And this is why we do manufacturing pilots...

Hello everyone!  

Early on in this week, we received our first batch of Brainwave Pros from our contract manufacturer in China. This was a very highly anticipated event, as these boards are one of the last long(er) lead time items we were waiting on to ship printer kits.

The production Brainwave Pro

When we receive a new batch of boards in, Mike, our electrical engineer, goes through an elaborate procedure to "bring up" a new board.  It basically involves monitoring key circuit parameters while slowly energizing a circuit with a current-limited power supply.  The idea is that if something goes wrong, the current limiting portion of the power supply will hopefully prevent a catastrophic failure that involves magic smoke leaving our electronics.

When this batch of Brainwave Pros were brought up, we noticed that the boards drew more power than normal.  This is a very bad situation as it typically indicates that there is a short on the board.

Back side of the board.  Note the vertical SD card reader.

Long story short, after some serious troubleshooting kung fu, our rockstar electrical engineer located the problem.  The Brainwave Pro board features protection components called transient voltage surpression (TVS) diodes to clamp the back EMF force from the motor during a sudden disconnection.  In other words, although not recommended, the board can survive an accidental disconnection of the motors while energized.   Unfortunately, after the technical hand-over to our contract manufacturer, somewhere along the line someone added an extra decimal point to a part number by accident while compiling the work instructions and bill of materials.  As a result of this transcription error, the wrong, under-spec TVS diodes were installed onto the boards, and these diodes were blowing and shorting out during normal operation of the board.

We have now tested a board by removing the offending TVS diodes and the board functions just fine.  We have ordered replacement TVS diodes of the correct specification from Digikey and installed them to verify that the board does work with the correct TVS diodes installed (and that protection of the driver chips does work).  At this point, we have instructed our contract manufacturer to purge all TVS diodes in their stock and refresh their stock with proper TVS diodes, sourced from Digikey (China).  

Unfortunately, the goal of the manufacturing pilot was to verify that boards from our new CM will energize and program correctly without intervention on our part.  We failed that test, and for peace of mind, we will have to do another round of manufacturing pilot.  (For all we know, they may also be putting on these diodes backward - we don't know what we don't know).  The delay is expected to be an approximately two week time frame.  Even though the risk is very small, we just cannot afford to have non functional boards land.  The cost of reworking a board here in the US to replace the offending diodes by a professional board house can be almost half the cost of the boards, brand new (material and labor) from our CM.

In the meantime we have 24 boards state side that will require rework to function.  But that's a lot better than having over 400 boards that require reworking.

-=- Terence & Mike.


More Kossel test prints

We've been testing and tuning printer profiles.  Here's some more pictures of test prints.  

One thing that is really cool, is that my friend Chris Gilroy had been working on an open source modular ergonomic keyboard.  He's been helping me log real world test case of the Kossel by building all his keyboard parts on a prototype Mini Kossel (reprap).

It's one thing to print cool looking artistic parts.  It's another thing to be printing functional assemblies that can be fitted together - and in his case, snapped together.

Enjoy the images!

-=- Terence


Stuck In Customs - and getting schooled on Chinese Export Laws

We've recently had another minor set of setbacks to the OpenBeam Kossel program, and to some slight degree, the OpenBeam product line in general.  We now have 3 shipments with clearance delays and 1 shipment with more serious issues.

Customs clearance delays is one of those risks in businesses that everyone just accepts. They can happen for a wide variety of reasons - your cargo container can just be lucky enough to be flagged for X-Raying - no different from a random TSA search.  Except, depending on the port of entry, X-Raying a container can take anywhere from days to weeks.  Another risk with ocean shipping is that the captain of a cargo ship has every right to scuttle containers off his deck to preserve the life and safety of his crew.  Watch the video below of a container ship navigating the Pacific Ocean during a storm and watch how the ship twists and bow.  We buy insurance on every ocean shipment and just hope that our number doesn't come up.

In our case though, we are finding a very consistent reason for the delays:  Our shipments contains either ball bearings, or mentions ball bearings.  Apparently this is enough to cause all sorts of additional paperwork.  

These delays are annoying, because they throw off our schedule and we are already trying hard to play catch up.  (In the case of the DHL shipment shown above, it contains all the production hardware for the Kossel.  I have been hand machining parts to build prototypes, but this is not feasible for production, obviously.  They also contain the ball bearings used in the ball joints - each Kossel contains 48 ball bearings in just the ball joint linkage itself).  

