I made a parallel axis tripteron

Within the realm of mechanical design, I usually find that the best way to understand a concept is to apply it myself. But if you're anything like me, you'll know that often the compulsion to understand a concept can outweigh the lack of any real practical application for it. So when I came across this novel conception of a parallel axis trypteron robot, my fate was sealed. Now, I don't want anyone getting the wrong idea, so I'll get this out of the way now. I'm not an engineer, and when it comes to robotics, control theory, and software development, I'm about as qualified as this bag of gravel. My goals on this project are firmly rooted in the cool factor of the final object, rather than any kind of submicron precision, so I'll definitely be turning this bad boy up a notch. But if this concept grabs you like it grabbed me, don't hesitate to check out the description for links to all my sources, as well as the 3D files and code that I hacked together for this project. And feel free to roast every aspect of it in the comments below. So what even is a trypteron? Well, first of all, it's actually a triteron, and that's just Greek for robot with too many parts. Second of all, it's a motion platform with three degrees of freedom that looks way more complicated than it really is. You see, these linkages, or wings, only transmit motion in one direction while allowing free movement in the other two dimensions. With three sets arranged perpendicularly, the head is fully constrained within our current three-dimensional reality, until we decide to move one of those anchor points. Now, it's obviously a bit of a leap from a basic trypteron like this one to the parallel axis variation we'll be slapping together in this video, but it's going to be a lot easier to understand the actual mechanics of why once we've got the thing in front of us. Just ask me how I I decided to make the bulk of the components for my trypteron out of aluminium. That's mainly for ease of machining, but it can't hurt to keep everything fairly lightweight, and rigidity isn't too much of a concern in this particular application. This application being no application. But aluminium isn't all that cheap, and I'd hate to get everything machined, only to find some critical design oversight that renders all of my machining efforts useless, my material scrap, and my life ruined. So to start off with, the easiest way to get it out of the computer and into reality, by far, would be with a 3D printer. Unfortunately, my 3D printer is a steaming hot pile of garbage. Okay, maybe that's a bit harsh. It was only 200 bucks, and I have managed to get a couple of functional prints out of it, but it's hard to get excited about the process, when every print is essentially a coin toss. Right, here we go. Damn it. Oh come on. So, with this fairly complicated project looming over me, I was finally ready to bite the bullet and buy a professional machine, when someone far more versed in the ways of YouTube infamy offered some sage advice. Do not buy one, just email someone and get a free one. So I did, and what do you know, it worked. A few days later, I had a brand new X1 Carbon and AMS unit sitting on my front doorstep. Once I had the printer set up, it wasn't long before I had a full set of components to play around with, which turned out to be a really good thing. I managed to find more than a couple of issues with the first iteration of the design, like this anchor point collision. I won't bore you with all the details, because this is supposed to be a machining channel, and we still haven't done any machining, but I did a lot of printing before I was happy enough to jump into the machining, or at least, jump into the side projects required to jump into the machining. To start off with, the charm of hot chips flying unfettered from the milling machine, and the thrill of staring right into the eyes of death by flying BT-30 collet chuck, has finally just about worn off, so I quickly knocked up an enclosure to keep the bulk of the onslaught contained. I know the execution here is a bit half-baked, but as they say, an enclosure on the mill is worth two in the bush. And look! Sliding doors! The Scaredy Cat screen was paying dividends almost immediately, as I got cracking on amount for my brand new shop-made cycloidal reducer-driven fourth axis, TM. Yeah, this isn't my proudest creation either. But what did you expect from a side project? Plus, we're only going to be machining aluminium, and I'd really like to make sure my brand new shop-made cycloidal reducer-driven fourth axis, TM, is actually worth using before I give it any more of my time. Of course I'm still working on the name, and with that, I was ready to jump into the machining. And when I say machining, of course I mean programming. Because, despite your preconceptions about the wonders of CNC machining, nothing can prepare you for the mountain of work required to craft a program that doesn't want to destroy your part, your machine, or just you personally. And things only get worse when you add a fourth axis into the mix. I decided to start on the linkages, because there's eight of them. So if I mess up the first one, I can chalk it up to a pre-production run experiment. But if it goes right, I can blame it on my unlimited skill. Of course, when dealing with new equipment, you're always going to have to perform pre-production run experiments. And while my findings in this case did indicate a positive outlook, they also highlighted the need for a proper flood coolant system. Well, I say proper flood coolant system, but what I really mean is a mishmash of Tupperware, vinyl tubing, pneumatic push connectors, and an old aquarium pump. It's the same, but different. I swear I'll get around to sealing the enclosure and adding some proper drainage, eventually. So if you like proper drainage, you should like the video. And if you subscribe, I'll know you'll want to see more. With coolant sorted, and the program 80% proven, I finally started to feel the benefits of automation that everyone keeps banging on