The Fabbing of the Badge
How the Hackaday Superconference Badge badge went from 3 to 300
It started with an email. It always does.
The badge project manager (Chris Gammell) asked if I could build 300 badges. It would be for the Hackaday SuperConference in November. We were approximately 8 weeks out. That seemed like more than enough time.
Files
The next step was fairly typical. Incoming bombardments of zipped-up CAD file attachments. As the person responsible for assembly, I needed to pay special attention to:
- Gerber files, including the drill files.
- The not-so-standard, comma-separated-value (CSV) file that lists parts names and locations (XYRS).
The former files are first sent to Rocky, my PCB fabricator in Shenzhen. Then I sent the gerbers (specifically the “paste” layer) to my solder paste stencil manufacturer.
The latter file informs the pick and place machine about what parts go where. My goal with my company (Small Batch Assembly) is to make sure this is always done correctly. Placing parts onto boards for low runs of electronics.
Discussions
Chris, the board designer (Voja Antonic) and I had a Skype call to discuss potential assembly issues. The most common errors come from the paste layer gerber files. We focused on that first. These pad sizes can be adjusted by hand to get the correct amount of paste, which drives the quality of assembly.
The pin pitch (spacing between pins) is the most crucial element. Above 1 mm is not a concern. Below 1 mm, the paste, placement and inspection (after the fact) are crucial to success.
An example on this board was the PIC18 microcontroller (U1). It is an SSOP package with a pitch of .65mm. The LIS3DHTR Accelerometer (U4), was an LGA16 package with .5mm pitch. As low as .45mm pitch parts are plausible, but it does push the envelope of what is possible with my machines. Success is based upon stencil registration accuracy. For parts like the LGA16, pins that are unexposed at the edges does not lend itself to inspectability.
The front side held another challenge: the USB Connector (CON1). During discussions, Voja explained past issues with assembling a similar surface-mount part. Since USB connectors need to tolerate a great deal of mechanical abuse, he swapped to a version with alignment pegs. The quasi-through-hole version upped our confidence in the mechanical reliability, but automatic placement was not guaranteed. I checked each one manually and nudged them into their holes before they went into the reflow oven.
Yield is defined as the total number of functional boards as a percentage of the total created. It was the accelerometer and the USB Connector that impacted our 90% yield. This would be unacceptable for a high-run manufacturing board. For a low-run pilot build, it was fine. We had time to rework some units and created extra to account for this scenario.
Let’s Panelize, Shall We ?
Bundling multiple copies of the PCB design is called panelization. This reduces labor and lowers cost during assembly. Dealing with panels is faster than dealing with individual boards. This benefits each step of the process: paste application, part placement, and reflow. I normally leave this to the PCB fabricators, for a more standardized workflow. I do this is only after careful guidance from my customers. To that end, I created a fabrication guide drawing for my fabricator, Rocky:
I prefer a mixed v-score and inside mill panel. Having seen this fabricator do it before, I was confident in his ability to follow my drawings. “V-score” utilizes a triangular bit to define the outline of the PCB. The partially-milled boards can be snapped apart after assembly. There areas marked “route” above are the inside mill cuts, completing the board outlines. I specified a 5-up arrangement which optimizes placement and fit on my equipment. I talked to Rocky via Skype to make sure his engineers understood my instructions. I felt a reasonably confident they understood my requirements. These face-to-face interactions (even via a screen) can be crucial to success.
Off To the PCB Fab
42 days before Supercon (on September 23rd), finalized designs went to the fabricator. The supplier promised delivery by October 10th, 25 days before the conference. Much like Voja explained in his post, we needed to buffer our schedule. The badges needed to be in Pasadena well in advance of the conference date (November 10th). The final through-hole assembly and testing of 300 plus badges was still required.
I only had about a week for production after the boards arrived. Time-wise, this was now cutting it very close. There would be very little time to recover in case something went wrong with assembly.
Part Ordering
We needed to be ready when the boards arrive. The focus turned to ordering and receiving all the parts. This task fell on Chris to order the parts and I would let him know as things arrived. The supply chain included shipments from Bulgaria, China, and the US (North Dakota).
We kept in touch as invoices and delivery notifications trickled in. It appeared the parts would get there on time.
Boards Arrive!
The boards arrived 3 days earlier than promised. This allowed me to use the weekend to produce the first articles for verification test. Moreover, it added to our time buffer, which increased peace of mind.
The first panel went through assembly, I shipped it to Voja in Belgrade for verification and test.
Here is a video of the first boards’ placement:
Production Is Good To Go !
The next Thursday (22 days before the conference), I got word from Voja. The first panel checks out. Production can begin.
After all of this build-up, production was an uneventful march. 60 panels must get pasted, placed, and reflowed. In reality this happened twice, because they are double sided. I was sure to do inspections along the way and each panel passed visual inspection, despite some small issues.
With 15 days left before the conference, production was complete. It took 30 hours and more than 50 thousand parts placed. At the Design Lab in Pasadena, Voja and his assistants performed final assembly. This consisted of de-paneling, soldering battery holders, and attaching an acrylic frame. Finally, they flashed the badges with the bootloader and application firmware.
Read more about the badge construction and details here.
Usually, I don’t hear much about where boards that I assemble end up. This badge was different. I relished the opportunity to see the final product as part of a technology conference. Seeing all of the creative puzzles and hacks posted on Hackaday made the journey worthwhile.