But perhaps the biggest lesson that we've learned, is the export troubles we have in bringing our latest shipment of linear rails (and glass plates, power supplies, etc) over from China.

Early on in the prototyping process, our friends at Western Tool and Mold were doing us a huge favor by helping us with some cargo consolidation.  My Dad would help me order my engineering sample parts, arrange for delivery to Western Tools.  Western Tools would hold them until I have enough parts, or when I have an order heading towards Seattle, and just drop the parts in.  That's how Johann's first Mini Kossel prototype was built; I essentially called in a favor and managed to find reasonably priced linear rails and other goodies.

Obviously, this process isn't sustainable, so I've asked a friend of mine who runs a contract manufacturing and optical instrument business in China to help with managing the domestic Chinese supply chain.  (Eventually, they will also be kitting the subassemblies, as well as take over the OpenBeam packaging business).  In an attempt to save time, we ordered all the components and had them shipped to our friend in Fuzhou, China, while we were hammering out the trade agreement.  We figured that once we have everything collected, we'll call Expeditors and handle the ocean shipment much like how we handle our regular shipments from our other suppliers now.

Turns out, due to the way tax laws work in China, it is very problematic for that shipment to leave.  Exports have to be declared (as the companies that export have to pay taxes).  Long story short, because my friend's company didn't purchase the parts, they can't declare the export (because at the end of the year when they are audited, there will be no corresponding money trail to support exporting the materials in question.  And without the export declaration, we cannot use a freight carrier like Expeditors.  

To get around the restrictions, we are going to break the packages up so that they fall under the "Small Parcels" guide line - individual shipments, each under $2500.00, and ship via a courier service such as DHL or UPS.  An expensive lesson for sure, but at least it'll get the parts here faster.  (IE:  we originally budgeted and scheduled for ocean freight, there will be no schedule impact because of this).

Image courtesy

Image courtesy

So, how have we been dodging this bullet all this time?   Most of these sellers we are buying from are small Mom and Pop shops.  None of them speak English; most of them don't even have English named banks that can accept TT or a Western Union payment.  They've been using a myriad of small parcel services in China to get their goods across the border into Hong Kong (which is a free port and has no trade restrictions).  We've had a couple of isolated incidents where the selected courier service will only deliver to the ground floor and we had to bribe the delivery personnel to deliver the goods upstairs  (Western Tool's entire factory is in an industrial building, 8 stories above ground.  How they got all their machinery up in there is a story for another day).  

Moving forward, our supply chain consolidation will have us buying parts through our friend in Fuzhou, so this will not be an issue.  By kitting the components in China, we will also be importing a completed product, which we can classify under "Hobby, Toys and Models" category, and avoide some more customs headache.

It's been a long, long road to get here.  I'll just be glad to be back to doing engineering and building printers instead of looking up HTS codes.

-=- Terence

The OpenBeam Kossel Famly - Part 1 - Architectural Decisions

Here's a blog post that's a long time in the making:  The introduction of the OpenBeam Kossel Family.  Due to the amount of information being presented, we've decided to break this into 4 segments, covering the architectural decisions, the Kossel Pro, the Kossel Reprap, and finally part availability and how to go about building one.

Terence's personal OpenBeam Kossel Reprap and Kossel Pro, built from T0 plastics.  Photo courtesy Chris Gilroy, Solarbotics Ltd.

The OpenBeam Kossel family is a branch of Johann Rocholl's excellent Kossel 3D Printer.  Since the beginning, we worked very closely with Johann and kept him well stocked with extrusions, linear ball bearings and other miscellaneous goodies.  The result was the Mini Kossel that replaced the legacy Kossel, which made its debut at the Seattle Mini Maker Faire a year ago.  Then, as one day, as I was driving, I came to the realization that the new vertexes Johann had designed for the Mini Kossel were basically a custom extrusion.  With that, I decided to redesign the Kossel Pro to merge the design tree with Johann's Mini Kossel, and here we are today.

