Addendum: bunch of folks have (correctly) pointed out that the pulleys/belt path are not parallel and going to cause all kinds of grief. They were right :) Since this video has aired, I've torn apart the mechanism and replaced some of it with ballscrews (for different reasons, but it did fix the belt issue :) ). Whoops!
Your videos have been a great help to me, thanks, keep up the good work:) Many people already mentioned this, but I had the same problem with Mach 3 losing/ignoring z offset and I made it a habit to re-zero the z multiple time to get rid of this "bug". switching to LinuxCNC using a mesa card and linear encoders years ago gave me a peace of mind and accuracy:D
That isogrid looks fantastic. I've newly discovered your channel and feel bad pointing out a problem 🤦♂️, hopefully not too late but have to say there's an issue with your belt layout. The belts need to be parallel to the axes, if not you will find that motion isn't linear throughout the range of travel. Particularly towards the ends of travel where the angle between the belt and the axis becomes greater. It's hard to communicate without a diagram but when the belt and axis form a triangle shape, the angle of that triangle becomes greater towards the end of the axis and for each mm of travel of the belt, the travel of the axis is less and more importantly less than it is in the middle of the travel. I hope that makes sense, it's something I had a problem with on a CoreXY printer build
Thanks! And no worries, you are absolutely correct :) I ran into these issues a little while after publishing the video ☹ I spent some time fiddling and re-adjusting so all the belts were better aligned (parallel and on the same level as much as possible). But there were some fundamental design issues that prevented me from getting it perfectly dialed in. Between that and the extra friction in the linear rails due to bad mounting... I decided to just scrap it and replace the system with a traditional ballscrew setup. Video incoming on that someday once I work up the motivation to finish the electronics. Oh well, lessons learned :) I probably should have spent more time reading up on CoreXY builds first to see what kinds of issues folks have.
Looking good, just tighten those belts! Cartesians are particularly vulnerable to backlash in timing belts. Another trick is to deck off 5-10 thou from the top as your first op; will make sure all vertically machined surfaces are perpendicular to the top face regardless of stock warping. Added bonus: if you have a second op on the backside you have a prepped datum surface
Oh, and if you're fed up with Mach and not averse to compiling control software yourself, PathPilot is GPL and Tormach grudgingly sends the source code for it if you ask nicely ;)
Yeah definitely, those belts are so saggy right now :) The printed tensioner really wasn't sufficient, bottomed out as soon as I started to tighten. Good point regarding decking off the top... I haven't really done that yet and I'm not sure why. Should probably start doing that in general so I have a known surface. Interesting regarding PathPilot! Did they just "skin" LinuxCNC, or would I also be buying into / getting stuck with Tormach-only functionality (e.g. functionality that won't work without also having a Tormach machine)? It looks quite nice so that sounds appealing :)
Linux CNC is like the engine. What people fiddle with, ie the gui, is only loosely connected and you could even run multiple of them at once! So pathpilot is more than just a gui, but you can think of it as a gui that controls linuxcnc Checkout “probe basic” for a path pilot alike clone. In some ways better, but less mature
Uhh... yeah it will! To be honest I'm not sure what I was thinking, now that I'm rewatching. Oops :) Some of the features were quite shallow (backstop for linear rail) so perhaps I was concerned they wouldn't get machined to the proper depth. Dunno! Past-me makes some questionable choices :)
We've been thread milling virtually every hole for years. Taps are never loaded into machines. The cam profile calls for 3 passes on each thread mill. I don't know if that's necessary, but that's what they are all set at. There is considerably less or a retract, and the linear (well sort of) speed is considerably higher.
For a machining a big plate like that one thing you might consider is gluing it to the table. You can use super glue and double sided tape to make something somewhat removable.
