Musings, deliberations, and end results.
Felix Jen – 01 July 2022 – 15 min read
While the first post in this two-part series focused on the design elements, this second part focuses on the business side of the design process. Here, I hope to tease out some of the axioms of design from a business standpoint by walking through the PCB production process. I aim to follow a single PCB, Project X, from its manufacturing to its final assembly.
This post will be fairly different from the first in the series, taking a more product-oriented approach rather than discussing the axioms in the abstract to hopefully shed more light on some of the topics.
To begin, we will need to understand the PCB that I am going to be working with. I will be presenting a hypothetical completed PCB, Project X, that is currently finalizing its design, but is open to changes. Project X is a standard two-layer PCB that most are familiar with, but given its internal complexity, discussions are being made to determine whether it should be made into a four-layer PCB instead. From an end-user standpoint, Project X will be going into a multi-part assembly, destined for global distribution with the target markets being primarily the United States, China, and Europe.
Project X will not be exposed to very harsh environments, but it will be exposed to a wide range of environments at the end of the day. It will be eventually fully encased in an enclosure, but contains a variety of user-interfacing components such as switches, LEDs, and buttons. While the company behind Project X would prefer that ultimate failures be prevented, they understand that such user-facing products will inevitably suffer some form of failure due to wear and tear. They will offer a time-limited warranty on the ultimate product that Project X enters, but they would prefer to avoid immediate returns of the product.
Ultimately, our goal as both a designer and as a company is to be able to produce Project X for as little as possible, by making the appropriate compromises where necessary and making the correct decisions of what not to compromise on.
For the safety side, Project X does not need to undergo any form of military or biomedical certifications since its application is exclusively intended for consumer electronics use. As well, Project X will not be exposed to any kinds of dangerous levels of voltage or current, and will be almost always kept at a moderate ambient temperature.
Our factory that we’re working with is a fairly large production facility with plenty of equipment and connections to suppliers. Therefore, while we are able to acquire any materials or components we would like, our factory (fab) already maintains a regular supply of existing parts for its other customers—some larger and some smaller than us.
Let’s dive into an in-depth look at the PCB production process from start to finish. The following is a list of the steps that will be taken to produce Project X and some of the considerations that we should take along each way. I would like to thank Strange Parts for their amazing video on the PCB production process as well and would highly recommend giving the linked video a watch for a more visual look.
A typical two-layer PCB typically starts as a “laminate.” This laminate is consists of a sandwich of copper and a “substrate” layer. This laminate is purchased from suppliers in large sheets known as “panels,” often multiple feet long in each dimension. The laminate, when used for production, is cut into smaller, more manageable chunks though larger productions and projects may use the whole laminate panel. Our Project X is a reasonably small project about the size of a hand, so we’ll definitely not be needing the whole panel.
We should discuss some of the specifications for the laminate. First, looking at the outer copper layers of the laminate, we should be able to determine the thickness of the copper we want in our final product. The thickness of the copper ultimately determines how much current Project X can carry in its traces without heating up due to internal resistances.1 Since Project X is not going to be carrying a large amount of current, we can likely go with a fairly standard copper thickness of 1oz/ft2. However, we do have some fairly thin traces within the center of the PCB across critical components, so perhaps, we should go with a thicker copper layer to prevent heating. We could certainly increase our copper thickness to 2oz/ft2.
We should also consider the thickness of the substrate itself. The thickness of the substrate, for our purposes, primarily factors into the assembly of Project X into the final product.2 We aren’t handling the ultimate product, so we can go with a fairly standard substrate thickness of 1.6 mm.
Next, we’ll look at the materials of the substrate we have to work with. In modern PCB production, FR4 fiberglass is the standard material of choice for its excellent dielectric properties, mechanical properties, and price. However, not all FR4 is created equal. FR4 is merely a NEMA grade designation for glass-reinforced epoxy laminates. It is not a specific brand of material and the sources of FR4 will ultimately vary in specification greatly.
One of the biggest specifications which can be chosen is the “glass transition temperature,” otherwise known as Tg. Typically, three Tg grades exist: (1) low between 130-140° C, (2) medium between 150-160° C, (3) high between 170-180° C or higher. The Tg grade is used to determine the dielectric properties of the substrate as well as ultimately its heat tolerance. As Project X will be undergoing reflow assembly and exposed to a wide range of temperatures in the manufacturing stage, we should consider the Tg grade of the substrate to be fairly important. Higher Tg’s will tolerate higher temperatures without warping (bending) or delaminating (ripping pads) but will also have a higher dielectric property. Lower Tg on the other hand will increase the likelihood of warping and delamination.
Why don’t we use a higher Tg then for all projects by default then? This brings us to our first axiom:
Quality costs.