The OpenBeam Kossel family consists of two forks, a Pro branch that features high grade mass produced components, as well as a reprap branch that features printed parts.  Great care went into designing both to ensure subassembly level part interchangeability.  Part commonality is an important design concept when designing these printers; because of the part interchangeability between the Pro and Reprap line, the same core vitamins kit will build both machines.  This allows Kossels to print their own repair parts in the field and offer Reprap machine builders an upgrade path.

Design Table showing the different Kossel family configurations

Mechanical Design Decisions:

In a nut shell, the mechanical design decisions came down to a simple premise: No compromises.

We wanted the Kossel family to be mechanically reliable printers, first and foremost.  After looking at various "Maker" style linear motion systems, we settled on the use of Chinese made copies of recirculating linear ball rails, coupled with proper cleaning and lubrication techniques.  It took a little bit of trial and error to get the lubrication correctly; we ended up designing custom fixtures that allows the balls to rotate through their motion range while applying the grease into the ball track.  This allows us to perform the grease application in a cost effective manner and offer the linear rails with grease preapplied.  

The Kossel Pro's linear rail, mounted to the vertical OpenBeam extrusion.  Note the self-terminating belt loop feature; this is a design that is ported into the new Kossel Reprap branch as well.  Photo courtesy Chris Gilroy, Solarbotics Ltd.

The rails add a lot of rigidity to the frame of the printer, and ensures that the printer's extrusions are true.  While norminal flatness tolerance for a hollow extrusion is about 1.5mm per meter, the rails are much stiffer and flatter, as a result of the grinding process used.  It also eliminates a huge variable in building the printers, as builders are no longer required to adjust the preload on the linear carriage.  Given the light load on the linear rails, the nature of induction hardened tool steel and ball bearings (the rails are so hard, they have to be cut by wire EDM when we order them), we expect these rails to last the lifetime of the printer.   

Kossel Reprap's full ball bearing pulley.  Photo courtesy Chris Gilroy, Solarbotics Ltd.

For the idler, we chose to use an actual timing belt pulley for the idler.  This results in much less wear and tear on the belts and eliminates the possibility of the belt skipping off flanged bearings.  It costs a little bit more to do it this way, but we believe the trade off in reliability is worth it.  For the Kossel Pro, we elected to design a ball bearing mount block and tension the block through a screw and nut mechanism.  This allows the top of the printer to stay flat relative to the base of the printer.  Future printer designs may use this space for a spool holder.

Finally we elected to use a nice, thick borosilicate glass plate for the build surface. Borosilicate glass, unlike laser cut acrylic, is temperature stable.  Reprap builders have found out long ago that repeatedly thermal cycling one surface of a sheet of acrylic (as heated plastic is melted onto it) is a good way to cause said plastic to warp.  By picking thick borosilicate glass as a build surface, we can be sure that the build surface is always flat.

Electrical Design Decisions:

Mike Ziomkowski, our electrical design engineer, will be writing a blog post in detail on the Brainwave Pro.  

Deltabots are unique machines:  with less mass on the end effector, they are fast and nimble, and execute GCode faster than a traditional 3D Printer control board can be fed through the FTDI interface.  Deltabots also have a problem with traditional board placement:  there just isn't that much usable space inside a triangular shaped frame to fit a rectangular board.  On top of all this, we really wanted a board that would handle competently 24V drive, and 1/32 microstepping for smoothness.  

Why 24V?  Motors are happier at 24Vs; accelerations are snappier, lower current will yield the same amount of holding torque.  Heaters are also happier at 24V; less current is needed to bring the same amount of wattage into the heating element at higher voltages. 

Early on, we toyed with the idea of creating a hexagonal control board, but we quickly abandoned that idea when we realized that the board had to sit between the power supply unit and the heated bed - in otherwords, between a toaster and an oven.  

Proto B Brainwave Board, mounted to the inside of the Mini Kossel Reprap's lower triangular frame.  Note the vertical SD card reader slot.  Photo courtesy Chris Gilroy, Solarbotics Ltd.

So we designed a board from the ground up to mount vertically:  We  used a vertical full size USB-B connector for connectivity to the outside world (MicroUSBs are just a bit too fragile, and min-USBs are being phased out).  We used the AT90USB microprocessor, bypassing the FTDI USB chip, for much faster USB communcations.  We then mounted all the motor and end stop connectors on the other side of the board.  Finally, we designed the board to fit within the 75mm height vertex of the OpenBeam Kossel family.  75mm edge-to-edge, 60mm on center hole spacing is the same dimension as the OpenBeam NEMA17 motor bracket mount.  This allows us to mount the extruder motor in the base of the unit for a very compact and clean installation and use the aluminum structs as a heat sink for the motor.

