@akak5128 : Use this as your inspiration, and take the first step... or you could follow my example and end up an old man who is still thinking about it! Seriously tho', if you watch iforce2d's earliest videos you will see that he started with the most basic tools, and progressed via some fascinating projects to this one.
Yea we saw this coming, and even encouraged it recently ;) For your light gantry, if you're not familiar: look into core-xy geometry. BOTH motors are static then and no motor is being moved around at all. The 2 static motors move the carriage with 2 individual belts, usually running on 2 levels (above/below each other). Initially seems strange, but essentially every motor is responsible for a diagonal movement, so if they turn the same way it moves front/back, and if they move oppsite, it moves left/right (or the other way around, not sure). This is usually used for 3D-Printers that need very light heads for exceedingly fast printing speeds (Voron and such). The Z-Motor of course will always be moved around with the head, but there are lighter/smaller ones than what you're using that will do just fine, like a nema 14 'pancake' type (usually round, 85g or so).
Was looking for this comment. I am currently designing my own PnP machine and have gone with core-xy kinematics. Im haping the light gantry is going to fetch me so good picking speeds.
Good to see more Kiwi content. Similar projects but your a big step ahead of me.
4 หลายเดือนก่อน +1
It's unbelievable what you're building again. I'm very excited to see how it continues. But your experimental setup is very promising. I'm looking forward to the sequel... I've seen all your videos several times and it's time to start again because they're just so good! 👍🙂
Watching you use a robot you designed (CNC) to build your next robot (PnP) is both brilliant and fascinating! The test sequence at 13:33 was entertaining, and really demoed repeatability of precession. This current setup looks very solid, and should lead to many more good experiments. A longer version of a rail with your fancy embedded belt might make for good rail launch design for planes. Could program launch force and timing for a precise and repeatable launch profile.
8:20 LinuxCNC!! Love the experimentation on the Z axis. A codestrip could also be used for position. A solenoid/spring/dampner on the Z axis would also be a cool idea, but what you have is good. When I developed my PnP machine, I remember having a lot of issues withe the straightness of the shaft and nozzle as turning it would add a slight offset to the part in placing. But, I didn't use a camera. Can't wait to see how you integrate the camera for part positioning with LinuxCNC. I also worked on a vacuum system where I had a pump only run between a low and high threshold of vacuum in a tank so the pump is not on all the time. You can see how I did that in my videos and how I created the vacuum sensor. It's difficult to source the sensor (mpxv6115vc6u).
I see you used the "O codes" of LinuxCNC, that's only the second time I've seen someone do that (dm17ry being the other). At first I was considering something like that too, but I think I'll end up making a Qt application to bring it all together in a more user-friendly way, that would cover the "integrate the camera" part too. Yes, I'll also be using a large vacuum reservoir so the pump only needs to turn on infrequently, that will probably be a 'dumb' pure hardware pressure switched valve. It will drain into a second smaller reservoir through an electric valve with a faster switching rate, to keep the vacuum behind the nozzle as constant as possible, with feedback and adjustment at the controller. For vacuum sensors I found the XGZP6847D all over the place on AliExpress, they are I2C which I prefer over analog, they can do -100 - 0kpa at 150Hz update rate.
@@iforce2d Yes, I agree. I wish I used an open source pick and place, and even back 12 years ago when I did that project, there was OpenPNP available, if I remember correctly. I will be interesting to see your Qt alternative. Thanks for the part number on the pressure sensor. I see they are only about $5. I may have to change to that in my pressure sensor board. My toughest part of the pick and place was actually film removal. I resorted to just putting weights on the ends of the film. I hope you find a better way than I did.
any time i think about the optical system on a pick and place, i think about an adafruit livestream i watched about 10 years ago, when Phillip described it as taking an upskirt photo of the component
On the first version with the rope use a spring that pushes the carriage down so by by backlash and no weight added except the spring. Edit: no mechanical added so less chance of failure
11:17 This remote motor design is pretty similar to what the 'ArcDroid' Plasma CNC cutters do. They use what appears to be like bicycle brake line & a return spring on the head itself. That may work out better for you since backlash is eliminated, and its one less 'tube and cord' that needs to be routed.
