Thomas, Thank you for building the AR4 and highlighting some of its features. I did want to touch on a few of the comments you made. In the overview section of the manual I do mention pre-applying solder to wires being joined together - I find it easier to pre-apply solder or tinning the ends - and then bring them together with the iron rather than trying to hold the wire ends together and hold the solder and hold iron at the same time, not having 3 hands - it helps to pre apply the solder. None of the wires going to the breakout board call for tinning so I'm not sure about your comments on that? On page 143 I do have some groups of wires soldered together to help group them going to the 3 screw terminals as its a tight fit and difficult to maneuver them into position without them being connected but you could certainly use ferrules in this step if needed. At one point in the past the design did utilize plug connectors and building a wiring harness. Given different applications, motions and articulations all the low cost plugs I sampled proved to fail over time. With this small arm and the limited cable routing and cable arc lengths, direct solder with the solder joints in strategic locations using heat shrink proved to be more reliable, less bulky, less costly and not as time consuming as crimping and inserting numerous plug pins. The primary goal of this project from day one has been to realize a usable robot and at a very particular price point. I've worked with industrial robots my entire career and its always been a frustration that any usable robot was out of reach. There is an absolute threshold on the price as most educators have a budgetary limit so I have made it priority number one to stay under that limit. This drives many of the design aspects, The motors for example are off the shelf variants and cannot be fully customized without adversely affecting the price. There is a limit as to what OMC can customize on these coming from the factory. The machined components are kept to an absolute minimum and as many as possible are cut from flat sheet material. This also drives the bearing designs and tensioning designs as custom machined threads and spindles for locknuts for example is not in the cards at this price point. It also drives some of the belt selections. I did notice the belt tension on your build was far too loose resulting in a good amount of play. Here is a quick video of the forced deflection on my build. (well it looks like I cant post a link in comments but I will post in on my channel or the forum) - Making the statement that "This robot arm is neither accurate nor precise, rigid or smooth" seems a bit harsh considering. The AR4 project is always under improvement, its not perfect but like I mentioned to you in an earlier email the robot is designed for pick and place, welding and handling operations. A great deal of effort has gone into this project, I know you have a great deal of experience and a wealth of knowledge in all manner of machines and mechatronics and your design suggestions are certainly well founded but I wanted to explain these aspects of the project, and hope you will convey these points to your audience. Thank you.
I had tried many different robot arms, but no one beats AR4. Another big advantage is that you can discuss problems directly with Chris Annin. His manual is the best and most detailed I have ever seen for such a product in this price class. I got manufactured custom-made motors at StepperOnline, and I know how many motors needed to order to get them to a fair price. I don't find some of your criticism fair compared to the price for an AR4. Last .. sorry for my bad English, I'm from Denmark 🇩🇰 😊
Thanks, now I understand why there is no plug connectors: Quote: "At one point in the past the design did utilize plug connectors and building a wiring harness. Given different applications, motions and articulations all the low cost plugs I sampled proved to fail over time. "
This is the first I’ve heard of this arm, its incredible! I’m setting up a workstation for my robotics projects later next month, I wish I could set up the arm to keep my tools organized! It’d be a fun project
4:29 The reason why a tinned stranded wire will fail in a screw terminal is twofold: 1. Tinning a stranded wire will introduce a weak point where the solder wicks into the wire. If the wire doesn't move, this isn't a problem. But if it's part of a moving system, that wire is now much more likely to break where the solder wicked up. Now, I'm told you _can_ get away with tinning as little of the wire as possible, but I've never seen anyone pull that off. 2. Solder is prone to "creeping" under pressure, it's not really meant to hold its shape when it's squeezed. Over time, the soldered wire will slowly deform, loosening the connection with the wire, meaning you have to go back and tighten the screws. I've actually encountered this after about 2 months, and it is a huge pain. Better to crimp on a ferrule to each of those wires and never have to worry about them until something happens on the other end.
Instead of ferrules I like to solder male breadboard headers to the end of wires if they're going to be screwed into terminals. I think ferrules are probably the better solution for most cases but I don't have a crimp tool and many ferrules are way too thick to fit into very small terminals
@@catdisc5304 ferrules are sized based on the wire diameter, if a wire with ferrule is too large to fit in a screw terminal there is probably a good chance the wire is too thick anyway and the wire can handle much more current than the screw terminal. Ferrule crimping tools are very cheap and simple to use, much simpler than other connector types.