Top view of the mini kossel.  This is a fully functional and wired printer with very clean lines and wiring.  Photo courtesy Chris Gilroy, Solarbotics Ltd.

There were lots of head bashing and threats of strangulation to get everything fitted into such a small package.  In the IATA kossel, the screws used to hold the corner vertex to the extrusion does double duty to hold the standoffs for the electronics. 

In keeping with the philosophy of designing for a "family" of printers, the Brainwave Pro's horizontal hole spacing matches the IATA /. Air Kossel's vertex hole spacing.  This allows us to build the smallest of the Kossel family - the IATA kossel that fits inside regulation hand carry luggage, without issues.

Other Improvements:

As we set out to design the Kossel, we found places where improvements can be made. One example is the J Head hot end.  Considered to be the reference benchmark for PLA printing, we sourced all our prototype hot ends early on in the program from the original designer.  However, there were a few things we weren't happy with on the J-Head design:

1)  The fit of a heater cartridge into the J-Head metal hole is a very loose slip fit, with no mechanical way of securing the heater cartridge.  We weren't happy about buying a precision machined part, only to have to resort to rolling the heater cartridge in aluminum foil to get it to the correct diameter to press the heater cartridge in.

2)  There is no mechanical means for securing the thermistor.  Pulling a thermistor off a J-Head in operation results in a catastrophic failure mode; without the feedback thermistor reading the temperature correctly, a heater cartridge on full throttle can become red hot and easily melt and destroy the J-Head housing.

3)  The default way of securing the PTFE liner with a hollow PEEK set screw is problematic; it creates sharp corners that make blind feeding of filament into the J-Head during filament change next to impossible.  Most users get around this by disconnecting / unscrewing the bowden feed clamp and manually fishing the filament through into the J-Head during filament change.

To remedy these issues, this is what we did:

A)  We added a set screw for retaining the heater cartridge.  

Undercut pocket and set screw hole for thermistor and heater cartridge retention.

B)  We added an undercut pocket for securing the themistor.  The thermistor is installed in protective PTFE tubing to prevent short circuiting, crimped with ring terminals (or Molex SL connector, depending on SKU) with a certified crimp tool, and then potted into the hot end heater block with high temperature non corrosive silicone.  It costs more in terms of labor cost to provide this as a pre-assembled component, but from a user's standpoint it is far more cost effective for us to spend the money on the tooling and potting compound and amortize the cost across a production run of hot ends and ensure that every hot end is built properly than to source their own tools.

4 hot end tips, set up ready for potting. The ring terminal variant is used for the Kossel Pro's end effector.

Completed hot end tip with potted thermistor.  We sourced a non corrosive silicone - a lot of Reprap builders uses "red engine silicone" which degases acetic acid and will corrode electronics over time.  We use Loctite 598.

Assembled Kossel Pro end effector.  Photo courtesy Chris Gilroy, Solarbotics Ltd.

C)  We changed the design of the PTFE retention liner.  Instead of a hollow PEEK setscrew, we custom designed our own retainer piece that features a large chamfered feed lip to feed the filament into the J-Head.  (This piece can also be swapped with another piece that features an M5 threaded hole to accept a push fit connector for direct bowden drive).  We had to design a special tool to secure this piece into our J-Head, so we designed the tool to be 3D Printable.  

Chamfered feed lip makes for much smoother feeding.  We designed a custom wrench that is 3D Printable to secure the PTFE retainer.

Additionally, these hotend parts are made by a local startup aerospace shop on their Haas VF3 machining center.  The threads are thread milled and holds a much tighter tolerance than traditional hand tapping with a regular tap.  As a result, we've logged over 200+ hours of leak free PLA printing without having to use any PTFE tape or thread sealant to assemble these prototype hot ends.

Points A and B are relatively small evolutionary changes, however, we were pleasantly surprised at how big an improvement C really is.  By having the big chamfered feed hole, filament can be fed blind into the hot end subassembly.  Nobody likes to have to disassemble half their end effector to change filament, so that is a very welcomed improvement.  

