Hey everyone, hope you’re enjoying the warm weather!
You may recall that a little under a month ago we started testing the new parts on a half-dozen printers. Since then, we’ve retrofitted the full TikoWall with new parts and scaled up our evaluation process. Last week we mentioned (in the comments) that our target RTM date was May 9th, however now that we’re here, we feel there are still some fixes to implement and data to collect before taking the big plunge.
Therefore, while we press forward with testing, we’d like to share something even cooler – the process itself, and why it takes so long. Many of you have asked us exactly how and what we’re testing, so today is your lucky day. Without further adieu, let’s go for a tour of the TikoWall!
How to Make a Reliable 3D Printer
Step One: Identify Failure Modes
A long long time ago, we started by identifying all of the possible ways that Tiko could fail. Traditionally, that is done with an FMEA (Failure Mode & Effect Analysis), which is basically just a list of all possible ways a component/system could fail.
As popular as it is, an FMEA requires an active imagination, and it’s all too easy to miss a whole bunch of non-obvious failure modes. So the first thing we did was to completely re-think the traditional FMEA, and created something we call a Matrix-Driven FMEA (MD-FMEA)… and it’s pretty rad.
In our MD-FMEA we list every single component inside Tiko as both a column and a row. Then, at each intersecting cell, we list all of the possible ways the two components could interact that would result in some kind of failure. ie, the connecting arms could pinch the power cable, the Bowden tube could snag on a motor, etc. It’s sort of like a multiplication table, and it helps us predict (and prevent) literally hundreds of possible failure modes that would have been missed in a traditional FMEA.
As if this isn’t cool enough, we made separate MD-FMEA’s for Mechanical, Electrical, Thermal, and Chemical failure modes We consider each category separately to even better catch any difficult-to-predict problems. We’re pretty serious about never being caught off-guard again. Now as we do this, we rank and color each failure mode by its likelihood and its severity, as follows:
The result is an MD-FMEA helps us rapidly identify an unparalleled number of failure modes and prioritize them for testing.
Step 2: Create Tests
Now that we know all of the ways a Tiko could theoretically fail, we systematically create tests for each of them. Basically, we come up with the conditions most likely to cause each failure to occur. Screws shaking loose? We make the motors vibrate. EMI interference between heater and board? Crank the power up and sweep through a bunch of PWM frequencies. These are just a couple of easy examples, but the test plan itself is quite complex.
We then arrange the tests so that we start with the most severe/likely tests firsts, and work our way to the most benign/unlikely, trying to save the possibly-destructive ones for last. Check it out, here’s a sneak peek of the TikoWall test plan:
With a test plan in hand, we then...
Step 3: Run Tests
Controlling one hundred WiFi devices from a single point is exactly as hard as it sounds. Obviously, controlling them one at a time would have been impractical, so we created an entire UI from scratch just so we could send commands / receive data to the whole TikoWall as quickly as possible. Check it out:
Each of these boxes represents a Tiko on the TikoWall. Notice there are 5 rows, with 20 boxes each, just like the wall. Using this UI, we send simple G-Code instructions and/or entire G-Code files to any/all of the Tiko printers, while requesting specific data from them to speed up the debugging process. We're also able to see which printers are online, whether or not they have up-to-date firmware, whether or not they’ve passed self-diagnostics, what auto-level plane they’ve detected, etc. All in real time. It’s an awesome tool.
Unfortunately, one of our biggest challenges is getting printers to simultaneously come and stay online. It’s incredibly difficult for a WiFi router to maintain communication with over one hundred unique devices, so we typically control the TikoWall one row at a time. It’s a bit time consuming, but we’re working on it.
Step 4 – Analyze Results
Every time a test is run, we have to check that each Tiko performed as expected. If not, we flag them with cards and begin debugging. Sometimes it’s a hardware problem, sometimes it’s a software/firmware problem, and getting to the bottom of it can take several hours. Once we do find a problem, we design/implement a fix, and repeat.
As you can imagine, this makes testing a rather time-consuming process, but it’s worth it. An ounce of prevention is worth a pound of cure, so we’re out to find every possible failure mode before these Tikos are out in the real world.
The Results Are In…conclusive
So the big question is, what have we found so far? Although the half-dozen printers from the last update ran perfectly, we now have a hundred up there… So did any new issues crop up?
Well, as can be expected, a few problems arose. Luckily these issues are minor and relatively quick/easy to solve. In fact, all will be solved by next week. To give you a quick run-down:
Carriage Ball Joint Neck – Turns out the neck on the carriage ball joints is a bit thick, which sometimes restricts the connecting arm's range of motion, causing problems at the extremities of the print envelope. Otherwise, they work just fine. It’s a relatively straightforward adjustment to the mold, and we should have new & improved ones in next week.
