Low Cost Pick and Place Machine

[ProtoVoltaics]

It seems like the big topic is speed.

We are at about 750 CPH with the angular vision (upward looking camera) only being used on the parts that need it. We are at about 300 CPH with angular corrections being used on every part.

Please understand that this isn't the final specs of the machine. We literally just got everything working pretty flawlessly about 2 weeks ago  . This is an ongoing project and we are still currently working to make it faster. Speed was the lowest thing on our list of "things to do" because what is speed if the machine doesn't work.

As other have said we are wasting a lot of time over the upward looking camera, and we have very slow movement in the z-axis. These are both true. We are working on these as I write this post! We have a new head design that will use a solenoid for the z-axis making this much faster. Also, the new head design will remedy the issue with the A-axis belt backlash. The reason we are waiting over the upward looking camera is because when we tell the PnP to rotate the part 3 degrees we don't know if it's actually going to rotate 3 degrees because of the backlash in the belt. With the new head we will be able to tell it to rotate 3 degrees and know that's what's going to happen, so with this we could move off the upward looking camera much faster.

First make it work, then make it fast.

[matseng]

We have a new head design that will use a solenoid for the z-axis making this much faster.

Is it possible to stop a solenoid in "mid-flight" to accommodate for parts with varying height?

[ProtoVoltaics]

We have a new head design that will use a solenoid for the z-axis making this much faster.

Is it possible to stop a solenoid in "mid-flight" to accommodate for parts with varying height?

No, the solenoid will go until it is stopped by the part.

[mikeselectricstuff]

We have a new head design that will use a solenoid for the z-axis making this much faster.

Is it possible to stop a solenoid in "mid-flight" to accommodate for parts with varying height?

No, the solenoid will go until it is stopped by the part.

That sounds like a major fail.
You need variable height, both to deal with board height and also when picking from random strips, trays etc. 

[ProtoVoltaics]

We have a new head design that will use a solenoid for the z-axis making this much faster.

Is it possible to stop a solenoid in "mid-flight" to accommodate for parts with varying height?

No, the solenoid will go until it is stopped by the part.

That sounds like a major fail.
You need variable height, both to deal with board height and also when picking from random strips, trays etc.

The z-axis screw is still there. We are able to move up and down to variable heights. The idea of the solenoid is so when we arrive over the part we can shoot the solenoid down very quickly to do the pick ups and placements.

Sorry I re-read my previous posts, I should have said "We have a new head design that will use a solenoid in conjunction with the existing z-axis"

Yes, we must have the a way to move up and down other than the solenoid, or we wouldn't be able to do head changes and many other things we are currently doing. Sorry for the confusion.

[rx8pilot]

It seems like the big topic is speed.

This is true....but CPH is not the only factor in speed. For a high-mix task (making small numbers of many PCB's), setup time can easily be longer than run time. That has been my own personal challenge. I needed a solution that could offer very fast change from design to design to be useful. This would allow me to do 10pcs of 3 PCB's for a last second demand from a customer and immediately follow that with populating a few prototypes. I wrote a ULP and some other glue software to make a direct path from Eagle to the PNP machine so that programming is extremely simple regardless of what is on the PCB. For physical setup, the goal is to keep 80+ parts in the machine all the time that covers most of my designs and upcoming prototypes. Changing only a few parts per run will ensure a super fast changeover.

That is something that should be considered for an entry level machine. Since you will not likely be able to fit 80+ feeders, the rest of the design needs to be very easy to setup. If the user spends a ton of time setting up the machine it does not matter how many CPH it can muster - if the PCB could have been build by hand during the setup time.

Once again, I am still hoping to better understand the target customer for this solution. I cannot think of a use case for it that makes sense other than it is neat to have a pick and place machine. Anyone?

I have a Quad 4000C that is just getting up and running... this machine was around $6k or so.

At that price point, what sort of shape was the machine in when you obtained it? How much work was involved in getting it running? Thanks.