Interestingly, the motivation for a launch service provider like SpaceX to make everything as light as possible is not to save on fuel but rather to increase the potential payload mass. The more you can carry the more of the market you can serve. The maximum amount of propellant a rocket can carry is something decided by the size of its tanks rather than the mass of the vehicle (I mean in cases when you already have a rocket which you are modifying, in the initial design phase of course vehicle dry mass matters). You generally don't spec tanks to be much larger than what you would need for a max capacity launch (typically measured by max payload mass to LEO or SSO), and after the tank design is finalized there's no amount of mass you can shave off that will increase you propellant capacity so any future mass reductions are either passed on to payload mass (or to a different subsystem which needs more mass budget). As a related side note, the cost of rocket propellant as compared to the cost of launch operations and the rocket itself is relatively tiny, so paying for the propellant itself is never really a constraining factor unlike with aviation where fuel costs inevitably dwarf everything else.
pretty sure he was talking about payloads of customers being shaved in weight. if you're buying a slot on a rocket with multiple customers in the launch your weight directly effects the price
Servos have encoders internal. If you want additional position information, extract it directly from the mobile platform. The sum of torques from your motor belt tension and the serpentine platform belt will stress your axle bearings considerably. I would consider adding a bearing near the serpentine pulley or moving the motors to the other side. Finally, how stiff is the orientation (angle) of the top strip to in-plane and out-of-plane torques? It seems to me that flex there is what will limit your position accuracy -- especially without direct part position measurement.
I realize it's already been said a couple times I the comments, but linuxcnc is pretty great. You can get up and running with just a parallel port and your stepper driver, essentially the same hardware you have for running mach. I've found it to be much more stable and it's pretty straightforward to set up. Additionally, you have the option to upgrade hardware later on, adding something like a Mesa card if you want to increase your speeds by generating a faster pulse train.
Oh interesting, I didn't realize you could control the stepper drivers directly? I had assumed you needed a mesa card or equivalent. Looks like I need to do some reading, that definitely lowers the barrier to tossing out Mach
@@BreakingTaps yeah, you can use the parallel port to got yourself going. You'll run into limits on the speed of steps you can send out, but you already have that limitation running mach.
Just buy the Ethernet Mesa... not too pricey and easy to swap out computers later. I didn’t need many outputs so just use one of the $5 “5 axis BOB” boards that you can pickup from eBay. Mesa breakouts are way better but started to feel expensive...
Very cool. You may want to watch out, steel core belts can have the steel cores break from fatigue when they repeatedly go around a small pulley. If it suddenly goes mushy, check the belts.
You will want to add some corner brackets to the tall legs. Otherwise it will wobble all over from the inertia. The vertical bolts into the 8020 just aren't stiff. Put the biggest triangle you can fit without interference
I've tried to find a follow up video but I couldn't. did you ever show a video of the finished version (in action) and an analysis on how accurate it got?
Way late to the party - adding on, the x and y idlers aren't in line with the drive pulleys, which can cause a cosine error when the gantry is close to a motor as the belt pattern forms a more and more severe angle that isn't a square 90.
Yep, you are unfortunately correct :( Someone else commented on that right after I published, and I proceeded to smack my head against the desk. Whoops :) I haven't finished this particular project yet, but the axis have since been disassembled and rebuilt with ball screws so it ended up being moot
I imagine that rigidity of the setup is low because of the long belt but could be improved by a wider belt if it’s a problem. Would it be easier to mount linear encoders like a DRO system than the rotary encoders on the shafts, thereby removing errors in belt stretch and flex?
Oh yeah, definitely :) I've since changed this over to ballscrews for a different reason, but I wasn't too worried about rigidity at the time since this is for an essentially non-contact application (not 3D printing though)
the belt ends on the fixed end of the upper bar wont move - no wear to expect there. of course the loose end (the one with the single wheel) does have a relative motion to the bar - so it can be subject of wear if touching the upper bar.
