The little compliant clamp spring is so cool! You could probably make the cam spring part of the clamp and make the whole thing one compliant mechanism by having a spiral spring going off of the servo to the middle of the clamp part :D awesome video as always!
For an idea of sensing where the board is, you can use a photosensor/thru beam to detect the incoming edge from there you are just counting steps until the trailing edge passes. You will know exactly how long the board is and where it is at so you can place it in the middle of the machine
Make a notch in the rail and two vertical slits. So the clamp has two guides to slide up and down. And clamped down it sits in the notch to be flush with the top of the rail. So no additional height at all. Great solution. Love it. Looks super cool. Greetings from Munich, Tom.
As for the side "thingies" holding the clamp - intead of putting the glue, you can model them to have C-like form hanging over the aluminium L and you can fix them fron the bottom with the screw againt the bottom part of Aluminium L profile. Also, to improve the guiding of the claming part you can use some kinf of tongue-groove design :) Anyway - great work *thumbs up*
Honestly, I would have *two* pieces of angle, both pointed *down*, with the board and belt sandwiched between them. During movement, the top piece is held clear of the board. When clamping, the top piece is pulled down onto the board. That way, *all* boards in the system are held firmly in place at once. As a bonus, the working clearance is literally the thickness of the aluminum of the top angle. Another option, more complicated to build, would be to have a single piece of angle on either side, pointing up but the sides pitched in, clamping boards by sliding one angle toward the other so that the edges of the board are being clamped. The pitch on the angle would vector the horizontal force to point in *and* down, holding the board firmly in place. Both options can be powered from a single servo that pulls one end of a belt to tighten the clamp. A spring on the other end of the belt provides some give to the system. Both options would use springs to hold them open. Does that make sense?
I work at an electronics factory and work with SMT machines. The positioning problems with loading in boards is generally solved by using a stopper that is swung up from below. The conveyor will then run, pulling the board along and the stopper will prevent the board from going too far. When ejecting the board, the stopper swings down and the conveyor runs until the board is out.
About adding a sensor to detect when the PCB is fed into position: a simple optical gate would do - shine a LED down from above the board (a little finger holding an IR LED could poke out over the board) to a photosensor below the rail. Then you could write some code to pull in a PCB, detect when it's passing into the work area, and stop the feed the instant it clears the sensor at a known point so the machine can "know" where the leftmost edge of the board is. Should also make it far easier to pick up on fiducials for final board locating if the machine has a reference point for the left edge of its work area. A second sensor at the exit end could be used to let the machine know when the current board has cleared the work area and it's safe to pull the next one in.
The first idea that came to mind, squeeze the rail profile that the pcb is up on together. that is, have a stepper motor that moves one rail closer to the other. and thus holds the pcb with friction then you do not need to have anything over the rail at all. Anyway just an idea, I'm sure your solution is just as god.
Three gotchas with compliant printed mechanisms are hysteresis, creep, and fatigue. I doubt any of those will be a problem with this mechanism (if made in something kinda springy like PETG or Nylon) but if they are you could make it so the servo arm instead holds a little piece of spring steel shim stock as the blade flexure and then you’ll have practically infinite number of cycles, no creep issues, and uniform clamping pressure over years of usage.
I second this, with some minor changes. A solid cam on the servo and two pieces of stock spring steel in the clamp, one driving down and one driving up should remove the fatigue issue. Granted there will be wear but so little and the cam is a small part to replace. Or even a single spring holding it open with a cam drive to close it.
The spring cam design is a super excellent idea. As has been mentioned by another person, the repeated flexing may create fatigue or creep in the spring, so it might be good to have it printed in PETG or better yet Nylon on a production version. The difficulty with tightening the belts is ever present... I highly recommend having an idler that moves outwards, and thus pulls the belt taut, via the tightening of a screw. It's what I've done on all my 3D printers because it lets you easily adjust for the belt stretching out over time, and means you don't have to work super hard to pop it on there in the first place. Anyway, excellent channel. I discovered it via the algorithm from your previous video and binged the whole thing. Great content and, continuing onwards, good luck with the project! :)
That Cam sprung shape is great! A bumper that rises and lowers similar to the clamps could serve as home position; rise bumper, convey in until board detected, clamp, lower bumper, run placement, unclamp, convey out.
I still love your presentation style and excitement. It's so encouraging. Keep up the great work. Looking forward to seeing a actual full panel placed from your feeders.
If you ever need a way to roughly locate the board on the conveyor before using the camera I propose this: A mechanical switch or an optical sensor built in your clamp that can register holes on the sides of the pcb panel. Thinking about it, even a set of metal "brushes" riding on some exposed tracks could do the trick
Congrats on an elegant solution for the clamps (A spring loaded cam body). In looking at what you've done I think the top of the clamp needs to be closer to rail height. Possibly notching the side rail a bit with and the clamp cad adjusted to make it closer to rail height so it would not stick up so much in the closed position. The clamp open position is a never-mind because the p-n-p head is parked during board movement. I'm also wondering if a mild cad rework to raise the gantry rail or extend the z movement amount might be a thought. Good work. Your videos are, as always, an entertaining presentation of your engineering process.
Stephen: A couple of ideas. The aluminium brackets are 90 degrees. How about bending them in a bit, to maybe lets say 80 degrees, perhaps cut them down a bit so they are not so tall, and then clamp the board by moving one of the rails closer to the other, pinching the board side on between the rails? That way, nothing sticks above the rail brackets at all. As for the converyer system, why not just have multiple belts to catch a bord from the next/previous. Should work fine as long as the belts move more or less in unison. Added benefit is that you can put some space in between one belt and the next, for a led and a phototransistor combo to detect when the board passes between, typically at the entry and exit points of the belt inside the pnp.
Hi, this might work for you. I was thinking I would try using the clamping idea for a positioning stop as well. But make a smaller one and flip the cam 180 degrees. Maybe incorporated it into the same clamping body. The pcb board would come up and hit the edge of the stop that is closed, one on each side of the rail. Then you could close the clamps and raise the stop. After the pcb board passes by you can then close the stop again to stop the next pcb board. I had a similar stop design on a Bosch conveyor. The Boche stop dropped from Below to allow a pallet to pass. Enjoying watching your project.
Since you don't need them to move independently you may be able to just use one port on the board for both servos? Nice solution going over the rail with the clamping piece, very low profile.
You could extend the belt mechanism to outside of of the main frame. If you design printable pieces on the end that accommodate for the pulley, no hacksawing the metal brackets is needed. Nice vertical tensioning system. You could use some pieces of gt2 belt on the moving part to even make it grip even more so you can reduce the length of the mechanism. Nice work, as usual. 👍
Is having the PCB riding on top of the rails going to mean that the thickness of different PCBs now needs to be taken into account for placing components? Really great project with lots of interesting engineering, I hope you make something from all this work.
Pcb thickness is a usual variable to enter on a pick and place if it is not using a top clamping system. Obviously a lift clamping system does not require this but I see both methods in actual industrial assembly systems.
Now double up the servos (all parallel, moving at the same time) and make them move another set of shorter L profiles from the ends that rest above the belt for moving and clamp down on the whole length. This moves the clamping mechanism out of the "hot zone" of the machine to the sides, allowing for more clearance.
You already have a way to position the PCB...move your camera into a position mid way between the rails and closer to the exit end. Using OpenCV have it move the PCB until it reaches the centerline of the camera field of view then stop moving forward and use your clamps to hold it in position ready to place parts! Nice job by the way on your clamp.
maybe optical sensors for simple board edge. Locating just one on the input combine with the stepper data you could find length and you would know where both edges are within range to find your fiducials. Looking awesome so far for sure. Keep up the amazing work its Turing out really awesome.
Nice work ! Maybe you could use a flexion sensor like the FS2L055 to detect when the board is passing the clamping system on the conveyor belt. With the size of the board as a parameter you can then know when to clamp on the middle of it. If it too complicated, Time of Light sensor ? VL53L0X from STMicroelectronics. Or more industrial one : GTB6 from Sick (you can detect the thickness of a paper with this so maybe it will be overkill).
Put a slight chamfered edge on the clamp lead-in so it can't catch on the square leading edge of the board if you get a warped board. Nice Job on the design. I'm following this closely and subscribe to the Git. I have too many projects, right now, to start an Index but I plan on jumping into it in the not too distant future. BTW, thanks for the incentive to switch to FreeCad. I was just gearing up in 360 but was not a fan of their new "screw the little guy" attitude. I can't afford their expensive subscription service so I was looking for a good alternative. You came through at just the right moment.
I think using the top vision camera would be a great way to make sure the board gets into position correctly. The top camera can just move to the right position, so you wouldn't need a manually adjusted sensor.
Not sure if anyone has made this suggestion or not, but you could just use another cam to stop the board in a particular place. Once the parts have been placed, the locking cams and the placement cams could be lifted and the board could then be sent out. so as the board is coming in to the pnp, the placement cams would be down and the locking cams would be up. the belt brings the board on for a known distance. it would then bump the placement cams, and the belt could even drag a little. after that, the locking cams would go down to lock the board in place.
If you use a phototransistor and diode pair mounted on the pick and place head and reflective tape attached to the extrusion under the belt enough of the reflective tape is exposed. When the pick and place is in 0,0 position, the board can be moved to it and when the beam breaks it clamps the board. You could later adapt this further using a motor drive on the conveyor to set the width based on the width of the board waiting to enter the conveyor should you be working with different width boards.
From my point one good improvement it to make the rails a bit longer then the cnc mounting space, and the motor what drives the pulley for PCB to be mounted in the exterior of the frame, in this way you can make the chain what's next can be connected to the soldering zone. And for detecting of the posicion for pcb you can use a line of photo sensors under the pcb ( when the last sensor triggered you can know the the position ) one way it by calculating time when photo sensor has been triggered and the speed of the pcb moving pulley.
So as a matter of interest can't you use shafts to allow a single stepper and servo to operate the conveyer and clamping mechanism, freeing up drivers on the mainboard for other functions?
Nice idea, but i'm a bit concerned with the life expectancy of those springy cams. Wouldn't be better to replace them with a real small metal spring? Like those found in retractable pens. I was thinking about the spring as a buffer between a rigid cam and the moving enclosure/clamp. The tendency to lift sideway could be removed by making the clamp move inside 3d printed rails.
Could leave it clamped for couple days and see if it's same strength. If each panel is only there 10 minutes you could estimate lifetime. More stress is faster creep in the plastic so for edit I would say make beefy but with low strain for same overall force.
What are you going to do with PCBs smaller/bigger than your current test PCB? Maybe spacers with some kind of bed underneath (stop it from falling through)? Micro switch could work for detecting the board at a set position.
A commercial pnp machine actually does have the aluminum rails above the board and the board rides on the belt below. To clamp the board, everything happens from below. The servo would have a cone shaped registration pin that both handles clamping and board registration. Registration is best done with a hole that is created during manufacturing, based on your CAD because board edges are often unpredictable and leads to misregistration from board to board, whereas holes drilled based on your CAD should always be predictable and reliable. Remeber, the boards you are ordering from PCWay right now are very low quantity order. When you start getting into bigger orders, your boards will be manufactured in panels, possible grouped with other orders and often the edges are different from one panel to the next because the manufacturer uses pin registration also. Anyway, the idea is to have the servo mounted so it can be adjustable in the x and y axis and the servo would have a cone shaped pin that it raised up into the hole and pushed up against the aluminum rail. This achieves both registration and clamping by having 2 servos on perpendicular corners of the board. Anyway, my 2 cents.
Board position sensing - they sell "3d printer optical limit switches" on amazon, if you're not opposed to more Dremel action on your rails. But a simpler approach may actually be as simple as a photoelectric switch and some reflective tape.
Great video! PLA has a bit of a memory effect when under load so over time when the spring cam thingy less and less spring force. Give PETG or maybe even nylon filament a try for the springy cam! The spring force should maintain much longer! (PETG is also slightly more flexible than PLA so it might fix your too strong issue you currently have)
Nice! Much better :) Neat solution on the cam mechanism. Definitely functional. Could also have something similar that's inside the rails and lives below the belt, that just clamps at the front edge. At this point I reckon it's a winner - polish off the design and move on :)
how about just putting the entire gt belt loop on top of the extrusion and let the clamp push the teeth of the belt in to each other to lock the whole belt?
If you want easy Z 0 absolute, you can do the clam from the down to the top of your aluminium plate, and it can move belt to the top, like this for a 1mm or 1.6mm thickness, you will have the same 0 z!
maybe add a simple slotted IR sensor to the end of the conveyor so you just have it running until a PCB goes through it. That way you are also not in any way dependent on the PCB color or general makeup of the PCBs on the panel. If you know the panel dimension you would then have a very exact reference to start from to find fiducials. Bonus point it also then knows when a panel has left the conveyor and its safe to move to the next one without dumping the placed panel on the floor.
How about making one of the aluminium rails spring loaded whereby a servo pulls the rail slightly away from the board to allow it travel on the belt, but once stationary, allow the rail to spring back and clamp the board. This keeps the servos out of the zone of travel of the head.
@Stephen How about actually using differential geometry to achieve all without actuators and only those two motors? Rotating aluminium rails into 45 degrees angle will prevent board from popping off the conveyor during the handoff even if only 10% of the board is actually in the PnP. Imagine looking at the conveyor receptor and it would look like this (left and right rails represented by less-than and greather-than respectively): Now, we don't need locking mechanism (no part is the best part) since we can actually leverage that angle of the rails to jam the board into the cavity simply by stopping one of the motors one step before the other one and that will in turn tilt the board by a fraction of the millimetre. Start the same motor forward one step before the other one to unlock the board. In case friction itself is not enough, make dents one after another as big as board thickness in the rails corner outwards. That will effectively make board's opposing corners slide into dents and prevent it from moving, but in order to be reliably unlocked dents has to be rounded (using something like a dull phillips screwdriver). Furthermore, having belt grooved on both sides would increase drag and reliability whichever approach you take. To add, your thrill is contagious, keep it up! Regards
Love it. Just curious, do you think this could operate without the clamp? Unless the head crashes into the board i feel like the belt stiction is enough work holding for placing
my thought was to build a larger robust spring into the clamping bracket, and use an eccentric cam on the servo, even if that spring was a piece of spring steel that slotted into the bracket.
Not sure if the servos pointing inwards (i.e. being under the PCB itself) is best - could interfere with some bigger parts on the underside of a 2-side PCB. can you redesign it, so that the servos and cam assembly are sort of "outside"? Otherwise: nice design, quite simple. I like it :)
The big machines often have the "rails" sticking out the side, the bearings can be on the outside of the rail, OX ====== XO Like this, X is the extrusion and the O is the bearing, steppers will be inside like now, the L channel is still on top as now, just the bearings are moved to the exterior edge, this should also make it easier for other systems to feed the boards in, like a stacker on the input
I have a suggestion maybe you can take inspiration from Ender 3 x-axis belt system as your converyor, 20x20 profile is much sturdier and has easy way to attach your servo clamps conveniently
Hi, I just had some thoughts about your feeder design (posting it here so you read it ;) ) You built in a limit switch to notice if there is still tape left. Nice idea, but isn't there normally some empty tape leading and some at the end of the reel? If I'm right, you'ld need some kind of diffefent method to ensure there are enough parts left
The mechanism is clever and you are on the right track, however chose your servo wisely (ie. do not cheap out) as the mechanism is relying on the servo shaft to hold against the spring tension, and I would be surprised if this did not dramatically reduce lifespan of very cheap servos
Don't worry about the clamp being too rigid or performing way better than it needs to. On industrial machines, clamping is done by pneumatic pistons and also, there's a lever system that increases the torque of the piston when it clamps the panel for PCBA, selective soldering or conformal coating. So yeah, the idea is to avoid any movement of the panel during processing. Also, pay attention to the position of the panel when it reaches the work position. You should consider adding a stopper and let the conveyor move a bit more in order to make the panel hit the stopper levers and have an almost perfect work position without any deviation on X axis of your PnP machine. That stopper option is also used on the industrial machines, even they are using servos for the conveyors. Watch this video closely to see how to conveyor works and what elements does it has: ruclips.net/video/_7TuFLDaZwM/видео.html I can't wait to see you build a selective soldering machine with 3D printed parts and other assemblies from aliexpress or wherever. :D I've use to work with a selective soldering machine and it was so cool to make the programs and tune the parameters to make the solder joints as standards requires. :D Cheers!
