The OpenBeam Kossel Family - Part ii - Kossel Pro and Mini Kossel Pro

In this blog entry, we will talk about the Kossel Pro and the Mini Kossel Pro, and the high level design decisions that drove it.

Overall Design Goals:

When I started the project with Johann many moons ago, I understood the magnitude of the task ahead:  To commercialize and make it easy for people to build a revolutionary printer.  Deltabots represents a big deviation from the Reprap world, and to gain acceptance, we'd have to do a good job to make it easy for people to build the machine.

After some brainstorming, these were the additional design goals for the Kossel Pro kit:

  1. We wanted a machine that can be built by kids and students, and by people with relatively little technical training.  We looked at the typical reprap printer build processes and identified a few major technical hurdles:
    1. Wiring.  Wiring is tedious.  There are multiple ways to wire something wrong, only one way to wire it correctly.  And often, the incorrect way can result in magic smoke being let out.  If crimping is involved, the majority of people have trouble crimping good crimps onto cables.  Bad crimps, by the way, are one of the worst problems to troubleshoot.  The failure mode is often intermittent and, in the case of motors without properly hardened electronics, destructive to the control electronics when the wiring fails.  The crimpers that I use are official Molex crimpers.  They only crimp CGRID III and SL connectors, cost about $360 and only crimp 22-24 and 34-36AWG.  There's another matching $360 crimper that crimps 26-28 and 30-32AWG wires.  Even for an engineering geek like me, that is a painful purchase.  
    2. Soldering.  Although soldering classes are par for course for MakerFaires, we feel that soldering also adds a technical challenge to building the machine.  It is also a very irreversible process:  Desoldering something is often much, much harder than the original soldering in the first place.
    3. Firmware.  Editing a config file, downloading libraries (often from multiple github respositories) and compiling the code to flash onto the controller board is actually quite a challenging exercise. 
  2. We wanted the machine to be easily serviceable in the long run.  This means components that are subject to wear and tear should be easily removable for servicing.

This is how we went about solving the engineering challenges:

  1. We provide all our kits with fully crimped cable harnesses.  All connectors used on the Kossel Pro are locking, polarity keyed connectors.  In addition, the Brainwave Pro control board features robust protection – there is reverse polarity protection on the board, and the motor drivers are hardened against back EMF.  Although not recommended, you can disconnect the motor while it is running at full current and not blow the driver chips – and it’d be difficult to accidentally disconnect a motor anyway, given the locking connectors on the control board.  For quality control, we specify that the crimps meet manufacturer’s pull test specifications, and we control the insulation used on all our wires (UL1007).  This ensures that there is no funny business with our cable vendors skimping on parts.
  2. Our production printer will also ship with all the connectors populated, and the end effector board assembled.  This eliminates the soldering requirement and gives us a mistake-proof path to wire the printer.  With all connectors polarity keyed, it would be impossible to wire a printer backwards in a way that causes harm, provided that all the connectors and cables are built correctly.
  3. We will supply the printer board with firmware preloaded, for those buying a turn key kit. We can do this because the auto-calibration routines are now robust enough to compensate for individual machine tolerances, and the wiring / polarity keying of the cable harnesses ensures that every printer, every motor and every circuit will always be wired the same and thus should behave the same way when assembled.
  4. The End Effector is a self-contained unit housing the hot end, auto-levelling probe and cooling fans.  It can be disconnected from the machine by undoing 6 screws, a wiring harness and a push fit connector.  Small enough to fit inside a coffee mug, it is relatively easy to cross ship to a customer should the printer fails and needs to be repaired quickly.

No compromises on hardware specifications:

Early on in the design process, we solved the issue of retraction with the long bowden drive, and were able to move the cold end extruder and the heavy motor off of the end effector platform.  Most 3D Printer designers would then take this as a liberty to skimp on the ball joints.

Not us.  We knew that although the bowden drive works, much work is being done on novel filaments such as NinjaFlex, TPEs and other flexible filaments.  We know that we are setting the bar for a reference design for delta printers, and there are folks out there who would be looking at our platform to use for pick and place machines, sorting machines, frosting extrusion machines, etc.  We also knew that if we designed the ball linkage system to competently handle a heavy extruder motor on the end effector, the thing will run forever swinging just a bowden drive end effector.  So we set off to build the best ball joints we can.

There are other manufacturers injection molding the universal ball joints for the ball linkage arms.  We are the only ones that are putting ball bearings in each of the moving parts in the ball joint (48 total!).  It is more expensive to build it this way, but by doing so, we aren't just reducing the friction in each joint; we are increasing the load carrying capability of the joint and increasing the mean time between failure (MTBF) of the joint.  And, being mindful of the experimenters out there, we kept the ball joints very customizable.  We include the parts to build the gluing fixture to build our ball joint linkages with each ball joint kit, and with a length of OpenBeam, you can build ball joint linkages to any custom lengths.  