Just to clear up: It's not the fuel cost that makes the difference on rockets but rather the payload increase and thrust to weight ratio (TWR), that's where they save money or more so where they can charge more.
this made me really confused as to how the thing moves. don't get me wrong, the video is awesome. it's just the movement system that is really confusing. It's like a corexy design, but with 1 belt
Question regarding the kinematics. Doesn't turning one motor move the platform at a 45 degree angle? While the other moves at a -45 degree angle? So couldn't you just declare that the diagonals are the x and y axis, thus making each motor independent and simplifying the kinematics?
Do you have enough a height for a chuck for drill bits? I have Taig mill and have a 3/8 arbor that goes in a collet then has a threaded end for a Jacob’s chuck. Makes drilling pretty easy.
Ehhh, it's tight depending on what you're holding the work with. Only 9" travel in Z, so if you're using a vice that runs out pretty quick with a drill and chuck (even stubby drills). Not as much of a problem for sheet material, but definitely not a strong suit of this machine in general.
Do you by any chance have any literature references for such a T-Bot system? I built one for my bachelor's project, but I cannot manage to find a single usable source for this mechanism.
Yeah, I was being a bit loose with terminology. Servo tends to cover a lot of categories: brushless DC, brushless AC, PMSM, etc. Usually associated with an encoder but doesn't necessarily have to be built into the motor unit. Strictly speaking these are "brushless dc motors" right now and will become servos once I add the encoders :)
@@BreakingTaps neat, looking forward to more info about this. I just use steppers for my 3d printer, cnc, etc, but have been curious if servos would be worth the cost.
Yeah I suspect for small stuff like this it doesn't matter much, a stepper would likely do fine. I mostly wanted to play with ODrive to get some experience :) But I have some much larger servos (with integral encoders) that I'm planning to put on my CNC when I find the motivation. That should be a considerable upgrade since it'll really benefit from the properties of a servo and closed loop
Sorry! bunch of folks complained about that (because, yunno, the entire world is metric except us haha), I even did metric in other videos and just forgot this time around :( Definitely on a big post-it note attached to my monitor though for all future videos. Amusingly I mainly think in metric in most places _except_ machining units. Sigh :(
mach3 is only as good as your comports! some comports can lock into a fault state easily; what's the baud rate? any controllers or direct connect to step/dir?
Can confirm Mach is terrible, incredibly Mach4 is even worse, if anyone learns anything watching CNC on youtube ut should be to NEVER use Mach for your CNC control.
![Lil Uzi Vert - Chill Bae [Official Music Video]](http://i.ytimg.com/vi/D7F42F_JM3U/mqdefault.jpg)
Addendum: bunch of folks have (correctly) pointed out that the pulleys/belt path are not parallel and going to cause all kinds of grief. They were right :) Since this video has aired, I've torn apart the mechanism and replaced some of it with ballscrews (for different reasons, but it did fix the belt issue :) ). Whoops!
I would love to see an update on the completed project. What kind of speed can you get with this mechanism? Thank you so much for sharing!
Love seeing H/I/T bot designs, even better to see someone building it from scratch, can't wait to see more!
Your videos have been a great help to me, thanks, keep up the good work:)
Many people already mentioned this, but I had the same problem with Mach 3 losing/ignoring z offset and I made it a habit to re-zero the z multiple time to get rid of this "bug". switching to LinuxCNC using a mesa card and linear encoders years ago gave me a peace of mind and accuracy:D
I've never seen this kind of design! Thank you for the awesome video!
