Four months ago we were screwed.
But that's putting the cart before the horse - which as we all know is a no no. First, a brief background: I'm part of a volumetric imaging company, Looking Glass. We invented volumetric printing the previous year as a way to seed the ground for our big dream of dynamic volumetric display. You know, the Princess Leia technology of sci-fi dreams.
And four months ago, we'd just spent all of our personal money and meager company earnings developing the dream - a high-resolution, full color-depth true 3D volumetric display. The only problem was, this display of the future was the size of a refrigerator and made a volumetric image just a little bit bigger than an ice cube. It would be years before we'd be able to turn it into a commercial product.
Then, on the precipice, we had an idea.
This is the unfolding* story of the L3D Cube, as it happens. My hope is that by being a nearly current play-by-play update of how this story unfolds, the narrative fallacy that plagues all post-game product launch accounts can be mitigated if not altogether avoided. If additional details of any step along the way - technical, business or otherwise - would be useful for folks please let me know and I'll expand the section best I can.
*unfolding on Kickstarter now. I will be updating the later steps of this Instructable as the present catches up to the future.
Not interested in all of this strategic product development stuff and just want to build your own cube from scratch? The links below are how to make it happen!
All of the hardware & software described here is completely open source under a Creative Commons Attribution-Sharealike 2.5 license.
That being said, a word of warning: it will be hard to build a single high-quality cube from scratch, since it involves large even-temperature reflow ovens which are usually only available for large scale production. That's why we think there's value in us providing a kit, because we deal with that nonsense. But if you want to dive in, here are the detailed designs:
Cube hardware: https://github.com/enjrolas/L3D-Hardware
If you're looking for details on how to start creating on these volumetric LED cubes, more technical details on Alex's "How to Draw Sweet Graphics for LED Cubes" Instructable here.
Step 1: The Idea
Did you know the top Technology Instructable of all time is a volumetric display? It's called a 8x8x8 3D LED Cube.
Before discovering this, we'd felt pretty isolated. There were only a handful of hobbyists experimenting with high resolution volumetric display, using techniques like spinning LED arrays and highspeed projector systems. Work on volumetric display was generally considered so expensive and complex it was performed only in places like the MIT Media Lab or secret departments of Samsung.
But then one day we found that there were actually millions of people reading about how to make beautifully simple volumetric displays with arrays of LEDs. While these LED cube displays were of far lower resolution than the displays of our dreams, it didn't matter - they were true volumetric displays. That said, each of these 3D LED cubes would take legendary soldering skills and between 100 - 200 hours to assemble (and an equal amount of time to program) -- but a few people would do it and they would become heroes to millions of others.
Excited to find we weren't alone in the DIY volumetric display seas, we tried to buy an LED cube kit. Only it turned out there weren't any.
And that was the idea. We would make the first color 3D LED cube kit with easy to share volumetric programs that could run on any cube in the world. And we'd make sure they could be assembled by anyone in under an hour. To pull this off, we'd need to make some hardware improvements and create a way for cubes to connect to each other over WiFi, so we got to work.
The L3D Cube was born.
Step 2: Making Prototype 1 & 2
The first discussion about this cube in our small engineering team (Alex, Samtim, Angus, Alvin and me) started the second week of July 2014. The following is the first email I could track down, where we started planning for the first prototype:
July 11, 2014: First email within our small lab about what would become the L3D Cube
"Alex and I have chatted about this topic a good amount in the past couple days, and seeing this instructable and the Youtube videos of people's cubes (with millions of views) has me convinced that the LED cube is the Cupcake of volumetric display. There is huge demand, but no one supplying good kits or assembled 8x8x8 cubes with a good centralized collection of sketches to run/improve. It's all pretty scattered.
So we're going to do a little experiment. Samtim and I just chatted and on Monday he's going to make a simple PCB layout with cut-out zones for an 8x8x8 RGB cube using the WS2812 LEDs (these are the LEDs that can be put into a string up to 1000 LEDs long and each be controlled with a single wire). In the lab we'll just need to design a dead-simple acrylic frame (roughly a 10cm cube) to hold the 8 PCBs. Then we're going to make a few volumetric applications and put some cubes up for sale on our site within the next few weeks. There are loads of improvements we can make to this simple cube (Samtim suggested some FPC designs and Alex mentioned ITO coated glass) but I think we can start real simple.
Alex, any other things for us to keep in mind for this? I assume this is just controlled by an arduino, just like the array Matt assembled."
The two key improvements over previous cubes we had in mind were 1. using addressable WS2812b RGB LEDs (also known as Neopixels) to simplify the assembly process and 2. integrating WiFi connectivity by default, for sharing and more complex streaming applications. These weren't terribly complex ideas, but if we were first to pull them all together perhaps we could use that momentum and community to fuel our more expansive vision of a personal volumetric display future. So we started to move at hyperspeed and had a fully-functional prototype one week later. Until we melted it.
