On the home front, we are moving towards Amazon fulfillment of spare parts and individual kits as upgrades for the Reprap community and expect to be online in the next few weeks. (Remember, we went through a similarly painful transition to move OpenBeam onto Amazon and to smooth out supply chain issues. Except, back then it took MONTHS and it was our only source of income). Our current projection for being able to resume Kossel Pro shipment is in April. (The big schedule time bomb called Chinese New Year in the middle of all this does NOT help things).
Kossel Pro Documentation
Like most (stereotypical lazy) engineers, we’ve pushed documentation to last. (There’s 2 reasons for this; one is we really needed to build the printer a few times before coming up with the optimum recommendations for how to build the printer. To do so requires production parts, which weren’t available early on in the project). Unfortunately with the shipment out to our backers and subsequent time spent on support, we have not been able to spend time on documentation.
We are diverting as much of our resources as we can towards documentation, and we are using this time period where our subcontractors are working on kitting and prep work to get as much work on the documentation done. We have now dropped the password protection on KosselPro.com and we will host all documentation there, marking clearly what is still work in progress.
My boss at my day job always told us: “If you really want to learn about your product – get a bunch of dudes to build you a hundred of ‘em. Then line them up in a room, and bang on it”.
Obviously, both with my day job (designing and building scientific instruments, not exactly high volume manufacturing) and with the Kossel Pro, we really can’t afford to do that. Our December shipment of 100 printers was the largest Kossel Pro deployment into the wild, and we’ve since learned a few things about the printer. As much as we try to test for and mitigate real life scenarios, these tests are still devised and conducted by engineers who had spent the last two years obsessing over the project. The real world, as we found out, is highly unpredictable. We’ve been doing preliminary analysis and debugging on our machines in the field and these are some of the common issues we’d like to address:
- Heater cartridge inconsistency and tuning
- Control board failure due to cleaning of hot ends while energized
- Probe deployment issues
- USB port strain relief
- Auto Levelling accuracy
Heater cartridge inconsistency
We purchased the heater cartridges from another manufacturer as “surplus stock” – the supplier had screwed the pooch and shipped them 24V heater cartridges and they use a 12V drive system. Our specifications calls for a 24V, 40W heater, and working backwards, we needed to deliver approximately 1.667A at 24V (24V x 1.667A = 40W). To draw 1.667A at 24V, the resistance of the heater cartridge needs to be approximately 14 ohms (Ohm’s law of V=IR; V=24V, I = 1.667, solving for R gives us about 14 ohms).
However, we are seeing, especially in failed units, as low of a resistance at 8 ohms. At 8 ohms, the heater circuit will dump a whopping 3A of current into the hot end, for a power dissipation of 72W - nearly double the design limit.
Unfortunately, when we did the firmware tuning, we did not think that there was such a wide range of heater core resistance, and even knowing this problem replicating it is still difficult. We are going to build a fixture to do 100% inspection on our heater cartridges’ resistance and we are going to have to start batching heater cartridges.
We are in the process of talking to Chinese heater cartridges manufacturers for a custom crimped and assembled heater cartridge assembly for the Kossel (we crimped the current ones ourselves). Some of the things we will be controlling are the tolerances on the outer diameter (they were ALL over the place with our current batch of heater cartridges – we did 100% test fits with our J-Heads before shipping, and on a few of the units, we had to ream out the J-Heads prior to shipping). We’ll also be specifying a UL compliant, but thinner insulation on the heater cartridge wires. This will prevent the breaking of the crimped connector issue that we’ve seen with some of the heater cartridges, as we’d be able to properly crimp the connection wires (which involves one set of crimps compressing around the insulation for proper strain relief.
Control board failure due to cleaning of hot end while energized
We've had a few failures of the Brainwave Pro boards in the field. Most of these boards failed the following way:
- Print had peeled off the bed, or some sort of catastrophic failure on a print resulted in PLA smeared all over the board.
- User heats the hot end up and while the board is energized, attempted cleaning with a metal wire brush.
- Board dies.
In one case, the board’s failure also caused the MOSFET controlling the heater cartridge to fail short. This is an especially dangerous situation as it results in the hot end overheating and will destroy the hot end. It is also a HUGE fire risk.
We believe that cleaning the hot end with a metal brush caused a shorting of the 24V power for the heater cartridge to the thermistor contacts. We followed the standard Reprap electronics design and this line is not protected on the Brainwave Pro (or any other Reprap electronics out there, for that matter). However, most printer builders run at 12V. Applying 12V to a microprocessor might make it unhappy, but it appears to be a fairly survivable scenario for the microprocessor. Exposure to 24Vs though seems to be universally fatal to the microprocessor.
We are still investigating and we will be putting additional protection onto the next run of the Brainwave Pros (in the form of clamping zener diodes to protect the microprocessor lines). In the meantime, the recommended procedure for cleaning out a clogged end effector is as follows:
- Invert the printer, if necessary for ease of access to the hot end. To be absolutely sure, you can unclip the glass and leave the HBP in place. Nothing should fall.
