So I made my first Ladybugbot and it worked pretty well. The next step was to make ten of them, and I wanted the new ones to be easy to make and reliable. After playing with the ladybugbot for a bit, I came up with a list of features and design specs.
++ Assembly needs to be snap-together. This will speed everything up and improve the reliability of my vibrating motor mounts (hot glue gets all over the place and has a lot of variability in how well it holds the motor and transfers vibrations). I also noticed that one of my motors got hot and melted the hot glue, which promptly gunked up the motor shaft. No more hot glue.
++ I need to make my own circuitry. I’m using circuitry ripped from a junk DeskPet right now, but it costs ~$15 for a legitimate, off the shelf Deskpet, and that will not abide, man. I’d also like to make it do more than be a R/C toy.
++ I’d like to have high product throughput. It takes about an hour to print out the casing and ~15 minutes to rip out a Deskpet’s guts, cut up some toothbrush heads, solder in my electronics and hot glue everything in place. I’d like to speed up the assembly time, and if I could get better than a product every hour, that would be ideal.
++ I need good yield. I’m getting a ~88% yield coming out of my makerbot
++ It should be cheap. Fancy angled toothbrushes cost $4 each. That and the electronics are the biggest cost driver in the ‘bot right now. I should be able to print my bristle heads.
Here’s the current Ladbybugbot Cost of Good Sold (COGS). I’m placing ABS cost at $50/KG and makerbot time at $1/hour (it costs ~$1000, will last for ~1000 hours and doesn’t cost much to operate).
1x Deskpet electronics (circuit board and battery): $15
2x Toothbrushes @$4/ea: $8 total
2x Vibrating motors, taken from little toys @$2/ea: $4 total
1x casing (~1:30 hour print time, 30g plastic): $3 total
assembly: 15 minutes (not valued, like a sucker)
assembly materials (hot glue, solder, etc): $0.10
1x box and packing materials: $.20
total: $30.3
Well, this is pretty high, but there’s a lot of low-hanging fruit here. The deskpet electronics are an obvious start, and the toothbrushes. If I can print the toothbrushes for $1, I can drop $7 from the COGS, and if I can make my own boards for ~$5, that’s another $10. I found vibrating motors on Alibaba for $.50 each, dropping $3 off the price. I bet I can get 30 minutes off the print time, saving another $.5, and that’s pretty good for a first pass, bringing the COGS down to a (theoretical) $9.80.
If I factor in my ~87% printer yields, every failed print costs me ~$3, so I have to distribute that cost across the ‘bots I can sell, adding $.38 to the cost of every bot.
I’d like to be making ~$5 off each bot I sell, meaning that this ‘bot, sold through a distributor, is probably coming in around ~$20 after the first pass of cost-savings. I’d like to drop that by $5-$10.
So, here’s the BOM I’m shooting for after the next redesign/sourcing effort
1x electronics kit: $5
2x printed bristles @$.5/ea : $1 total
2x Vibrating motors, sourced from alibaba @$.50/ea: $1 total
1x casing (~1:00 hour print time, 25g plastic): $2.25 total
assembly: 5 minutes (not valued, like a sucker)
1x box and packing materials: $.20
total: $9.45
One of the nice things about Hong Kong is that a design iteration takes a day. I can get circuit boards made in 8 hours, pick up parts from an electronics distributor in the city, and print a couple test ‘bots while I’m out getting all the other stuff. I can send out a new design in the morning and be testing it by evening.
So I got to working. I put together my first circuit design and picked up parts
It worked like crap–it was very insensitive to light, I couldn’t find a fitting surface mount pot in hong kong to save my life, and the vibrations would pop the motor solder joints loose in record time. I designed and breadboarded a simpler version of the electronics, and then fired up eagle and designed the next iteration of the board.
In the meantime, I tweaked the ‘bot mechanical design so the motors would snap in. I went through an iteration trying to make a snap holder for a circuit board and found that it didn’t print well. I also experimented a lot with integrating bristles into the printed case design, rather than buying toothbrushes, or even printing bristles separately. I also went through a couple iterations figuring out ways to hold the circuit board in place. After eleven iterations of the design, I finally got to a design that worked half decently.
It took me about 25 design and prototyping-hours and about $150 to get to a (ugh, mostly) stable circuit board, bill of materials and mechanical design. That’s astoudingly fast and cheap for a product, but it could be infinitely better. Looking back at my progress through the design cycle, I had very steady progress from my first through third prototypes. Each iteration took about an hour. After my third prototype, I started designing my own custom circuitry. I think this is where I went wrong. Let me explain:
Low-cost 3D printers are a powerful tool, and mine played a central role in the development of this ‘bot. After my third iteration, I switched from using the rapid-prototyping tools that I had ready access to and built with a mix of rapid-prototyping services in the area, parts from catalogs and my own rapid-prototyping tools. Suddenly, my iteration time and cost skyrocketed. On an absolute scale, it wasn’t a lot–any product engineer will tell you that iterating a hardware design in a day for $50 is amazing. Still, though, I was suddenly going through iteration upon iteration without much to show for it. I’d reuse the electronics from my first prototype in ten different prototypes, pulling the electronics out of one and swapping it into another.
The reason I was able to get initial results so quickly is that I was able to use commodity electronics–someone else’s cost-optimized design that already did most of what I wanted. In this case, I used a $.75 circuit board from a low-cost toy. The board had a rechargeable battery, IR receiver, could receive commands from a remote controller, had a small microcontroller to control everything, and could turn two motors off and on. I’m sure it took hundreds of man-hours to perfect the design and production of the remote control electronics that I used. Of course, I could create a custom board to fit my design, and sure–there are some things I’d like my ‘bots to do, but that I can’t do with these off-the-shelf boards (sense the surroundings, communicate with other bots). Maybe I’m looking at this the wrong way, though–I’m setting an arbitrary design and doing a lot of custom engineering to create something that fits that design. Why not take a more opportunistic approach and base my design around low-cost, commodity electronics than I can get my hands on? Bristlebots are a great example–all I need to make a new robot design is a mechanical shape to hold everything in place and form bristles to propel the ‘bot. Add my own vibrating motors and the electronics from a $1.60 commodity R/C car and I’ve got a radio-controlled bristlebot.
I don’t think it’s very interesting to just 3D print more plastic shells for commodity products, but it’s interesting to use mass-produced goods as lego-like building blocks to make other more interesting products. It’s not always a perfect fit, but I believe that there’s a lot of clever new designs we can make by building off of other low-cost products. I’m scratching around the tip of something that I believe is very powerful, very interesting, and very big, but exactly what that something is will be a topic for another blog post.
Until then, here’s a sexy naked picture of some ladybugbots:
