The focus of this week's blog update will be what, specifically, has and has not changed since we ran into our motor issues. One of our backers expressed concern that we'd gotten rid of the feedback loop, one of the whole reasons they'd backed us, and they were pretty upset that we'd gone and effectively changed the product without even consulting our backers. Since we haven't actually done this, we thought we'd clear things up for everyone! We'll also have a quick summary and discussion on Innovation Night and Maker Faire TO, and events in general...but we'll save that for the end.
When we first asked our IGG backers to support us, we promised an extrusion machine that was different from all of the others for three very important reasons. The first is that, unlike any other filament generation system on the market today, we include *everything* you need to recycle plastic. Spooler, extruder, and grinder. The second reason had to do with the extrusion system itself - we use MixFlow technology, which is now patent pending as being a novel breakthrough in extrusion technology. Every other plastic extrusion system we know of, including all of our competitors, use the same type of plastic extrusion that's been around for decades. In short, a screw is churned through plastic to ensure it is all melted uniformly and consistently. While this works well on massive industrial machines, it just doesn't meet expectations at a more desktop friendly level - so we totally re designed how extrusion works, and designed ProtoCycler around this innovation which we call MixFlow. Last but not least, ProtoCycler is "fully computer controlled" - this means both that there is automatic operation, where the settings are pre-programmed, as well as a closed feedback loop, which monitored the filament to ensure it was always within diameter tolerance.
Upon successful completion of our campaign, we decided to try something even more adventurous - Intelligent Extrusion. There's a previous blog post about it, but effectively we used the current or "amperage" feedback from our main motor to teach ProtoCycler how to extrude new plastics it had never seen before. The current feedback on a DC motor allows you to see exactly how hard it's working, so we could put in an unknown plastic and continue raising the temperature until the motor was loaded correctly. Too much effort would mean the plastic was too cold and hard to push through the nozzle, whereas too little effort would mean the plastic was overly melted and may be degrading or bubbling. Using these principles, we were able to extrude PLA "blindly", with ProtoCycler learning every aspect of the extrusion as it went along - this is how we extruded our glow in the dark spool at the We are Wearables event.
Alas, things did not turn out well on the motor front, as almost everyone now knows. Our DC gear motors didn't hold up to our endurance testing, and when we looked into things further we found out that to get ProtoCycler working at the level of reliability we wanted would cost *significantly* more than the current unit - prohibitively so. We'd found motors that could provide our level of torque, for a cost that was feasible...and then found out that they could only provide our level of torque once, or they would fail over time. To put things in perspective, we'd added a 3x Factor of Safety to our torque specs...but we would have needed closer to a 10x Factor of Safety for them to work. This meant getting a motor three times more robust, which would cost 3 times as much, use more power, take more space...etc etc.
We ultimately made the decision to switch to a stepper motor, which has none of the reliability issues that a DC motor has. This ended up lowering our energy usage and costs slightly, and has a host of other benefits like increased longevity, quieter and more reliable operation, and increased compatibility with RAMPS boards and other RepRap ready hardware, for those who want to one day make their own ProtoCycler. Sadly, this also meant losing our *current* feedback - the system that measures how hard the motor is pushing to try and learn about new plastics. It did not, however, require losing our *diameter* feedback - the system that ensures the filament diameter is kept within tolerance. In fact, by adding the stepper, we ended up making the diameter feedback system more robust...we were already re-doing the whole control algorithm, and weren't quite at the performance we'd had with the DC motor, so we effectively doubled the effectiveness of the diameter feedback to compensate for all of this. And, we're hoping to get Intelligent Extrusion back as well - check out our Hackaday update for more info on that front.
To get into the specifics, below is a rough outline of how the old and new diameter feedback algorithms work. You'll notice that while one or two sections are missing from the original one, it is an ultimately much more robust algorithm:
Old (diameter) feedback loop:
- Record the current pulling speed as a section of filament (which we'll call section X) exits the nozzle
- Wait for X to reach the (puller) diameter measurement sensor. The puller sensor is the one located after the cooling fan, right before the pulling system, that measures the final diameter.
- When X reaches the sensor, read it's diameter. Based on this diameter and the recorded speed it was pulled at, estimate a flow rate.
- Use the current flow rate estimate to adjust the speed at which filament is pulled
- Repeat as above! -everything up to this point is included in the new loop, as you'll see below.
- If there is a sudden surge or drop in current, immediately adjust the flow rate estimate to reflect that...assuming a sudden surge will temporarily increase the flow rate, and vice versa.
