Exiting times at the ZT Automations Shop - we are closing in on FCS, or First Customer Ship, on the preassembled Kossel Pro. It had taken over 80+ engineering hours to get to this point, along with multiple fixtures along the way.
The first shipment will be going to a fellow engineer in Florida. We chose her as the FCS candidate for two reasons; Florida is about as far as we can go from Seattle for ground shipping, and will put the maximum amount of transit stress on the machine being shipped. We also hope to be able to get useful feedback on packaging - not that we think that anything will break, given how we prep these things for shipping, but if something does go wrong, we'd want to fix things before sending the rest of the printers out.
Now for the bad news. We've found significant errors in the pre-assembled printers (one of the reason why its taken us this long). So each machine will have to be reworked and inspected before they can go out the door.
I take full responsibility for the rework required; we did not have good enough documentation (and also things were in a state of flux) when the pre-built machines were assigned. That being said, looking at what issues crop up on these pre-built machines have given us a tremendous amount of insight on what needs clarification on the printers (and what needs to be addressed in post-built checkup). For instance, one of the printers had an issue with probe retract after G29. Upon further inspection, it was found that the new rev K-Head heater block had been clocked in such a way that the heater block and heater cartridge hits the glass plate on the probe retract move and causes the motor to skip. (It actually hit with enough force to bend the K-Head's heat break - well, at least the electronics and motion control system is nice and strong).
Here's what we've learned from working on a bunch of Kossel Pros, and all the improvements that we have made to these pre built machines, and how we plan to push these as an update out to machines in the field.
1) As noted in our November update, the manufacturing discrepancy in the injection molded linear rail cartridge appears to be the primary source for auto-levelling issues. We now have a set of laser cut shims and a kit containing these shims. I've ported the Excel spreadsheet that I've been using to calculate the shim placement to an online calculator here. With these shims we have consistently been able to reduce the tilting bed symptoms of the auto levelling routine. If you decide to try shimming, please let us know how it worked out for you.
2) On the recommendation from the team at MatterHackers, we have been gluing the ball joint arms to eliminate another source of potential inaccuracy in the printer's construction. Up until now, we have been performing a 100% inspection on the arms after gluing using the same gluing fixture as a go/no-go gauge; we slip the glued arm back onto the fixture and note the amount of resistance it takes to put the arm back on. Anything that drags on the arm gluing fixture, we reject.
However, we wanted to really see how well we are doing with the gluing operation. So I designed up a new fixture for *really* measuring the center-to-center distance on the ball joint. As the parts have lots of ball bearings and can free rotate, it can be quite difficult to get an accurate reading. Our fixture rides on a linear rail guide and uses a spring to apply the same amount of force on the swing arm being measured every time to ensure accuracy. We then coupled this to a Mitutoyo dial indicator and tapped into the digital readout; now, our 100% inspection also allows us to see just how far we deviate from the 300mm center-to-center distance that's programmed into the firmware.
Furthermore, since we are doing 100% inspection, we can bin the arms based on where they fall on the bell curve. Our arms are now binned in 0.1mm increment; majority of the arms are within +0.05mm to +0.15mm of 300mm. The full set of data can be found here.
One of the things that had really surprised me is just how precise the ball joint linkage have to be in order to build an accurate machine. Getting this fixture in place will give us better insight into our gluing processes, as well as help us understand how inaccuracies in the ball joint arm length translates to performance issues in the real world with the printer.
3) Firmware tuning - every Kossel Pro that we ship will have its firmware tuned; the PID controls tuned to ensure even temperature output, as well as the extruder constant tuned to account for manufacturing variances in the extruder's hobbed pulley. Here is a fixture that we built for extruder calibration:
So far, our default values have been within 2% of actual measured values on the extruder constant; but we're about to learn more as we compile the tuning data on these 10 printers.
4) Other upgrades. As these are the last machines to leave our shop, we've installed every known upgrade to them. We are shipping these machines with our latest tungsten-disulfide coated K-Head parts. Tungsten Disulfide is an aerospace approved process that puts a high temp non stick coating onto metal parts, and it really works! Other changes: We've tossed in a spring loaded lanyard onto the printer to keep the feed tube suspended in place. PTFE tubing have been substituted in on the bowden feed tube. There are complaints about push fit fittings not holding onto PTFE tubing correctly, so we designed a new one from scratch, and solved some of the complaints that we have about material feeding on the SeeMeCNC EZStruder with regards to filament snags. Our bowden tube adapter uses the same, highly reliable push fit fitting that is found on E3D hot ends and have been rock solid in our testing, as well as completely eliminating snags in the EZStruder's push fit adapter.
Our first machine is now entering burn-in testing; as part of the test, the machine prints its own shipping restraints that lock the linear rails from moving during transit. We have assembled the laser cut crate and it works great and should provide the protection to allow the machine to arrive in one piece. We expect to ship it next week and we will update the blog with a shipping schedule for the balance of the machines once we confirmed that the machine had reached its destination.
We've were dealt some very major setbacks in the last 6 months - our dealings with our last contract manufacturer / sourcing agent bordered on outright fraud and blackmail and tied up a huge chunk of cash, as well as countless hours of my time to unjam. I ended up having to fly to Hong Kong for a long weekend to lay the groundwork to resolving the issue once and for all. For now, please know that we have not given up on the project. We've emerged from our crisis as a much stronger company, and we have some exciting R&D projects in the works, as well as long term development plans on the Kossel that we'll share in due time.
-=- Terence T. & Mike Z., ZT Automations