This is one of my favorite videos you've posted in a while (which is saying a lot)! I *love* that cam mechanism. I was literally jumping with excitement. SO COOL. Also, I want your swaaaag. Buying a hoodie right now. With the slots in the L-bracket, if you sell kits of this, would you provide L-bracket with the slots pre-cut?
Love your videos. Can you use the cam on the pick and place head as a detection ? Like have the head move to a specific position and when it detects movement in a specific area on the image ?
Aren't this type of holders reduce the working space on plate? Maybe you can hold it from downside by some wacuum table like a reverce aerohockey? Sorry for bad english, but i hope that this will help :)
Optical sensors can give you inaccurate stop positions. I know because I sell an AOI for PCB assemblies they uses this. We have to include a function to inspected the leading edge of the pcb to adjust the X axis position of the inspection program. Mechanical end stops at typically more accurate. Usually a pin that rotates on a shaft up to the stop position and then rotates back down after the operation is complete.
that is a cool design, have you stress tested it, to look how many time it can be used before it might wear down? (I hope that stress test is the right word here, I'm no engenier)
I would have just used pins to pin both sides of the board. They could just go through the side. One optical endstop could detect that the board is in place.
You could have the conveyor belt make a U-shaped dip below the aluminium extrusion instead of going over AND under to continue it on the other side. The resulting 20mm gap (+radii) shouldn't matter to panels being fed into the machine.
I think you do not need to have an L-shape rail at all. Just have a 3D printed some-millimeters heigh side guard. this would also allow for the clamps to be placed much lower. Maybe just having some rectangular plates mounted (able to rotate) on the 3d printed side guard, heavier on the outside, so it will bring the inside part upwards by gravity and is pushed up by a servo to clamp down would be sufficient. On the ends, you might also get away with some 3d printed parts instead of cutting into the rails. I think you might want to add an automatic feeding tower and some storage rack to both ends, so you prolong the autonomous period of your PnP. Also having automatic width adjustment based on the PCB format would be a nice addition for a larger spectrum of machinable PCB sizes. If you use the tool head camera to determine the optimal position of the PCB on the belt for production it would be possible to (of cause later) optimize the traveling distance of the toolhead based on the PCB layout, the available feeders and their content.
The Lshape is important for vertical stiffness. Without it the rails will be floppy. Test it out for yourself - ruclips.net/video/WdevPq0b554/видео.html
@@olivercouch1651 You are right. What I wanted to say is you do not need it as side boundaries as they are used now. Because it makes the structure quite high and leads to the clamping mechanismn to get quite bulky.
Have you ever been to a dairy farm, or seen a octopus? You already have a vacuum system, correct? Multiple vacuum suction cups under the board would hold it tight w/o any worry of hitting anything above the board.
Use silicon oil under the belt in order get a better life time for it. Otherwise it is rubbing too much against it. Silicon oil is normally used for treatmills, so that should do the job.
That cam mechanism is super clever! Nicely done.
The little compliant clamp spring is so cool! You could probably make the cam spring part of the clamp and make the whole thing one compliant mechanism by having a spiral spring going off of the servo to the middle of the clamp part :D awesome video as always!
For an idea of sensing where the board is, you can use a photosensor/thru beam to detect the incoming edge from there you are just counting steps until the trailing edge passes. You will know exactly how long the board is and where it is at so you can place it in the middle of the machine
Make a notch in the rail and two vertical slits.
So the clamp has two guides to slide up and down.
And clamped down it sits in the notch to be flush with the top of the rail. So no additional height at all.
Great solution. Love it. Looks super cool.
Greetings from Munich, Tom.
A Stephen Hawes video is mostly just him celebrating things working and it warms my heart XD
Thanks for taking community ideas into consideration. You're the man! Thank you for the good vibe you provide to all your subscribers! 👍🏻 👍🏻 👍🏻
As for the side "thingies" holding the clamp - intead of putting the glue, you can model them to have C-like form hanging over the aluminium L and you can fix them fron the bottom with the screw againt the bottom part of Aluminium L profile. Also, to improve the guiding of the claming part you can use some kinf of tongue-groove design :)
Anyway - great work *thumbs up*
I was also thinking screws through the aluminum for securing the servo motor as well instead of glue.
