As Blue Clover Device's hardware designer, I get a lot of requests and questions from the team, as well as clients, for projects and explanations.
One recent humdinger from our marketing manager: “Hey, could you write a blog post on good PCB design practices for manufacturability, production and reliability?”
I’m always happy to help, but I was unsure as to how, or if, I could pull it off.
The internet is so full of easily searchable information on PCB design, that I doubted myself and the significance of a possible contribution, and seriously wondered if there was anything I could offer to other PCB designers.
“OK”, I thought to myself. “What are the main things I worry about when I design the types of PCBs I work on?”. “What do my checklists look like? What kinds of little loops run in my head while I'm laying out a board? How can I quantify and justify the artistic, intuitive and perhaps bizarre aspects of what I do? What design practices have I picked up from my experience working in research, start-up environments and now at an ODM?” Most importantly, “What have I learned over a modest hardware engineering career of incredible success and equally fantastical @#$! ups?”
First pro-tip: Ask yourself questions. Start an internal dialog. Just try to keep it in your head and don’t vocalize it, as people find that annoying.
Now that that’s out of the way, let’s talk PCBA design! In this blog post, I'll cover the following:
- A History of PCBs
- Board Layout
- Board Size, Shape, and Part Placement
1. History of PCBs
As a millenial, PCBs have always been a part of my life. From the first TV remote I took apart to building and repairing PCs with my dad as a kid, green circuit boards with a mix of through-hole and SMT were already in every consumer electronics device on the market. Although PCB and IC technology has advanced tremendously since then, to an untrained eye, PCBs and motherboards from the late mid to late 90s look the same as they do now. The basic PCB design/manufacturing principles and methodology is still the same for most consumer electronics. However, we design PCBs atop the shoulders of giants. At our disposal we have about 100 years of development of this technology, from the first patents in the early 1900’s, to the first mass produced PCBs in the 1950s for personal radios, to the development of the RS-274 Gerber standard developed in the 1970s, which is still (more or less) used today.
As much as I would love to write an entire blog series on the history of PCB design and manufacturing, it would be a deviation from my official mandate, and beyond the scope of this semi-educational blog… though I do encourage you to check out the following links as a primer to all of this:
“Open your mind to the past. Art, history, philosophy...and all this will mean something” - Capt. Jean Luc Picard
Before one considers designing a PCB, one first needs to design and finalize a schematic. It is so important in fact, that all modern PCB design software requires a schematic to be completed (well, at least started). This is so the software can convert the wires in your schematic into ‘nets’, which is a set of minimum point to point connections that is equivalent to the human-friendly schematic. This is how the software knows what components need to be connected to each other on the PCB. Besides the simple connections, here is where you define the nature of those connections. For example, a connection providing power to a component may require wider traces and larger/more vias than a connection for a low voltage signal. It is always good practice to define your connections here, so you don’t accidently route a 6mil trace to provide 3A of power to your high power LED. Often you can define individual connections, or define classes of connection properties and assign them to individual connections. Differential pairs, controlled impedances, busses, power and grounds are also defined in the schematic.
Analysis and simulation are a big part of electrical design, and there are many tools available, ranging from integrated tools in your schematic/pcb software, free tools based on SPICE, to proprietary tools for simulating microcontrollers (although I would never recommend completely relying on MCU simulations for your designs. One would be better off purchasing a demo board). One of my favorite free tools when designing specialized analog or other functional circuits with discrete components is:
It may be dated and a little crude, but the fundamentals are there and you can easily save and share designs with no pay-wall, sign-ups or other nonsense.
Ok, so once your schematic is designed, or perhaps before it is finalized, it is important to at least start thinking about your BOM. There are a few things to consider before selecting specific part numbers and packages:
Does the pitch of your component match the design rules of your chosen manufacturer? If you are doing a 6/6 or even a 4/4 board, you may still be using packages that have a pitch smaller than this.
Are you assembling this PCBA yourself by hand and not using a pick-and-place? Are you going to be doing some board level modifications (soldering/desoldering components) yourself? If so, what package can you comfortably work with? Most people can handle an 0603 package with a bit of training. Personally I can solder 0201 packages without any microscope or lenses, and can even do 01005 if I have to (hopefully I have a microscope to at least inspect my work). However I would still try to avoid using too many of those tiny packages if I knew I was assembling an entire 50+ part prototype myself. BGA components are also possible by hand, but are tricky to do, and impossible to inspect without some kind of xray machine. Also once you start using BGAs, you may have to incorporate micro vias and blind vias, which makes the PCB more expensive to manufacture. Sometimes however, you don’t have a choice, and require a BGA component.