That isogrid looks fantastic. I've newly discovered your channel and feel bad pointing out a problem 🤦♂️, hopefully not too late but have to say there's an issue with your belt layout. The belts need to be parallel to the axes, if not you will find that motion isn't linear throughout the range of travel. Particularly towards the ends of travel where the angle between the belt and the axis becomes greater. It's hard to communicate without a diagram but when the belt and axis form a triangle shape, the angle of that triangle becomes greater towards the end of the axis and for each mm of travel of the belt, the travel of the axis is less and more importantly less than it is in the middle of the travel. I hope that makes sense, it's something I had a problem with on a CoreXY printer build
Thanks! And no worries, you are absolutely correct :) I ran into these issues a little while after publishing the video ☹ I spent some time fiddling and re-adjusting so all the belts were better aligned (parallel and on the same level as much as possible). But there were some fundamental design issues that prevented me from getting it perfectly dialed in. Between that and the extra friction in the linear rails due to bad mounting... I decided to just scrap it and replace the system with a traditional ballscrew setup. Video incoming on that someday once I work up the motivation to finish the electronics. Oh well, lessons learned :) I probably should have spent more time reading up on CoreXY builds first to see what kinds of issues folks have.
@@BreakingTaps Check out the voron 2.4, excellent corexy machine
this guy is very underrated
Looking good, just tighten those belts! Cartesians are particularly vulnerable to backlash in timing belts. Another trick is to deck off 5-10 thou from the top as your first op; will make sure all vertically machined surfaces are perpendicular to the top face regardless of stock warping. Added bonus: if you have a second op on the backside you have a prepped datum surface
Oh, and if you're fed up with Mach and not averse to compiling control software yourself, PathPilot is GPL and Tormach grudgingly sends the source code for it if you ask nicely ;)
Yeah definitely, those belts are so saggy right now :) The printed tensioner really wasn't sufficient, bottomed out as soon as I started to tighten. Good point regarding decking off the top... I haven't really done that yet and I'm not sure why. Should probably start doing that in general so I have a known surface.
Interesting regarding PathPilot! Did they just "skin" LinuxCNC, or would I also be buying into / getting stuck with Tormach-only functionality (e.g. functionality that won't work without also having a Tormach machine)? It looks quite nice so that sounds appealing :)
Linux CNC is like the engine. What people fiddle with, ie the gui, is only loosely connected and you could even run multiple of them at once! So pathpilot is more than just a gui, but you can think of it as a gui that controls linuxcnc
Checkout “probe basic” for a path pilot alike clone. In some ways better, but less mature
this belongs on the list of cool things and channels I found on youtube this year
If you bend a curved surface to be flat when machining, wouldn't it just recurve itself after unclamping?
Uhh... yeah it will! To be honest I'm not sure what I was thinking, now that I'm rewatching. Oops :) Some of the features were quite shallow (backstop for linear rail) so perhaps I was concerned they wouldn't get machined to the proper depth. Dunno! Past-me makes some questionable choices :)
We've been thread milling virtually every hole for years. Taps are never loaded into machines. The cam profile calls for 3 passes on each thread mill. I don't know if that's necessary, but that's what they are all set at. There is considerably less or a retract, and the linear (well sort of) speed is considerably higher.
For a machining a big plate like that one thing you might consider is gluing it to the table. You can use super glue and double sided tape to make something somewhat removable.
Interestingly, the motivation for a launch service provider like SpaceX to make everything as light as possible is not to save on fuel but rather to increase the potential payload mass. The more you can carry the more of the market you can serve. The maximum amount of propellant a rocket can carry is something decided by the size of its tanks rather than the mass of the vehicle (I mean in cases when you already have a rocket which you are modifying, in the initial design phase of course vehicle dry mass matters). You generally don't spec tanks to be much larger than what you would need for a max capacity launch (typically measured by max payload mass to LEO or SSO), and after the tank design is finalized there's no amount of mass you can shave off that will increase you propellant capacity so any future mass reductions are either passed on to payload mass (or to a different subsystem which needs more mass budget). As a related side note, the cost of rocket propellant as compared to the cost of launch operations and the rocket itself is relatively tiny, so paying for the propellant itself is never really a constraining factor unlike with aviation where fuel costs inevitably dwarf everything else.
pretty sure he was talking about payloads of customers being shaved in weight. if you're buying a slot on a rocket with multiple customers in the launch your weight directly effects the price
Servos have encoders internal. If you want additional position information, extract it directly from the mobile platform. The sum of torques from your motor belt tension and the serpentine platform belt will stress your axle bearings considerably. I would consider adding a bearing near the serpentine pulley or moving the motors to the other side. Finally, how stiff is the orientation (angle) of the top strip to in-plane and out-of-plane torques? It seems to me that flex there is what will limit your position accuracy -- especially without direct part position measurement.
