Low Cost Pick and Place Machine

[ProtoVoltaics]

Hi All

I work for a very small group called ProtoVoltaics. We make electronics and do custom projects for people including prototyping.

This has been keeping us busy, and placing a bunch of boards by hand is time consuming. We looked online for pick and place machines, but what we found were machines that: needed the user to check, and correct misplaced placed parts, could place small parts but not IC's, and were either too slow or too expensive.

It didn't take us too long to decide we would make this our next in-house project. We have been working on our machine for a few months now, and we are ready to grow awareness of our project. As of now we have only been using Hackaday.io, and our blog to show our updates. However, there are many other great places to talk about projects like this such as eevblog.

We are looking to start a discussion about our PnP machine to find out what people would like in a PnP machine. Maybe you have a PnP already, but the CV isn't robust enough, or it can only place a few hundred components per hour. Maybe you have never seen a PnP machine before, and just want to know how one works.

We look forward to seeing what people have to say. We invite you to check out the project details on our hackaday.io/protovoltaics page, and to view videos of our projects/progress on our YouTube https://www.youtube.com/channel/UCLqKBaRZHeLOnbVsihLs8_A/videos

[rx8pilot]

Why do you have multiple offset cameras?

[ProtoVoltaics]

Why do you have multiple offset cameras?

We use two cameras one looking in the X direction, and the other in the Y direction. Doing this allows us to correct for any XY offset error when we pick up and place parts.

We can see the middle of the pick head, the middle of the parts to be place, and the middle of the area on the PCB where the part will be placed. This works great for small parts like the 0603's used in the video, but we think we will have to make some changes to do IC's (we haven't tried yet). Another reason for doing this is to increase Components Per Hour (CPH). Since there is no upward facing camera we only have to move between the feeders and the placement area.

[kjs]

We can see the middle of the pick head, the middle of the parts to be place, and the middle of the area on the PCB where the part will be placed. This works great for small parts like the 0603's used in the video, but we think we will have to make some changes to do IC's (we haven't tried yet). Another reason for doing this is to increase Components Per Hour (CPH). Since there is no upward facing camera we only have to move between the feeders and the placement area.

0603 small parts? These are monsters.... We place as a standard 0402's and sometimes 0201's with our Swiss Essemtec machine and it handles all types of IC's including BGA's very well. No issues with placement or angle since we got it around 5 years ago. Small footprint and well thought through. At CHF25k the price was right too.

[DanielS]

0603 small parts? These are monsters....

Well, OP's company seems to be looking for a replacement for hand-placing components. Populating components smaller than 0603 by hand would be extremely tedious, as if 0603 was not already bothersome in any significant volume.

That said, 0402 is a very nice size to be able to pick-and-place since it fits almost straight across SOIC-8 leads to reduce sprawling around op-amp feedback loops and other sensitive paths that need to be as tight as possible.

[ProtoVoltaics]

We can see the middle of the pick head, the middle of the parts to be place, and the middle of the area on the PCB where the part will be placed. This works great for small parts like the 0603's used in the video, but we think we will have to make some changes to do IC's (we haven't tried yet). Another reason for doing this is to increase Components Per Hour (CPH). Since there is no upward facing camera we only have to move between the feeders and the placement area.

0603 small parts? These are monsters.... We place as a standard 0402's and sometimes 0201's with our Swiss Essemtec machine and it handles all types of IC's including BGA's very well. No issues with placement or angle since we got it around 5 years ago. Small footprint and well thought through. At CHF25k the price was right too.

Yes, I agree. 0603 parts are much bigger than 0201 and 01005 parts. I was simply comparing 0603 parts to IC's in my comment before. Also, yes, a $25,000 CHF ($27,000 USD) machine should be able to do large and small very accurately. We are looking to do this for less than a fifth of that price.

Thank you for mentioning your machine. I will see what I can find on them, and maybe I can learn something from what they are doing.

[ProtoVoltaics]

0603 small parts? These are monsters....

Well, OP's company seems to be looking for a replacement for hand-placing components. Populating components smaller than 0603 by hand would be extremely tedious, as if 0603 was not already bothersome in any significant volume.

That said, 0402 is a very nice size to be able to pick-and-place since it fits almost straight across SOIC-8 leads to reduce sprawling around op-amp feedback loops and other sensitive paths that need to be as tight as possible.

Yes, this machine is a replacement for tweezers and a microscope for us. 

[kjs]

If it's just a replacement for tweezers and microscope it makes a difference. When we looked into it we wanted a machine which can do medium volume production and the prototype work. Advantage is that once the prototype assembly worked you know that production will work too.
We looked at many of the cheaper machines and none came even close. The price I mentioned was list price but the market was down which gave us a pretty hefty discount on it. Often it is cheaper in the long run to buy something more expensive.....