about. Now, there's no denying that the simplicity of Dan's design is fantastic. But as you might have already spotted, there are a couple of things I've decided to do a little differently. I'll go into a bit more detail on that once we're ready to put everything together. But for now, the only relevant detail is that I might have accidentally, on purpose, overcomplicated things a little bit. The weight reduction slots, angled faces, and compound surface geometries on these parts are almost entirely gratuitous. Mind you, so is almost everything I do around here. But they sure do make for an excellent opportunity to take the brand new shop-made cycloidal reducer driven fourth axis for a spin. I have to say, despite its imperfections, the fourth axis is an absolute game changer. Being able to access almost all the features of a part, like this one, and this one, and that one, and this one, and a single setup, is getting close to effortless. A fifth axis would definitely be better. And of course, no matter how many spinning majigs I manage to compulsively acquire, I still have to load the stock and clean up the nubbins. But if you're looking for truly effortless machining, the sponsor of today's video, PCBWay, is the obvious choice. Because as well as custom PCBs for any occasion, did you know that PCBWay offers an extensive range of CNC machining services? From laser cutting and sheet metal bending, to 3, 4, and 5 axis machining capabilities and more, in pretty much any material you could need. A lot of people commented on my last video wondering why I didn't get PCBWay to make the cycloidal discs for this bad boy. And I'm sure I'll get some more in the same vein on this video, so I'll try to do a better job of explaining. First of all, my vanity. But more importantly, PCBWay's services are there to take care of the jobs that you haven't got the time, equipment, or patience for. I mean, who really wants to spend days cutting dovetails and slots in 4140? For this project, I definitely could have used the budget they provided on some custom PCBs, but I'm fresh out of knowing what the f**k is going on here. So instead, I thought we could use it to celebrate hitting 100,000 subscribers in true non-genier fashion. I'm not saying there's anything wrong with the standard issue silver play button, it's just a little generic and maybe kinda tacky. Now this, made of grade 5 titanium, with a surface roughness of 1.6 microns, is more like it, and it only took 12 days, from placing my order, to holding this shiny, heavier than I expected, object of vanity. Massive thanks to PCBWay for saving me the trouble of sourcing titanium and breaking who knows how many emils. So to find out what PCBWay can do for you, click the link in the description and get a free quote today. And of course, thanks to you, who, for some reason or another, seem incredibly interested in what goes on in this shop. With the flood coolant pumping, and the cycloidal drive cycling, I managed to run almost all of the linkages without incident, though I never could get the chatter on this radius to completely disappear. I guess my stick-out game just isn't as strong as I thought it was. As always, it was the final part, where things went awry. I'm still not entirely sure how or why, but I've got a bunch of theories. Somewhere between the roughing and finishing operations, the y-axis origin had shifted, as though the machine had forgotten where it was. It could have been my poor probing routine, or too limp a grip on the stock, or maybe the entire fourth axis had shifted during the admittedly overzealous roughing. But whatever it was, I'd tried nothing, and I was all out of ideas. So I ran the part again, and got exactly the same result. Well, it was a little better. At least this time, the contour was continuous. And being that I was out of material, and the blemish is nothing more than cosmetic, I decided to move on. With all the linkages ready for op 2, it was time to head to the manual mill to give the unfinished radiuses a tickle. Oh yeah, how rude of me. Please say hello to the newest member of my growing machine family. I actually picked up this milling head at the same time as I got the lathe, with the honest intention to unite the two in holy matrimony. After it had been sitting in a crate doing a grand total of bugger all for the last few months, I finally accepted that that was never going to happen. And if it did, the anguish of having to every time I wanted to use the mill would surely drive me to an early grave. So I called in a favour to help me pick up the matching table, as I should have done in the first place. Of course you should always be careful about the favours you call in, because you never know how you might be paying them back. Then I could finally mount the linkages up on the rotary table, and finish off the nubbins. The parts with the off-centre chamfers also had their through holes and bearing seats a little closer to one side than the other, which definitely caused a bit of confusion. But I decided to address them with a bit of hand filing later, and focus on the parts with reliable datums. This little thing has turned out to be way more capable than I expected it to be. Next up were the cheeks, and by then I was definitely feeling a little more confident. For starters, I'd reached a pretty good recipe of feeds and speeds for this setup, so despite the jankiness of it all, I wasn't quite so nervous watching it run. Also, I'm pretty sure this billet stock was a different alloy, more suited to machining than the square bar I'd used to make the linkages, and the larger cross section of the stock meant the stick out wasn't so much of an issue. Once I had a program I was happy with, all I had to do was jump back into CAM, scale my model to a factor of minus one, invert all of the tool orientation angles, and I had a set of toolpaths perfectly suited to creating a mirrored part. It's simple stuff, really. Unfortunately, I wasn't able to use the CNC to drill the threaded holes in these parts, as they're tilted in two directions. But a simple 3D printed fixture