Continuing around to the top of the printer, our belt tensioning idlers are fully integrated pulley blocks with a matching timing belt pulley inside, running on dual flanged ball bearings.  By creating an individual pulley block that moves for belt tensioning, we can allow the top of the printer frame to stay level while adjusting the individual drive tower's belt tension.  Over the lifetime of the printer, we can expect quite a bit of up and down motion on each of the drive towers.  We are proud to use EMS Grivory, a 47% fiberglass reinforced nylon, for all the injection molded parts, and our pulley rests on a pair of flanged ball bearings for low friction operation.  

Compare this to the X-axis idler on a Makerbot Replicator - a simple steel shaft, rolling on an injection molded ABS piece (the same material as Legos and your computer keyboard case).  I'll give you one guess which will fail first.

Makerbot X-Axis tensioner.  Single press fit steel rod, riding on an ABS holder without a bearing surface.  Great design - if you are in the business of selling replacement parts or trying to upsell users into upgrading.

The Kossel Pro End Effector

The end effector represents the bulk of our engineering effort (and the primary reason why the project is late).  Leading up to the Kossel Pro's design, we ran various delta printers (and a couple of Cartesians in the form of Mendels) and documented the failure modes that we've been seeing.   In no particular order, these are some of the considerations that have plagued our various prototypes (including our cartesian robots):

  1. Cables and connectors wear, come loose, fatigue, etc, due to the high temperature of the hot end and the constant motion of the end effector.

  2. The bowden cable setup is extremely sensitive to thermal jamming while printing PLA.  PLA expands with heat; in an ideal hot end, the plastic should go from solid state to molten state in as short of a melt zone as possible.  If the hot end starts overheating, the melt zone elongates and the pressure required to drive the filament through the hotend increases.  This becomes problematic when the bowden feed tube is long; additional friction will cause a feeding malfunction.

  3. For thin walled part, the plastic should hit the part and then immediately drop below the glass transition temperature of the plastic.  Otherwise the part will sag and deform.  Achieving uniform cooling - from all sides of the nozzle, is a challenge.

  4. The mass of the end effector shall be kept as light as possible, precluding the use of a direct drive extruder system.

  5. The auto-levelling probe shall be made robust and repeatable.  We've also had a few accidental deployments where during a long print, the probe accidentally deployed and destroyed the print.

What we did different with the Kossel Pro's end effector:

We paid a lot of attention to proper strain relief and connector crimping.But even with the wires pre-crimped, constant motion will still put strain on the cable harness.  So, we designed a custom cut PCB to go on the bottom of the end effector assembly for the heater cartridge and the thermistor to connect to.  The latter connects via screw terminal, since thermistor wires are made from a notoriously difficult to solder to material and are too thin for crimps to hold properly without a very expensive crimp tool.  We will provide the thermistor pre-potted into the hot end heater block, and with ring terminals pre-crimped and tested for our turn key kit builders. 

For cooling, the hot end is sucked up into the end effector assembly (see below).  This means that we would need very low profile screw terminal blocks for the end effector PCB.  They don't exist; so we designed our own.  The internal duct work also captures nuts that mates to the PCB and the wires are pinched down by a screw head and washer onto a ENIG (immersion gold plated) contact patch.  

Because the end effector board is bolted to the assembly, there is no flexing of the cables relative to the end effector.  Additionally, there are mount points for zip ties to attach the incoming cable bundle to the end effector as well to relief strain on the connector as the end effector moves around.  And since we have a circuit board underneath the end effector assembly, we decided to bling it out a little bit and added an LED ring light to the bottom as well.  It can get awfully hard to see what the printer is doing, with a big print head hovering over the print.

2, 3)  As mentioned above, bowden fed hot ends are very sensitive to thermal jamming when printing with PLA.  To compound the problem, there are often a lot of confusion about the role of the fan mounted on a hot end.  Most operators recommend turning the fan off for the first few layers, for better adhesion.  The problem occurs if the printer is not moving fast enough, or if it sits idle with hot PLA in the barrel.  

To mitigate this, we've found sub-miniature, 20mm diameter fans made by Sunon.  These are standard DigiKey parts, available around the world.  The 20mm size allows us to place the fans where it matters, instead of having to muck with fancy duct work.  There are three fans.  One fan is wired to the Fan-Vcc - selectable on the new Brainwave II, while the other two are under PWM control by the microprocessor.  The  fan that is constant on blows into a cooling channel that directs airflow directly across the vents of the J-Head to prevent thermal jamming.  Using a FLIR infra-red camera, we're seeing that the cooling is indeed VERY efficient, even for a 20mm fan.

The other two fan blows into a plenum that shapes the airflow into a ring and blows it down and out around the extrusion head.  Because these fans are on PWM control the microprocessor can throttle back if necessary to improve adhesion.  We will be borrowing the FLIR camera to keep fine-tuning our printer profiles.