[ProtoVoltaics]

Just added a new video of the Auto Tool Changing.

https://youtu.be/syRGDpmOsxs

[eneuro]

Just added a new video of the Auto Tool Changing.

Is it possible integrate this thing with oven-I mean take PCB from stack, pick and place and pass for "cooking" oven to run some kind of automated line with final visual inspection before soldering?

Another thing-is it possible play with this machine in manual mode eg. by inserting G-code like CNC machinists but with those cameras "hawk eye" support? It could be helpfull to make corrections if needed...

Also working on my own small machine to automate PCBs assembling (small PCBs up to 5cm x 5cm) but with integrated custom iron plate oven with glass window on top to monitor soldering progress.
I've already done custom kinetics module for Linux CNC which allows me drive my machine using G-code in manual mode, so that is why asking for such possibility in your product 

As we can see my Eneurokins Linux CNC kitetics module translates G-code to machine geometry coordinates and vice verse.
BTW: this is not trival cartesian transformation since CNC machine oryginal concept was something like this to manufacture parabolic mirrors molds, but smaller versions I hope will be fine for pick and place and will allow put PCB to custom oven while another PCB is prepared for soldering 


So, do you have G-code support in manual mode like CNC machinists used to do sometimes for simple tasks ?

My goal is use my pick & place in manual mode first then try automate using OpenCV for realtime video processing 

[ProtoVoltaics]

Just added a new video of the Auto Tool Changing.

Is it possible integrate this thing with oven-I mean take PCB from stack, pick and place and pass for "cooking" oven to run some kind of automated line with final visual inspection before soldering?

Another thing-is it possible play with this machine in manual mode eg. by inserting G-code like CNC machinists but with those cameras "hawk eye" support? It could be helpfull to make corrections if needed...

Also working on my own small machine to automate PCBs assembling (small PCBs up to 5cm x 5cm) but with integrated custom iron plate oven with glass window on top to monitor soldering progress.
I've already done custom kinetics module for Linux CNC which allows me drive my machine using G-code in manual mode, so that is why asking for such possibility in your product 

As we can see my Eneurokins Linux CNC kitetics module translates G-code to machine geometry coordinates and vice verse.
BTW: this is not trival cartesian transformation since CNC machine oryginal concept was something like this to manufacture parabolic mirrors molds, but smaller versions I hope will be fine for pick and place and will allow put PCB to custom oven while another PCB is prepared for soldering 


So, do you have G-code support in manual mode like CNC machinists used to do sometimes for simple tasks ?

My goal is use my pick & place in manual mode first then try automate using OpenCV for realtime video processing 

Our software does support G-code in manual mode. I have gotten the conveyor question many times, and there will most likely be a version we make with a conveyor. At the moment we don't have one because we are simply trying to get everything else finished, so we can Have a fully functional machine to take to the National Maker Faire in Washington DC. 

[ProtoVoltaics]

Hey guys, Just wanted to post an update here. We were working on speeding the machine up, and we got it to a point where we would use it for ourselves  . We have a few more things we are working on now to get it even faster, and should be done with that in the next couple weeks. Currently we are placing at a little over 850 CPH. Here is our latest video, I forgot my camera at home so this video is taken with my phone.

https://youtu.be/C7Pi4akL5og

I think the really cool thing is that it aligns to every piece it picks up, and every pad it's placing on. You can see in the video it will stop over top of the PCB, and then align and go down. This will be very beneficial to us when placing more than one of the same board at the same time. If the offsets between boards are different our machine will still work, because the position it goes to doesn't have to be dead on. It will just align to the pad it needs to and place.

Also, I haven't mentioned here, but the feeder we are using was homemade too! We actually just got those completed last week!

[EEVblog]

0603 small parts? These are monsters....

No, they aren't.
PnP machines have traditionally had a bit of a hump to get from 0603 to 0402, so much so that PnP companies market their ability to handle 0402 on their machines. Many older machines need upgrades or option to handle 0402.
A low end PnP machine limited to 0603 is not really a limitation. It's common for designers to not go to 0402 unless they have to. Less risk in production.

[ProtoVoltaics]

0603 small parts? These are monsters....

No, they aren't.
PnP machines have traditionally had a bit of a hump to get from 0603 to 0402, so much so that PnP companies market their ability to handle 0402 on their machines. Many older machines need upgrades or option to handle 0402.
A low end PnP machine limited to 0603 is not really a limitation. It's common for designers to not go to 0402 unless they have to. Less risk in production.