made short work of getting them set up in the mini mill for drilling and tapping, and a second allowed me to clean up the nubs in no time flat. I decided to tackle both of the belt clips at the same time, following the same recipe for roughing and finishing as the previous parts. This time I also managed to fit in a parting operation at the end, so after a quick deburr, all they needed was a swift bit of splitting, and a few threads, and then I could transfer their mounting holes over to the cheeks. The next piece of the puzzle, and the main difference between Dan's Triteron and my own, is what I'm calling the carriage. His choice to use dual rails does a great job of reducing the complexity of the overall system, but it also means that the capacity for y-axis travel is reduced when the other axes are at full extension, and vice versa. Now, arguably the biggest advantage of this motion system is that the y-axis can actually just be as long as you need it to, and if it were the length of a warehouse, a little bit of lost travel at either end isn't going to be the end of the world. But I decided to build the Triteron around this combination V-slot extrusion and linear slide, and look, I'll be the first to admit that the use of V-slot extrusion is a little bit lowbrow, especially within the realms of machinist YouTube. But in this case, being able to route the belts through the slots and hollows, and easily attach the motors and belt tensioners, allowed me to look past my disdain for the material. The thing is, I could only find it in 500mm lengths, so to slip past my slider-sized sourcing sorrows, I designed this part to have the second and third rails separate from the first, with the only downside being this monumental cam program that came along with it. My confidence in the setup so far had been on a steady incline, so I figured it was just about time to give it some more beans. I upped the feed override to 150%, and it was absolutely ripping. Right up until it ripped the workpiece from the chuck. Yeah, I'm not sure what I was thinking either. Grabbing onto only the very tip of the stock, when I could have just flipped the jaws and saved myself a world of hurt. See, my roughing operations were only leaving a fifth of a millimetre, and with no finished surfaces to indicate off of, let alone the patience to dial this part back in, I decided to just have another go. This time, I did flip those jaws around, and I also added something of an index to the base of the stock. That'll hopefully keep the part located, and also speed up the process of getting it set up in the first place. So after mounting the last piece of suitable stock I had left in the workshop, and dialing back the feeds and speeds to a proven level, the part was slowly but surely unveiled. Unfortunately, it soon became clear that there were some pretty obvious inconsistencies in the position of tool paths at different angles. This was the same issue that I'd encountered earlier on the linkages, but this occurrence narrowed things down quite considerably. After the first incident, I made sure to tighten every fastener on the machine I could lay my hands on, and even added some extras. So for the sake of my sanity, we're going to assume that our issue is rooted in either my probing routine, or the positioning accuracy of the cycloidal drive. Now I've got some potential solutions for both of these issues on order, but that's going to be a whole other story. So for now, we're just going to turn this thing up one more notch, run a blend operation with minus 0.2mm stock to lead, and pretend like nothing ever happened. Once that was done, it was back over to the manual mill to touch up the nub, punch some holes, and add some threads. Running the smaller rear spacer went by without a hitch. Then it was on to the final piece of the puzzle, which I'm calling the head. It took me a while to wrap my head around this part, because the only bits of alley I had that were big enough, were only just so. And being parts I scavenged from an existing machine, they already had some rather inconvenient pre-existing holes. I decided I could get around the presence of the holes with some care, but the tiny sliver of excess stock I had to play with, was definitely not enough for the chuck to hang on to. Especially if the goal was to avoid sending a roughing end mill through one, or all of the jaws. And just to be clear, that was definitely the goal. So I decided to make a dovetail fixture. Well that's not gonna work. And then I immediately decided to make a slightly larger dovetail fixture. It has the same indexing features at the base, that we added to that stock before. Minus one for, uh, weight reduction. And this slit, with the help of a couple of cap head screws, allow it to clamp down on a part, once I've added a matching dovetail to the stock, of course. Then it was just the same old story of roughing, followed by some semi-finishing, and then some finishing, with a ball-nosed end mill. Now, the step over on the finishing passes is pretty coarse, but I'm sure you can understand, that with a max spindle speed of 3000 RPM, I was trying to avoid waiting around for days, to get to the end of the program. I'm actually pretty pleased with the final pattern, even if it is a bit cuspy. I cleaned up the remaining faces with the fly cutter, then used a vise, on top of an angle block, on top of a 1-2-3 block, on top of a tilting table, to drill the holes, and add the threads required for the wings to attach. Now, there's a good chance you've already noticed this, but remember when I said, I'd have to take extra care to orient those pre-existing holes? Well, that didn't go too well, but I think I have a solution to cover it up. Hilarious gags aren't the only things, that have been coming off the bed of the X1 Carbon. While I've been busy making chips, it's been working its hands-free magic, on a whole bunch of other little bits and bobs, that we're going to need to put the tripter on together, including the motor mounts, belt tensioner, and a whole bunch of spacers. Just quickly, while Bamboo Lab