Not sure if I missed the price, but a quick look at the link says its about $1200 for the kit, and $800 for the motors. Not bad all things considering, could certainly help a mom & pop shop get some basic automation going
what kind of automatization can you achieve with this accuracy and stability? Plus all the effort of building it, which is almost impossible for a normal person who is not a tinkerer.
Ive been looking at cobots for my shop, and was quoted about 4K for one with the full setup and service. I keep looking at these projects, and almost every one of them ends up on github, and that to me is like water skiing on the moon.
Having been a developer for a big robot company in my past, that rail design was awesome game changer for us. There are still zero locations, but yes, they are reduced. Next big change for us was 6 DOF Force sensor with a 3D Laser camera...Best job EVER!
I was surprised to see that is the official way to do it, it is well known now with 3D printer failures that you do not tin wires that go into screw terminals.
@@conorstewart2214 also there are 2 types of screw terminals. One have additional sheet metal inside terminal to spread force over whole connector, it also acts as a spring so this connector have harder time becoming loose. And there is a second one, where wire is kept in place by screw directly. Second one can be used only with solid wires and ferrules, as stranded wires are simply getting crushed and cut inside by rotating screw.
I find it's more secure to just ditch the ferrule personally, if you don't use the right one and it's not properly crimped, the wire can slide out while the ferrule is left behind in the screw terminal... Ferrules aren't used for 230V wiring in houses and stuff either, and they will never budge
@@PureRushXevusyou do not use ferrules on solid wire. Their main purpose is to hold the ends of stranded wire together so the current flows through all strands not just those few the screw terminal caught. For the very small wire made of delicate strands the ferrules provide protection from the screw damaging and literally cutting off some strands. Thick solid wire doesn’t need ferrules. They also provide some strain relief for thin wires.
Nope, not one. You probably need multiple. For example to make "helicopterview" shots of parts while they are almost breaking in your tensile stress machine. And also one for your promo shots. One to feed your filament recycler and most importantly one for pooring a cup of coffee or tea during your recordings.
Axis 2 and 6 are designed OK. Other axes makes me say "what was he thinking?". Trapezoidal belt for axis 1, 3 and 4. Leadscrew on axis 5... Yes, it moves center of masses down and reduces required torque, but it would be much better to grab beafier gearboxes with higher reduction rate and drive joints directly. As for electronics, he could install TMC5160-BOB in the back of each motor, connect all of them via SPI, and control everything directly from RaspberryPi.
@@ferriswhitehouse1476 Not much. TMC5160 costs as much as DM542, but much more advanced and can correct step loss. And cost could be cut with other changes like dirt cheap magnetic encoders instead of expensive optical ones and only one very simple and cheap MCU would be required. Also wiring would be simplified - TMC5160 can be daisy-chained via UART, so only 1 UART wire and 2 Power wires would be required.
@@TheNamelessOne12357 okay sure for the electronics ill agree. It was released a few years ago so maybe those options weren't available, I'm not sure. Maybe a user could do those upgrades on their own. For the mechanical aspect though I would imagine its a pretty big cost jump. If you put more weight up on the ends of the arm then everything below it needs to be beefier as well. So upgrading all gearboxes and motors and frame components.
@TheNamelessOne12357 Chis is a Robotics Engineer by day job. The AR robots he designed for free basically, to be as cheap as possible. I asked him about the gear ratios and he said he made it this way so its speed is useful. If you look at the arduino/teensy code, it's quite advanced now. The windows app is just a UI to interface with the microcontroller that does all the IK calculations like ATAN2 etc, and synchronisation of joint moves. The serial on the microcontroller is agnostic to the UI platform. It could be Android, pi, windows. It's just the UI was designed for windows. -Ken
I've been watching Chris's channel for a few years. I love that his work is getting more mainstream traction! I've been wanting to make an arm for myself since I've seen his channel. His videos are amazing to help understand how things work!
Oh boy that backlash. I guess I'll wait for AR5. Feedback: Hall effect sensors are more accurate than limit switches. There are even hall effect encoders to check the actuall position.
They could, especially if they are machining all those parts anyway, even just small engravings on the surface, like a line on each part, just match the lines, would be good. However this likely isn’t a priority, the machine has a lot of other mechanical issues that would be more important to solve than making it easier to assemble.
Take a look at the Parol 6. It’s more expensive at around $2300 in kit form but the movement repeatability tests I’ve seen look impressive and it’s open source.
