Cinnamon ][ Lessons Learnt From Launching A Kickstarter Product

Cinnamon II the ultimate retro smartwatch
Cinnamon II The Ultimate Retro Smartwatch Kickstarter Project.

The Cinnamon II is an Apple® ][ compatible wrist watch. It’s features 32k of memory and a 1 Mhz MOS6502 virtual machine. It achieves hardware compatibility with the early 1977 and 1979 machines. It includes an Applesoft compatible BASIC interpreter, a MicroSD emulated drive and an onscreen keyboard display for user interactions.

The inspiration for the project came from Aleator777’s incredible Instructables project and my previous work on emulating the Apple ][ hardware on low speed embedded platforms. It also represented a fantastic opportunity for me to work with the ultra low power ARM Cortex microcontrollers.

This post is going to focus on the lessons I learnt while developing the Cinnamon II and what it takes to get a hardware product onto Kickstarter as a maker.

 Product Design Process

The first step of designing a product like this is to really understand just what you are attempting to build. For the Cinnamon II I had the pleasure of having a huge amount of community feedback on the original Instructable project.

It’s cliche but it really pays to seek out criticism on the idea, there was a variety of ways in which the Instructable fell short. Firstly the original watch is considerably larger than the photos suggest, it’s very chunky being over 3 inches across. Secondly the original battery life was on the order of only a few hours. Thirdly and which is where I stepped in, it actually wasn’t actually an Apple ][.

The first question for me and any design process, is just how big is the product actually going to be? Enclosure size makes a huge impact on all elements of the final design. Sizing the enclosure came down to understanding what was going to be the critically limiting component.

For the Cinnamon II this critical component was discovered by searching for dimensions of various critical parts such as the battery, screen, button caps and even standard watch band sizes. For this project the screen turned out to be the critical dimension.

Once realizing the screen was going to be the most important element of the design I set out by doing a series of hand drawn sketches around the specified screen dimensions. My sketches were mostly influenced by physical constraints such as the band size and the human wrist but also influenced by design features found early microcomputers.

Sketch Of Case Constraints.
Sketch Of Case Constraints.

 Electronics Design Process

Choosing the processor on which to base the project was subject to several difficult constraints. It had to be fast enough to emulate the Apple II, thankfully I had run previous experiments involving emulating the platform on AVR’s. None of the AVR’s were anywhere near fast enough (perhaps 3x too slow @16MHz). So I wasn’t going to be using an AVR, there’s a variety of faster chipsets around, things like the MSP430, ARM Cortex, and many others.

The next two constraints were crucial, the chip had to be very power efficient and physically small. In the tiny form factor of the watch, something like a LQFP144 was not going to fit. After much research I decided to use a STM32L1 (ARM Cortex) device from ST microelectronics. It was fast, available in a tiny 64pin package and incredibly low power.

The STM32L1 was perhaps slower than I’d like with a 32MHz maximum clock. But the ARM instruction set is faster than the AVR and gcc offers better compiler optimizations for the platform.

Despite being only twice as fast as the AVR, the ARM Cortex-M was able to achieve three times the performance on my optimized 6502 virtual machine. Based on this I’d suggest with optimizations the ARM Cortex-M is approximately 50% faster than the AVR instruction set at the same clock speed (fast flash memory helps a lot).

Interesting to note, you can overclock STM32 chips and I think they are perhaps very conservatively rated. Due to a glitch involving a bad PLL multiplier I accidentally ran the STM32L1 at approximately 50MHz. The watch successfully booted up and ran everything in double speed with no notable corruption.

 Phase One - Development Boards And Breakouts

Once I had settled on the chip I was going to use, it was a case of hunting for a prebuilt development board. At first it’s often far cheaper to get a development board than to look at fabricating PCB’s. Thankfully ST Microelectronics offers STM32-Discovery boards for most of their product line.

There was a variety of support components required, items like the lithium charger, and the main switch mode power regulator. Selecting these components took a while, I began by searching Sparkfun for similar boards and noting the power regulator chips that had been used. This allowed me to spot some popular chips (increased availability and lower cost).