did send me this machine, and a buttload of filament for free, this isn't a paid sponsorship, so they didn't get to see this video, before it was uploaded. And I was genuinely planning, on buying one of their machines already. Of course, if you're ever putting this much money into anything, you should always do your own research. But if you do decide on a Bamboo Lab printer, or just need to restock your filament, consider using the affiliate link in the description below, to help keep these projects rolling. So with the machining done... What the hell? God damn it. So, with the machining done, and all the parts printed, it was time for assembly. And what better opportunity, to explain the features of my design, than the first time it was ever put together. I definitely wasn't already editing this video, before I got around to planning this sequence. Let alone filming it, and crossing my fingers, that no one would notice the continuity errors. What would make you think that? As I touched on earlier, the whole idea of this carriage component, is to allow the second and third axes, to be stacked on top of the first. Which will allow a more complete range of motion, with the limited main slide I'd chosen. And most of the seemingly over complex features, like this big old pocket, are there to allow me to fill it up, with a whole bunch of pulleys and spacers, so that the belts can get to where they need to go. The rails go on next, retaining those shafts, and connecting the rear pulley mount, to the main carriage. Then I'll add the two sliding wing mounts, along with their matching belt clamps. Hopefully all those pulleys are starting to make sense. Next, I'll attach the carriage to the main slide. But not before pointing out this handy slot, and pair of grub screws underneath. Which will let me attach the third belt. It's going to be a lot easier to install the belts, without the wings and the head in the way. So we'll get the motor mounts and belt tensioner on next. The two motors at the front, bolt on with a spacer, and the third motor goes in one of the rear positions. Dealer's choice here. The shell is mirrored, and the remaining slot will be occupied by our control electronics. By the way, these motors are a bit overkill. But once you've experienced closed loop control, it's hard to want to deal with anything else. And the integrated drivers, will make this whole setup nice and compact. After adding all the required pulleys and shafts, to one half of the shell, I can carefully line up the other half. Then this cover will snap into place, and hold everything together. I followed a similar process for the forward set of pulleys, which I'll be able to slide forward and back, to tension the belts. Threading the belts in, was definitely the fiddliest part of this process. Really love a word with the bloke who designed this thing, right about now. With that out of the way, I could attach all the wings and their requisite thrust bearings. But not before prettying up the bolts a little bit. Finally, I attach the head. And just like that, we've got ourselves a trypter on. Of course, it doesn't actually do anything yet. So it was finally time to do anything but the code. But eventually, I couldn't put it off any longer. Nice. I toyed around with a couple of ideas, for getting this thing to move. But none of them were without their challenges. My first attempt involved starting from scratch. But I ran into brick wall after brick wall, trying to figure out how to move multiple motors simultaneously, and in coordination, while also being able to control acceleration. Which is kind of important, if you don't want your robot to shake itself to pieces. My second approach, and the solution I ultimately settled on, made use of an Arduino, running Gribble? Gribble? GRBL, for motion planning and control. But while the previously mentioned firmware supports a number of kinematic arrangements, unfortunately, the parallel axis trypteron, it's not among them. You see, if we want the trypteron to move in a coordinate system that makes intuitive sense, think, up-down, left-right, front-to-back, for any given movement, we're going to have to carefully coordinate the position and speed of all three motors. Think of a CoreXY machine, but we've gone ahead and stacked a z-axis on top. So, in order to move our head left-to-right, motor one and motor two need to move in opposite directions. To move up and down, motor one and two need to move in the opposite direction to motor three. And moving all three motors in the same direction will move the head forward and back. Now, I know I sounded confident saying that, but what you're not saying is the hours of mental strain, note-taking, and tutorials it took for it to click. And even then, only just enough to narrowly avoid this adventure becoming a tragedy, instead of the comedy we're all hoping for. So now I knew how to modify a standard g-code so it made sense to the trypteron. And I thought about hanging it on the wall as a sort of ridiculous push-button dancing display stand for that gorgeous silver play button we got. But it was pretty boring manually converting the files and it'd need to be plugged into a computer running a g-code sender the whole time, which would be more than a little inconvenient. But if I could solve the portability issue, I figured it'd make a pretty handy piece of camera gear. The answer was to overcome my irrational fear of serial communication and write a little bit of code for a second Arduino to allow some basic functions like jogging, adjusting the feed rate, setting predefined positions, moving between those positions with the help of this tidy looking keypad. Of course, it wasn't a completely smooth path to that outcome. As with any software integration, there were some very interesting quirks. But it works. Most of the time. Once I've added some imperial 4-20 threads, I can power it up using an NPF battery plate with the help of a boost converter. Now I can capture completely unnecessary 3-axis b-roll wherever I go. Who am I kidding? I'm never more than 100 meters from this shed.