Really great project, im always impressed to see people bring to life the 3D model from CAD to reality. Im thining about the tool changer, if the problem is the backlach of "rope" drivne linear motion, mayeb some kind of sprping that will allways work couter to the motor will help with that? Im really looking foward to see the whole project.
Kinda motivates me to get back to work on my PnP. I assume the noise you heard on that bolt is the cable streching, not slipping. It would obviously have similar results at the Z axis.
I like it! Plz explain how these work together: with 1M/sec, rpu3, linuxcnc, how big is one step? I've used the OPi+2E arisc form of linuxcnc with 1.5mm pitch leadScrews and never got near that velocity. Different hdwr, but huge diff. Thx TomP
I raed the .ini as 27mm.sec anbd * 60 is 1620 mm/min max vel. At 12:52 vid says "1m/s". Maybe 1m/min? Very quick anyway I grabbed a copy of you repo. Thx TomP
It's very unlikely you could get this velocity with lead screws, especially 1.5mm pitch. For 1m/s it would have to spin at 40000rpm which is about twice as fast as a drone motor. Belts allow more travel per motor turn. For example on the Y axis I'm using 40 tooth pulleys, the belt is 2mm pitch per tooth, so one motor revolution moves 80mm, meaning 1m/s is only 750rpm. The .ini file in the github is not relevant to this video. It's probably for my veeery slow solder bot, I can't remember. The relevant .ini settings for this PNP are: SCALE = 20 MAX_VELOCITY = 1000 MAX_ACCELERATION = 4000 The machine units are millimeters, so 'scale' means that 1mm is 20 steps. The other values are per second, so a velocity of 1000 means 1m/s. The acceleration of 4000 mm/s/s means this axis can reach full speed in (roughly) 0.25 seconds. The docs are here: linuxcnc.org/docs/html/config/ini-config.html
I got this from Stepper Online. www.omc-stepperonline.com/dual-shaft-nema-8-hollow-shaft-stepper-motor-bipolar-1-5ncm-2-12oz-in-0-49a-20x20x27mm-8hs12-0494h
I've found a vacuum sensor on the head worked best. I ran a esp32 as a mini wireless gcode server for it and a APM pressure sensor, I'm sure you have a few of them around. Are you sticking with Linux CNC or going to use openpnp? I'm going to retire my modified liteplacer soon and it's got good elm cameras on it, get in touch if you want to do a trade for any bits, I'm in Aus.
My first plan was a vacuum sensor on the head, now I'm thinking maybe somewhere further back might be good enough, but I don't really know how quickly pressure can 'level out' through a thin tube. So looking at your comment I'm curious about whether you tried other locations for the sensor and it was unsatisfactory. Here is what I'm thinking at the moment, my main concern is that the 140cm distance back to the sensor is kinda long, and that sensor might be a bit slow to feel a missed pick etc. iforce2d.net/tmp/pnpVacuumSystem.png The reason for the double reservoir is so that the vacuum behind the nozzle can be kept to a more consistent narrow range, without turning the fridge pump on and off much. It also lets me put the pump and large reservoir in another room, and still have a relatively short air line between the small reservoir and nozzle.
@@iforce2d it's common for the vacuum sensor to be mounted in the control board instead of the head (Look at the Lumen PNP). Also looking at your diagram you can combine the two valves on the head into one 3 way valve that vents to ambient pressure. This can be run on the head or even further back by the control board (again see the Lumen PNP setup).
@jon_raymond Sure, the question is how quickly does it settle over a long pipe. If I build a fast machine with a flyover camera I don't want it to be waiting for a vacuum sensor to settle. About the 3 way valve, eventually I might want an actual blow-off (from a positive pressure reservoir) in which case there will need to be a second valve that's normally closed.
@@iforce2d Valid points. I might mention that generally adding a second/more nozzles is the easiest way to substantially increase a machine's part placement speed.