Honestly, I would have *two* pieces of angle, both pointed *down*, with the board and belt sandwiched between them. During movement, the top piece is held clear of the board. When clamping, the top piece is pulled down onto the board. That way, *all* boards in the system are held firmly in place at once. As a bonus, the working clearance is literally the thickness of the aluminum of the top angle. Another option, more complicated to build, would be to have a single piece of angle on either side, pointing up but the sides pitched in, clamping boards by sliding one angle toward the other so that the edges of the board are being clamped. The pitch on the angle would vector the horizontal force to point in *and* down, holding the board firmly in place. Both options can be powered from a single servo that pulls one end of a belt to tighten the clamp. A spring on the other end of the belt provides some give to the system. Both options would use springs to hold them open.
Does that make sense?
I work at an electronics factory and work with SMT machines. The positioning problems with loading in boards is generally solved by using a stopper that is swung up from below. The conveyor will then run, pulling the board along and the stopper will prevent the board from going too far. When ejecting the board, the stopper swings down and the conveyor runs until the board is out.
I love this freaking channel. The innovation is through the roof. Makes me think of problems out of the box.
About adding a sensor to detect when the PCB is fed into position: a simple optical gate would do - shine a LED down from above the board (a little finger holding an IR LED could poke out over the board) to a photosensor below the rail. Then you could write some code to pull in a PCB, detect when it's passing into the work area, and stop the feed the instant it clears the sensor at a known point so the machine can "know" where the leftmost edge of the board is. Should also make it far easier to pick up on fiducials for final board locating if the machine has a reference point for the left edge of its work area. A second sensor at the exit end could be used to let the machine know when the current board has cleared the work area and it's safe to pull the next one in.
The first idea that came to mind, squeeze the rail profile that the pcb is up on together. that is, have a stepper motor that moves one rail closer to the other. and thus holds the pcb with friction then you do not need to have anything over the rail at all. Anyway just an idea, I'm sure your solution is just as god.
That's also a very good idea. Combine that with a rail shape that is more U than L like to prevent the PCB from being able to move upwards.
Three gotchas with compliant printed mechanisms are hysteresis, creep, and fatigue. I doubt any of those will be a problem with this mechanism (if made in something kinda springy like PETG or Nylon) but if they are you could make it so the servo arm instead holds a little piece of spring steel shim stock as the blade flexure and then you’ll have practically infinite number of cycles, no creep issues, and uniform clamping pressure over years of usage.
I second this, with some minor changes. A solid cam on the servo and two pieces of stock spring steel in the clamp, one driving down and one driving up should remove the fatigue issue. Granted there will be wear but so little and the cam is a small part to replace. Or even a single spring holding it open with a cam drive to close it.
The spring cam design is a super excellent idea. As has been mentioned by another person, the repeated flexing may create fatigue or creep in the spring, so it might be good to have it printed in PETG or better yet Nylon on a production version.
The difficulty with tightening the belts is ever present... I highly recommend having an idler that moves outwards, and thus pulls the belt taut, via the tightening of a screw. It's what I've done on all my 3D printers because it lets you easily adjust for the belt stretching out over time, and means you don't have to work super hard to pop it on there in the first place.
Anyway, excellent channel. I discovered it via the algorithm from your previous video and binged the whole thing. Great content and, continuing onwards, good luck with the project! :)
I was really excited as you were when found it works!!
That Cam sprung shape is great! A bumper that rises and lowers similar to the clamps could serve as home position; rise bumper, convey in until board detected, clamp, lower bumper, run placement, unclamp, convey out.
I still love your presentation style and excitement. It's so encouraging. Keep up the great work. Looking forward to seeing a actual full panel placed from your feeders.
If you ever need a way to roughly locate the board on the conveyor before using the camera I propose this:
A mechanical switch or an optical sensor built in your clamp that can register holes on the sides of the pcb panel.
Thinking about it, even a set of metal "brushes" riding on some exposed tracks could do the trick
Congrats on an elegant solution for the clamps (A spring loaded cam body). In looking at what you've done I think the top of the clamp needs to be closer to rail height. Possibly notching the side rail a bit with and the clamp cad adjusted to make it closer to rail height so it would not stick up so much in the closed position. The clamp open position is a never-mind because the p-n-p head is parked during board movement. I'm also wondering if a mild cad rework to raise the gantry rail or extend the z movement amount might be a thought. Good work. Your videos are, as always, an entertaining presentation of your engineering process.
Stephen: A couple of ideas. The aluminium brackets are 90 degrees. How about bending them in a bit, to maybe lets say 80 degrees, perhaps cut them down a bit so they are not so tall, and then clamp the board by moving one of the rails closer to the other, pinching the board side on between the rails? That way, nothing sticks above the rail brackets at all.
As for the converyer system, why not just have multiple belts to catch a bord from the next/previous. Should work fine as long as the belts move more or less in unison. Added benefit is that you can put some space in between one belt and the next, for a led and a phototransistor combo to detect when the board passes between, typically at the entry and exit points of the belt inside the pnp.
If you put a thin strip of rubber under the plastic part that grabs onto the PCB you'll probably need less force to hold the PCB in place.
That is straight up SICK. You are a mechanical engineering god.
Hi, this might work for you.
I was thinking I would try using the clamping idea for a positioning stop as well. But make a smaller one and flip the cam 180 degrees. Maybe incorporated it into the same clamping body. The pcb board would come up and hit the edge of the stop that is closed, one on each side of the rail. Then you could close the clamps and raise the stop. After the pcb board passes by you can then close the stop again to stop the next pcb board. I had a similar stop design on a Bosch conveyor. The Boche stop dropped from Below to allow a pallet to pass. Enjoying watching your project.
