different impedance readings. So all that comes into play.

Yeah and that's a trade off, that's a consideration the design engineer has to make in terms of how they're doing the prototyping, what the outlook is for them, and the turn times and the costing you know, there's other factors outside the actual design of performance on the circuit itself because if you do it, and you have it once, then you say this is my design so at least you will have more consistency amongst different manufacturers because you see, I need these materials, I need backup I did it - do it. But you'll have consistency in the final product but you more I mean most certainly, will not have consistency in the turn times. The available materials that different guys have, especially when you start getting to the more the exotics, and the high-frequency stuff.

So that could start playing into effect and people charge different amounts for it based on the lead time, what they have in stock, what they want to charge, etcetera as that gets complicated but at least, it'll be more close to a similar design on revision vs. if you say, look: I'm just gonna let the manufacturer do it and tailor it towards what I can get quickest and at best cost that'll still give me my main factors and whether they're controlled impedance or stackup height or whatever - and let them do that. So that's kind of the two different ways that people can go about designing.

My impression Jon, before you go on, is that a lot of designers do kind of hand off that stack off to their manufacturers. Do you think that's true?

Yes, we do get it kind of both ways. In some cases, we just get a stackup, for example, if they want it to be 062 plus or minus 10% that gives you the layers that have the impedance requirements, and then we go and generate the stackup and manufacture the board. To me that's probably the more straightforward way because you're guaranteed you get what you want.

Sometimes they're specific about what they want. They call out the dielectric spacings, the core material is everything and now you have to build that stack up, then plug in their numbers and model it, and then it usually doesn't come out the way they thought it was going to. And again, we touched on two reasons why.

So you kind of get a little bit of both. But what I was gonna start saying is, that we also offer a service, a stackup service that you can come to us at pre-design.  You've got your board all laid out, you're ready to do your routing, and you can come to us and say: hey, I have a six layer and eight layer design, this is the material that we want to use, and you can tell us a little bit about your design. Which layers are plane layers, which ones are the signals. We don't look at the reference to this is, ninety ohm diffs, and 100 ohm diffs, and then we can go ahead and model that stack up at that time. We can come back and tell you what size traces to use for the single-ended, the tracing space for the differential pairs, the copper weights, everything. We can come back and give you that complete stackup.

So now, that's using our materials. Our DK values, our stackup software, and then when you go to Roger design - if you use those numbers - then when we get back your design and your stack up - the project’s done.

Which I think's a really great model, because then you're doing this partnership - the designers telling you where they're trying to get - you're actually informing them, from a manufacturing standpoint, best practices, and I love that whenever that happens. I wish it happened more.

That's right, that's free of charge - again right at the beginning stage - to me that's the smoothest way to do it. And then you have a stack up you can actually send in with your data package and you'll be guaranteed you'll get what you want.

That's awesome, what a great service I love that.

You've talked a little bit about it, is there anything you want to add? The distribution of copper I get. I used to specialize in RF and microwave boards and that issue you talked about where they model it without the plating ending up on the outer layers right. The inner layers it doesn't matter, but the outer layers, you have to do multiple planing cycles and then it's completely outside of the range of what they simulated and I'm like: I don't know why, and without a fundamental understanding of the manufacturing process it's easy to see how that could get missed. Is there anything else you wanted to talk about? I'm gonna ask you guys three or four or five tips and tricks to give people who are listening some takeaways. But before I do, is there anything else you wanted to add relative to stackup in regards to manufacturing or distribution of copper?

Maybe just a brief... oh sorry Jon do you want to go?

I was just gonna touch on, you were talking about outer layers and then inner layers. If they want to use heavier copper, I like to point out that that's great on plane layers because when you have a heavier copper, your z-axis is higher and now when you go to put the prepreg in, you have to have enough resin to fill in there. And if you don't have enough resin then it can cause delamination or other manufacturing issues. So again, to point out, we get a lot of that too. We get a lot where they want 2 ounce copper on the inner layers, and they'll mix their traces and planes together and they'll be putting 4 mil traces on 2 ounces of copper. That doesn't work, that doesn't work at all.

Yeah and then you have a trace that looks like this [gestures] right like or this - they're not this any more there cuz that's a hard if... yeah it's not a good idea.