All in all, this system worked great. There is a (non trivial) delay between the filament exiting the nozzle and being pulled to size, but MixFlow ensured that it was consistent enough in the short term that the feedback loop could respond quickly enough. That being said, the motor would occasionally slip or lock up - creating gaps or blobs in the filament, respectively. With a lot of tuning and the current feedback in place, we were able to estimate what this effect would be relatively well, and predict what would happen based on the current. Ultimately, the system used both a delayed but known reading (from the sensor), combined with a guess that happened in real time (from the current measurement), to maintain the filament diameter.
New (diameter) feedback loop:
- Record the current filament speed as a section of filament, which we'll again call X, exits the nozzle.
- Wait for X to reach the puller sensor, and read it's diameter to calculate a flow rate...use this flow rate to adjust the speed at which the filament is pulled -everything up to here is the same as the old feedback, as stated above.
- In addition to the above old methodology, a second sensor has been added right at the nozzle. This sensor can see the filament being extruded in real time, and can adjust instantly. If it sees a thicker section, it tells the motor immediately to adjust - no prediction based on current measurements required.
- The diameter at nozzle is also recorded and compared later, to compute the draw down ratio. This is used to calculate what the diameter should be at the nozzle, if it's to be kept accurate at the puller. Like the flow rate, this is continuously evaluated. In other words, if the filament is getting twice as skinny between sensors and we want 1.75mm at the end, we want 3.5mm at the nozzle.
- An augmented PID loop is in charge of the nozzle feedback, which is much more robust than the "P" algorithm used on the old current feedback.
This system has a few differences, but also quite a bit of similarity. In both cases, MixFlow creates relatively consistent filament regardless. Any long term drift is accounted in both the old and new system by the continuous flow rate re evaluation - for instance, if your hopper is near empty, the flow rate will fall, and the puller will automatically slow (on average) to account for this. The only real change between algorithms is what happens if there are any unexpected short term changes in diameter. In the old system, the current sensing would be used to guess at what might happen - it did a great job, but wasn't perfect...no guess ever is, and it was missing the "I" and "D" of "PID". In the new system, ProtoCycler can see these changes in real time and adjust accordingly - no guesswork required.
We think these changes make ProtoCycler a better product, all around. It maintains diameter more reliably, lasts longer, is more compatible with existing electronics boards should you want to make your own, and has a number of other small improvements. Yes, shipping times have slipped somewhat significantly because of this. But we believe strongly that a delay in shipping is worth getting the best product possible into the hands of our backers and future customers, and so far the overwhelming majority of feedback we've received since making this decision has supported us (thank you!!!).
That all being said, we do value your support and feedback immensely, and to us there is no better way to engage and talk with all of you than to show up in person, with ProtoCycler, and do live demos. It's a great way to show our backers, customers, and supporters first hand where we are and exactly how things work, a *lot* quicker and more effective than answering e-mails or writing blog posts, and a great way to see what the community wants - and Innovation Night and Maker Festival Toronto were two great examples! We had a ton of people come out and see how ProtoCycler works first hand, and talked with quite a few at length about what changes we've made, how they all work, etc etc. To all who attended Innovation Night, or the main event on the weekend, thank you so much, and we hope you had an amazing time! Your enthusiasm and curiosity are amazing, and we cannot be more thankful for your kind words, critical feedback, and endless support.
Ultimately though, fulfillment is increasingly our only priority, and so we'll only be making two more appearances before we ship units - Maker Faire NYC and the TCT show in the UK. These are both huge shows, if not the two biggest 3D printing shows in the world before the year is over. It will be our first time showing off ProtoCycler to the eastern half of the US, and the first time we've ever been in Europe - so it will be an amazing chance to reach an incredible amount of people and show them what ProtoCycler can do, without taking more than a few days away from fulfillment. And, they're both within a week of each other, at the end of this September. More details to come on both, but Google should have some good answers if you just can't wait for us to start posting more about them ourselves.
In the meantime, please continue to reach out to us with any comments or concerns you may have about our development progress or anything else. It may take us quite a while to respond as we are incredibly busy moving forward with fulfillment, but it is important to us to make sure everyone is as educated as they can be about ProtoCycler and what it can and can't do. If we could, we'd be showing up to all of the events, answering all of the e-mails, and units would have been delivered months ahead of schedule. Sadly, running a business like this with just 3 people is incredibly difficult, and a very serious balancing act is required! So thank you all for your patience and understanding, and the continued words of encouragement. Together, we're showing the world that plastic waste is actually the next big thing - and we couldn't be happier to have that opportunity :)
-The ReDeTec team