I realize it's already been said a couple times I the comments, but linuxcnc is pretty great. You can get up and running with just a parallel port and your stepper driver, essentially the same hardware you have for running mach. I've found it to be much more stable and it's pretty straightforward to set up. Additionally, you have the option to upgrade hardware later on, adding something like a Mesa card if you want to increase your speeds by generating a faster pulse train.
Oh interesting, I didn't realize you could control the stepper drivers directly? I had assumed you needed a mesa card or equivalent. Looks like I need to do some reading, that definitely lowers the barrier to tossing out Mach
@@BreakingTaps yeah, you can use the parallel port to got yourself going. You'll run into limits on the speed of steps you can send out, but you already have that limitation running mach.
Just buy the Ethernet Mesa... not too pricey and easy to swap out computers later. I didn’t need many outputs so just use one of the $5 “5 axis BOB” boards that you can pickup from eBay. Mesa breakouts are way better but started to feel expensive...
Very cool. You may want to watch out, steel core belts can have the steel cores break from fatigue when they repeatedly go around a small pulley. If it suddenly goes mushy, check the belts.
I could see a design like this being very useful for the gantry on a plasma cutter
Really nice camera slider.
You will want to add some corner brackets to the tall legs. Otherwise it will wobble all over from the inertia. The vertical bolts into the 8020 just aren't stiff. Put the biggest triangle you can fit without interference
I've tried to find a follow up video but I couldn't.
did you ever show a video of the finished version (in action) and an analysis on how accurate it got?
LinuxCNC for life! It takes a bit to learn, but once you have you can NC anything.
Way late to the party - adding on, the x and y idlers aren't in line with the drive pulleys, which can cause a cosine error when the gantry is close to a motor as the belt pattern forms a more and more severe angle that isn't a square 90.
Yep, you are unfortunately correct :( Someone else commented on that right after I published, and I proceeded to smack my head against the desk. Whoops :) I haven't finished this particular project yet, but the axis have since been disassembled and rebuilt with ball screws so it ended up being moot
@@BreakingTaps that's one way to solve it! Keep being awesome!
I imagine that rigidity of the setup is low because of the long belt but could be improved by a wider belt if it’s a problem. Would it be easier to mount linear encoders like a DRO system than the rotary encoders on the shafts, thereby removing errors in belt stretch and flex?
Oh yeah, definitely :) I've since changed this over to ballscrews for a different reason, but I wasn't too worried about rigidity at the time since this is for an essentially non-contact application (not 3D printing though)
the belt ends on the fixed end of the upper bar wont move - no wear to expect there.
of course the loose end (the one with the single wheel) does have a relative motion to the bar - so it can be subject of wear if touching the upper bar.
if anyone asks those are vernier holes for that rail. Let's you adjust its offset precisely by choosing which holes to use. :)
Iso grids have angled walls. Its not the triangles. Its the angled walls and there need to be holes in the vertices
Nice video. Mic-6 or Alcoa's original tooling plate is actually a 7000 series aluminum.
Ah that's good to know, didn't realize. Crazy how different it cuts from 7075!
Possibly a scale discretion(XYZ) wise?
Best bet to model is to create one side and then mirror and adjust/double check distances.
Just to clear up: It's not the fuel cost that makes the difference on rockets but rather the payload increase and thrust to weight ratio (TWR), that's where they save money or more so where they can charge more.
