DURR: The shell
Posted in Manufacturing , on
We have a few hundred iterations on the casing. But mainly, we used our Makerbot for rough estimates on height and diameter, and tried a few different moulds for stability and materiality, before ordering accurate test samples from Shapeways.
Same with the strap fasteners, we went back and forth on standard watch-fasteners, button studs and custom-designed friction-fasteners.
We ended up with the self-made friction-fasteners.
We decided to use the strap itself as a back-cover, to keep the amount of parts down.
When we found the perfect dimensions at last, we had decided to make 50 pieces to test if there is a market.
With that low a volume we figured we could order them from Shapeways and dye them ourselves:
We used dye from Rit, after reading this post from Colleen Jordan at Wearable planters.
What we learnt from this is more interesting than photos of the process.
In every step of the way we encountered problems. Every article you read online about manufacturing says that you can't underestimate the amount of time manufacturing takes, and our case was no exception. What might differ though, is that we've done EVERYTHING ourselves, so if and when we go bigger, we know where we are likely to meet problems. Yes' it's taken a huge amount of time, but we haven't spent an enormous amount of money.
We've had to manually glue a small attachment to every button because of a misalignment, test-dye each sample a dozen times to find the correct shade, and calibrate each single IC because of an inconsistent timer circuit. If we'd had a large manufacturing scheme from the beginning we'd be screwed.
One of our principles is that we want no investors, be it the VC-kind, angels or crowd-sourced. We'll bootstrap everything, and if we can't afford big-scale things, we'll do it small. This means we don't owe anyone anything, but it also means we can't afford high-volume manufacturing. As a consequence, every added component challenged our ability to afford making DURR at all. When we do small-scale we don't get the same high-volume discounts as bigger companies. So we had to keep the component-count low, and tweak what we had to fit. All in all, we ended up having 15 components, and that's not just the electronic components, that's also including the leather strap, both the strap fasteners and the casing.
Making Durr Part 1: Electronics
Posted in Manufacturing , on
So we just released the first 50 of Durr for sale, and we'll be writing a bit about how we made them.
In this first post, we'll talk about how the electronics were made.
Prototyping
When we got the idea initially, it was a test we did for fun, to see for ourselves how we experienced time in different situations.
We found an old Arduino we had lying around (unfortunately no photos were taken), and wrote a simple program that made a hooked up vibration motor go off every 5 minutes. Then we strapped it to our arms.
Wearing it on our wrist for the rest of that day, we realised we had to make a better prototype. Not just because it was clunky to wear, but because it was really cool.
The speed of perceived time really did change
, and the thing worked way better than expected.
So we looked up how to shrink down arduino projects, and found out we could use arduino as an ISP to program smaller microcontrollers. Like the ATtiny45, which we bought a few of.
It has more than enough space to hold the code needed, and can run on 3V provided by a button cell. For anyone interested, this article by High-Low Tech is almost everything you need to get started. So we reused our old code and fed the ATtiny45 with it, and with some ad-hoc soldering, a Makerbot-printed casing and some electric tape we got this:
It wasn't beautiful, but of a more decent size at least. So we figured, we'll make a few beautiful ones to sell! But more testing was needed.
We wore it for a couple days, and then noticed a problem. It ran out of power. Two days isn't nearly enough for a product to be sold. We needed to cut down on the power consumption of the ATtiny. After a lot of research, we figured out how to make the chip sleep for most of the time, only waking up once in a while to check if anything was happening. It now lasted for two months on one CR2032 battery!
Next, we had to test the time between vibrations. We had a feeling that 5 minutes was optimal from the first prototypes, but we asked a few friends to be test subjects on 10 minutes, 15 minutes and with different vibration patterns. Long story short, we were initially right, 5 minutes is perfect. You'd be wrong thinking that it's stressfully often. It's not. Promise.
Making and ordering the circuits
Deciding on the right vibration pattern and time gaps, we sat down to make circuit boards and order parts. We've been playing with electronics before, but only for one-off prototypes and installations. So we'd never done PCB design before, which we had to learn from scratch. The components we bought from DigiKey, whose UI everyone seems to hate but we really like. Anyway. For the first testrun we tried Fritzing , which is really simple to use, but a bit limiting. You can actually order PCBs via the EDA , which we did, and got the ones below:
They did work, but not as good as we wanted. And we'd forgotten buttons to turn it on/off! For testing purposes though, it was allright.
So we redesigned the board, this time in
Kicad
, which was a huge pain to install and get the hang of. But more accurate than Fritzing.
Before ordering we printed the circuit on paper to verify that everything fit. It did.
This time we ordered the PCB's from OSHpark , which batches orders from customers to keep the prices down. They have a great finish, and they're purple. Which is cute. At this point, we figured the PCB was too tall with our DIP components and that the battery holder was too clunky, and we once again redesigned the board to use a few SMD components. We printed the final PCBs with Seeed studio, both because they had black finish available, gold-plated ENIG pads and really thin boards.
Soldering the electronics
We had to do all the soldering ourselves, and let's just say high-temp, lead-free solder is challenging to get to be friends with SMD components. Components next to a Norwegian krone:
We managed though, and soldered the 50 boards in triplets:
Bugfixing the microcontrollers
Just when we thought we were done with the electronics, we found out the internal clock in the microcontrollers were inconsistent. Shit. At this point, we were unable to use an RTC or crystal to keep the time consistent, so we had to manually calibrate each chip. That was no fun. In the end they were tweaked to be all good, and we'd learnt a lesson.
In the next post we'll be talking about how we did the casing of DURR.