Wires Shaking Loose – We used simple pin headers for connecting the heaters/microswitches to the PCB, and after extended periods of vibration, they sometimes came loose. The solution is simple: we’re switching to locking connectors. We’ve implemented these on one board… so just 99 left to go. :)
Bowden Tube Pops Out of Fitting – The new extruder fitting wasn’t sized properly, allowing the Bowden tube to occasionally pop out from the stresses of extrusion. The connectors have already been adjusted to better clamp the flare, and we’re just waiting to receive them here at HQ.
Extruder Motor Insufficient Torque – In a few of the printers, the extruder motors either didn’t have the torque to push the filament through, or we had to overdrive them. The added heat softened the filament, making it unextrudeable. Tiko has a relatively strong extruder motor, but a tall gear ratio for quick loading/unloading of filament, and it looks like we've sacrificed a bit too much torque and got too close to that fine line, so we’ll just switch to a torquier gear ratio and call it a day. We can still test the extruder in the meantime, we just have to give it cool-down periods. Lame.
Base Substrate Adhesion – Not a real “flaw” per se, but rather a particularly inconvenient mix-up where we ordered 100+ units of the base with an incorrect substrate material, which has poor adhesion compared to the material we’ve been using (in small scale) for the last few months. It’s not impossible to print on it, but it has effectively made print-testing en-masse impractical, because the prints keep coming off the beds. The replacements are due to arrive next week. In the meantime, we’re running pseudo-prints (basically, giant messy cobwebs) to keep the kinematic assembly/liquefier/extruder busy with something. The more hours we put on them, the better the data. We’re just trying to find Tiko’s weak spots, so we don’t need to print physical objects to do that. Oh, but how we want to!
And that’s it!
So, no, Tiko is not yet ready for shelves, however the great news is that all of the old - much more serious - problems are gone. We haven’t seen a liquefier jam in over a month, a carriage has not gotten stuck, the accelerometers have proven bulletproof, and the improved extruder inlet receives filament with ease.
Basically, we’ve solved the big problems, not seen any new big ones, and now we’re ironing out the easy stuff. It’s really exciting to be at this level. So, where do we go from here?
90% Done, 90% Left to Go.
It seems like we've been testing forever now, but as you can see it’s a time consuming process and we’re not yet finished, especially since we need to implement the fixes and re-test the new & improved parts.
We’ve also run some full-on print tests, but because of timelines and the base adhesion issue, we’ve only put a few hundred print-hours through the TikoWall. That just isn’t enough to call Tiko “ready”, so we need more time. We’re confident that two weeks is enough. Meanwhile, we’ve got other things happening too…
Another thing worth mentioning is that we’re almost two weeks into official regulatory testing, and we have not experienced any significant issues.
We have to add some safety markings near the motors/liquefier, and implement a watchdog timer into our firmware to pass the EFT (Electrical Fast Transient) test, but otherwise it’s been smooth sailing. Not to say we got lucky – we spent months reviewing designs with regulatory officials and pre-testing Tiko at accredited facilities. Because of this we caught a number of problems long before committing to formal testing. Ah, we love it when a plan comes together. :)
The final results won’t come in for another week or two, but it’s looking pretty good!
This is generally an exciting time here at Tiko, as we’ve recently hired no less than seven new people – a few interns for the summer and some full-timers for post-delivery operations (marketing, customer service, etc) bringing us to a total of 18 people on Team Tiko (not everyone is in the above pic). There’s a cool sort of chaos here as everyone works towards the common goal of delivering this beast of a printer. Lots of teamwork, scrums, collaboration, and mentoring. Ah, spring is in the air!
We hoped to finish testing last week, but it looks like we have about two weeks left to go. However, given the results thus far, we know we’re just dialing in the last few bugs before “The Big RTM”. As a result, we’re already preparing RTM documents for a variety of components so we can hit the ground running ASAP.
With about around weeks of testing left, and a six week RTM-to-first-shipment turnaround time, we’re setting a tentative first-ship-date of July 1st. These are relatively conservative numbers, and the six-week manufacturing lead time isn’t particularly volatile, so once we give the go-ahead to manufacturing (and to all of you, via an update), that date will be locked in. Oh boy, we’re getting goosebumps just thinking about it!
Shipping – Important Information
This is probably a good time to mention the shipping survey. There are a little over a thousand surveys still uncompleted, and we continue to receive many change-of-address requests. If you’re simply holding off until we announce The Big RTM, it’s all good, just keep in mind we’ll announce a cut-off shortly before the shipping start date, and any surveys not filled out will be pushed to the back of the queue.
Meanwhile, the change of address process is quite straightforward and you don’t actually need our intervention. Just head over to shipping.tiko3d.com and update your address, and we’ll get it automatically.
Boy, that was a beast of an update, but hey, it’s been a busy month! The madness isn’t over just yet - there’s still a great deal of testing/retrofitting/retesting to do before we can RTM everything. We can’t wait to finally fulfill our promise – not just of shipping a printer, but of turning this whole industry upside down. You know how your coffee machine, inkjet printer, microwave, etc just... you know... work? Well why shouldn’t your 3D printer? That’s why we’re here, that’s why we’re testing, and that’s why we'll never stop until we’re number one!
Viva la Revolución!