As always, it's been a busy month for us - so busy that we haven't had time to get many of our pictures from the old blog back up, but unfortunately that's taken a back seat on production. The bad news is that we're having issues with two parts from our new vendors (we used to CNC them ourselves), but the good news is that we're continuing to improve our in house production and are still producing as much as we can to get ahead.
To be specific, we're having an issue with one of our two grinder wheels, and with our torque sensor shafts. We are hoping to have corrected grinder wheels in house by the end of the week and are in the middle of endurance testing our solution for the torque sensor shafts. Both took a bit of time to solve - the torque sensor was an "in house" fix, and the grinder wheels was mostly our suppliers re making them - and we've been moving forward to produce as many other components as possible in the meantime to minimize any delays.
Based on our discoveries with the torque sensor shaft, we've also held back the ~50 units from our last run and are in the middle of applying the change to all of them so that they're "future proofed" - these are the units we were supposed to ship in April but we'd much rather ensure there won't be any issues with them before we ship them out!
Below, in detail, what went wrong and what our fix is for both, as well as an overview with where we are in production right now.
This one is, to put it mildly, incredibly annoying. We're not 100% sure what happened but the issue has persisted despite two attempts at getting us the right wheel, which has been incredibly stressful for us and is of course equally annoying for all of you!
Our grinder wheel samples came in fantastic, and so we went ahead and ordered a set of "right" and "left" wheels. It's important to note that the wheels are not identical - the wheels are offset so that they mesh together, and of course the teeth have to face "inwards" on both of them. The first set of grinder wheels - the "right" set - came in perfectly. They're nitrited for long lasting protection, black oxide coated to look good, sharp as a tack, and exactly to spec. You'll see how they look in the pictures below.
Then came the "left" set...and they were backwards. Specifically, the teeth zigged when they should have zagged. A picture is worth a thousand words - in the below picture the "green" squares mark the teeth on the right wheel, that go inwards as they should. The blue squares are how the left wheel should progress, so that they mesh in time with the right wheel to grab parts and grind effectively. Instead, the red squares are how the teeth are actually aligned...backwards from how they should be.
We spoke to our supplier and they got right on making another set. In the meantime we were keeping ourselves busy with both the torque sensor (see below) and moving ahead with other sub assemblies. Finally, yesterday, the new set came in. We were, to put it lightly, not impressed...they'd taken longer than expected to remake and are completely unusable for a number of reason. Again, pictures are worth a thousand words:
The new set that came in has the teeth the right way...but it's the wrong colour?! We were pretty confused too, and dropped everything to get to the bottom of things asap.
Here's a close up of the surface finish. It's rough instead of smooth, and the cutting edges of the disc are all roughened - dulling them completely. It turns out that, for a completely unknown reason, they were sand blasted. We don't know why, and have some pretty serious words with our supplier on the matter already. While re-nitriding and polishing the cutting edges could, in theory, salvage these wheels, the problems unfortunately didn't stop there.
Unfortunately, of the 3 random wheels we pulled off to inspect the surface finish, all three of them were physically damaged with multiple burrs. It looks like they've been dropped on each other during packaging, and have been damaged in countless areas all over each wheel. The burr above may look small, but it's *massive* relative to the tolerances the grinder needs to have to reliably cut plastic. This means that putting them together would be extremely difficult, and they would be almost impossible to turn...as they'd be gouging themselves as they spun.
As we mentioned earlier, these all just came in yesterday - already significantly delayed - and they're effectively scrap. The culprit in question seems to be the finishing company, who both failed to finish them correctly and then mishandled them badly. We've already spoken to both our vendor and the finishing company and the good news is that our vendor has been making the wheels since we confirmed the teeth alignment, so they have quite a few lying around waiting to get finished (we normally hold off on shipping inventory as we don't have space to store hundreds of grinder wheels at once!). This means the only delay from this point forward will be finishing the wheels without sand blasting them or damaging them. We may skip the black oxidizing to save a few days and get things moving as quickly as possible, meaning you might get a unit with slightly different coloured grinder wheels. There is no functional or reliability change in doing this as the oxide coating is only for appearance - it's the nitriding that strengthens and protects against rust. We're speaking with the finishing group to see how long the oxide would take and will make a judgement call on whether to skip it for this batch or not based on the time estimate they give us. Without oxiding them, they should ship out to us this Thursday - and we're going to be on site to QC the wheels to ensure that the third time's the charm with the "left" grinder wheel from our new vendors!
Torque sensor alignment / calibration
This problem has taken a little longer to solve as those who've been following closely will know. Calibrating our torque sensors was one of the reasons that our "pass rate" to date has been less than stellar, causing us to hold back more units than we'd like each shipment for further QC and repairs. The bad news is that this got worse with our new torque sensor shafts...but the good news is that we now know the root cause of the problem and will be able to fix it completely going forward in production, which should significantly increase our pass rate and, therefore, our shipment rate.
The issue that we've had for a while is that some units would have below acceptable consistency in the auger pressure, causing pressure spikes / gaps, which directly influences the control algorithm and therefore the filament itself. Before the move, we were brute forcing this to get units out the door - units that failed this QC check would get pulled, the sprockets visually aligned closer to the chain and sensor idler, and then re-tested and shipped out once they passed.