So just keep in consideration, from a copper distribution standpoint in layers. You have to nest prepreg in between them, it definitely makes a difference. So, if you're dealing with half ounce copper, no problem - you can pretty much do whatever you want. When you start getting above one ounce, then it starts changing the ballgame. So, from a proper distribution standpoint, just take that into consideration when you have - I'm going back to impedance - but when you have impedance on the outer layers and you're referencing to a plane layer underneath. Try to leave it all solid plane without mixing it with signals. That makes a big difference.

Because now, you have a nice, consistent, even, solid dielectric spacing between the two - so that's a definite plus. Like a six layer, for example, where you have power ground on layer 2 & 5 and then 3 or 4 signal layers, you have to use a lot more prepreg to nest in between there. So again, try to pull up most of your dielectric spacing between those two areas because you're going to need  more of it to nest the prepreg. So that's a little bit about copper distribution.

All right, all right guys, so let's talk about some real practical takeaways right now for designers and engineers who design boards that are listening today, from a manufacturing standpoint. I'm sure that you see some of the same oversights being made on a consistent basis. Can you give us three to five tips and tricks; things that designer should look out for when best design for manufacturing practices that you guys see. Mihir, why don't you kick off?

Sure, well mine has a bit of a tie-in more on the design side, because that is more of my background especially that's right now, but there's really two main design areas when it comes to stackups and manufacturability. It's the whole RF analog side and then this digital - high-speed digital - side and they're kind of characterized by two very different, but very heavy driving factors. On the RF analog side you generally find your designs more influenced by the necessity for a low dielectric constant, low signal loss, low leakage, and then generally these have a lower layer count so you really need a low and uniform dielectric constant and all these other things.

So your choice of exotic material is gonna be far more important. But you don't necessarily need to work - that's gonna be more of an important bigger part of your cost, and a factor in your design decision. It's just more important to the design. Whereas with a lot of more high-speed digital stuff, these are usually a way higher layer count, and they have all these other things like burying blind vias, really, really tight  traces, and just all these crazy ICs that have like a hundred pins of BGAs that needs all sorts of fan-out etc, and so your costs on that was gonna be way more driven towards the actual manufacturing time and the complexity, and to a lot of people it sounds obvious probably, on this podcast, it is.

But I mean you'd be surprised even as you're designing stuff, people really don't fully understand that buried and blind vias, while they're so easy to throw in on in Altium and just say, this is great, everything routes up perfectly. It does add a lot of cost and time. You could manufacture it, but it's seriously gonna impact your design when you have to do board back, so that's gonna be far more important than generally your choice of material. But obviously, as layer count increases that cost is going to be driven up too. So things like that. You have to take into consideration the differences in the designs and things that engineers are looking at when they're designing them and how that plays out usually in cost and lead time.

There's a lot of trade-offs aren't there?

Yes, that's right.

What would you say Jon?

Well let's tackle unbalanced stackups for a second. Because we get a lot of that-

Pretzels?

Yeah pretty much.

So again, I mean one of the things you need to take into consideration is, you want to have a symmetrical stack up. A lot of time too - especially if they're using hybrids - so they'll put a thick ten core Rogers on the top, and then something thin on the bottom. And again you want to have a balanced stackup, otherwise you're gonna end up with a warped board that to me is a very, very key thing, is to keep it symmetrical. We'll get those stack ups, we'll have to go back and tell them: listen is it possible, the chance of warpage, and try to explain to them.

They need to be symmetrical, so that's something to take into consideration from the get-go. Another one that we get a lot of, and I think I touched base on it a little bit; is to take into consideration the copper weights you call out in the trace and space that you're routing. Because it makes a big difference. So you know, if you're going to be doing a three mil trace with a three mil space, we have to start with quarter ounce copper and then, on the outer layers we have to plate on the surface. If it's on the inner layers, you can do small trace and space on half ounce copper. But once you start getting to two ounce copper and above, you need to be around six and seven mil tracing space.

And we get a lot of that, where we have to go back and tell them: listen your design has four and four you're calling out for one ounce copper, two ounce copper, it's not possible, so we're gonna have to go ahead now and reduce the copper weight, or even worse, that they have to stick with a heavier copper. They have to go redesign their board and lose time.

Explain that, it may be obvious, but explain why that's impossible? I've run up against this a whole bunch of times, but explain because it may not be as obvious as it is to you and me Jon. Why can't you take two ounces of copper and do a four ounce or 4 mil trace or three mil trace, what happens?