One big factor for me these days is whether the part’s eCAD library exists online. Defining landings for packages and tying them to symbols is an arduous task. If you are not picky about the part number that is needed for the BOM, such as an LDO for a specific voltage, search to see if there is a part library that is available that fits your requirements on one of the following 3 sites:
Of course it is always good practice to verify that the library is correct before you use it.
Some other useful resources for BOM selection (and even schematic capture) are tools like Texas Instrument’s WEBENCH and Murata’s simsurfing tool. On WEBENCH you can select design parameters for your circuit, and it will suggest ICs that fit your requirements. Not only that, it will generate entire schematic reference designs for that specific IC and your set parameters, and recommend an entire BOM (one engineer I have worked with interestingly called the group of discrete components in a reference design ‘popcorn’). Not only that, you can also download the entire schematic straight into your eCAD software (it supports a few different platforms).
Murata’s SimSurfing tool allows you to enter electrical parameters for passive components, and view highly accurate data and characteristic plots on those components. It is much easier than trawling through hundreds of datasheets, while looking for that special component. However, you are stuck with Murata (which isn’t a bad thing, they produce high quality components imho).
Lastly before deciding on a component, it is important to see if it is an active component and widely available. Believe it or not, many quick turn manufacturers, even in China, require that the part number is available on popular US distributors like Digikey or Mouser. Many of them will order parts from there despite being on the other side of the Pacific. This is due to the convenience of their service (quick shipping, nice search interface, digi-reel service), and because having a backup reduces the risk of not sourcing the component locally. Also, if many distributors are carrying the part, it is less likely that the part will become obsolete. Obsolescence sucks.
A site like https://www.findchips.com/ is a great place to do this kind of search. Fun fact, findchips is owned by supplyframe, wich also owns hackaday and tindie! Their office is really close to ours here in San Francisco.
4. Board Layout
So you have your schematic and BOM compete. Now it’s time for the fun part. Simply place your components on your board wherever they fit and run the auto-router. Done. Wasn’t that easy?
Did you really think I was going to suggest that you use the auto-router to completely route your board? Seriously. I’m glad you are still here actually. I would have likely stopped reading already. If you are new to PCB design, then let me clarify: do not use an auto router. Route your boards yourself. Auto-routers are becoming really great, and eventually we will get to the point where an auto-router is able to route a board completely. But we are not quite there yet. There are at least 100 different rules of thumb that you follow when laying out a board, and some very important decisions that need to be made constantly that only come from experience (@#$#ing up). I’ll try to cover as much as I can here, but you should do some extra research!
Now that we have cleared that up, let’s continue with our board layout.
A) Board Size, Shape and Part Placement
Sometimes your design is driven by the mechanical design or industrial design alone...But Hopefully the product design team consulted you, the hardware engineer, before the final size requirements were established. In this consultation, you did a schematic, selected your components, and roughly laid them out on the proposed PCB shape. Things were tight, so you decreased the package size of the resistors, capacitors (where possible!) and ICs. You went from 6/6mil trace width/spacing to 4/4mil. In your head you knew things would work, and if not, you have the luxury of adding layers and blind/buried vias. You report back to the product design team and give the green light. When management is ready to proceed, everything will be smooth sailing.
Unfortunately the above scenario would be ideal. It is not always the case. I have worked for hardware startups where the electronics were an afterthought. Seriously. They raised a bunch of capital or got upper management excited for a sexy, small widget. They designed the mechanicals, the look, the feel. Then they turn to a hardware engineer (sometimes on contract, never involved until this point) and they say, "OK, time for you to design us our electronics to fit in this thing, with A-Z functionality."
Now of course, anything is possible with the right amount of imagination, intelligence and most importantly, budget. I am going to assume that if you are reading this blog, developing your own purpose built microscopic ASIC using the latest and greatest technology is out of the question. In that case, you are limited to off the shelf components, and the PCB manufacturing capabilities of your prefered board manufacturer.