4)  Miniaturization was a key concern.  Even though our end effector packs quite a punch, we only grew the radius by 10mm compared to Johann's original minimalistic design, and the plastic parts clock in at around 20 *grams*.  We even spec'ed in to smaller end stop switches (although the end stop holders are designed to use both)

5)  Finally, we've added a safety detent to the auto levelling probe to enable the probe to be locked from deployment.  For long prints, this offers an additional peace of mind.

Systems Integration:

Since we decided on custom electronics, we took a very close look at the hot end / end effector subassembly, and decided to integrate all the required wiring into one single wire harness.  On the Kossel Pro, a single wiring harness carries all the power and control signals for the heater, the hot end thermistor, the constantly on fan (for cooling the thermal transition) and the microprocessor controlled fans.  We left an extra pair of signal lines for the auto-levelling probe to use, but we also broke out a second copy of the auto-levelling probe connector port for future expansion (such as the use of FSR sensors for bed levelling).  

Because of this high level of systems integration, the Kossel is a very field serviceable printer.  The complete hot end and and end effector assembly can be swapped by undoing 6 screws, 1 wiring bundle and 1 push fit connector.  Best of all, it is built around open source hardware - in this case, the venerable J-Head hot end.  This makes the Kossel extremely easy to service in the field - worse case comes, we can easily cross-ship a complete subassembly and have the end user install the new end effector and back to printing in no time.

No detail too small:

In Johann's original Mini Kossel Reprap design, the vertexes features a notch in which  one can drop a nut into and slide onto the end of the extrusion, allowing the printer to be easily modified and accessories to be easily bolted on after assembly.  Naturally, forming this notch in an extrusion would be somewhat problematic; the best we can do is to flip the vertex on its end and perform expensive secondary and tertiary operation machining.  

So we did the next best thing we can:  we extruded the cutout for the nut into the extrusion, and we injection mold these little end caps to stick into the OpenBeam hole in the end of the extrusion.  This allows us to preserve the geometry for the nut clearance cutout.  

Top Vertex extrusion, OpenBeam (in clear anodize for contrast) and injection molded end cap

We've also paid attention to vibration dampening and added OpenBeam feet to the bottom of the triangle.  This, along with the supplied 3M Bumpon adhesive feet, serves to isolate vibrations from the machine's space frame from conducting onto the desk that the printer is sitting on.  

Another OpenBeam innovation is the use of laser cut nut spacers.  We include a sheet of laser cut nut spacers in each of our kit.  Design credit on this really should go to Matthew Wilson, who pioneered this on his BrainScan test fixture on the original Brainwave board.  This eliminates the blind assembly process by constraining the nut in the extrusion grove in a fix position.  It costs a little bit more to do this, but the results and the ease of assembly for the user makes it all worth while.  I would love to see other manufacturers follow our lead in this aspect - just remember, you saw it here first. :-)

Original design credit: Matthew Wilson


Finally, we are a big fan of well labelled wiring jobs, so we are providing a pre-printed sheet of wire identification labels.  

The Best Printer we can build, and one that we use ourselves.

I (Terence) am a tool snob.  I spent hours and hours browsing for a multi-tool for my Dad's birthday before settling on an American made Leatherman Charge TTi with titanium handles.  Years ago on a trip to Japan I bought a folded steel sashimi knife, forged the same way samurai blades of a bygone era were forged.  I then bought waterstones and learned to sharpen and hone that blade by hand, and started sharpening my wood working chisels that way too.  When I took my Solidworks advanced training course a few weeks back, I brought along my own mouse and plugged that into my workstation every morning at the start of training.  

I tell you this, because as a design engineer, tools are the vital link between our ideas and reality - they help bring our ideas to life.  Good tools make the task at hand so much more enjoyable.  

For the Kossel Pro, I wanted to create a printer that I myself would use and be proud to use.  We've been pretty open about the development process, our ideas and design philosophy, and our goals.  We hope that in our communications, our passion for advancing the state of the art, as well as our commitment to quality comes through.  We will never be the cheapest printer out there - we do not believe in a rat race to the bottom. Instead, we strive to deliver a well made machine at a reasonable price.

The Kossel Pro you see here represents our first step into the world of 3D Printers.  For now, I have kickstarter backers to take care of - whose generosity and trust have allowed me to chase my dreams of building a good machine.  As the machine's launch approaches, this is just the beginning of a journey.  We have lots of ideas and plans for the Kossel in the form of future upgrades, and we'll be back to working on those ideas in earnest after our Kickstarter rewards have been fulfilled.

If anyone is interested in reserving their place in line for a Kossel Pro, the preorder page is now active.  The Pro, with heated bed, will retail for $1249 at preorder and comes with everything except the spool of filament to assemble the printer and start printing.  

And once again, thank you for your support of Open Source Hardware.

-=- Terence & Mike and the Kossel Development Team