Since you don't need them to move independently you may be able to just use one port on the board for both servos? Nice solution going over the rail with the clamping piece, very low profile.
You could extend the belt mechanism to outside of of the main frame. If you design printable pieces on the end that accommodate for the pulley, no hacksawing the metal brackets is needed.
Nice vertical tensioning system. You could use some pieces of gt2 belt on the moving part to even make it grip even more so you can reduce the length of the mechanism. Nice work, as usual. 👍
You: want the finger to hold the board in place even when the servo is turned off
my first reaction: bi-stable compliant mechanism :)
I love your energy and creativity - this is a great channel for teaching and empowering others 😎
Is having the PCB riding on top of the rails going to mean that the thickness of different PCBs now needs to be taken into account for placing components? Really great project with lots of interesting engineering, I hope you make something from all this work.
Pcb thickness is a usual variable to enter on a pick and place if it is not using a top clamping system. Obviously a lift clamping system does not require this but I see both methods in actual industrial assembly systems.
Now double up the servos (all parallel, moving at the same time) and make them move another set of shorter L profiles from the ends that rest above the belt for moving and clamp down on the whole length. This moves the clamping mechanism out of the "hot zone" of the machine to the sides, allowing for more clearance.
You already have a way to position the PCB...move your camera into a position mid way between the rails and closer to the exit
end. Using OpenCV have it move the PCB until it reaches the centerline of the camera field of view then stop moving forward
and use your clamps to hold it in position ready to place parts! Nice job by the way on your clamp.
You don't need to cut the frame. Just build another module and push the pcbs over the edge onto the 2nd conveyor belt;)
Very clever clamp. Well done!
Great design, super easy, barely an inconvenience....cheers.
Impressive idea! You are an inspiration to those who keep push the maker spirit forward.
maybe optical sensors for simple board edge. Locating just one on the input combine with the stepper data you could find length and you would know where both edges are within range to find your fiducials. Looking awesome so far for sure. Keep up the amazing work its Turing out really awesome.
Nice work ! Maybe you could use a flexion sensor like the FS2L055 to detect when the board is passing the clamping system on the conveyor belt. With the size of the board as a parameter you can then know when to clamp on the middle of it.
If it too complicated, Time of Light sensor ? VL53L0X from STMicroelectronics. Or more industrial one : GTB6 from Sick (you can detect the thickness of a paper with this so maybe it will be overkill).
Put a slight chamfered edge on the clamp lead-in so it can't catch on the square leading edge of the board if you get a warped board. Nice Job on the design. I'm following this closely and subscribe to the Git. I have too many projects, right now, to start an Index but I plan on jumping into it in the not too distant future. BTW, thanks for the incentive to switch to FreeCad. I was just gearing up in 360 but was not a fan of their new "screw the little guy" attitude. I can't afford their expensive subscription service so I was looking for a good alternative. You came through at just the right moment.
Yo that worked surprisingly well, nice one! Definitely stealing this idea!
Those clamps are awesome! Loving your videos!
Try using the camera vision to detect when the board is in place. Might be harder but one less part is always good.
I think using the top vision camera would be a great way to make sure the board gets into position correctly. The top camera can just move to the right position, so you wouldn't need a manually adjusted sensor.
This is amazing! Best channel I've stumbled across in a while. Love your enthusiasm :D
Not sure if anyone has made this suggestion or not, but you could just use another cam to stop the board in a particular place. Once the parts have been placed, the locking cams and the placement cams could be lifted and the board could then be sent out. so as the board is coming in to the pnp, the placement cams would be down and the locking cams would be up. the belt brings the board on for a known distance. it would then bump the placement cams, and the belt could even drag a little. after that, the locking cams would go down to lock the board in place.
If you use a phototransistor and diode pair mounted on the pick and place head and reflective tape attached to the extrusion under the belt enough of the reflective tape is exposed. When the pick and place is in 0,0 position, the board can be moved to it and when the beam breaks it clamps the board. You could later adapt this further using a motor drive on the conveyor to set the width based on the width of the board waiting to enter the conveyor should you be working with different width boards.
From my point one good improvement it to make the rails a bit longer then the cnc mounting space, and the motor what drives the pulley for PCB to be mounted in the exterior of the frame, in this way you can make the chain what's next can be connected to the soldering zone. And for detecting of the posicion for pcb you can use a line of photo sensors under the pcb ( when the last sensor triggered you can know the the position ) one way it by calculating time when photo sensor has been triggered and the speed of the pcb moving pulley.
4:44 was actually sick af
What a great idea and implementation! Thanks for putting my comment in the video too!
PLA will not maintain its tension, it will creep fairly quickly. Would love to see if anyone has experience printing springs with different materials.
So as a matter of interest can't you use shafts to allow a single stepper and servo to operate the conveyer and clamping mechanism, freeing up drivers on the mainboard for other functions?
Nice idea, but i'm a bit concerned with the life expectancy of those springy cams.
Wouldn't be better to replace them with a real small metal spring? Like those found in retractable pens.
I was thinking about the spring as a buffer between a rigid cam and the moving enclosure/clamp.
The tendency to lift sideway could be removed by making the clamp move inside 3d printed rails.
Could leave it clamped for couple days and see if it's same strength. If each panel is only there 10 minutes you could estimate lifetime. More stress is faster creep in the plastic so for edit I would say make beefy but with low strain for same overall force.