Ultimately, the higher the Tg our substrate, the higher tolerance our supplier would have to be. Higher Tg tends to be less popular, despite being a “better” material in terms of heat resistance, simply because higher quality costs money. There are fewer suppliers which can produce higher Tg substrate, and since, as most projects don’t need such high temperature tolerance or reworkability, it’s an unpopular choice. Therefore, when we examine our material choice, we should consider whether a higher Tg is necessary for our project. For Project X, we’ll go with a medium Tg.
Now that we have our laminate decided (1.6 mm, 1oz/ft2, and medium Tg), we can start to move onto the rest of the process.
As mentioned in our overview section, we are primarily designing with a 2-layer PCB. However, we can still use a 4-layer PCB if we believe the design so requires. Producing a 4-layer PCB proceeds similarly to a 2-layer but instead of starting with a laminate, we start with a pre-preg layer. This layer is a layer of copper that is placed directly on top of a thin epoxy fiberglass layer. These layers are then sandwiched together to form a 4-layer PCB. We have similar physical and material considerations for our pre-preg layers like we do our laminate, so those won’t be repeated.
One of the key benefits of moving to a 4-layer for Project X would make layout significantly cleaner and easier. We have two additional internal signal layers to work with, and much more surface area for our traces to proceed. We can also use the pre-preg layer to increase the amount of current we can carry in our traces without heating up.
Then why don’t we immediately default to a 4-layer PCB? This teases out second axiom:
Time is money. Stuff is money.
For every project in manufacturing, we pay for two things: parts and labor. The more parts we need, the more money we need to spend. The more labor we need, the more money we need to spend. A four layer PCB not only doubles the amount of substrate layers we ultimately need, we also need to image two additional PCB layers. This is a significant cost for our project. Therefore, we should consider whether we should default to a 4-layer PCB here, if a two-layer PCB can simply suffice. Based on our design complexity, we should probably go with a two-layer PCB to both reduce production complexity and cost. If Project X was significantly more complicated or space constrained, we would probably swap to a four-layer.
At this stage, we have selected our laminate and we have provided our Gerber Files to our fab and they have finished looking those over. We’re onto Outer Layer Etching. In this step, the outer copper layers of our laminate panels are etched chemically to form our traces.
Modern production facilities follow a multi-stage photomasking process of copper etching. This process of photomasking is actually incredibly important in the whole production process is repeated multiple times in various production aspects.
First, the copper layer is covered with a layer of wet “photoresist.” Photoresist is a thin UV sensitive layer which is inert to the chemicals used in the etching stage. Next, a 1:1 representation of the Gerber file for the copper layer is printed onto a transparent film in negative format. This creates a black and white image of the Gerber on the film which is overlaid on the photoresist-covered copper. A strong UV light is shown on the laminate to “cure” the photoresist that is exposed. This copper laminate is then washed to get rid of uncured photoresist.
What we are left with is a copper layer with cured photoresist where there should be copper. Our laminate panel is then dumped into an extremely harsh chemical etching solution which aggressively eats away any unprotected copper that is not covered with photoresist. The board is then washed and photoresist “stripped” off in a chemical process.
The same photomasking process happens on both the top and bottom layer of our two layer board, imaging and etching both our outer layers at the same time for efficiency.
Our panel, now with the outer layers etched, is put into a drilling machine. This is where we begin to drill the plated holes on our board. According to our Excellon file, the CNC router will punch our holes which are ultimately designed for plating.
I’ll take this moment to discuss our third axiom:
Opportunity cost is a cost.
Opportunity cost is defined as what you give up in exchange for the choice you make. For example, if you were given the choice between a hotdog and a taco for lunch and you choose the hotdog, the taco is your opportunity cost.
Why am I bringing this up now? Each manufacturer only has a limited number of CNC routers at their disposal for drilling holes at one time. Therefore, if they spend more time drilling holes for Project X, they must either take on less projects to maintain lead time, or push other projects back. Both of these options cost the manufacturer money in the long run, and you can be certain that they will pass on those costs to you. That said, the general idea here is that the more holes you put on your project, the longer it’ll need to spend in these drilling phases, and the more it will cost you.
Now that the plated drilling is done, the panel undergoes chemical plating. A thin layer of copper is deposited on the laminate. This copper adheres to any of the existing copper as well as the substrate itself. This effectively bridges the top and bottom layers through the plated holes on our board that we just drilled. Copper will fill their way into those holes, which is why we only drill the plated holes before this step.
After spending some time in the plating bath to get all the holes plated as well as copper thickness increased, the board is once again cleaned and washed to get rid of any caustic chemicals.
Any of our unplated holes are drilled in this stage. As well, the board gets its profile cut in the CNC router. Our third axiom once again applies here, the more complex the profile shape, the longer the board will need to spend in the machine. Therefore, the more it’ll cost you to produce. Thankfully, Project X has a relatively simple rectangular outline and breezes through this stage.
Now that our board has all the holes drilled and plated, we can start to add on our soldermask layer. This is a protective UV cured epoxy layer which forms a tough corrosion-resistant coating on top of our board. It also serves to resist any solder on the PCB which helps to prevent solder pooling and bridging on fine pitched components.
