So, update part two. If you missed it, yesterday we put up an update covering some things going on at Mosaic, as well as an introduction to our calibration process, and software. You can find that update here: Link
Today, we’re going to dig into hardware, manufacturing, shipping, and timelines. Without any further ado…..
There have been some really interesting challenges presented to our team over the past month.
Something important to understand - we’re not a 3D printer manufacturer. We’re not working with gantry systems, extrusion systems, and leveling beds. When we run into issues, there’s no hitting up Google to find a solution someone else has already come up with. There’s testing, iteration, failure, and trying again.
When you think about The Palette’s functions in their most raw sense The Palette needs to accomplish two tasks. First, construct high quality, multi-component filament. Second, deliver these segments of filament to your 3D printer’s nozzle at precisely the right moment.
Over the past few weeks we’ve been having consistent issues with getting reliable splices within our minimum acceptable tolerance. To give you an idea, a splice that is over 1.9mm in diameter will not feed through a printer, causing a failed print. This has never been an issue since about 14 months ago, and came out of the blue quite recently.
After two weeks of breaking the systems down, and testing them one by one we figured out what was wrong. Normally, when you’re testing variables, you test them individually – checking the results of your tests against your varied inputs. We went through each one of our systems, analyzed them, and compared our test results. From stepper motors, to stepper drivers, to drive modules, to splicing tools – each system had its error points and variability quantified. Changing up any single one of these systems did not give us results that were within our acceptable tolerance level.
That’s because the issue wasn’t with any single system – there were two independent issues that came up with two separate systems. This caused a rather frustrating experience for a few of our engineers, as figuring out two independent issues in a system’s heavy product is not an easy thing to do.
So what were the problems? And how did we fix them?
Last month we received a notification that one of our suppliers was shutting their factory down – the supplier for our drive gears. That’s OK, because we have two back up suppliers ready to step in. We also had a few hundred gears on hand from our original supplier, so it wasn’t set to delay our critical path at all.
Now, The Palette has five drive gears inside. Four on ingoing drives, one on an outgoing drive. When we were running splices we had consistent results with ingoing drives two and four, and inconsistent (bulging, large) splices on drives one and three.
It turns out that the gears from our supplier were inconsistently sized – ranging from 10.6mm in diameter to 11.2mm. The outgoing drive, as well as drives two and four were all in the 10.8 area, where one and three were in the 11.2 area. This caused an over driving of the filament on drives 1 and 3, leading to bulging splices that would not feed through a printer.
Here is one sample gear, measuring around 11.12mm
Another gear measuring around 10.85mm. These two gears in the same Palette would cause issues when it comes to splicing, either overdriving, or underdriving the filament.
Here you can see the two ranges of gears separated after being measured.
We then proceeded to measure samples of 40 from our batch of gears, and found this representation to be relatively accurate. About half the gears were sized at 10.9mm, the other half around 11.1mm.
Because of this variance, we’ve added a new step into our QC processes. This step is to measure drive gears and ensure all gears that are installed into a given Palette are the same size. This variance is something that’s never been an issue before, but is now something we know to look out for.
So, that’s problem one figured out. But consistent gears were not the only issue we uncovered from one of our suppliers….
Now, when a splice is made it immediately goes into a forming channel. This channel is made of a Teflon machined tube that acts to re-form the splice. This channel ensures that the splice is cooled/formed within a certain tolerance level, the tolerance level necessary to feed it into a printer.
This channel has a diameter of just under 1.9mm, meaning that in theory, no portion of any splice should ever be over our accepted diameter of 1.9mm. As I’m sure you’ve experienced before, theory and practice are two different things.
We ordered a few Teflon sizes smaller than 1.9mm to test the results we would get with smaller sized channels. We ordered 1.85 and 1.8mm ID channels from the machine shop we work with – this was a few weeks back.
We tested them out the day they arrived - on the first weekend of March. We were still getting splices that were too large. To be perfectly honest, this made absolutely no sense to us. Some of the splices we were measuring reached as high as 2.1mm in diameter. So we decided it was time to measure what the true size of the ID of the channel was…… and we found some really interesting results.
Although it was very difficult to measure a hole that small, we consulted a metrology expert at our machine shop who was able to help us out. We learned that in one dimension (horizontally), the channels measured to be the size they should. 1.9, 1.85 and 1.8. Vertically, all of the channels measured approximately 2.1mm, a massive variance.
We called up our machinist and spoke to him about the issue. He explained that this part is made in two operations with different endmills and that if not carefully controlled, it can lead to this oval shape that we experienced. He then did some research and found an alternative way to machine these features that would ensure the channel was within spec.
So we sanded down parts of the splicer, to even the vertical area out so it was no longer oval shaped. We then ran the Palette with these modified parts in place and found that the splices coming out were within specification. This proved that the oval shaped channel was the other factor leading to the oversized splices. This, along with the new drive measurement process has led to substantial improvement in splice quality. We’ve now fed over a thousand splices from multiple Palettes through printers with no issues at all. We have just received new prototypes to essentially match the sanded parts and they have also operated flawlessly. This has proven that the new approach to machining these parts will ensure they are within the specified tolerances for all future orders. The first 50 sets of parts have been ordered and will be in later this week.
Now, remember at the beginning of this section we mentioned that The Palette has two jobs. We just covered job number one – forming multi-component filament. Unfortunately, because this was such an issue over the past few weeks, we’ve been pushed back for testing and calibrating Palette job number two. Sending filament to your nozzle at the exact right moment.