Hi, robotic engineer here. I made this robot a while ago when he was at the AR3 version, this made me learn how important precision machining and zero backlash gearbox is important for robots. Also his kinematic his questionable. While it does work, you need to develop your kinematic with singularities in mind and all the possible configurations of the robot for that position. Instead he limited the joint limit values to avoid having more than one configuration, but this is not optimal, dangerous even.
@@tec4303 I'm no expert, but I would assume either failure of a joint from higher force loading than anticipated or risk of toolpath collisions from having to work around over-constrained IK paths.
@@tec4303 even with limited joints range, its possible to have multiples configuration in certain positions which can lead to the robot changing configuration mid-path. A linear movement (MoveL) is usually made from multiple targets very close to each other along a line, if the robot calculate that there is more than one solution to reach these targets, it has to choose which one to use and stick to it for every targets following that line, or else the robot might twist its wrist 360 degree suddently at maximum speed to keep the linear speed and that is dangerous.
I found this project some years ago, and have been interested in printing it for a while. but there are some things that's been in the way (except cost), and one is that i would love to see it using Slip Ring connectors instead to be able to move freely without risking any cables getting tangled. But awesome build and nice to see the process!
A point on the software, what you are provided with looks to be on pair with a pretty standard industrial robot arm controller system. Which is great, and it gives all the function you need to make it move. However, what you need to learn is less about specifically this software, or any industrial robot's software, but the systematic knowledge of robot. More specifically to the problems you get at the end where robot doesn't move in the right way, it will all make so much sense after you learn the basics of robot kinematics and trajectory planning.
Me: I have a busy day job and I just started a 3D printing side business. I don't have the budget, shop space, or the time for any distractions. Tomas: Want to build a kit robot? Me: Yes!
"Doesn't have those cheap R/C servo motors..." Needs an end-affector, installs a cheap R/C servo motor.... lol But I do appreciate the honest review. Robotics is hard. Chris does amazing work, truly. The AR4 is a great start but still needs refinement. The software is probably its best feature. It's very capable. I look forward to more robot reviews! :)
did watch all the videos from the creator.. awesome guy, awesome robot! not an easy task to get a kit out for 2k$ and write the software and give it for free... great guy!!!
Very, very interesting at a reasonable price. I hope this is successful enough to lead to a major revision that can address the issues you noted with repeatability, precision and rigidity because I'd be quite interested even at a slightly higher price.
You could try addingfriction to the axes to damp out vibration. Either using a much thicker grease, or 3d printed brake pads against the cylindrical parts
The Ender 3 of robot arms! Been following the project a while but can’t quite dedicate the studio space to it yet. I’d love to see if you could make it sort and inventory small parts in your studio with the vision capabilities (or barcodes)
I’d love to see you try and do some welding with this! (Background: I took a year of welding at a local community college but my hands are too shakey (damn essential tremor!) to get clean welds. I have all sorts of project ideas though, and it would be great to see how well this robot fairs in the task!)
I've been thinking about building a robot arm. Basically a "3rd hand" setup. Push a button puts it in free move mode. Then push the button again and it locks in place. The most generic example is soldering. Position the arm to hold the wire in place and then I can solder it together.
This is a cool idea. I could see something like this but severely simplified. The switch could be on the tool holder which would dis/engage three (potentially one) servo which locks the arm in place. A hefty servo, like the 25kg in this video, could provide the force to prevent a ball joint from slipping. A momentary button could be used for a squeeze, move, release operation.
Me too! I have also been trying to come up with a simple work light. As soon as you touch the light the brakes release... move it to a new location, release it... the brakes lock. Yeah... I know... just use an anglepoise lamp, and get over it! 😜
Seeing it all assembled and tested it doesn’t look as impressive as it initially did. That is a huge amount of wobble and flex in it. That 2 kg payload rating is very generous, if it wobbles that much lifting a small camera how is it going to handle 2 kg? 2 kg may be the theoretical maximum based on motor torque but usable payload capacity seems to be much lower. It is still impressive being the work of one guy but in terms of functionality and practicality it isn’t as impressive. The motors having encoders but not being able to recover from skipped steps is odd, why isn’t even basic closed loop control added? Based on the capabilities of the robot the teensy likely isn’t being stressed much, it should be more than capable of running closed loop control on the motors. Maybe this is planned for a future update?
Great video. I am too apprehensive to buy this kit until I build my skill sets. However, I appreciate the detailed walk through of the build. I am interested in the kit and can see that years of thought and skill went into building it.
