The current trend in hardware startups seems to be about outsourcing everything as quickly as possible. If you have a refined product offering and consistently ship hundreds of thousands of units annually, yes, outsourcing makes perfect sense. But if you’re creating a new product, you won’t really know your market or have a realistic volume estimate until you’ve shipped your first units.
More importantly how will you learn to effectively design for manufacturability? I strongly believe that every startup should be hand assembling at minimum their first 1000 units.
We’ve already admitted that we do things differently. We want to create a quality product that actually matches the needs of our customers. We want the ability to iterate and refine our product offering continuously during the early stages of volume manufacturing. We’re happy to slow things down and build a real company, not just hype.
And we’re not talking about Arduinos, breakout boards, 3D printed plastics, and a ton of hot glue. We’re talking about small scale manufacturing with custom PCBs and injection molded plastics.
Is this even possible on a budget of $25k? The short answer, “yes”, and we’d like to share with you the details. It will take more than a few articles to cover all the details, so for today we wanted to share with you our basic methods and setup.
Custom Printed Circuit Boards
The cost of high quality, low volume PCBs has plummeted. As of today, Seeed Studios offers 10 2-layer boards, with 10cm by 10cm maximum dimensions for just $4.90 USD. That’s just $0.49 per board at quantities of 10. For a hobbyist portal, they even offer great quality and great pricing at initial run quantities (1000-10,000 boards).
There are hundreds of board manufacturers out there happy to take small volume orders. There’s really no excuse not to be developing on custom PCBs even in the early prototype stages. I tend to skip right past the breadboard and just spin up a custom dev board as quickly as possible.
Currently, Seeed Studios is our fab of choice for prototypes and initial runs. Our last prototype was at my door within 5 business days from the date of order and also included a custom stainless steel stencil.
Pretty much every new product developed today is going to involve surface mount components. Even a simple device can contain anywhere from 50-100 parts. Obviously hand population of 50 components per board in unrealistic past 100 units.
Board population services certainly exist, but at low volumes the setup cost tends to be prohibitively high. With our first product we ran the numbers on outsourcing our first 500 units. We saw quotes in the range $20-$30 per board when you factored in the setup costs. That’s $10,000-$15,000 in population services. Can’t we just buy an SMT pick-and-place for that price?
It turns out you can. After a bunch of online research we settled on the NeoDen4. Our final purchase price ended up just under $10,000 USD including all our required attachments, upgrades, and freight. (Full disclosure: we technically already own the NeoDen4, so we’re not including that in our $25k for this project).
So far our experience with the machine has been great. It’s nowhere near as fast as those you’d see in a high volume setup, nor can it handle the smallest of components. We stay away from BGAs and use 0603s whenever we can. The NeoDen4 can handle 0402s, but overall if we can spare the space I’m happy to work with larger components. At our volumes, speed isn’t really an issue. For example, we’re currently running a panel with 6 devices with about 60 components per device (about 360 total components). Without optimizing for speed, it spits out a completed panel in about 15 minutes.
Do you absolutely need a pick-and-place? Not necessarily, but it was a good investment. I hand populated our first 100 TallyFi’s and used that cash to buy the pick-and-place. With that capability in house we can now easily make minor changes, custom runs, and build inventory as needed.
Of course, there is still solder paste application and reflowing. I hand squeegee the paste using a stainless steel stencil (about $20 with the Seeed Studios order). We explored low cost reflow ovens, but to be honest we still use a modified toaster oven that runs custom control software I wrote years ago. It gives us full control, and I haven’t seen anything in a reasonable price range that can match the performance.
I’ll save the details of low cost tooling for low volume injection molding for another time. Just be aware that a simple quality tool can be had for under $5000 USD.
At low volumes all you really need is a desk and a spare evening for assembly and packing. With the boards populated we load the firmware, run a software self diagnostic, and assemble the unit. In full time production mode I can put out anywhere from 50-100 units in a single day.
That may seem like nothing, but the goal is just getting the first 100 or 1000 units into the world as quickly and cheaply as possible. We just want to show that it is possible without a huge investment as long as you have the time to invest in a simple setup.