We’ve been working on software since The Palette’s inception almost two years ago, and have been developing our two closed loop feedback systems (pinging and ponging) for the past year, and six months, respectively. These systems are working well, but we’ve been aiming to get a few good weeks in of final testing before we moved our full team’s effort into production.
Unfortunately these systems weren’t able to be tested in parallel, as our splices were not feeding through the printers. Because of this, we’re now heavily engaged in testing the systems component of The Palette to ensure that both feedback systems are functioning reliably and acting as they should. We cover timing at the end of this update, if we do push back a week or two this testing is the reason.
So right now, we’re pushing to optimize the firmware and software algorithms inside The Palette in order to make sure the pinging/ponging systems are all ready to go when the first batch is made. Check out this picture of two Palettes (bottom right) printing on two standard machines (Wanhao and CraftBot), while one (top left) is going through splice testing and optimization.
Everyone on our team has been working to not only discover the root cause of these issues, but to put our mitigation strategies into action to solve these issues as soon as possible. As always, we want to communicate to our backer community some of the challenges we’ve been facing with the launch of our manufacturing operation.
So the hardware is now being tested, and the first run of Palettes are starting to be constructed. Read below to get a deeper look into our manufacturing operation, and its progress.
There’s a lot that goes into a manufacturing operation. We’ve sourced all our parts, which are either in our office, or arriving this week. We’ve now started to build out our sub assemblies, which will be bolted on to our sheet metal skeleton, and turned into a Palette.
It may be counter intuitive, but putting together all of the sub assemblies to make The Palete itself isn’t the most time intensive process. The most time intensive processes are creating each of the sub-assemblies, as well as going through the final quality control check.
As mentioned in our update yesterday, we’ve hired a few assembly team members to begin building the sub assemblies. Over the past few weeks they’ve been prepping and building these assemblies, which will soon be bolted onto the skeleton, and turned into full machines.
Our sheet metal skeletons are arriving this week, and when they do everything will begin to be bolted onto their frame. Right now, our main focus is mainly on electrical, soldering, and pre-fabricating Teflon pieces. Soldering is a very labour intensive process – everything from the grounds on the screens, to the scroll wheel connections, to some of the custom wiring harness sizes that are required.
Check out some of the manufacturing progress in these pictures:
An encoder in its mid-processed state.
The rest of the batch currently sitting in this mid-processed state
The encoder completed, and a sampling of more completed encoders
Our first batch of 50 completed Scroll Wheel plug in points, complete with ferrites to allow The Palette to pass certification standards
A batch of fans in their mid-processed state
The rest of our steppers came in the other day – approximately 2100 steppers in this picture (enough to fulfill every Kickstarter unit + more!)
A sampling of our Teflon drive pieces. These will be attached to in-going drives, and connected to the merging module.
CNC Shield with The Palette's extra stepper drivers installed.
Front view of The Palette’s Box. 400+ of these are set to arrive at our office this week!
We’ve been working overtime to fully layout and document our quality control processes. These processes include the development and building of numerous rigs to allow us to complete not only our FAT (Final Acceptance Tests,), but also our ICT’s (In-Circuit Tests).
ICTs are the tests that are completed during the assembly process. For example, when all of the soldered connections for The Scroll Wheel’s are completed, an ICT will be performed to ensure the connections are strong. When an ICT fails, it allows us to catch a production issue with a specific part, or sub-assembly, before the rest of The Palette is assembled. This means that we will have a much lower failure rate on the FATs at the end of the process.
Now comes the million dollar question – are we set to ship 40 units at the end of this month?
In all honesty…. The only answer we can give at this point is maybe. We had a team meeting late last week to go over what needs to be done before then, and came to two conclusions.
1) Although we have eaten up all of our buffer time, we may be able to get out 40 units by the end of this month. This will be very difficult to do, given the final optimization on pinging/ponging and the tight quality control we have in place, but if everything goes our way it is within the realm of possibility.
2) If we don’t get all 40 out, we will either ship a partial order, or this order will be pushed to the first/second week of April.
Following the shipment of the first 40 units, the production team will continue to produce Palette’s full time so that we can work to fulfill all remaining orders as soon as possible. We will develop a much better understanding of our maximum output capacity over the next month as we ship the first few dozen units and will follow up with a complete delivery schedule then. We will do everything we can to maximum throughput while ensuring quality remains constant so that we can get all outstanding orders fulfilled as soon as possible.
A quick note on documentation:
The Palette isn’t a 3D printer – it’s a product none of you have used before. Just like 3D printers have little tips and tricks, The Palette does as well. We’re in the process of creating our documentation package to distribute information from our brains to yours.
In this guide we’ll educate you on everything from how to slice a multi-extruder .gcode/.x3g file, how to use our software program, how to setup/run your Palette, as well as some quick fixes that will cover 90% of the issues you may run into.
We were required to heavily modify this document to comply with certification regulations which has slowed the process of getting it out. This guide will be released when the first Palettes leave our facility. Right now, it's being cleaned up, and having visuals added so that everything is clearly communicated to you when you receive your Palette.
To wrap up We’ll be in touch more over the coming weeks with updates on the assembly process, QC processes, as well as a more solidified timeline as we put everything into place. We’re around to answer any questions you might have. If you’re curious about our software and what we’re up to as a company, please check out our update from yesterday – Here.