This past weekend had been a physically draining weekend. Mike Ziomkowski and I spent just about every daylight hour reorganizing the OpenBeam storage lockers to make room for the Kossel Pro extrusions that had been sitting in my parent's garage since last November. (Yikes, had it really been that long?). We were joined by my wife Rachel on Sunday as we inventoried every last plastic part and OpenBeam kit into the storage locker, in order to compile build readiness data for the upcoming fulfillment activities.
Every part ever used on every OpenBeam and Z-Designs kit that's sold, or currently in development is assigned a part number. (Due to the tight integration of our parts in the printer, and the long friendship between the two of us, Mike and I have decided to share joint administration of our part numbering system and our document control system.) We are using Arena PLM, a cloud based Product Life cycle Management software, to handle our document control and engineering change order management system. Exactly how we use Arena, and how we use it to facilitate an engineering transfer of our manufacturing to an off-shore contract manufacturer will be the subject blog posts later, but for now, let's just say we're getting a bit more organized.
For the kid who in college ended up on the front page of the school news paper for having a uniquely messy and cluttered dorm room (My old college roommate reads this blog - Hi Louis! :) Remember when we hung stuff from the ceiling because we ran out of storage space? :-D ), let's just say it had been a long personal journey of growth to get to where we are.
This is what a Kossel kit looks like inside our PLM system. Shown here is the configuration with the heated build platform, which is by far the most popular model. Every ZT-KIT part number represents a kit that we sell on Amazon.com and every ZT-ASY number represents a combination of parts that have had work done to them (such as kitting and sealing them into a plastic bag). -KITs, are for sale, -ASYs are not.
As you can see, the Kossel Pro is essentially a kit of kits. We kept things this way to make the printer as modular as possible. In fact, we will no longer sell printer kits directly after pre-orders closes - After the preorders close, we will be partnering up with Solarbotics Ltd, in Calgary, Alberta and MatterHackers Inc in Orange County, California, for selling full printer kits, and we will only be selling the sub-assemblies on Amazon. We are doing this because as engineers with day jobs, neither Mike nor I can offer the support that this truly deserves, and we'd trust these two companies to provide great support to our users.
When you drill deeper down into the Bill of Materials, each kit is made up of a collection of -ASY sub assemblies. We have started kitting the sub assemblies, as mentioned two weeks ago, and included in this time is training for our kitter on some basic soldering skills as well as wire crimping skills. We have made fairly good progress, but there's still a bit to go.
Here's a link to the Google Spreadsheet listing the status of the kitting in details. At the current rate, we are hopeful that we'd be able to complete First Article Inspection on all -ASYs by this Sunday. After completing all the FAI on the -ASY numbers, we'll then have to kit them up into the final kit (which is a much quicker process, typically we're just building a ULine mailer box, putting the parts into it with some Geami wrap, and then sealing it off with a print-on-demand label.)
Once the FAIs are approved, we expect our throughput for kitting to be about 20-40 kits per week. We'll do batch shipments of the kits to our backers and to our preorder folks. Early bird backers will be the first to receive their kits, followed by Kickstarter folks, followed by pre-order folks, in that order.
Now, onto the engineering update:
Our injection molder delivered early! More importantly, the fixes that we have done to the mold completely fixed the issues at hand. Here is a change log of the parts.
As molded originally, the Auto Probe had two issues: First of all, the standoff that's acting as the L-arm of the original hex key occasionally have trouble hitting and actuating the pin plunger on the switch. This was due to the round shape of the stand off and the gap that the pin falls into being a little bit too wide.
We've also significantly increased the down force on the spring deploying the probe. This, combined with tightening the tolerance on the slip fit between the pin and the pin body, was done to improve the centering accuracy of the probe. (The Reprap allen key arm was notorious for wandering all over the place, when the probe gets deployed.) This unfortunately meant that the spring force required to lock the probe into the up position also got correspondingly larger, and the probe would no longer reliably retract with the original spring that we've specified.
OpenBeam's printer's selling point is the auto levelling probe, continuing on Johann's tradition of making Delta printers easy to use. Obviously, having an unreliable probe wasn't acceptable.
So, this is the change we made to the plastics:
We significantly increased the size of the spring cavity for the return torsion spring, to allow for a MUCH stiffer spring to be installed. Now when the probe is retracted, there is a nice solid *CLICK* as the probe locks into position.
We also added guide ribs to center the stand off when the probe deploys. Now the switch is actuated 100% of the time.
Here's a picture of the actual assembled probe, in deployed and retracted positions:
The other part that we made significant improvement to is the hot end clamp. We weren't happy with how loose the fit was to the M5 nut and the washer. So, we added crush ribs to the inside of the hexagonal feature:
The features highlighted in blue are called "Crush ribs" in industry parlance. As the name implies, it's a thin piece of plastic (rib) that gets crushed in a press-fit situation. It's job is to collapse and provide just the right amount of an interference fit to make things tight. I had some reservations in using crush ribs with this plastic (remember, all black plastic parts on this printer is 47% glass for reinforcement!) but it worked really well.
As a "Hail Mary", I also increased the width of the channel for the serrated flange lock nut that the Metrix Create Space crew have been using in their Reprap Kossels. Now, the push fit connector is just about bullet proof.
Machining the Vertexes
We've had a bear of a time with the vertexes. I failed the initial run's first article inspection, because some of the holes were off.
At OpenBeam, we've been taking extrusions for granted; we found a really great supplier and haven't looked back since. Unfortunately, our volumes were too small for our overseas vendor to handle, so we had to go with a domestic supplier for the Kossel vertex extrusions. I visited the vendor in Chicago last year in October and gave them the business.
Unfortunately, we didn't specify to the vendor how to cut the vertexes, or how accurate the cuts needed to be. I had seen a carbide tipped chop saw on my tour and assumed that's what they'd cut the extrusions with. Afterall, that's what my overseas vendor does, and they get incredible, breath-taking accuracy with that saw, and it's wicked fast.
We originally had the vendor quote machining the part to completion, but the dollar amount they wanted for the fixture alone would have bought me a small CNC mill. Because the vendor had already quoted turn key machining, we thought that it was understood that the edges of the extrusions are critical. We've been holding the part on its end, referencing the machining operations on a face that we assumed was orthogonal to the extrusion axis. Turns out, our vendor decided to cut the parts on a bandsaw. And not just any band saw - looking at the cutting marks, it's apparent that the band saw had a broken blade and the blade started drifting during the cuts.
So, our machinists devised another work holding method, called an "Expanding Mandrel":
In an expanding mandrel setup, a nut forces the metal open to "expand" and grab the insides of the part. Because the insides of the part is formed by the extrusion process, it tends to be fairly accurate. The two band saw cuts on each end can wander, but the part will still be held true.
Here's a video of the setup in action:
Cycle time is about 8 minutes. So, rounding up a little bit, we can reasonably assume that we can produce about 16 sets of parts per 8 hour work shift. Tom is still dialing in the fixture to get it better, although the test fit I got was fairly good:
Other project administration stuff:
We shot some assembly footage, but I was pretty disappointed at the video quality. As a small business owner, you learn really quick what you are good at and what you need improvement on, and video production is definitely in the latter category. One of the biggest challenge is that we are shooting our video from a hanging camera looking straight down, and it's awfully hard to see what the camera is seeing and what is in focus while I work. (For the technically inclined, I used the table to set the focal plane when in fact I should be setting the focal plane elevated off the table. The depth of field is greater past the focal plane than in front of it).
That's it for this update. Here's a few more random pictures taken during the last week. Enjoy!