After getting some ideas from existing products, part availability and performance specs. I settled on a two chip design based on the TPS63031 3.3v switched mode buck/boost regulator, and the MCP73837 lithium charger, Looking back I’d replace the MCP73837 with a cheaper USB only device. However the TPS63031 was a fantastic device, incredibly resilient, efficient and very tiny.

Once I had an idea of the support components I ordered a bunch of breakout boards from Sparkfun, Adafruit and others. These breakouts were great for initial testing and they provided a reference design that could help when laying out the printed circuit board.

 Phase Two - The First Printed Circuit Boards

Prototype Printed Circuit Boards
Prototype Printed Circuit Boards.

Once I had experimented with the breakouts and ST Discovery board I felt confident that there was no major issues with the design. I looked at doing the first set of printed circuit boards. These first boards focused on testing the overall system integration as well as issues such as device footprints etc.

I think for the first set of boards, it’s not particularly important to consider the final enclosure/board size. I choose to design a relatively spacious board with plenty of test points and wide tracks etc. The extra space also allowed me to do a two layer design. By designing the first board with large tolerances and only two layers I was able to leverage one of the many board prototyping services available.

Board batch protoyping services are fantastic value for money, and I highly recommend leveraging when available! For early board revisions you can get boards made for a tenth of the cost of spinning up your own panel. This does come with a potentially significant increase in lead times but if you stick with things like the default board thickness for the fab etc you can cut down on a lot of delays. I have heard good things about ITEAD Studio, DirtyPCBs, Seeed Studio, and OSH Park.

For my multilayer, high density boards I have used Gold Phoenix PCB, but prices can become quite expensive when scaling up but the lead time is very low. Looking back, you can save a huge amount by being smart about PCB’s, for short runs they end up being a very significant cost.

 Phase Three - Final High Density Boards

First Revision Printed Circuit Boards
First Revision Printed Circuit Boards, Including Errata.

The final phase is taking the verified design and scaling it down for the final printed circuit board. If you are lucky you’ll have an enclosure large enough that you won’t need to scale down at all. But if you do, for an example a wearable, there’s a few ways to make the process much easier.

Firstly a perhaps counter-intuitive statement for first timers, the smaller the part, often the better. When you move away from hand soldering toward reflow soldering, smaller parts have some very significant advantages. The effect of surface tension is increased at small scales. This actually makes it far easier to solder tiny parts as they will pull themselves into alignment.

I often see people worrying about whether or not they can solder SMT devices etc. Surface mount is terrifying to a lot of makers and this is unfortunate because in a lot of ways surface mount is easy! You will need some different tools but now days they are incredibly cheap. Just grab a Yihua 858D hot air gun, some leaded paste, toothpicks, and nonmagnetic tweezers for a total of approximately $50 USD.

With a basic hot air gun you’ll be able to solder discrete parts down to below 0402 and some of the tiniest QFN packages you can imagine. Use all surface mount parts and go as small as required. They reduce cost, and so much difficulty.

Given the focus on small parts, be careful to consider the track tolerances, most PCB fabs will charge extra to produce high tolerance boards. Often it makes a lot of sense to be careful to maintain the maximum clearance possible. You’ll get a better yield in terms of non defective boards and you’ll save a ton of money.

As for manufacturing the boards, if you are doing less than perhaps one hundred boards (depends on size), panelizing boards will not save money. You’ll pay far more in fixed costs for the fabrication masks than the board costs. It can be a necessary evil for pick and place etc, but it will cost much more.

Also to note I used an interesting trick to reduce the cost of board manufacture. My particular circuit board fab, wanted to charge more for complex routing. For the watch I needed the board to have several corner cutouts to fit the screw mounts of the case. Instead of doing an irregular shaped board I was able to order a rectangle board but add drilled breakaway corners, saving quite a bit of money. Easy trick for complex shapes where perfect finish isn’t necessarily required.

Using Breakaway Tabs To Reduce Prototyping Cost
Using Breakaway Tabs To Reduce Prototyping Cost.

 Case Design Process

Probably the most foreign part of the product design process to us makers is designing and manufacturing the enclosure for the product. A really good case can make all the difference in the world. It takes a bare circuit board and turns it into something real.