The boards we sell are designed using 0603 that is why all our videos show 0603 placement. We can do 0402 currently, but our main focus now is very accurate placement of IC's, rather than anything smaller.

[rx8pilot]

0603 small parts? These are monsters....

No, they aren't.
PnP machines have traditionally had a bit of a hump to get from 0603 to 0402, so much so that PnP companies market their ability to handle 0402 on their machines. Many older machines need upgrades or option to handle 0402.
A low end PnP machine limited to 0603 is not really a limitation. It's common for designers to not go to 0402 unless they have to. Less risk in production.

True. Most PCB's will be just fine with 0603.

[ProtoVoltaics]

Shortly after my last post on here we successfully placed our fist IC. Very exciting, we hope to have a video of the machine making a full board up soon.

[Alphatronique]

Hi  you may look into ebay  that one was sold few week back
http://www.ebay.com/itm/MyData-Automation-TP9-2U-Pick-and-Place-Machine-/261851567952

and i have personally send one to scrap last week  ,there day machine that not handle 0201 was commonly send to scrap
you quick find that even if machine miss 1 part on 100 ,time you spend find and debug it have prohibitive cost (good comercial machine have 1 /10,000 )

few year back i put lot of time and money into try make one ,end result was that i learn that cost less to buy a working old one
and now i put my time doing project ,not project that will permit to later make my real project ;-)

have fun

[ProtoVoltaics]

Hi  you may look into ebay  that one was sold few week back
http://www.ebay.com/itm/MyData-Automation-TP9-2U-Pick-and-Place-Machine-/261851567952

and i have personally send one to scrap last week  ,there day machine that not handle 0201 was commonly send to scrap
you quick find that even if machine miss 1 part on 100 ,time you spend find and debug it have prohibitive cost (good comercial machine have 1 /10,000 )

few year back i put lot of time and money into try make one ,end result was that i learn that cost less to buy a working old one
and now i put my time doing project ,not project that will permit to later make my real project ;-)

have fun

Yeah, making a pnp machine has been very difficult. It's been a fun process, and we are very confident in our machine. Our current machine is very accurate and very repeatable. We haven't run it long enough to see where we are in terms of error, but we will get there.

We haven't set a price for the product yet. We know it wont be as cheap as liteplacer, but we are around 3 to 4 times faster at the moment, we have feeders, and we are capable of IC placements.

We should have the pick and place ready for orders in the next 4-8 weeks. We will continue to update on where we are in terms of progress and price.

Public feedback is always welcome. We are eager to know what you all would like in a pnp machine. So far public feedback has been very helpful, and we have been able to incorporate many of the things they have asked for such as: automatic feeders, automatic IC placement, placing more than one board at a time, auto tool changing (heads for smaller/larger parts), larger size for increased placement area.

[Alphatronique]

you place your first chip yesterday and you already in sale pitch mode ;-)   

let time to build couple of hundred board and see how it perform and how mutch babysitting it require ,how software is usable
 for teach new part , new data coming from different cad system ,different part  batch from different supplier  ,solve the feeder issue

was very active in the pick place word , in many forum and i see miraculus low cost home made pick palce apper every 30day or less
but in 10 year  only one i see that ramp to production was the neodem an maddel ,and i a barely usable machine
even the table top 30K$ mancorp machine was not so usable

if you really what a helpfully machine for home prototype  ,take a delta 3D printer ,put laser pointer on it
and let laser spot tell were to place part to humam that place it whit Tweezer ,and of course tell witch part on a LCD display
that low cost ,faster that most home made pickplace ,more reliable and avoid placement error   and you got 3D printer in bonus 
  that same way used for vintage troug hole  and whit only a light lamp

it may look that i rude but on that it have no free lunch ,and you may now got very good reliable used commercial machine that do 0603 and up for very decent price
and have machine that have >20 year of research and development put on it and that not require any babysitting

[ProtoVoltaics]

I don't find any of what you said rude, Alphatronique, I very much appreciate your feedback. Yes, we placed our first IC, but we have been placing other parts for quite some time now. In the low cost pnp market I haven't seen any that can do IC's reliably with vision. For us being able to place IC's was just us seeing our machine as a step ahead of whats out their currently as opposed to saying "ok now that we can place IC's lets go to market."

[JackP]

The issue with that is (and people like rx8pilot will hopefully back me up) it does not allow you to be doing other things. It may be somewhat faster than normal manual placement, but you are still tied up. With a decent old PnP machine, you do not need to be constantly tending to the machine, so can write firmware or design a daughter board etc. With enough money, you can automate the process further, to autoload boards into the machine. Good for through hole boards in a factory workspace, yes.

