For flatness, we inspected a random sampling of extrusions on a granite inspection plate. Granite inspection plates are precision ground blocks of granite. Their job is to set up a reference plane upon which measurements can be taken. This particular plate we used is made by the L.S. Starrett company. They, along with Brown & Sharpe, Mitutoyo, are considered to be some of the finest metrology instrument manufacturers in the world. The plate we have is a Class A inspection grade pink granite table; the table is flat to within 10 microns across its 36" x 48" surface.
To inspect for flatness, we randomly pulled 5 samples of each color out from the different shipping containers. (it is CRUCIAL that the sampling be as random as possible, and not pull the top 5 pieces of box 1. Years ago on my engineering job, our inspection department is known to take short cuts. They inspected a lot of medical battery cases from one box and accepted the shipment. Little did they know the vendor had shut of the injection molding machine to go to lunch halfway through the production run and there was a batch of undersized parts from the machine warming up after lunch. The parts didn't get caught and quarantined and the production line went down as a result.) We place the bars onto the granite table and check for signs of the bar rocking. Then we take feeler gauges to probe the maximum size feeler gauge that can slide between the granite and the aluminum without lifting the bar off the table.
What is simply amazing to me is that 3 of these 5 bars don't exhibit any rocking at all, and I was hard pressed to slide the 0.04mm feeler gauge - the thinnest in my set, by the way - in. On the worst warped piece, I can slide the 0.1mm gauge in. I repeated the test by putting a dial test indicator on the aluminum and pressing down onto the bar to see if I can measure any deflection. On over 50% of the parts, the total deviation is only 0.02mm from when I press down on the extrusion.
Allowable warp on the extrusion is 1.1mm per meter, per Aluminum Extrusion Council's guidelines. These extrusions are, at worst case, an order of magnitude flatter than what the AEC allows. This is a far cry better than the 3-5mm warp that we were seeing from our other extrusion vendor! 3-5mm warp, in a 1 meter length, is very very visible to the naked eye. It certainly makes me question the other supplier's quality control procedures.
Next, we check for accuracy of cut lengths. To measure length, we stood the extrusions up and used a Mitutoyo Digimatic Height Gauge to measure the distance between the granite reference surface and the top of the extrusion. This method is somewhat faster and less prone to parallax errors than measuring with a pair of digital calipers. The cut lengths are most critical on our new Mini Kossel kit, so we pulled 3 sets randomly to give us 36 samples to measure. (This was before we merged the design, at one point I still had 12 pieces of 240mm in my design for the Mini Kossel). I've placed the metrology results in an Google spreadsheet here.
The results are incredible, once again. The calculated standard deviation - a measurement of how tightly clustered the spread is, is a mere 0.03mm. As a comparison, my hair clocked in at 0.08mm. If you were to assume that the distribution of the cut lenghs is Gaussian - and that's a pretty safe assumption to make - 99.7% of all samples fall within a six sigma (6x standard deviation) spread around the mathematical mean. In other words; given the standard deviation of 0.04mm; 99.7% of the measured values will fall within 0.04*3 or 0.12mm of each side of average. To be safe, I can claim +/-0.3mm tolerance on the cuts (which is damn good, by the way, that is approaching machine shop tolerance for machined parts), and in selling 1000 pieces of extrusions, I may get 3 returns for being out of specification on cut length tolerance. That is seriously impressive, and certainly something that I can build a profitable business upon.
I measured the precut kit lengths too, but as you can see from the data, that was a strictly academic exercise. In fact, I stopped half way through and just packed up to go home.
Armed with this data, I am happy to announce the following:
A) Extrusions, in the form of 1 meter long bars, are now on their way to Amazon fulfillment centers, and will be back in stock shortly. We can now guarantee the new extrusions to 0.5mm per meter flatness.
B) We will be re-entering the Kossel extrusion supply business, with the following difference: we are contributing members to the Open Source Hardware community, we offer free shipping on our extrusion kits (free second day shipping for Prime households), our extrusions are competitively priced, available in black anodized, and we guarantee their cut accuracy to +/- 0.3mm.
C) We have removed our biggest road block for building pre-cut kits; dealer backorders are now being processed, and there will be a few kits showing up on our web store. In a month or so, when our next shipment arrives, we will start selling precut lengths a-la-carte with Amazon Prime fulfillment as well.
D) With our supply chain issues resolved, we have signed up with MakerShed; even though I will not have a booth at Maker Faire this year, OpenBeam will be. Look for OpenBeam kits in MakerShed soon!
E) With our confidence in our new supplier, we have already placed a very large order (bigger than ALL the previous OpenBeam orders, combined to the American supplier, including from our original kickstarter campaign). This, along with the competence of our new supplier and the fact that they only need 2 weeks instead of 8 to turn an order around, hopefully resolves our supply chain problem and having to hang an "out of stock" sign for months at a time on our web store.
F) We still have a handful of extrusions from our old vendor; we will be getting rid of these in a fire sale to recoup the money sunk into it. Basically, these are extrusions that may have flatness or cosmetic issues. We don't feel good selling them at full price, so we'll just sell it at a large discount to benefit our local maker community. Details will be announced soon.
We are happy to say that after an incredibly long and dark few months, the light at the end of the tunnel wasn't from an oncoming freight train and we are happy to be on the right path again. With the production issues behind us, our next order that is about to leave for its ocean journey is for a whopping 13km of extrusions. This single order (and not including the 3+km that we just purchased as a "test run") is more than all the extrusion orders to the old vendor combined, including the material for our first kickstarter campaign, and we expect to be placing an order of similar size regularly to ensure that we stay in stock. We are happy to be growing, and we thank you, our loyal customers, for supporting Open Source Hardware.
-=- Terence, Rachel and the furry monster puppy.