July 19, 2014: Prototype 1 was completed
It didn't end up being made of panels, as noted in the email, but instead used 64 thin PCBs of Samtim's design, each hand-soldermasked and baked in a little oven. These LED sticks plugged into a simple PCB baseboard and were controlled by a Spark Core - essentially an arduino with WiFi that some of our friends had designed.
Our failure rate on these homebrew surface-mounted LEDs was high - around 5% - but the system worked. The first program to run on this cube, which we called "LyteLab" at the time, was Fireworks, written one morning by Alex on the train ride from the island he was living on in Hong Kong.
This Prototype ended up melting when we hit it with too much current.
July 31, 2014: Prototype 2 completed
I took this second unit to San Francisco to show around at sushi bars and crashed parties. People seemed to love the cube, but would they buy it? A sales experiment seemed in order.
Step 3: Experimental Selling on Amazon
So we posted the cube for sale on Amazon right before I hopped on a plane. We'd never done anything like this before and didn't know what to expect. I thought there was a less than 10% chance someone would randomly buy the cube on Amazon. To make sure this was a real experiment in the wild, we didn't tell any friends and made no twitter or facebook posts.
By the time I got off the plane, someone on Amazon had bought our only functioning cube prototype. An hour later, we sold another unit. Since there was only one cube in existence, this posed a problem. I rushed to pull down the Amazon inventory count to zero. Angus & Samtim back in the lab then rushed to make a second cube by hand before the ship deadline imposed by Amazon a few days later.
The cubes were and still are designed to be assembled without any soldering in under 30 minutes. Only a screwdriver is needed. This and the full color depth were the most obvious differentiators from the traditional cubes that took months to assemble, and it looked like at least a couple folks picked up on that. We hoped the sharing of 3D cube apps would become more important as more people got cubes, but at the time of these first sales, that system of sharing apps didn't exist.
In these first two cubes, we had really limited quality control procedures. Basically we assembled the cubes, and if they got through a basic torture test program of flashing the LEDs at full intensity for a couple hours, we put them in a box. However, that testing procedure didn't account for the bell curve nature of component failure. And yes indeed, the first customer received a cube with two LEDs that blew, almost certainly because we fried the LEDs in our homebrew soldering operation.
But all was not lost - because the LED reeds are all identical, if an LED on one reed happens to blow out, a new LED reed can be swapped in, in a few minutes. So we sent him a replacement LED reed, he swapped it out, and all was good in the world.
Step 4: Prototype Revisions 3 - 6 & Maker Faires
Then we started to make more cubes, pushing through a new revision every week or two. And the cube got a name - Alex texted me a drawing of an LED wearing a sombrero with the name "L3D Cube" scrawled underneath. For some reason, that stuck. The following are some notes about each rev, as we crawled from the swamps of one-off prototype onto the land of a real scalable product.
Rev 0.3: August 27, 2014, 2 units
- spark core in the front does not permit plugging in a microUSB cable - change location to the back of the unit
- the acrylic guide on the baseboard did not work to prevent accidentally incorrect plugging of LED sticks
- logo and annotations laser etched
Rev 0.4: September 3, 2014, 6 units
- complete BOM in 5, 100, and 1K pieces, coming in around US$120 - US$130 depending on quantity
- move Spark Core to back, with microUSB connector pointing towards back of unit.
- the Spark Core microUSB will also be our primary power jack to avoid confusing jack duplication - eliminate barrel connector
- add slots into baseboard aligned with PCB sticks, to prevent incorrect orientation of the LED sticks - Angus to design layout and send to Samtim
- add a large rectangular hole for the optional arduino connections & a hole for the Offline-Online slider switch
- can be run off of computer USB for low power apps or can plug into a USB adapter 2A for higher power applications.
- will ship with a USB-A to microUSB cable and a 2A power adapter, US plug style
- includes auto-brightness scaling code and small changes to the baseboard, including bus monitor circuits and two 220uF capacitors per Alex's email
- include microphone and 3-axis accelerometer
- current limited to 1.8A
- logo should be on the front
- include 3mm holes near the top of the case to hang the case (like in clubs/rooms/etc)
- slide switch or connect to cloud button, confirmed. Slide between "Offline" and "Online" modes.
