Build Better PCBs, Faster: A Practical Guide to Speed Up Your In-House Electronics
Let’s set the scene: You’ve just got your novel circuit design working on a PCB. Stress levels start reducing. You solder another one. It works too. You buy a coffee. It’s delicious. Excited, you finish the builds and present to your team. The CEO impresses investors with the updated prototype. The product team immediately starts gathering feedback about the new device. You look at your design, and smile.
Life is good.
…until you remember you need to make 300 of these this month 🤦♂️ The fantasy breaks. Your coffee runs out. You remember that as an engineer, making something work is only a small part of the picture. Lame.
In the world of consumer electronics, Design For Manufacturing (DFM) is talked about a lot. Care must be taken to ensure that the materials, construction and testing of your hardware can be carried out by contract manufacturers in a reliable and cost effective manner. Less discussed though, is the long and tricky period of rapid prototyping before DFM becomes front of mind. It’s not uncommon to go through 20/50/100 in-house prototypes before initiating discussions with CMs and suppliers. The speed at which you do this can be the life or death of the product — and in extreme cases — the company.
If it ain’t broke, d̶o̶n̶’̶t̶ you still might need to fix it
Above is an early prototype of a brain stimulation headband I worked on in 2018. We were proud of these headbands. Still are! They brought in our first lot of invaluable customer feedback, and were even demoed at Berkeley SkyDeck Demo day in front of the best VCs and angels in the valley. But there was one problem → they took ~3 hours to build, and we needed over 100 of them. With only two full time engineers at the company, these builds became a headache. Towards the end, we despised making them, despite loving the product itself. How did this happen?
Answer: Electronics. Well… Mostly the electronics
80% of issues arose from difficulty soldering just three components — the micro USB connector, the micro-controller, and this ridiculusly tiny USB communication chip. It’s best to stick with the largest and most abundant components for as long as you can, and there are much more reliable ways to flash an MCU than with a consumer friendly micro-USB port (read: pogo-pins, or even headers).
Speaking of soldering, the first few batches were surface mount soldered with just tweezers and a heat gun. If you don’t enjoy picking up grains of sand and moving them to exact locations, then you’re probably not gonna enjoy this, either. And neither will your productivity.
But it’s not all about mastering soldering skills. Where and how you store your components has a huge impact on build time. Initially, we lumped everything into a plastic box and individually searched for each component as needed. This is a tempting (yet ultimately damning) method to deal with small scale inventory, and cost us a lot of time searching for, and counting, and recounting, and recounting, and recounting our stock.
And finally, the box. The beautiful, white 3D printed box. What began as an idea to speed up the assembly time (such as including ridges for the battery to sit against, custom dimensions to snugly fit the PCB, drill holes, a slot for the switch. All quite thoughtful actually… in theory) wound up as an untimely dependence on a forever-dysfunctional 3D printer.
Lesson: 3D printing is the best manufacturing method in the world, as long as you only need 1.
Time to speed things up
There’s almost always a better of doing what you’re doing. Our biggest areas of improvement were as follows.
An inventory is like your own personalized online catalog of electrical components; the go-to when you want to know how much of any component you have, where it’s located, how much it costs, what its datasheet is, and whether you need to order more or not. When those Digikey/Mouser boxes arrive, take the time to add them to an inventory. It’s a small time investment and will quickly pay itself back; usually by the first order. There’s a variety of options out there — paid and unpaid, basic and advanced feature sets. If you’re just starting, don’t worry about the intricacies for now, you only need basic features to start saving time, such as:
- An up-to-date count of each component you have
- Detailed information on each part you add (datasheet, footprint, cost, manufacturer, etc)
- Ability to group parts into separate BOMs or projects
- Fast search functionality
For this, we use PartsBox which includes these features among others.
One of them is a handy ‘build’ function which deducts components from their total amounts based on the BOM. The UI is clean and simple, and the search functionally is exceptionally fast. There’s also a Low Stock report, which keeps a running list of any stock that’s dipped below a certain (definable) level. Since using PartsBox, we’ve shaved ~70% off the time it takes to order components — mostly because we rarely need to count them anymore. There’s also support for purchasing in paid plans, but we’ve not needed this so far.
Let’s break this into three sections: Soldering, assembly, and reflow. Here’s how we sped each of them up.
A standard solder paste syringe will take you through one-offs and board repairs, but relying on this will quickly leave your hands sore if used in a production run. Using a PCB stencil quickly addresses this problem, as well as being more accurate and much, much faster than doing it by hand. If you’ve not come across stenciling before, check out this SparkFun tutorial video to see how it all works. It’s not much of an extra expenditure either — a stainless steel stencil costs around $20, and this $17 ChipQuick solder paste from Amazon worked a charm.
To align the stencil with the PCB, its common to make a frame for the PCB to sit in, then place the stencil over the top. Personally, I’ve found better results with this following method.
First, Dremel (or CNC mill) a footprint of the PCB in a piece of cardboard or thin plywood. Here’s an example using a Dremel.
Next, tape the stencil over the milled platform, making sure it aligns accurately with the pads on the PCB. Keep in mind that you’ll need to lift the stencil up, so don’t tape it too tight!
This forms a kind of trap-door setup, where you can easily lift up the stencil, place the PCB into the footprint, lower the stencil down, apply the solder paste, and repeat. It’s blazingly fast ⚡ Using this method, we were applying solder paste to a new board every 50–60 seconds.
I’m sorry, tweezers, it’s not me — it’s you. You’re slow at getting components out of their packaging, and then I need to think about the orientation after I’ve fumbled with the diodes and dropped them a couple times en-route to the board.
Hand-held pick and place setups save the day here; making it super easy to pick up components directly from their packaging, with the same orientation. every. single. time.