@@iforce2d The length of the "pipe" doesn't really matter, it's the volume of the pipe. Tiny tubing will react much quicker than larger tubing because it's less volume to pressurize as long as flow doesn't become an issue. My vacuum sensors are mounted on my PCB and I'm sure they'd react quicker if mounted on the head, I'm not sure it's worth the tradeoff of more wiring to the head. Seemed more important to me to mount the solenoids near the head. I use some of that 3mm ID silicone tubing where my nozzles move but I switch to nylon 3mm OD for the run back to keep the inside volume down and it's more durable. I ran out of steam before I really tested the operation, but I've largely settled on how I built it.
The gantry doesn't add much weight versus the extra stiffness it adds. It makes things much easier to align and fasten at the ends, and I would still need a decent chunk of material at each end to mount the motor, pulleys anyway. I like that all the screw holes are machined in a single setup on the CNC, so I know they're very accurate and there will be less risk of racking since it's all one piece. If there was only a linear rail, fixing it at each end and maintaining good alignment seems like a frustrating task. If I want to attach extra bits (lights etc) along X axis, having the gantry will make it easier .
![[TH] VCT Masters Shanghai Playoffs DAY 2 // PRX vs G2 | EDG vs TH](http://i.ytimg.com/vi/T5j2XVcDFs0/mqdefault.jpg)
I have to say over and over again, the stuff we can do today at home, is mindblowing.
and quite inexpensively, too.
@akak5128 : Use this as your inspiration, and take the first step... or you could follow my example and end up an old man who is still thinking about it!
Seriously tho', if you watch iforce2d's earliest videos you will see that he started with the most basic tools, and progressed via some fascinating projects to this one.
So true... 😅
Yea we saw this coming, and even encouraged it recently ;)
For your light gantry, if you're not familiar: look into core-xy geometry. BOTH motors are static then and no motor is being moved around at all. The 2 static motors move the carriage with 2 individual belts, usually running on 2 levels (above/below each other). Initially seems strange, but essentially every motor is responsible for a diagonal movement, so if they turn the same way it moves front/back, and if they move oppsite, it moves left/right (or the other way around, not sure). This is usually used for 3D-Printers that need very light heads for exceedingly fast printing speeds (Voron and such).
The Z-Motor of course will always be moved around with the head, but there are lighter/smaller ones than what you're using that will do just fine, like a nema 14 'pancake' type (usually round, 85g or so).
Was looking for this comment. I am currently designing my own PnP machine and have gone with core-xy kinematics. Im haping the light gantry is going to fetch me so good picking speeds.
Good to see more Kiwi content. Similar projects but your a big step ahead of me.
It's unbelievable what you're building again. I'm very excited to see how it continues. But your experimental setup is very promising. I'm looking forward to the sequel... I've seen all your videos several times and it's time to start again because they're just so good! 👍🙂
I've been looking forward to this video for so long. As always you don't disappoint. Thanks for taking us along your build journey!
Watching you use a robot you designed (CNC) to build your next robot (PnP) is both brilliant and fascinating! The test sequence at 13:33 was entertaining, and really demoed repeatability of precession. This current setup looks very solid, and should lead to many more good experiments.
A longer version of a rail with your fancy embedded belt might make for good rail launch design for planes. Could program launch force and timing for a precise and repeatable launch profile.
Impressive accuracy you have with it, and fast as well.
Can't wait to see this picking.
5:17 ... 🤭 OMGoodness 🤭
Nice one Chris. I was explaining to friend overseas this is Kiwi can do stuff!
8:20 LinuxCNC!! Love the experimentation on the Z axis. A codestrip could also be used for position. A solenoid/spring/dampner on the Z axis would also be a cool idea, but what you have is good. When I developed my PnP machine, I remember having a lot of issues withe the straightness of the shaft and nozzle as turning it would add a slight offset to the part in placing. But, I didn't use a camera. Can't wait to see how you integrate the camera for part positioning with LinuxCNC. I also worked on a vacuum system where I had a pump only run between a low and high threshold of vacuum in a tank so the pump is not on all the time. You can see how I did that in my videos and how I created the vacuum sensor. It's difficult to source the sensor (mpxv6115vc6u).