Ah yeah, that distinction makes sense. Cheers!
this made me really confused as to how the thing moves. don't get me wrong, the video is awesome. it's just the movement system that is really confusing. It's like a corexy design, but with 1 belt
Question regarding the kinematics. Doesn't turning one motor move the platform at a 45 degree angle? While the other moves at a -45 degree angle? So couldn't you just declare that the diagonals are the x and y axis, thus making each motor independent and simplifying the kinematics?
Yeah I suppose you could!
Do you have enough a height for a chuck for drill bits? I have Taig mill and have a 3/8 arbor that goes in a collet then has a threaded end for a Jacob’s chuck. Makes drilling pretty easy.
Ehhh, it's tight depending on what you're holding the work with. Only 9" travel in Z, so if you're using a vice that runs out pretty quick with a drill and chuck (even stubby drills). Not as much of a problem for sheet material, but definitely not a strong suit of this machine in general.
Do you by any chance have any literature references for such a T-Bot system? I built one for my bachelor's project, but I cannot manage to find a single usable source for this mechanism.
Can this be used for plasma CNC machine? Is it a good configuration for plasma table?
Are you any good with an ' etch a sketch ' ?....cheers.
If it’s a servo motor it should already have an encoder, though, right?
Yeah, I was being a bit loose with terminology. Servo tends to cover a lot of categories: brushless DC, brushless AC, PMSM, etc. Usually associated with an encoder but doesn't necessarily have to be built into the motor unit. Strictly speaking these are "brushless dc motors" right now and will become servos once I add the encoders :)
I should note that these motors do come with hall sensors pre-installed, but those are pretty poor resolution compared to proper encoders :)
@@BreakingTaps neat, looking forward to more info about this. I just use steppers for my 3d printer, cnc, etc, but have been curious if servos would be worth the cost.
Yeah I suspect for small stuff like this it doesn't matter much, a stepper would likely do fine. I mostly wanted to play with ODrive to get some experience :) But I have some much larger servos (with integral encoders) that I'm planning to put on my CNC when I find the motivation. That should be a considerable upgrade since it'll really benefit from the properties of a servo and closed loop
Awesome video. Can you, pls, share what motors did you use?
He did. Pointed it out in the video and it’s in the description.
I see a lot of people just dropping the software that comes with these things and using their own. There's even one written in NodeJS LOL
Is there a more specific term for the belt? because i only find the belts for ones pants on google xD
How did you do the isogrid?
I'm looking for a program to generate them.
Giant etch-a-sketch
I'm a pretty big fan of LinuxCNC…
Is this mill your AvidCNC?
Yep! Benchtop Pro 2x3 model
@@BreakingTaps Thanks!!
T bot was my nickname in college
You gonna put a laser into it and make a kapton tape super duper capacitor?
this is a 6040 cnc? wich spindle do you have on?
AvidCNC Benchtop Pro 2x3, with a CNCDepot S30C spindle. So basically a super-sized 6040. Still aluminum frame, but weighs about 300lbs :)
Nice! Can you also blend in metric units into the video? Hard to follow, if you have to google everything
Sorry! bunch of folks complained about that (because, yunno, the entire world is metric except us haha), I even did metric in other videos and just forgot this time around :( Definitely on a big post-it note attached to my monitor though for all future videos. Amusingly I mainly think in metric in most places _except_ machining units. Sigh :(
I like the ´hidden´ Subsribe button
mach3 is only as good as your comports! some comports can lock into a fault state easily; what's the baud rate? any controllers or direct connect to step/dir?
Could you make some Alu parts Out of an stl-file i would provide? We had to talk about costs, but would it be Generally possible?
Can confirm Mach is terrible, incredibly Mach4 is even worse, if anyone learns anything watching CNC on youtube ut should be to NEVER use Mach for your CNC control.
💯
Real engineers use metric.
Real engineers aren’t affected by the system of measurement.
@@EnlightenedSavage No? They just drive a Martian probe right through the planet.