OFF TOPIC:

It was very dirty and needed service. I learned  by doing and it took about 3 weeks (on and off) to get it to full function. Lots of rubber and urethane parts were replaced and I did significant updates to the control PC and power system. Cleaning, lubing, aligning, and inspecting. The next and last step to being useful to me is making PCB holders for odd shape PCB's. The machine can only clamp onto square boards with a dimension bigger than 3" or so. I just need to finish the universal PCB holder trays. At this point, it is in amazing shape and I have piles of feeders. I know all the details of it and can easily service most issues myself (I am a machinist as well so the mechanics are easy to deal with)

[tggzzz]

That is something that should be considered for an entry level machine. Since you will not likely be able to fit 80+ feeders, the rest of the design needs to be very easy to setup. If the user spends a ton of time setting up the machine it does not matter how many CPH it can muster - if the PCB could have been build by hand during the setup time.

Very true.

I use similar reasoning to instantly ignore most "labour saving" kitchen gadgets - especially when cleaning time is included!

[eneuro]

The idea of the solenoid is so when we arrive over the part we can shoot the solenoid down very quickly to do the pick ups and placements.

I have no idea about what solenoid you are talking about, but it looks like reinventing a wheel 

I will probably use one of those linear motors-they are much faster than your CV system for the moment and I think that even when you improove CV code, I do not know how fast your cemeras are (how many frames per second CV now performs) high speed camera is a must or some kind of digital camera to be able shoot photo at given position very fast without motion blur effects.

Yaskawa Linear Motors in Motion


BTW: What kind of motors did you used for horizontal and vertical movement for the moment?

OK. It seams you are using balls screw, so... those linear motors are 50% faster and achieve 9cycles/second with speed up to 5m/s 

To outperform your PnP I need to be ahead and.. it looks like I'm thanks to advances in linear motors  technology 

Anyway, I like those feeders from Quad PnP linked above, but probably I will go into feeders placement which will allow replace all for given PCB at once, so very little time needed to setup for another boards and I'd rather put them on the bottom of workpiece table, but we'll see what happends soon


Another option is might be smart feeders which will take ICs from reels in parallel to running PnP and place at given position on working table or separate system to provide ICs sorted in th eorder assembly needs them- from many feeders on line ready to reach by main assembly head always at one place instead of jumping each time to different feeder position-loosing time to reach another part 

So, this will allow provide stream of ICs (parts) ready to PnP while they are will be already taken from reels, what ever we have and in the order machine wants them, so its only task will be optimize and go always to the same position for another IC part to PnP and since another system will prepare those parts in parallel with assembly head, there is no overhead and loosing time to move to given reel or any limits to the amount ICs per board since this will be task for other machine-lets say feeder machine, which will be always ready to give main assembly head information what kind of IC is redy to take and main assembly head will be able order any IC on demand (out of order) while placing previous one and it will be feeder machine task to replace current part with another requested if something strange happends-assembly head misses part or destroys, etc.

 

[ProtoVoltaics]

I have no idea about what solenoid you are talking about, but it looks like reinventing a wheel 

I will probably use one of those linear motors-they are much faster than your CV system for the moment and I think that even when you improove CV code, I do not know how fast your cemeras are (how many frames per second CV now performs) high speed camera is a must or some kind of digital camera to be able shoot photo at given position very fast without motion blur effects.

OK. It seams you are using balls screw, so... those linear motors are 50% faster and achieve 9cycles/second with speed up to 5m/s 

To outperform your PnP I need to be ahead and.. it looks like I'm thanks to advances in linear motors  technology 

The solenoid will be what shoots the pick up tip down to grab the part. This is the fastest solution for this, faster than linear rails. The ball screw only has to travel short distances now with the inclusion of the solenoid, so speed here doesn't really matter anymore. The ball screw will only need to move for head changing and varying part heights, board heights. I hope at some point we will be too fast for the ball screw and have to change to something faster!

I would much like to see your machine can you please provide a link?

[ProtoVoltaics]

BTW: What kind of motors did you used for horizontal and vertical movement for the moment?

We are using Nema 17 stepper motors at the moment.

[rx8pilot]

The Quad feeders are very simple and would be easy to make. The official ones are rather robust, but could be simplified to allow easier manufacturing. There are a lot of electronics that can simplified as well. These feeders are old skool but work very well.

[mikeselectricstuff]

The ball screw will only need to move for head changing

Ball screw is the wrong choice for Z axis. Rack & pinion or toothed belt would be better.
using a solenoid to patch-up an inadequate drive is a poor solution.
 