Rev 0.5: September 26, 11 units
- implemented hidden pushbutton on underside of baseboard as an easy wifi reset alternative to the Spark multi-button push reset
- changed the corner pieces and side acrylic to slide and lock, to allow for the box walls being constructed and held together as a first step
- using bulk order Spark v1 cores, US$20 per unit
Rev 0.6: October 8, 20 units
- corrected spacing of sticks on baseboard to 20mm, front-center to front-center for a perfectly cubic LED array;
- attempted altering design of panels and baseboard to allow for even stacking of cubes without large cube-to-cube gaps, abandoned temporarily
- capacitors and big resistors located on the underside of the base PCB, to make the topside view cleaner
- screws must be ferromagnetic to make assembly easier
- implemented hidden pushbutton on underside of baseboard as an easy wifi reset alternative to the Spark multi-button push reset
- hang holes added discretely to back of cube
- corrected logo on PCB, correct direction
- label Rev and date on the PCB Add website to back of acrylic, centered Logo on the acrylic should also say "created by Looking Glass" in small font
- adjust acrylic strip cut profile slightly to prevent fracture (happened once to Alex on the Rev0.5 cut profile)
- ensure all threaded corner pieces are properly threaded to prevent stripping
- think about serializing the prototypes
Alex showed the cubes at three Maker Faires within three weeks and I took our few samples to hackerspaces, fablabs, and fabcafes in Tokyo, Hong Kong, and Taipei. And bars - we took the cubes to a lot of bars.
We were selling experimental cubes as fast as we could make them, so it was time to automate part of the assembly process. That's what's shown in the video above - we found a small factory willing to pick-and-place the WS2812b LEDs onto our somewhat jello-like PCB sticks. Our error rate on LEDs fell to zero errors in post-bake testing.
Step 5: CubeTube: Volumetric App Sharing + Music
After the first few revisions of prototypes, we decided to build in some sensing capability into each cube, including a microphone for self-contained music-reactive applications. We also started to connect cubes, via a streaming computer, with other sensors and inputs - the Kinect, Structure, LEAP motion, Makey Makey.
Inspired by the Makerbot Thingiverse created by Zach Smith, we started to conceive of a similar site that would let volumetric apps be shared between cubes with the click of a button, wirelessly downloaded or streamed to one's cube. We basically wanted to ensure cubes owned by non-coders would continue to improve and become more interesting over time, while giving coders a platform on which to freely share their work.
As of this morning, we're calling this sharing gallery CubeTube, where a single button click loads new code onto cubes on a given WiFi network.
For anyone interested in creating on these cubes, Alex gives a detailed Instructable here
Step 6: Kickstarter Dreams
Since we sold the first experimental cube on Amazon, we'd been working towards launching the L3D Cube officially on Kickstarter before the holiday season hit in the US. We wanted to not only get people their first volumetric display - we also wanted to build a platform for people to share volumetric apps between cubes with the click of a button. For this, we'd need a critical mass of at least a few hundred volumetric nerds like us, and Kickstarter is how we decided to get there.
For the video, we wrote a script and shot everything over the course of two days. Alex and I have edited our own videos before, but because we're amateurs, getting from raw footage to a 2-minute video is probably 100+ hours of work, and lots of redone shots. We didn't have that kind of time for a pre-holiday launch. So, we hired some friends who make excellent music and interactive art videos in NYC, which of course was the best decision ever.
We also decided to make a huge 16x16x16 cube (in the video above).
We've seeded the campaign with some new social media tools, like Thunderclap and writing to a few press outlets, such as Wired and Gizmodo. But nothing that unusual - we believe if we reach critical mass in the first couple days, it will carry us through the holidays.
For the reward tiers, we chose pricing that is 2.1x - 2.9x the BOM.
And that brings us to the present, as of 5:17am EST Monday morning, November 24, 2014. I'll be updating this living Instructable as our L3D Cube Kickstarter campaign progresses.
Clicking the launch button now.
Step 7: The Kickstarter, Day 1 Metrics
(note: I'm updating this as the Kickstarter progresses. I'll prepare a clean synopsis and 'what we learned' section after all is complete, but this is a record of how the Kickstarter project progressed - I'm hoping this is helpful to other folks thinking about running Kickstarter campaigns on their own projects)
We launched on a Monday morning EST, a few days before people will start to leave for Thanksgiving in the US. This wasn't based on any scraped historical data - it was just a guess at good timing.
At 6:31am, we got our first backer. It was my brother.
At 6:52am, we got our second backer. It was Alex's Dad.
Then the first non-family backers started to stream in.
We hit about 60% of our publicly started goal in our first day. Here are some things we are learning about the first day.
- Thunderclap. This is a "crowd speaking" service that basically sent out a massive coordinated set of posts on friend's Facebook pages and Twitter feeds at the same time, 10am on our first day. (about 4 hours after launch). After a little analysis, I'm of the opinion that we won't use this again - it's just not that effective.
- We had a domain we setup to track converts from the Thunderclap, more or less - kickstarter.l3dcube.com - and it looks like including direct Kickstarter backers, just 7 backers, or a little over 5% of our first day pledges, came from the Thunderclap. Analytics for the first day or so up above.
- Press. We got none the first day. Our first press article just came through, around the 30 hour mark.
- Friends and family probably kicked in 15%-20% or so of first day pledges.
- We didn't have any advertising (e.g., Google Adwords, etc)
Based on this, I surmise that Facebook & Twitter doesn't drive nearly as many pledges as Kickstarter Creators might expect.