I see you used the "O codes" of LinuxCNC, that's only the second time I've seen someone do that (dm17ry being the other). At first I was considering something like that too, but I think I'll end up making a Qt application to bring it all together in a more user-friendly way, that would cover the "integrate the camera" part too.
Yes, I'll also be using a large vacuum reservoir so the pump only needs to turn on infrequently, that will probably be a 'dumb' pure hardware pressure switched valve. It will drain into a second smaller reservoir through an electric valve with a faster switching rate, to keep the vacuum behind the nozzle as constant as possible, with feedback and adjustment at the controller.
For vacuum sensors I found the XGZP6847D all over the place on AliExpress, they are I2C which I prefer over analog, they can do -100 - 0kpa at 150Hz update rate.
@@iforce2d Yes, I agree. I wish I used an open source pick and place, and even back 12 years ago when I did that project, there was OpenPNP available, if I remember correctly. I will be interesting to see your Qt alternative. Thanks for the part number on the pressure sensor. I see they are only about $5. I may have to change to that in my pressure sensor board. My toughest part of the pick and place was actually film removal. I resorted to just putting weights on the ends of the film. I hope you find a better way than I did.
👌great Job so far, wish you good luck for the rest 👍
Definitely following this. Liked and subscribed.
The one thing I’ve been desperately waiting for to come from you! Finally it’s here! 🎉
any time i think about the optical system on a pick and place, i think about an adafruit livestream i watched about 10 years ago, when Phillip described it as taking an upskirt photo of the component
On the first version with the rope use a spring that pushes the carriage down so by by backlash and no weight added except the spring.
Edit: no mechanical added so less chance of failure
Thank you for thinking about Headphone users. You've earned my subscription.
To be honest that particular noise might have killed even non-headphone users.
@@iforce2dYeah, me here. My headphone was broken a day before you upload the video, so unfortunately I cant blame you 😂
Very cool build 🙌🏼
11:17 This remote motor design is pretty similar to what the 'ArcDroid' Plasma CNC cutters do. They use what appears to be like bicycle brake line & a return spring on the head itself. That may work out better for you since backlash is eliminated, and its one less 'tube and cord' that needs to be routed.
Really great project, im always impressed to see people bring to life the 3D model from CAD to reality. Im thining about the tool changer, if the problem is the backlach of "rope" drivne linear motion, mayeb some kind of sprping that will allways work couter to the motor will help with that? Im really looking foward to see the whole project.
Great video I’m building one too
Nice with current control over gcode.
Kinda motivates me to get back to work on my PnP. I assume the noise you heard on that bolt is the cable streching, not slipping. It would obviously have similar results at the Z axis.
looking good 👌
ahhh mannn.I dont have room for a pnp. But now i need one in my life.
Very nice!
Yup! Here we go.
Parabéns
Perfect !!!!!
Ohh, fast little robot build have you, nice .. nice, heh heh.
Pneumatic for Z? You will need pneumatic for «twizzers» anyway.
I like it! Plz explain how these work together: with 1M/sec, rpu3, linuxcnc, how big is one step? I've used the OPi+2E arisc form of linuxcnc with 1.5mm pitch leadScrews and never got near that velocity. Different hdwr, but huge diff. Thx TomP
I raed the .ini as 27mm.sec anbd * 60 is 1620 mm/min max vel. At 12:52 vid says "1m/s". Maybe 1m/min? Very quick anyway I grabbed a copy of you repo. Thx TomP
It's very unlikely you could get this velocity with lead screws, especially 1.5mm pitch. For 1m/s it would have to spin at 40000rpm which is about twice as fast as a drone motor. Belts allow more travel per motor turn. For example on the Y axis I'm using 40 tooth pulleys, the belt is 2mm pitch per tooth, so one motor revolution moves 80mm, meaning 1m/s is only 750rpm.
The .ini file in the github is not relevant to this video. It's probably for my veeery slow solder bot, I can't remember.