I’d recommend the pumps with a foot-actuated pedal, so you don’t need to tense your hand while trying to drop the components on their pads. Like the trap door stenciling setup, this solution is remarkably cheap, especially if you follow this tutorial and make your own. But if time is of the essence, this electrical pick and place pen from Amazon should do the trick too.
The final step here relates to how you organize your components. Now that you can pick them up from the tape directly, its time to arrange them accordingly. Using a flat piece of cardboard, stick your components down row by row with double sided tape (keeping them in their packaging, of course). Now, you can effortlessly access everything with the vacuum pen.
This technique may just be the fastest way to assemble boards, bar an actual pick and place machine (but if you disagree, I’d love to hear from you!). Using this method, a single engineer was able to assemble 48 boards in a typical 8 hour day. Previously, the record was 6.
The last step in the process. Refining this step is more about quality than speed. Even the cheapest heat guns can get the job done, but the process hits a plateau rather quickly. Attempting to speed this up leads to:
(a) Scattering the components across the board because the fan’s too strong, or;
(b) Setting the heat too high, overheating the substrate and destroying some traces.
Next stop: A reflow oven.
Using an oven is an effective way to guarantee a repeatable, even heating profile for your boards. It only takes minutes, and lets you do other stuff in the mean time. We used this small cheap one from Amazon, and have had a great deal of success with it considering the price! We were able to set our own heat profile for our specific PCBs too, which helped reduce the number of solder bridges and tombstones components. The result was 12 soldered boards in just 7 minutes — a huge improvement for our capabilities in a small workshop environment.
Design for Testing
Yes, you can solder a wire to any pad on your PCB to test it. But is this the best way to do it?
The answer will always be no.
Especially for new PCB designs (or designs where >30% of the schematic has changed), be liberal with test pads, solder jumpers and breakouts. Make a larger, uglier board first that’s full of this stuff before committing to fabricating the ‘final’ thing. Make the assumption that the first design won’t be perfect — in fact, make the assumption that nothing will work at all — and include this sacrificial dev board into your timeline.
One trick that’s saved me a lot of time in the past is to isolate each sub-circuit with a solder bridge, so that its super easy to test everything in isolation.
Adopting this approach over the last year has softened the blow of setbacks from a design not working, made the root cause easier to diagnose and correct, and improved cohesion between the engineering team and other teams that depend on a working product.
Ok. Cool. Nice. But what about when the schematic is already locked in, and you’re confident in the design?
If your design requires numerous QA checks — voltage, firmware flashing/testing, IO tests, discontinuity tests, etc — a pogo pin testing jig is probably the best way to simplify the process down.
This jig allowed us to test 14 outputs without needing to solder headers onto the PCB breakouts. Plus, by soldering the pogo pins directly to the protoboard, we eliminated the chance of interference from the eventual breakdown that comes with a breadboard, and could be confident in the electrical connections. We used simple rubber bands to fasten the PCB down (unconventional, but it works!) and could mount / dismount the boards reliably in less than 10 seconds. A big time saver.
And finally, the PCB Fabrication
The consensus of the 2019 Fictiv State Of Hardware Report is clear: Electrical Engineers want their PCBs yesterday (paraphrased, of course……you get the idea). Despite the tariff-man taxing us more for our resistors and diodes coming out of China, Shenzhen remains the hub for fast and cost effective electronics; PCB fabrication included.
There’s a long list of PCB manufacturers which readily work with large corporates, all the way down to small scale companies and startups. Elecrow, JLCPCB and SoonEasy are examples of high quality PCB manufacturers.
Turnaround for a rigid 2-layer PCB order can be as fast as 4 days — and that includes delivery to the other side of the world. From personal experience, the cost per unit will be 2 →10x less than local PCB manufacturers (in the United States) depending on the quantity. There is also an exceeding amount of capability and for advanced PCB manufacturing techniques; including FPC, rigid-flex and high performance substrates (such as aluminum).
Since switching to a Shenzhen-based PCB fabricator, we cut our design iteration lead times from 2 weeks to 4 days (compared to working with US-based manufacturers). As for the economics, we are now paying between 2–10x less for PCBs too. Whilst local manufacturers continue to play a critical role, the message for hobbyists and startups is clear: Deal with Shenzhen companies directly when ordering PCBs.
Wrapping it up
This guide covered the following topics:
- The importance of being able to prototype quickly in a fast-paced hardware startup
- Common pitfalls which slow down development; tiny component selection, only using tweezers to solder, not inventorising, unnecessary 3D printing
- High impact methods to speed up in-house manufacturing; using Partsbox.io, trap door stenciling, hand-held vacuum pens, reflow ovens
- Designing for testing and better project scheduling; test pads, solder jumpers, breakouts, sub-circuit isolation and pogo-pins
- Faster, cheaper PCB fabrication in Shenzhen
These lessons emerged from a lot of mistakes, successes, trial-and-error and persistence at humm, where we’re working hard to bring the most accessible and effective brain stimulation products to market. And looking back, these lessons couldn’t have come sooner.
From first focusing on a feature rich brain stimulation headband (ft. that laborious white box!), to really listening to our early customers and understanding their biggest pain points with the product and the tech, we were able to pivot and start iterating on things they actually cared about. With the inertia of hardware, this process is never fun (in fact, its downright debilitating), but its essential in creating something people can integrate into their lifestyle. If you’re interested to learn more — and see how different the product is — feel free to check out our (refreshed) website.
I hope you find these lessons as useful as we did, and that they lead to faster, more productive prototyping! Let me know in the comments below if you implement any of these suggestions, or have suggestions of your own. Lets keep the conversation going 😎