The issue is that with our new torque sensor shafts, the pass rate dropped to the point that it became worth testing extruders before we assembled them...which is never a good thing, but forced us to examine the problem more closely.
Here's the test setup we devised to test units quickly and easily:
It's operation is pretty simple - a large wheel of constant diameter is bolted onto an extruder. It lifts a 15lb steel bar (left over grinder stock, no less) up and then lowers it. The torque sensor reading is monitored for consistency and accuracy. Here's an example of a "pass" and a "fail" - notice the little spikes in the fail in red, which you can see more easily in the magnified, lighter orange line in the same graph:
The problem was that our procedure for aligning the sprockets, chain, and idler was occasionally failing to correct the problem. We developed a new mathematical procedure...and still ran into issues. We machined a jig to exactly mount the sprockets correctly...and still ran into issues.
The issue turned out to be the play in the torque sensor shaft. It's held in place axially against two bronze bushings with a thrust lip, and is a pretty low tolerance part, as the bearing should stay in the chain the whole time. Unfortunately, the new shafts seemed to be just far enough on the "big" end of the tolerance that they could move enough to cause the bearing to pop out of the chain, causing the sensor to spike. The below GIF shows exactly what happens as the shaft can wiggle back and forth...notice that the teeth of the large sprocket and hub of the small sprocket both remain motionless, but the torque arm, idler bearing, and chain all move close to 0.5mm.
The obvious solution here was to shim the shaft so that it couldn't wiggle, so that everything would stay perfectly aligned and the bearing would never slip out. We did just that and, you guessed it, still ran into issues.
the problem turns out to be with the manufacturing tolerance of the idler bearings and chain themselves. The bearings are almost EXACTLY the correct width - the chain, to save cost in the international specification, can vary a decent amount. Some chains will gladly accept the bearing with plenty of wiggle room along their whole length. Others are right on the line, and some of them get too skinny to allow proper clearance such that if the bearing is just off, it will pop out of the chain. See below example where there is no clearance.
So, now we know the real issue - and don't have to worry about units failing our QC check and requiring tedious alignment after the fact. As units get "broken-in" by using them over time the problem would naturally disappear anyways - but we want them to of course work as soon as you open the box. For this reason we've gone through each and every unit we had in the shop - whether boxed up, or still in pieces - and have stripped out the drive section of the extruder to be repaired / upgraded.
What's our solution? Pressing the bearing into a collar that's got a heavy taper and is, of course, a bit narrower to allow sufficient tolerance across the entire spec for our chains. We've mocked a few up in plastic to ensure that the concept is completely sound, and then will start immediately machining collars in house for now to get units re-assembled and out the door as quickly as we can.
This is all happening currently - we should have the first collars machined by the end of tomorrow if all goes well with the testing through this evening. The plastic collars we laser cut to quickly check are actually holding up pretty well as well - but we know that over time the high duty cycle and high load of the collar means it will have to be made out of steel to last against the chain and sprockets. We're hoping that we'll be able to use custom race bearings as well, that have a thinner outer race, which is what we're testing later this evening and into tomorrow morning as it would drastically reduce the overall time to implement our changes.
This isn't our first delay - not even close. But we've learned a thing or two in the past about mitigating them as much as possible, and when we went back to the drawing board on our production and assembly process, we specifically made it flexible to account for problems like this. This means that we've been working on assembly and production full time like we normally would be - we've just been making more of some sub assemblies than others, sort of re-distributing our efforts. This means we have a ton of inventory built up on a lot of things, to minimize any cascading delays that affect customers further down the line.
This means we have units getting constantly moved through torque repair and ready to ship out - here's the next batch ready to be stripped for testing and re-build.
As we wait for the final confirmation on how to correct the torque sensor once and for all, the units queing up are literally taking over every rack and desk space in our office.
We can't even use our meeting room as the table is nearly covered in units - and everyone has a unit or two on their desk. We have a little over 50 units queued up and ready to go out as soon as the torque sensor idler is replaced.
We also have literally hundreds of other sub assemblies ready to go - this shot shows all of our spreaders and side panels ready to go into completed chassis.
We also have 60 UI panels ready to go - we'd make more, but there's no more rack space anywhere since we have so many completed sub assemblies ready for final assembly once we get our grinder wheels in.
Even trivial assemblies like spooler bases and shafts - which only take a few minutes each - are being built up by the literal hundreds. It's easier to make more of these in advance as they don't take up nearly as much space as, say, the UI panels.
All of this means that once we've gotten our new grinder wheels in, we'll be able to produce units far more quickly than normal, which should almost completely make up for the delay as time goes on. We did loose a little bit of time towards R&D and re-organizing the production schedule to mitigate the delay, and of course people who should have gotten their units last month are already a month late - but we're doing everything we can to move forward as quickly as possible and ensure the units we do ship work properly from day 1. We will keep everyone updated on social media and our other channels as the idler and grinder wheel are confirmed, and hopefully we'll be back into the full swing of shipping within the next week or so.
-The ReDeTec Team