There's two scenarios: one is when you give us a design that's 4 mil trace for the 4 mil space between trace and trace, and trace and pad. In order to finish - after etching that trace - we have to do what's called an x-factor. So now we have to increase that trace, X amount, might be one, two, or three mils depending on the copper weight. Because again, having copper it's a higher z-axis. So when you actually have a further distance to etch down to the base of the copper to get down to the laminate, you start losing the feature size as you edge the copper down. So we have to increase that feature size.

So if it's a 4 mil trace and it's 2 ounces of copper, we might have to increase that to a 6 or 7 mil trace. But if your air gap is 4 mils - now we're reducing that air gap down to 2 or 3 mils - which is not manufacturable. So that's where you'll be coming to the problem. And again you also have peel strength. I mean if you have a 3 or 4 mil trace on two ounce copper, I mean the chances of it actually peeling off the laminate is much higher, because you have a certain peel strength. So again, you're not gonna have a small trace on a heavy copper feature for various reasons.

And maybe even in more layman's terms, because this is what helped me understand it when I was doing - because you really don't learn this stuff when you're studying like for engineering or maybe, I don't pay attention.

No, you don't learn it, you don't learn it, you're right.

Simplistic, the thing is people, maybe we could even put this up on the video I don't know if you can add that or add a link? If you picture traces from the side view they're not straight up and down. Right?

Never.

They're at an angle, the reason they kind of look like they're little trapezoids - is because the top of the trace is under the duress of the edge - about the actual chemistry - a lot longer than the bottom. So as it edges down, the top is getting whittled away more than it is at the bottom. So you tend to add an edge like this - if your traces are really close together, you don't have that space in the middle. It looks like you have all the space in the world at the top, when you get towards the bottom of that z-axis, they're actually touching, so you can short out traces that's like the simplest example without getting too deep into everything.

Can you undercut in that scenario or am I thinking of it backwards?

Yeah, undercutting is a term you kind of get on the outer layers, but when you have the dry form you can kind of get it undercut but for the most part it comes to geometry. I mean, you have a very tight tracing space, and first of all, you have limitations on your gap. And even if you could increase the trace big enough, you're gonna end up like Mihir pointed out, with a very small trace on top and a larger trace on the bottom.

Yep it makes sense.

Oh yeah a lot of mechanical electrical kind of issues.

And as well, I have a friend in the industry who was in the board industry for 40 years, and he used to say: it looks good on paper, but he said physics trump's theory right? Like theoretically, it should work right, but he goes: but physics wins out every time. So Mihir, any more kind of practical design for manufacturing tips that you can think of, or Jon, either one of you?

I think we kind of - if people take at least a few tidbits from what they heard today - there'll be an immediate ROI on their time listening, to the success and speed of their design.

Good, well I'm excited to announce that Mihir and the Royal team, will be joining us at AltiumLive as our sponsors. They just let me know that today, so I'm very excited!

So I'm sure you guys will bring some sample boards, or some video and some great assets that they can look at. They can talk to you one-on-one, learn some more tips and tricks, face to face just gather information, which is sort of the magic of AltiumLive. Our goal is to just put the design community in a room with the supply chain, with people that are very knowledgeable, which are veterans in the industry and just let them rub shoulders and start creating new solutions or just collaborating for successful designs and take some of the pain out of it for all of us. So we're delighted to have you guys in San Diego in October.

I really needed an excuse to come to San Diego.

Right! I know, and it's on Coronado Bay, so we're staying at the Loews Coronado Bay Hotel so there's water on three sides of this hotel, and it's in October, which is like, October in San Diego is like heaven. It's like 73 degrees, on the water, so...

You already sold us!

Right it sounds like if you're not coming to learn some design stuff, at least tell your boss you are, and get a nice trip to San Diego... just kidding.

So anyways, we're glad to have you and I'm glad to to get to know you guys a little bit more. I know of Royal but I've never gotten to know you until this last week and so it's been a delight to get to know you both and learn from you. And thanks for sharing your DFM wisdom with our listeners, and we look forward to engaging with you more at AltiumLive, and we'll be sure to share many links.

I think I have eight links to share from Royal and you can dig more into what they do, who they are, and get to know them a little bit better as I have this week. So I'm sure you'll enjoy that. So Mihir, Jon, thank you again so much for joining today. Thanks for joining on our podcast.

Thank you, thank you Judy.

Well until next time please subscribe, join, engage with us at Altium we always enjoy learning from you and learning about what you would like to learn about. We're only making guesses unless you tell us specific topics you would like to learn about. So keep the comments coming. We look forward to engaging with you next time on the OnTrack Podcast. Until then, remember to always stay OnTrack.