What are you going to do with PCBs smaller/bigger than your current test PCB? Maybe spacers with some kind of bed underneath (stop it from falling through)?
Micro switch could work for detecting the board at a set position.
Your bonkersness and problem solving make me happy 🤓
Absolutely Brilliant! You have such a creative mind.
A commercial pnp machine actually does have the aluminum rails above the board and the board rides on the belt below. To clamp the board, everything happens from below. The servo would have a cone shaped registration pin that both handles clamping and board registration. Registration is best done with a hole that is created during manufacturing, based on your CAD because board edges are often unpredictable and leads to misregistration from board to board, whereas holes drilled based on your CAD should always be predictable and reliable. Remeber, the boards you are ordering from PCWay right now are very low quantity order. When you start getting into bigger orders, your boards will be manufactured in panels, possible grouped with other orders and often the edges are different from one panel to the next because the manufacturer uses pin registration also.
Anyway, the idea is to have the servo mounted so it can be adjustable in the x and y axis and the servo would have a cone shaped pin that it raised up into the hole and pushed up against the aluminum rail. This achieves both registration and clamping by having 2 servos on perpendicular corners of the board. Anyway, my 2 cents.
Looks like a winner
I just love your enthusiasm
Board position sensing - they sell "3d printer optical limit switches" on amazon, if you're not opposed to more Dremel action on your rails. But a simpler approach may actually be as simple as a photoelectric switch and some reflective tape.
Great video! PLA has a bit of a memory effect when under load so over time when the spring cam thingy less and less spring force. Give PETG or maybe even nylon filament a try for the springy cam! The spring force should maintain much longer! (PETG is also slightly more flexible than PLA so it might fix your too strong issue you currently have)
Can probably run both those servos off the same output if you're not already, which will free up one more header on your board.
Nice! Much better :)
Neat solution on the cam mechanism. Definitely functional. Could also have something similar that's inside the rails and lives below the belt, that just clamps at the front edge.
At this point I reckon it's a winner - polish off the design and move on :)
how about just putting the entire gt belt loop on top of the extrusion and let the clamp push the teeth of the belt in to each other to lock the whole belt?
If you want easy Z 0 absolute, you can do the clam from the down to the top of your aluminium plate, and it can move belt to the top, like this for a 1mm or 1.6mm thickness, you will have the same 0 z!
Damn your adding some really nifty stuff to your machine!
maybe add a simple slotted IR sensor to the end of the conveyor so you just have it running until a PCB goes through it. That way you are also not in any way dependent on the PCB color or general makeup of the PCBs on the panel. If you know the panel dimension you would then have a very exact reference to start from to find fiducials.
Bonus point it also then knows when a panel has left the conveyor and its safe to move to the next one without dumping the placed panel on the floor.
How about making one of the aluminium rails spring loaded whereby a servo pulls the rail slightly away from the board to allow it travel on the belt, but once stationary, allow the rail to spring back and clamp the board. This keeps the servos out of the zone of travel of the head.
@Stephen How about actually using differential geometry to achieve all without actuators and only those two motors?
Rotating aluminium rails into 45 degrees angle will prevent board from popping off the conveyor during the handoff even if only 10% of the board is actually in the PnP. Imagine looking at the conveyor receptor and it would look like this (left and right rails represented by less-than and greather-than respectively):
Now, we don't need locking mechanism (no part is the best part) since we can actually leverage that angle of the rails to jam the board into the cavity simply by stopping one of the motors one step before the other one and that will in turn tilt the board by a fraction of the millimetre. Start the same motor forward one step before the other one to unlock the board.
In case friction itself is not enough, make dents one after another as big as board thickness in the rails corner outwards. That will effectively make board's opposing corners slide into dents and prevent it from moving, but in order to be reliably unlocked dents has to be rounded (using something like a dull phillips screwdriver).
Furthermore, having belt grooved on both sides would increase drag and reliability whichever approach you take.
To add, your thrill is contagious, keep it up!
Regards
What about notching out the rail to make it even lower profile?
Brilliant design!
Nice Video its realy helping me to build my automaded 3D Printer cause i had the same idea with the conveyor
Love it. Just curious, do you think this could operate without the clamp? Unless the head crashes into the board i feel like the belt stiction is enough work holding for placing
I feel like the 3D printed spring will get brittle and break eventually. idk id make it stronger somehow.
maybe the best possible longevity can be found by changing how wide and thick the spring is while conserving clamping force.
my thought was to build a larger robust spring into the clamping bracket, and use an eccentric cam on the servo, even if that spring was a piece of spring steel that slotted into the bracket.
Not sure if the servos pointing inwards (i.e. being under the PCB itself) is best - could interfere with some bigger parts on the underside of a 2-side PCB. can you redesign it, so that the servos and cam assembly are sort of "outside"? Otherwise: nice design, quite simple. I like it :)
Very good engineering man.
The big machines often have the "rails" sticking out the side, the bearings can be on the outside of the rail, OX ====== XO Like this, X is the extrusion and the O is the bearing, steppers will be inside like now, the L channel is still on top as now, just the bearings are moved to the exterior edge, this should also make it easier for other systems to feed the boards in, like a stacker on the input
Hey man, love ur vids and your awesome passion for your projects! Much love ❤️
It came out really well. Isn't one clamp enough since the pcb is the exact size of the distance between the rails? It won't rotate.