- Bring the hot end up to 180 deg C
- Using a NYLON (ie, NON conductive) brush, carefully brush away stuck plastic, staying clear of the thermistor wires!
If there's any doubt / questions, just imagine you are cleaning the insides of your computer. You wouldn't go at it while the computer is energized with a metal brush, right?
We consider the permanent nature of the thermistor wires a feature, not a bug. We want the thermistor to be impossible to dislodge from the hot end – as mentioned before, disconnecting the thermistor from the hot end during operation is a fatal error that results in equipment destruction and a huge risk of fire. The unfortunate part about this arrangement is that if the thermistor wires do break the tip will have to be replaced. It comes back down to not having the space to do a removable / servicable solution on the J-Head’s heater block. Our next hot end (whose R&D is being accelerated, to deal with some of the issues that had come up) will feature user servicable temperature sensors.
Probe Deployment Issues:
There are a few issues and comments about the probe’s design and its deployment:
- The probe standoff loosens, and eventually flexes enough that it doesn’t push down on the switch.
- Concerns about the general slop and positional repeatability on the probe when deployed.
- Related to 1 – the probe impacts the plastic housing (and switch) too hard, drastically shortening the life of the switch.
Here’s what we are doing to improve things:
- We are revising the assembly instructions on the probe to call out the addition of Loctite onto the threads of the probe to stop it from loosening. We are going to try to source the M2 FSC with Loctite preapplied for the next manufacturing run
- We are going to be talking to our vendor about changing the probe arm from round to a square profile. This should resolve a lot of the positional repeatability issues on the probe.
- Unfortunately, this is not something that we can address. The spring force is required to overcome the 75 gram force actuation limit on the switch (the lowest we could spec, by the way). The nature of hooks law and how compression springs work results in the slam.
For those who are really concerned, the only work around we can offer is that you can deploy the probe by hand. The switch in question is freely available from Mouser for less than $2.00, and the next generation auto levelling system does away with the auto probe all together. In the grand scheme of things, this is way too far down the priority list for us to address.
USB strain relief
When we designed the Kossel Pro, we chose a full size USB B port for robustness. During my time working with the original Brainwave, I’ve sheared my share of MicroUSB ports off that board.
However, what we didn’t factor in is the weight of the USB connector (on the cable end). We are learning that a vertical USB port really requires additional plastic around the connector to provide strain relief. One of the warning signs is that we see the sheet metal peeling from the shield of the connector. This forms a sharp barb that catches the edge of the USB B connector, making it impossible to remove the cable (this happened on one of our engineering test printers as well as one of our user’s printers).
We've also found, on the engineering test printer, that due to the cable not being able to be removed, the additional stress on the board will cause intermittent connection on the USB port, to the point where the board is no longer usable.
To fix this, and to head off a major recall, we are planning, in the next week or so, release a 3D printable part that can be added to frame to provide additional strain relief to the connector. We are certainly happy to catch this train wreck before more issues appear.
Auto Probe Accuracy
The G29 Auto Levelling routine that the Kossel Pro uses is directly from the Reprap Marlin firmware. It was written and improved upon by Johann Rocholl, and it’s used by delta printer builders all over the world.
We now have reasons to believe that G29 doesn’t actually work very well. One of the biggest issues that’s been pointed out, by the software team at MatterHackers, is that the least-square-sum method that is used to compute the bed tilt really requires double floating point precision to work, and Arduino can only handle single floating point precision. As a result, a lot of the precision is discarded and what we are seeing is actually rounding error.
It should be pointed out that this is NOT an OpenBeam Kossel Pro issue, this is a general delta printer issue. Dudes a lot smarter than me, such as the software team at MatterHackers, are looking into different solutions and options. Ultimately, the changes will be committed back into Marlin, and everyone in the delta printer community will benefit.
We would like to thank all our backers and preorder folks for their patience – this project wouldn’t have been financially possible without our backers – and frankly, also without the support of all the people that have been buying and using OpenBeam for the last two years. (OpenBeam essentially bankrolled a lot of the development of the Kossel Pro, to bring the project across the finish line). We are aware of the shortcoming on the documentation and we are fixing them. We are aware of some of the minor design tweaks required to polish the printer and we are fixing it for the upcoming production run (and making the parts available to our existing backers – we’ll figure something out).
At the just concluded 3D Printer World Expo in Burbank, more than one person had come up to us and complimented us on the engineering on the Kossel Pro – that it is one of the more polished open source hardware printers out there, and that the craftsmanship put into it by all our vendors – our machinist, our injection molder, our OpenBeam extrusion supplier, is very apparent. We are certainly proud of the engineering that we have done, but over the course of the project, we’ve also identified shortcomings with the Reprap project, and we look forward to fixing and addressing some of these issues we have identified. We look forward to continuing to carry the torch and innovate, and designing and building better tools to enable makers out there.
Thank you for your support of Open Source Hardware,
-=- Terence and Mike