We hope you've enjoyed this update; we'll be back soon with another update on the improvements that went into the Kossel Pro.

-=- Terence & Mike, and the Kossel test team.


The test fits shall continue until everything is fixed

Hello everyone!

We've had an interesting weekend.  Our T1 plastics were supposed to be delivered last Friday so I could have the entire Memorial Day weekend to work on printers, but they were held by customs.  The reason?  I was a bit too descriptive on my part naming.  The ball joints were called "Deltabot Ball Bearing Joint A" and "Deltabot Ball Bearing Joint B".  Because the part name contained the word "Ball Bearing", US Customs decided that it was now a bearing housing instead of just a regular plastic part, and held it until they could get my Federal Tax ID on the shipment so that I can be taxed accordingly.  

The tax rate is 2.9% value of the shipment, which for 30 pieces of engineering samples, came out to less than a Big Mac and Fries at McDonalds.  The cost of losing a long weekend, however, is far more damaging.  This goes on to illustrate just how messed up our tax laws can be.  Of course, in final production (after Kickstarter campaign), the kitting of the ball bearings, plastic parts and packaging will be done in offshore and the completed assembly will be declared as hobby / construction toy part, which can be imported tax free.  The fact that I can import finished goods tax free, but get taxed per individual line item (4.9% on bearings, 3% on screws, etc) shows how backwards our tax laws can be and how it does not encourage small business owners to keep jobs in the US.  Interestingly also, if I had just named the part "Deltabot Joint", chances are it would have flown straight through customs clearance.  Either that or my house would be raided for weed by the DEA - you never know. :-P

Mrs. OpenBeam was conveniently out of town on a business trip, which makes such annexations of common living areas slightly more OK.

The good news is that MOST of the issues have been solved now at T1.  The bearing press fits, crucial to achieve low friction, backlash free motion in the ball joints, have been fixed.  We've added cavity identification markings into the tool - each part shows a different number depending on which cavity it came out of - to better allow us to track problems with injection molding should they arise in the future.  

Bearings now press fit into all the ball joints nicely with just the right amount of resistance.  This is a major win.

Bearings now press fit into all the ball joints nicely with just the right amount of resistance.  This is a major win.

Other fixes included cleaning up some flash issues on the end effector, and better processing to bring the parts within tolerances and flatness specifications.  At Western Tools, their injection molding machines actually monitor each shot's parameters and continually compares it against the programmed machine settings, and should the processing parameters drift, the machine will alert the operator and stop operation.  This means that once we dial in our manufacturing processes, parts will be stable and consistent.  It doesn't stop us from performing incoming inspection on each batch, but we are less likely to have issues this way.

The snap fits worked perfectly the first try.  We got REALLY lucky.

The snap fits worked perfectly the first try.  We got REALLY lucky.

Of the remaining problematic parts, only 3 parts remain:

*  There is still some fit issues on the ball carriage rail's top and bottom halves.

*  Some of the geometry for the self-terminating loop isn't quite right, and it's hard to install the timing belt as a result.

*  There is still a bit too much slop in the nut clamp.

These are relatively minor, and we expect that it'll take 2 weeks or so to fix.  While we can greenlight production on the rest of the parts, some of the parts share the same mold and may have to wait until mold changes are done before the part can be ran.

I was driving to work this morning thinking that this phase of the project is taking longer than expected, when I realized that as far as complexity goes, the Kossel actually has more plastic parts that had to interface with other mating parts perfectly than my previous two products I've designed for my day job.  The last product I designed, a microscope stage top incubator, only had 9 plastic parts and didn't have the tolerances required on this project.  My other project was designing the mechanical of a low end florescent microscope, and there, only one set of parts (the hinge) required high precision.  And both of these projects were funded from a very generous R&D budget that didn't blink twice about expedited delivery charges and overnight shipping.  

In the meantime, I'd like to share with you some more pictures.  We have pictures of the Mini Kossel and the Kossel's heated beds, thermal images of the hot end on the Kossel Reprap showing the benefit of our ducted fan cooling, as well as thermal images of how uniform the heated beds are.  Also, I'd like to show some pictures of test prints that we've done on the Kossel Reprap as part of our work tuning the slicer profiles.


-=- Terence