[rx8pilot]

I would much rather have a slow automated machine than a fast manual machine for sure.

I applaud the effort on this project, you have obviously seen the challenges involved in PnP. The challenges in making 'affordable' PnP are even greater since this is a precision endeavor and precision costs money. There are some clever ways of using low cost parts and pre-made sub-systems like motor drivers or vision, but it is still very difficult which also costs money.

I am not suggesting anyone give up on making PnP affordable, but perhaps re-examine why it needs to exist. For my use, I needed a low-cost PnP solution to help with prototypes and small batches that are all not good to send to commercial CM outfits. That means it needs to be precise, but not fast. It also means that it needs to be very easy to setup, hold any type of component, and hold a large variety of parts (for me anyway).

Take away the need for super speed, you still have a requirement for precision so the cost can only go so low. Ease of use to be able to change from job to job is not cheap either. This requirement is why modular feeders were invented. Modular feeders will add cost even if they are only semi-automatic dumb feeders.

If it takes a long time to setup or  struggles with small parts to the point where an operator has to be watching the whole time, it will quickly become a neat toy. The only reason PnP exists is to make PCB's fast enough to be a profitable business, right? Businesses don't buy toys, but they do sometimes need an entry level option - in the case of PnP it is a very fine line.

I am crossing my fingers that my own investment in PnP actually pays off by getting me to the next level of business. Only time will tell.

[Alphatronique]

Hi

actually the  "semi-automatic dumb feeders" was one of the most impotent thing  ,if machine cannot pick it right  it cannot place it    following the shit in ,shit out principle of physic

most low-cost machine try to compensate for bad feeder by vision but this requite to teach of each part ,and unfortunately it have more variation on part that it look
just like ceramic cap have good color variation even on the same reel ,not to talk on end contact variation  ,same for resistor each manufacturer or event fabrication plant have slight variation of pad shape
so evil was in the detail ;-)     on my production machine a mydata  machine weight 2Ton was so rigid and mechanically stable that i place all day long thousand of 0402
part whiteout any vision assistance whit exception to board fiducial )  ,this permit to make Small batch attractive since no vision training invoke and no subject to ambient light variation

that remember me that one of local small shop have a mancorp  MDC machine  ,every Monday morning machine was not usable since out of calibration and have pour yield
after investigation it find that for clean-up floor the cleaning man move the table on witch machine was located and lack of frame rigidity un-calibrate machine every time 

have fun ..

[ProtoVoltaics]

I completely agree, a pnp needs to be more accurate than fast, and needs feeders to support it. We went into this project stressing the need of feeders. We went through 3 versions of our feeder very quickly, and we actually CAD'd a new version last night. Here is a video on version 2 of our feeder.

https://youtu.be/uNBV_dabrDA

The newest version we just CAD'd up has nothing that comes above the pnp table, it is a completely hanging system. All of the versions have been hanging systems, but they have come up above the pnp table by about 3 inches. This was a problem for our camera attached to the front end of our pick head.

Also, I agree that if you have to babysit the machine there is no need to get it. We made this machine to free up time not to spend it doing the same task in a different way. I won't say we are at that point now, but I think we are close, and we will continue in that direction till it has been achieved.

We do plan on having a conveyor system in the future it's just not something we are working on near term. It's all still a work in progress, and will continue to be long after we start selling the product.

If anyone is in the Washington DC area we hope to see you at the DC National Maker Faire. We just got our invite this morning! 

[mikeselectricstuff]

That feeder looks very expensive, in some areas unnecessarily so. A ton of machined parts. I can't see that being sellable for much less than traditional P&P machine feeders. A low cost machine must have low(er) cost feeders.

Interconnect also looks a total mess - 3 connectors, and a consumer audio & power jacks on a device that will see vibration  And all that internal wiring.

How about something like a PCB-mounted D9 or D15, with ribbon interconnect between feeders  - that would save a lot of internal wiring.

The long path from the sprocket wheel to the feed point may have issues with crumpling on plastic tapes, causing parts to jump out.

Your "feed-then peel" action is likely to make parts jump out of plastic tapes. needs to be a single smooth action.

I'm not sure that design will scale well to large parts like QFPs due to tape radius around the sprocket, and tall parts (maybe also heavy inductors) could also be problematic without good support & guidance of the tape - a wide tape needs support on both edges.

You may also find that worm gear gets clogged with dust from paper tapes.

It ought to be possible to friction-drive the tape take-up from the main motor  to avoid the seperate motor - a simple belt from your middle gear perhaps?

IMO needs a few more iterations.