I have a suggestion maybe you can take inspiration from Ender 3 x-axis belt system as your converyor, 20x20 profile is much sturdier and has easy way to attach your servo clamps conveniently
Hi, I just had some thoughts about your feeder design (posting it here so you read it ;) )
You built in a limit switch to notice if there is still tape left. Nice idea, but isn't there normally some empty tape leading and some at the end of the reel? If I'm right, you'ld need some kind of diffefent method to ensure there are enough parts left
Couldn't you do detection of the PCB with the downward camera. Just have it move over there when the convyer belt starts moving.
The mechanism is clever and you are on the right track, however chose your servo wisely (ie. do not cheap out) as the mechanism is relying on the servo shaft to hold against the spring tension, and I would be surprised if this did not dramatically reduce lifespan of very cheap servos
Don't worry about the clamp being too rigid or performing way better than it needs to. On industrial machines, clamping is done by pneumatic pistons and also, there's a lever system that increases the torque of the piston when it clamps the panel for PCBA, selective soldering or conformal coating. So yeah, the idea is to avoid any movement of the panel during processing.
Also, pay attention to the position of the panel when it reaches the work position. You should consider adding a stopper and let the conveyor move a bit more in order to make the panel hit the stopper levers and have an almost perfect work position without any deviation on X axis of your PnP machine. That stopper option is also used on the industrial machines, even they are using servos for the conveyors.
Watch this video closely to see how to conveyor works and what elements does it has: ruclips.net/video/_7TuFLDaZwM/видео.html
I can't wait to see you build a selective soldering machine with 3D printed parts and other assemblies from aliexpress or wherever. :D I've use to work with a selective soldering machine and it was so cool to make the programs and tune the parameters to make the solder joints as standards requires. :D Cheers!
This is one of my favorite videos you've posted in a while (which is saying a lot)! I *love* that cam mechanism. I was literally jumping with excitement. SO COOL. Also, I want your swaaaag. Buying a hoodie right now. With the slots in the L-bracket, if you sell kits of this, would you provide L-bracket with the slots pre-cut?
Just remember to change the material to something more temperature resistent when you start hot air soldering
Love your videos. Can you use the cam on the pick and place head as a detection ? Like have the head move to a specific position and when it detects movement in a specific area on the image ?
Can you solder QFNL-Package IC'S?
Planning to start a own product and need to know how to manufacture
Aren't this type of holders reduce the working space on plate? Maybe you can hold it from downside by some wacuum table like a reverce aerohockey? Sorry for bad english, but i hope that this will help :)
I thought about that but i think it wouldnt work because there could be holes everywhere in the PCB
Are you planning on making a automated pcb mill later on?
ingenious design
The merch should have your gold logo the one with a soldering iron
Use optical sensor to detect when the panel has reached the desired point
Optical sensors can give you inaccurate stop positions. I know because I sell an AOI for PCB assemblies they uses this. We have to include a function to inspected the leading edge of the pcb to adjust the X axis position of the inspection program. Mechanical end stops at typically more accurate. Usually a pin that rotates on a shaft up to the stop position and then rotates back down after the operation is complete.
that is a cool design, have you stress tested it, to look how many time it can be used before it might wear down? (I hope that stress test is the right word here, I'm no engenier)
I would have just used pins to pin both sides of the board. They could just go through the side.
One optical endstop could detect that the board is in place.
You could have the conveyor belt make a U-shaped dip below the aluminium extrusion instead of going over AND under to continue it on the other side. The resulting 20mm gap (+radii) shouldn't matter to panels being fed into the machine.
Can you try suction from underneath by a vacuum , love your content
Amazing video as always
Excellent travail, continuer comme cela, c'est top
Something the would be awesome is a pcb stack loading system for full automation
I think you do not need to have an L-shape rail at all. Just have a 3D printed some-millimeters heigh side guard. this would also allow for the clamps to be placed much lower. Maybe just having some rectangular plates mounted (able to rotate) on the 3d printed side guard, heavier on the outside, so it will bring the inside part upwards by gravity and is pushed up by a servo to clamp down would be sufficient.
On the ends, you might also get away with some 3d printed parts instead of cutting into the rails. I think you might want to add an automatic feeding tower and some storage rack to both ends, so you prolong the autonomous period of your PnP.
Also having automatic width adjustment based on the PCB format would be a nice addition for a larger spectrum of machinable PCB sizes.
If you use the tool head camera to determine the optimal position of the PCB on the belt for production it would be possible to (of cause later) optimize the traveling distance of the toolhead based on the PCB layout, the available feeders and their content.
The Lshape is important for vertical stiffness. Without it the rails will be floppy. Test it out for yourself - ruclips.net/video/WdevPq0b554/видео.html
@@olivercouch1651 You are right. What I wanted to say is you do not need it as side boundaries as they are used now. Because it makes the structure quite high and leads to the clamping mechanismn to get quite bulky.
Have you ever been to a dairy farm, or seen a octopus? You already have a vacuum system, correct? Multiple vacuum suction cups under the board would hold it tight w/o any worry of hitting anything above the board.
Nice design! Subscribed for more :D
Use silicon oil under the belt in order get a better life time for it. Otherwise it is rubbing too much against it. Silicon oil is normally used for treatmills, so that should do the job.