designer’s notebook
Money-Saving Tips for Printed Circuit Board Design
Raising the topline with good execution will make up for the expense of expediting development.
There is a simple equation when it comes to counting profits. The fixed and overhead costs must be less than the revenue for there to be profit. The first units out of the gate owe the company for all the nonrecurring engineering (NRE) costs. The item will be in the red until all that is paid back by the margin between unit cost and unit price.

In many sectors of the economy, particularly commercial, the cost of goods sold (COGS) is close to the selling price, meaning little margin after overhead is accounted for. (PCB design is among the overhead costs.) Many units must leave the factory and find a consumer before the project hits the breakeven point. Product cycles are such that price erosion puts the squeeze on margins right from the beginning. Consumer hardware is a tough game, no doubt.

Competition among the players keeps us on the path of continuous improvement. Sitting still while others strive to grab your market share actually means moving backward, so let’s take it for granted that we have to keep reinventing our products. Those new features, whatever they are, will likely add to the bill of materials (BoM), which increases the variable costs. We can soften that blow with a few money-saving methods.

So much more margin is to be had for the first mover. No one else gets that premium pricing advantage. The risk of someone mirroring your company’s blueprint is always there, so that lead is perishable but worth having and keeping. A design win can lead to more design wins but only if we execute on the new product introductions that follow. Winning creates enthusiasm.

I’m not here to think about losing, so keep that energy going with attitude and willingness to do what it takes to stay on schedule. Monitoring progress against expectations will show if the product timeline is beginning to slip. Acting before it gets out of hand is the best way to approach this ongoing concern.

The one thing that might be known at the outset is the expected end-date. Working back from there, establish milestones for completion of library, schematic, placement, critical routing, simulation, etc. Each of those milestones should conclude with a meeting where we can check off all the items needed to proceed to the next gate.

To keep it real, let’s say the schematic-complete date is here but the schematic is not. Have that meeting anyway. Use that as a pivot point and figure out how to pick up your coworker by getting a lot of quality board layout work done in a short period.

Some of these things can be done in parallel with additional design resources if your solo efforts are falling short. If you know you need help to get to the tape-out date, it is up to you to get that help. The rate for a service bureau may seem high, likely more than your own rate. The payback comes if everything falls into place as planned. The majority of the revenue comes before the copycats release their versions. The first one in with a similar product might get away with it for a while and might even supplant the first mover. That’s not the norm. Be the original and stay the original by developing something better than the previous iteration. Spend money to make money or lag with a me-too product’s market share.

Let someone else create the library. Component footprints can be outsourced for a reasonable price. Most of them are available off-the-shelf and, if not, soon after you place the order by providing a data sheet. Some library generators also do schematic symbols and include step models of the parts.

You know the PCB cannot be any better than the underlying component footprints. A wise designer double-checks the new geometry prior to use. If you manage this correctly, you can retire from making footprints and just check them over before adding them to your library.

Tighten the loop between the product designer’s data and the PCB data. We get mechanical data in different ways. Typically, a few layers have shapes that define route and component keep-in/keep-out regions. Sometimes, this is okay; most of the time, the mechanical engineer has a few glitches in the data they provide for the first go-around. The things I have seen and accepted over the years could make this a very long story (FIGURE 1).

Computer Screen
Figure 1. Little did I know in 2012 that Chromecast would sell tens of millions of units. Reducing the layer count from eight to six helped enable a low price.
Here’s an example. The headroom under a shield was set to zero when it needs to be enough to permit parts that fit but flag parts that are too tall. Flagging all the parts as too tall is not right, but hey, I can work with it. One way is to ignore the design rule checks. Another is to edit the properties so the actual head room is enforced.

Both methods come with risk. By pointing out the glitches as they happen, you can get to a better place where the mechanical data are correct by design. Experience has shown a mechanical outline drawing is rarely available at the start. Some attributes may be in a state of TBD (to-be-determined).

As these items are ironed out, keep feeding back your edits to their baseline form-factor. The goal here is to reduce the pain of having to capture a slightly different outline down the road. Time is money, so streamline those occasions. The sooner you can get a handle on it, the lower the impact it will have. Schedule-wise, this goes hand-in-hand with cleaning up the schematic early in the game.

You are not being a pest when your aim is to clean up the process. Frame the request as good for the design integrity rather than complying with the particulars of the ECAD tool. Setting aside the gate swapping, we don’t routinely alter the netlist. By the same token, we should not make a routine out of “fixing” the MCAD data. Every change can be an improvement, but it’s even better if it doesn’t come with a self-inflicted setback. Dialogue with that team pays dividends.

Use one side of the board. Say it again! Use one side of the board for components. This is probably the biggest favor you can do for your assembler. If the form factor permits populating one side of the PCBA, you will come out ahead of the game versus putting parts on the primary and secondary sides. Most MCUs will function with this arrangement. Explore the possibility if it makes sense.

Single-sided boards make one pass through the reflow oven. In order to solder both sides of a PCB, one side must be populated with the tiny parts only. The first soldering pass on a double-sided board will use a higher temperature solder, so it doesn’t melt again when the other side is soldered. Even then, heavier components on the secondary (downward facing) side may require glue to ensure they don’t come free.

Limit the BoM to one type of technology. Again, we’re looking at reducing processes. Surface mount is the prevailing technology, and your best bet for an end-to-end solution for your component picks. Mixing surface-mount and through-hole parts is a cost driver (FIGURE 2). The wave soldering process is trickier with surface mount included. Like oil and water, surface mount and through-hole do not mix well.

Work station
Figure 2. Placement is fine-tuned using robotic technology to align an array of lasers in this San Jose factory.
Avoid or minimize high-density interconnect (HDI). This can be tricky, too. It is difficult to fill out a BoM without a few ball grid array (BGA) packages where the pin-pitch requires the use of microvias. These tiny vias are formed with a laser, and the fabricator creates them layer by layer. The result is every layer of microvias involves a pass through the lamination press. That’s time-consuming, which is to say, money-consuming. The press is a big, expensive machine that uses a lot of power. It can be the bottleneck in a smaller fabricator’s flow. It doesn’t matter how big they are; they all charge for sequential lamination because it’s so much more work.

Getting around all of that will have you shopping for chips that come in quad flatpacks (QFPs) or other packages with perimeter pins only. Finding a second source is another chore. Regarding the assembly line, they will still have to x-ray the board to find voids in the solder joint for the thermal pad in the middle of most of those packages. We’re often working with a system of boards where one of the distinctions between the boards is the required technology. Maximizing use of the cheapest boards may help downsize the main PCB of the system.

Creativity and innovation pay off if the execution doesn’t get in the way. It’s often better to get a set of features into the field and work on yields and costs as you iterate down a list of desirable upgrades. So-called agile development can help steer a product from concept to concrete with calculated steps. Feature creep is the point, but it is tethered to market requirements, so you have the product to win the short game and evolve to the cost-competitive model as the market dictates.

John Burkhert Jr. headshot
John Burkhert Jr.
is a career PCB designer experienced in military, telecom, consumer hardware and, lately, the automotive industry. Originally, he was an RF specialist but is compelled to flip the bit now and then to fill the need for high-speed digital design. He enjoys playing bass and racing bikes when he’s not writing about or performing PCB layout. His column is produced by Cadence Design Systems and runs monthly.