[rx8pilot]

why is a ball screw a problem? Seems that is a normal solution.

[mikeselectricstuff]

why is a ball screw a problem? Seems that is a normal solution.

My pick/place uses rack & pinion. I've not seen the inside of many others.

I was assuming this was the reason for the slowness being mentioned.
To get decent Z speed from a screw, you'd probably either need a shallow pitch or fairly fast motor rotation. You don't really need the force that a screw can provide, so I'd think a lower gearing ratio would be more efficient.

[eneuro]

BTW: What kind of motors did you used for horizontal and vertical movement for the moment?

We are using Nema 17 stepper motors at the moment.

Really? Those Stepper motor - NEMA-17 size - 200 steps/rev, 12V 350mA  $14.00   ones?

Do you overload them to get higher  power (RPM) while when we take 12V*0.35A= 4.2W ONLY of input nominal power ? Which is the mass of this PnP moving head?
With two motors you have less than 10W nominal electric power available or I missed something?
Those motors have 200 steps/rev, so get linear speed of 1m/s with 1mm offset/rev, we have 100 such small 0.001 m movements, so it looks like we need spin those steppers to 6000 RPM (100Hz) and make 20000 steps per second.
What could be theoretical speed in horizontal/vertical with this limited nominal power in those motors?
Maybe do you have average stats of head speed in XY and Z axis to estimate its theoretical maximum performance, but probably motors maximum available power must limit this speed whatever axis system will be choosen-it looks like you have less than 10W input power i XY, so when we take  ? head mass  moving back and forward we could try estimate maximum theoretical speed possible based on energy available only.

Anyway, I'd like to use maybe even two headers in my solution moving in horizontal position, so it could be nice use linear motors with two of such heads on one horizontal transmission line, so ball screw probably not a solution for me, so that is why I'm interested in linear motors instead

Idea is that while one of the headers will do PnP and asembly taken part, second PnP head will in parallel go for another part and vice verse, so with the same linear speeds two times more parts can be placed by one PnP machine with additional mentioned feeder machine 

Those linear motors  have builtin high precision encoders as I know, so  I can imagine those motors can be pushed hard based on distance to move and later slow down quicly (energy can be recovered back), so it looks like they can move much faster in average than ball screw lines, since longer distances can be made without need to achieve huge RPMs-just shoot moving part like magnetic gun 

[H.O]

Do you overload them to get higher  power (RPM) while when we take 12V*0.35A= 4.2W ONLY of input nominal power ?

It's not THAT simple!
A properly driven step motor is driven from a power stage with active current limiting powered with a voltage MUCH (usually 10-20 times) higher than the rated coil voltage of the motor (to overcome the inductance of the winding).

Unfortunately the datasheet of the motor at Adafruit that you link to doesn't have a speed torque/torque curve so I'll use another motor as an example, specifically the MOONS MS17HD4P4150 NEMA17 motor.
It has a rated current of 1.5A and coil resistance of 1.7ohm putting the rated coil voltage at 2.55V do you really think it will be able to put out 3.8W only?
Look at the speed/torque curve in Datasheet. When driven from a 36V supply at 1.5A it will do roughly 0.17Nm at 1200rpm - that's about 21W of mechanical power, not 3.8W.

Oh, and this is not overloading the motor, this is driving it properly.

[eneuro]

Oh, and this is not overloading the motor, this is driving it properly.

Maybe, those specs are not good information source-I've linked it just to ask topic owner if this is thhose motors used

[eneuro]

I would much like to see your machine can you please provide a link?

No problem, follow this thread from time to time and you will see one head version soon, since I need some help in manual SMD soldering, so will spend more time on this my PnP concept prototype project there:
Cheap recognition & pick & place & reflow oven solder machine for tiny PCBs 

[ProtoVoltaics]

We will be leaving today to attend the Orlando Maker Faire September 12th and 13th! If anyone is in the area we hope to see you there!

http://www.makerfaireorlando.com/

[ProtoVoltaics]

Hey everyone we we accepted by KickStarter and have launched KickStarters first PnP project campaign! https://www.kickstarter.com/projects/protovoltaics/pick-and-place-machine-by-protovoltaics

We had a lot of interest in the machine at the Orlando maker faire, so we decided to go ahead a launch our KickStarter, and have it run through the remaining faires we will be attending. The entry level price for the machine is $6000 fully assembled. Machines ordered on KS are due to ship in December.