If your board size/shape is already determined for you (or if its not):
- Place all connectors that need to be in specific locations.
- Lay out all components, starting off with functional circuits, grouping circuits together based on power, then communication/signal requirements
- Group all circuits together in the proposed board layout, top and bottom (if the design permits.
If the components do not not fit, then it is time to consider reducing the package size of your components. Ensure that all components are SMT, as they require less real estate. Keep in mind, that reducing resistor packages means that the power rating will most likely also go down. Ensure that your new resistors are still sufficient. Also, capacitors come in all packages. The smaller the package, the lower the voltage rating. Also, the real capacitance at the application voltage/frequency may be much lower. Check the datasheets, especially the DC Bias and Frequency Characteristic charts. Also take into account your personal re-work limitations. Going down to a 0201 or an 01005 packages saves acres and acres of board real estate...but can you rework them with your soldering skills and equipment?
Try steps 1-3 again. If they still don’t fit, then here is where you need to start modifying the PCB design parameters. Starting from the least expensive to the most expensive changes:
- Shrink the trace width/spacing minimums. Most of my boards start off as 6/6mil trace width/spacing. 4/4mil design rules can actually free up a lot of space. Check to make sure your manufacturer can handle this. Also, what thickness do your power traces need to really be? I use online trace width calculators to determine the minimum trace width based on the application current. 4pcb’s trace width calculator is usually my go to: https://www.4pcb.com/trace-width-calculator.html.
- Reduce the via annular ring and drill size. Once again, check your PCB manufacturer, and second, check to make sure that your application can handle this. Vias are a little more tricky to calculate, as the inner diameter plating thickness of holes and vias isn’t as well controlled. The inner plating thickness determines the maximum current that via can handle. In the end however, a smaller via frees up even more space. http://circuitcalculator.com/wordpress/2006/03/12/pcb-via-calculator/
- Blind/buried vias. Now we are getting into some serious PCB design territory. But your board still does not have enough space, and you need more room dammit. Up until now, we have used through hole vias. A single through-hole via, regardless of its layer, punches a hole through the entire board. This is the cheapest way to produce vias on a multi-layered board. Check with your manufacturer to see if they can handle these. But if they can, then you will free up even more space.
- Vias in pads. This is another trick that can save a lot of space...but you simply cannot put vias in pads. This is because the via will drain the solder during the reflow process, leaving little to no solder for the part that is actually supposed to be soldered. If you go this route, the manufacturer will charge you more because these holes need to be plugged before the boards are pasted and populated.
- Layers. Lastly you can always increase the layers of your board. I’ll give my two cents on this later on.
The second decision is how many layers. Really this decision should be binary. That is, can you accomplish the design with just two layers, or do you need multiple? The reason I suggest multiple, and not 4, 6, 8 etc, is that in my opinion, a hardware engineer should not limit oneself to a fixed layer count. Does a software engineer designing apple or android apps kick off their design by saying “This app will be less than X mb”? Do they limit themselves by a targeted file size? Not really, only to the extent that a large file size would be unreasonable (2GB for a simple chat app would raise a few eyebrows). Also, a good software engineer just does things efficiently to begin with. Like in software, arbitrary limitations in your hardware requirements can turn the design process into a game of making unnecessary compromises. In the end, your design will suffer for no good reason. For complex designs, less layers means more vias, less isolation, breaking up more ground planes/power planes, strange return loops, noise, poor thermal properties. The list goes on.
I am not aware of any quantitative formula to estimate a layer count. I guess such a function could be formulated based on connections (airwires), total landing surface area, board area, power requirements, isolation requirements, number of high speed signals etc...It would be a nice add on to a PCB layout software package.
Be cautious, however. Assuming that the inner layers of your board are used for ground and power planes, you will not have access to your traces if you go up to 6+ layers. For debugging this can complicate things.
In short, do not let the project manager or bean counter tell you that a board’s layer count should be limited. The layer count for your design is however many layers is necessary to fulfil your design requirements reliably and efficiently.
Since this blog post is getting long (and the Marketing Manager interjected), I'll conclude by saying that it's probably better to start by knowing which requirements you need to fill, and which you should push back on for the sake of good design.
About the Author
Brien G. East Jr. is the Director of Hardware Design at Blue Clover Devices. A Canadian expat, Brien is happy to embrace sunny California as his new home.