Awesome, I've been following Chris Annin for around seven years already. Still never got into making the arm since budgetwise I don't have a reason to buy it, but I love his work and this robot arm should really become more popular so maybe it will develop further and further into a really good and solid product. I think this is the best robot arm you can get as a hobbyist. Maybe one day, when I have a disposable income, I will definitely get one. Very happy to see you build one Thomas! :)
This is the only usable DIY robot as far as i could see on the net, and the designer is very good at it. Being open source is just a bonus, and the reasonable part prices are the cherry on top. No, i do not have anything to do with it/them, and i have not built one yet but i will. I work with KUKA robots...
I wonder whether you could add some type of physical damper between the arm and the camera. Two examples which come to mind are the suspended shock mount for microphones and sorbothane washers.
For those rising screw terminations, I've had _much_ success wrapping strand conductor around the insulation and treating the conductor as a ferrule. Torque it down until you can't anymore - and then some depending on hand strength, you'll have long-lasting connections. Works damn fine and all it costs is time.
very interesting, as it happens I'm currently building a large custom robot arm for my camera setup as an upgrade to my previous. Still a ways to go before I find out if I've made it nearly rigid enough. But I'm actually glad to see that these small kits are not perfect, at least I can tell myself it might be worth the effort ;-)
for camera moves, could you make the shot take twice as long, hopefully reducing shake, hopefully providing more data for stabilization to work with, and then play it back at double speed?
I don't know if I missed it in the video but does the tapered roller bearings not get any preload? I'm pretty sure tapered roller bearings are always supposed to have a preload applied to them.
I think that was what Tom was referring to at 12:55, when he pointed out the tapered roller bearings being preloaded by a flat thrust bearing. I took it that he thought there should be opposed tapered roller bearings, with preload, which would give better axial control.
it would be interesting to get the robotic arm to print something. How about something simple, like an object in vase mode? it's obvious that at the amplitude of the backlash, there's no way you can produce decent prints, but I think in vase mode, the robot could manage.
Great Video about a very lovely Project! I 've built a BCN3D Moveo back in 2018 and used Chris' Wiring diagram and software to get it to move. He helped me getting the Arm set up in the software and i had real fun building and playing around with this thing... That was my first big project which lead to me taking off in the DIY/Maker kind of stuff.
Looks very nice. Metal is just better than plastic for mechanics. If we could get low cost metal 3D printers I think they would be great for robot parts and both laser light and iron powder can be very inexpensive. There really is no reason for robots to be expensive.
This is very cool. Not the best use case, but as a demonstration of its motion control, could you fit a hot end and print a benchy? Would require post-processing to convert the GCODE to whatever this runs.
The software can also read gcode now. One could wire the extruder motor as J7. The nozzle should better be as large as possible to accomodate for the comparatively low accuracy.
Cool project. Really interested in what uses besides camera movement you can up with. I've felt like doing a project like that myself, but space is limited and i struggle to justify it aside from the fun of building it.
Forgot which channel, but there was someone who used an old German car manufacturer robot arm and got to work in a woodworking place and they had put an router on it. They did some cool stuff with that, though that has been some years ago. ( maybe they did some metal work as well)
Yes, but if you read the pinned comment at the top here, the designer states that TS's belt tensions were way too low and that contributed significantly to the observed overall wobble. Would be nice to see a followup showing the change, if any, after increasing belt tension as the designer suggested in his comment.
having a robotic arm is one thing, it's very rare to see the common man do anything useful with it, programming the desired movements is the difficult part. Ive done it for some applications, linear tracks on the base def makes many tasks much easier. Think hard about what you would actually need a robotic arm for, before you go spend time on it.Random repetative tasks around the shop are much more difficult to make worthwhile than very long production runs. (vs set up time)
for use as a camera robot: You could use a camera with manual focus and control the focus ring with a servo. That would add some cool depth of field effects in your camera movements. Also how fast is the arm? Because in film studios they use these kind of tools for fast movement during slomo shots
Excited to see if this inspires other designers or companies to make a competitor I think there's too many caveats for me personally to put $2k into AR4 but maybe AR5. I'd really love to slap something like this on rails or maybe even treads and have it do part removal or build plate swaps for print farm automation. The off the shelf print farm automation like Mosaic Array are way outside my price range.
Love the idea of this, but as the video shows and a lot of the comments talk about, this one project only introduces more projects to do, such as trying to fix or mitigate the backlash, making the whole arm more rigid, software improvements, etc.