We are back at work on the machine to get the speed up above 1000 components per hour and we plan on meeting this before the next faire. We will update the video as soon as we can. We have already increased the speed a lot from removing the belt that we were using to rotate the part.

Someone asked who are market was, we have had interest from people who do small batch runs and prototypes, maker spaces, academic groups have showed the most interest so far, and from businesses that typically have they work sent off, but are interested in something to help with surge protection. (when they need a few extra boards fast, and it doesn't make sense to send them overseas.)

[eneuro]

The entry level price for the machine is $6000 fully assembled.

Shipment included? How much does it weight fully assembled?

Anyway, I'm looking for my PnP price... lower than $1000, to have faster ROI  even making not too many PCBs a month, but placing SMD is boring, so yep I want make it cheap

You are talking about december.. this year or next? 

[ProtoVoltaics]

The entry level price for the machine is $6000 fully assembled.

Shipment included? How much does it weight fully assembled?

Anyway, I'm looking for my PnP price... lower than $1000, to have faster ROI  even making not too many PCBs a month, but placing SMD is boring, so yep I want make it cheap

You are talking about december.. this year or next? 

For the KickStarter Shipping is included. It's about 30 lbs with the control cabinet and packaging.

Shipments will start this year December 2015.

[ProtoVoltaics]

This new head design will eliminate all of the backlash that we currently have.  It also brings rigidly mounted cameras that will allow us to move faster.  Previously, we had to run slower than top speed just so our cameras wouldn't move off target due to rapid accelerations and decelerations.

Right now we have to waste a bunch of time over the upward looking camera while we confirm that our rotational correction has been applied.  This is because of the backlash in the system.  The computer vision tells us that we have to rotate, say 5 degrees, to have the part properly aligned, however when we tell the motor to turn by 5 degrees it really doesn't turn 5 degrees because our belt system has slack in it.  Some of the motor's movement is just to take up the slack in the belt instead of actually rotating the part.  We end up needing to use the computer vision to confirm that the part was actually rotated correctly and this means having to hang out over the upward camera until the part is actually fully rotated to the proper orientation.

The new head will eliminate all of this wasted time.  The new design has no belt, so we'll be able to take an image of the part, and then immediately start flying to the place location.  On our way there we should get the feedback from the computer vision saying what the angular offset is, and we can correct it on the fly.  This is the first major step to getting us to over 1000 cph (components per hour).

The new head has been mounted on our prototype, but we need a bracket to mount a limit switch to so that we know when we make contact with the pick-up tip.  We have time scheduled late tomorrow with the laser cutter, so we should be able to start running trials with the new head design on Thursday.

We're also trying a NEMA 23 motor on the z-axis to see if there's any speed difference versus the old NEMA 17.

The last change we applied today was a new feeder divider that will allow us to reconfigure the incoming feeders for the various types of SMT reels out there from 8mm to 44mm.

[eneuro]

This is the first major step to getting us to over 1000 cph (components per hour).

It looks like in my PnP I solved rotation of parts much easier and in parallel with assembly head (no need to rotate anything during pick & place) since... parts are already rotated by feeder machine, so this means very simply assembly head and... faster PnP, while two machines (feeder and assembly head) cooperate together in PnP task 

[ProtoVoltaics]

This is the first major step to getting us to over 1000 cph (components per hour).

It looks like in my PnP I solved rotation of parts much easier and in parallel with assembly head (no need to rotate anything during pick & place) since... parts are already rotated by feeder machine, so this means very simply assembly head and... faster PnP, while two machines (feeder and assembly head) cooperate together in PnP task 

If you assume the parts in the feeders are rotated perfectly in the feeders you will never be able to place small / low pitch parts. You will need a way to rotate these parts if you want any kind of reliability. Also, we ran into some parts in tray feeders that were not oriented the same way as others.

Our head is attached now, and we are going to laser cut the last piece we need now! Hopfully a new video tomorrow?