Here, for this process, we return once again to photomasking. A thin layer of UV cure soldermask is applied to the entire board. Next, the manufacturer prints our soldermask Gerber layer onto a transparent film and overlays this on our board. This sandwich of film and board is placed under a high powered UV light to cure the soldermask where it is exposed by the film. The board is then washed to get rid of any uncured soldermask. The same process is repeated for the other side of the board.
We are then left with a soldermask covered board with areas for pads and through holes perfectly exposed according to our design.
Next, our board moves quickly to surface treatment. Copper will naturally oxidize when exposed to air, in a short amount of time. This is why we need to create a protective barrier for the copper. On the other hand, we still need to be able to solder components to the copper, so whatever protective barrier we create needs to be conductive and take solder well. While there are too many surface treatment options available to discuss in this blog post (we’ll cover it in the future), we’ll discuss our two primary options: HASL and ENIG.
HASL or “hot air solder leveling” dips our whole board into a pile of molten solder and then uses a powerful blast of hot air to blast away the molten solder. Since we just applied a soldermask layer which aggressively repels solder, the solder will only stick to any plated areas that aren’t covered by the soldermask. This is a very efficient process that can be used for any board but has some major drawbacks. It’s a very imprecise process, with the amount of solder deposited varying greatly even run to run. Therefore, it is a poor choice for fine-pitched and small components, which can often have solder bridges created. As well, it’s a fairly aggressive process requiring the whole board to be dunked in a hot bath of solder. However, because of its speed and simplicity, this is a very cheap process.
ENIG or “electroless nickel immersion gold” is a pair of processes together which form a soft gold layer on the PCB. First, the PCB is put through a chemical process which binds a micron-thin layer of nickel to the copper of the PCB. Nickel binds readily to copper whereas gold does not, so this process is necessary to form an intermediary layer. Next, the nickel-plated board is immersed in a electrolytic plating bath where a few microns of gold of deposited on the board. This gold layer is then exposed to the air and the board is then washed to get rid of any chemicals. What’s left is a extremely smooth and pure surface layer on the PCB with extreme precision. Our soldermask has protected the rest the board which didn’t need the surface treatment, and any exposed areas increased in thickness by a few microns at most. The major drawback of ENIG though is that it is a fairly expensive process, involving gold and a lot of time.
Since Project X uses pretty small components, we’ll choose ENIG for this stage and eat the costs of production.
At this stage, we have a finished bare PCB. While we can continue our production discussion into the final assembly and testing stages, the focus on this post was to tease out the axioms we can glean from the process rather than provide an end to end discussion.
Throughout the PCB production process, I unveiled three of the critical axioms which are important from a cost standpoint and therefore a business standpoint.
Trying to improve something’s quality, all else being equal, will always increase its costs. This is a fairly inevitable truth. Higher quality materials will typically cost more. Higher tolerances will require things to move slower or requires higher skill which costs more. In the design process, it’s important to determine whether the design warrants the utmost quality in every possible decision and what compromises can be made. It’s our jobs as designers to understand the business constraints we are placed in and to make decisions around those constraints.
Every second of time that something spends in the production process costs money. If it requires someone to do something, you’ll have to pay for that. Every additional process a project has to run through to achieve something adds to the complexity and the cost of the project. The same is true every time we add “stuff” to a project. Nothing exists for free, so if we add even a single thing, it’ll add to the cost of the project. It’s again up to designers to determine what costs are worthwhile to achieve the business objectives of the project.
One cost that is often overlooked is the opportunity cost of production. While opportunity cost is closely intertwined with time, it’s more than just the raw time spent on a project or process. It’s everything that a manufacturer has to give up to focus working on your specific project. If they are giving up machine time for a much more lucrative project, you can expect their opportunity costs to be higher. If they have to delay a high profile job to satisfy your lead time, you can expect their opportunity costs to be higher. No manufacturer operates a charity, so you can expect those costs then be passed along to you as part of your production cost. While the exact opportunity costs are difficult for us as designers to control, it’s important we try to minimize their impact on the project, all else being equal. Choose efficiency to minimize downtime.
Special thanks to Maker Keyboards for edits.
All metals have a fixed resistance, determined by their cross-sectional area as well as the distance current needs to flow across the metal. Therefore, if we have a very long trace within our project and our trace is very thin, we will have a very high resistance for a given thickness of copper. As more current is pushed through that trace, we face an increase in power loss and therefore generate additional heat which may negatively impact our project. This can be combatted (almost linearly) by increasing the cross section of our trace, by increasing the thickness of our copper layer, or the size of the trace itself if feasible. [Back]
The thickness of the substrate also can tend to affect the capacitance and dielectric properties of a PCB. However, given the small amounts of current and voltage, we can expect this impact to be negligible and not a major factor in our ultimate design decisions. [Back]