Good "review" of this kit. Even with restrictions I think for this money you get an awsome kit. @12:33 you say the motors don't meet the backlash of 0.25 degrees but you measure the complete robot, not just the motor, so that means you also measure any play in the bearings. I use a Kuka KR180 and with that one, motions are also not always smooth. Depending on the load (I have a 30 kg spindle motor mounted as end effector) I can see some jitter during moves and expecially when the brakes are released from the motors there is some movement to compensate for changing load on the motors. The AR4 is still an amazing development and it does have some nice uses for parts picking applications
I would love to see a smooth camera arm (this but more refined) and a buildplate shelf system where the arm gets commands to put in a finished buildplate in a shelf system and it knows how tall you're prints are in order to avoid crashing in to already placed buildplates.
Thanks for the video of this awesome project! 👌🏼 So if the motors got encoders is it possible then to make a manual (by pushing the arm) movement and let the robot repeat it? Or even edit the spline to make it smoother? 🤔
Not really. The robot mechanics attached to the gearbox would be overloaded and fail before sufficient torque is achieved to back drive those gearboxes.
I always wanted a robotic arm for camera shots - I still do, but the footage you got is still too shaky. I am glad you acknowledged it. I am using a MUCH simpler rig for circular shots and even mine has some shakiness too. Nice project, Thomas! Loved watching it! EDIT: 15:24 GEEEZ! Ha ha ha, I thought what I was seeing BEFORE was the raw footage! Naahhhh, too expensive, too shaky. I need to figure out something different 🤣
I use my AR4 constantly and find that crimping on a ferrule is less reliable then properly tinning and then soldering the wires. However, in the robot's electrical box, it is better to crimp the wires to a ring connecter, then attach and secure around the screw on the terminal block. TLDR: use tin and solder around moving parts, crimp wires around non moving areas
@@Estop117 I test the robot in commercial kitchens. The robot is good for small automation. Take that and upgrade the smaller motors to high torque steppers and you have a robot that can move fry baskets and other heavy objects
A pair of opposed Angular Contact Bearings - the kind used in bicycle necks would still be overkill, but could replace those massive conical roller bearings in each joint. Cheaper, lighter, easier to design for. Fit for purpose.
Oooh I'd love to see Jeremy Fielding build one of these and see what his opinion is vs the one he built from surplus parts. Granted, Jeremy's is massive in comparison, probably weighs as much as a small milling machine, and probably cost 10x as much but still, I'm curious what he would think.
You know, for the price point, and being a one-man operation, this isn't too shabby. It also makes a good starting point. All the parts are there, even if they aren't the best for function. You could probably retrofit some more precise motors, better gears and belts, and rebuilt joint bearings, to get a more precise machine.
It's actually exceptional what he's pulled off. To reduce deflection and backlash to the degree people would want it would easily push this build north of 10k. Even industrial 6 axis robots have these same issues that sometimes have to be worked around especially in vision guided applications.
Knowing where those motors actually come from, the MOQ and NRE for custom wire lengths require a commitment I doubt Annin Robotics or SteppersOnline are willing to commit to.
I can solve your wire extension issues; Wago. They make inline splices for their 221-style lever nuts and I am _certain_ there are plugs you can just fasten on without soldering meant for industrial applications, _also_ in the 221-style.
You could mount some really long rods to the back of your table top for the robot arm to move all the way back and forth on for some killer shots...like face tracking for instance.
haven't built mine yet, but now it looks more complicated than I've imageined. In my application, the repeatability and shaking is not that big of a deal. WOuld love to see more videos about the Arm from you
correct me if I'm wrong but I though the black stepper drivers (seen at 3:31) can set to try to correcting missed steps a few times before sending an error output. (I was wrong there is only one input for step instructions) but you can get self correcting drivers and set them to correct missed steps.
Heh... 😏 I always thought these DIY Robot Arms could only really serve as Camera Dollys but apparently even that isn't a viable Project 🤣 So, in the end it leaves them only suitable as a physical Training Platform? How much use could one get out of them for that purpose considering the real Deals usually run on proprietary Hard / Software? 🤔
it fell like it will be less hustle to design your own , there is probably patents on proper mechanical solutions by manufacturers of those expensive ones, I wonder when do they run out !
I have built a few robot arms over the many years, but I have yet to see an actual use for them other then to learn and show off. Please inform me what actual home use would this arm do to make it worth the investment. I don't make youtube videos for a living so a camera pod is useless to me.