The relevant .ini settings for this PNP are:
SCALE = 20
MAX_VELOCITY = 1000
MAX_ACCELERATION = 4000
The machine units are millimeters, so 'scale' means that 1mm is 20 steps. The other values are per second, so a velocity of 1000 means 1m/s. The acceleration of 4000 mm/s/s means this axis can reach full speed in (roughly) 0.25 seconds.
The docs are here: linuxcnc.org/docs/html/config/ini-config.html
What NEMA 8 motor are you using I need something with a hollow shaft.
I got this from Stepper Online.
www.omc-stepperonline.com/dual-shaft-nema-8-hollow-shaft-stepper-motor-bipolar-1-5ncm-2-12oz-in-0-49a-20x20x27mm-8hs12-0494h
Waffle on !! 👍🏻😂😂
How about pneumatics for the Z axis?
I've found a vacuum sensor on the head worked best. I ran a esp32 as a mini wireless gcode server for it and a APM pressure sensor, I'm sure you have a few of them around. Are you sticking with Linux CNC or going to use openpnp? I'm going to retire my modified liteplacer soon and it's got good elm cameras on it, get in touch if you want to do a trade for any bits, I'm in Aus.
My first plan was a vacuum sensor on the head, now I'm thinking maybe somewhere further back might be good enough, but I don't really know how quickly pressure can 'level out' through a thin tube. So looking at your comment I'm curious about whether you tried other locations for the sensor and it was unsatisfactory.
Here is what I'm thinking at the moment, my main concern is that the 140cm distance back to the sensor is kinda long, and that sensor might be a bit slow to feel a missed pick etc.
iforce2d.net/tmp/pnpVacuumSystem.png
The reason for the double reservoir is so that the vacuum behind the nozzle can be kept to a more consistent narrow range, without turning the fridge pump on and off much. It also lets me put the pump and large reservoir in another room, and still have a relatively short air line between the small reservoir and nozzle.
@@iforce2d it's common for the vacuum sensor to be mounted in the control board instead of the head (Look at the Lumen PNP). Also looking at your diagram you can combine the two valves on the head into one 3 way valve that vents to ambient pressure. This can be run on the head or even further back by the control board (again see the Lumen PNP setup).
@jon_raymond Sure, the question is how quickly does it settle over a long pipe. If I build a fast machine with a flyover camera I don't want it to be waiting for a vacuum sensor to settle.
About the 3 way valve, eventually I might want an actual blow-off (from a positive pressure reservoir) in which case there will need to be a second valve that's normally closed.
@@iforce2d Valid points. I might mention that generally adding a second/more nozzles is the easiest way to substantially increase a machine's part placement speed.
@@iforce2d The length of the "pipe" doesn't really matter, it's the volume of the pipe. Tiny tubing will react much quicker than larger tubing because it's less volume to pressurize as long as flow doesn't become an issue. My vacuum sensors are mounted on my PCB and I'm sure they'd react quicker if mounted on the head, I'm not sure it's worth the tradeoff of more wiring to the head. Seemed more important to me to mount the solenoids near the head. I use some of that 3mm ID silicone tubing where my nozzles move but I switch to nylon 3mm OD for the run back to keep the inside volume down and it's more durable. I ran out of steam before I really tested the operation, but I've largely settled on how I built it.
what is the name of the program you use on first seconds of video for modelling ?
onshape.com
Have you come across the Opulo open source PNP project?
i need 4x4 ft kit
0:43 did he just say he wants a pick and place pick and place machine 😅?
Why do you even need a gantry for the X axis - why not just use the linear rail? I do this on my 3D printers and it works fine.
The gantry doesn't add much weight versus the extra stiffness it adds. It makes things much easier to align and fasten at the ends, and I would still need a decent chunk of material at each end to mount the motor, pulleys anyway. I like that all the screw holes are machined in a single setup on the CNC, so I know they're very accurate and there will be less risk of racking since it's all one piece. If there was only a linear rail, fixing it at each end and maintaining good alignment seems like a frustrating task. If I want to attach extra bits (lights etc) along X axis, having the gantry will make it easier .