There’s a variety of methods and tools that can be used to design a physical object. These processes usually full under the banner of Computer Aided Design (CAD). Many makers aren’t fully exposed to CAD, but that is starting to change with the 3D printing craze. There now exists a variety of competitive open source tools FreeCAD and OpenSCAD are two major players.

Solid modelling is one of those skills that really pays off to learn. However it is a big time undertaking and the online tutorials don’t really seem to meet the standards we’ve come to expect from electronics and programming, FreeCAD Tuorial, Introduction To CAD.

3D Model Of The Cinnamon II
3D Model Of The Cinnamon II Smart Watch.

As for the actual design process, there’s a few things that can really help the process. Firstly, upfront, decide on things like how thick you’d like the case walls to be and what the minimum feature size on your design can be. The minimum feature size will impact every element of the prototyping process.

Secondly if your case has anything resembling tight tolerances, especially on components that make up the critical dimensions. You’re going to want to have actual measurements for the components. With the Cinnamon II I had issues with the screen being slightly larger than expected, this was a source of much pain in the early prototypes.

Paint Curing Under Heat Lamp
Paint Curing Under a Heat Lamp, Showing Surface Porosity.

 3D Printing Prototypes

For the prototypes used in my kickstarter campaign I used 3D printing to do the cases. For low numbers 3D printing is really the cheapest and fastest option. There’s several major 3D printing technologies used commercially at the moment.

Extrusion, which is often used in maker printers results in relatively poor dimensional accuracy and poor surface finish. An extrusion printed case would have taken much further processing in order to get prototype ready. Laser sintering, is another common technology, it solves the dimensional accuracy issues and generally results in a better surface finish but it does result in significant porosity which can make painting difficult. Stereolithography is the most accurate of the three, it results in very accurate prints without porosity issues but uses a UV cure resin material.

For my prototypes I used Shapeways strong and flexible plastic printing, it was cheap, relatively low lead time and had great dimensional accuracy. I thought it would be a simple choice. It wasn’t, the Shapeways prints exhibited great porosity. I was originally planning on sanding the prints with a fine grit paper and then vapor polishing using acetone and then spray painting using a plastic primer.

That plan failed in more than one way, firstly nylon can’t really be vapor polished as it’s incredibly chemical resistant. Secondly, you can’t spray paint with that kind of porosity, the paint instantly beads on the surface of the plastic and results in a very splotchy finish. Painting raw sintered nylon is very painful and it’s very difficult to get a smooth finish, it’s like trying to paint a brick. In the future I’d like to experiment with Shapeways polished offerings.

In the end I used multiple coats of a thinned out acrylic primer to essentially seal the surface, and with some painstaking fine grit sanding and polishing I was able to get a fairly smooth finish I could spray over. Was far more time intensive than I would have liked.

 Urethane Casting Prototypes

Looking back I’d do the prototype cases using a combination of a detailed SLA 3D printed master and cast polyurethane. It’s not an approach I’ve seen used in maker the community however much of the props industry uses the approach heavily so it should be fairly easy to source the resins.

There’s a number of reasons for preferring urethane casting, you get a very good surface finish with no porosity. It’s much more cost effective than 3D printing, the molds can be reused up to ten times and the two part resin is fairly cost effective. The resin does not need to be painted, as pigments can be added prior to casting.

Looking forward, in the future I’m going to do experiments with DIY vacuum casting, hopefully should present a great way to scale up 3D prints. Some great examples of the urethane casting process: Urethane Casting A Lightsaber and Commercial Vacuum Casting

 Conclusions

All in all the Cinnamon II project was a lot of fun to do and incredibly rewarding. Learnt a lot of tricks about manufacturing prototypes, and spinning up ARM processor boards.

I think the big take away, is that recent technology advances have made all this stuff accessible to the average entrepreneur. Not too long ago, taking on a project like this would been prohibitively expensive. That’s no longer the case, dream big!

 If you are interested, The Cinnamon II is on Kickstarter

Apple and the Apple Logo are trademarks of Apple Inc.

 
48
Kudos
 
48
Kudos

Now read this

Turning The Arduino Uno Into An Apple ][

Emulated Apple ][ Running On a Stock Arduino Uno. I’ve always been fascinated by the early days of the computer revolution. Today we take tremendously powerful machines for granted but it was not always that way. As a personal project I... Continue →