Pick and Place Machine

I have been tracking the OpenPnP project from some time and feel much of the software has reached a level of maturity that I think it is worth exploring what it would take to implement a DIY Pick and Place Machine at the space.

http://openpnp.org/hardware/

I am curious to know everyone thinks.

1 Like

We have or had one. @artg_dms?

Aren’t they used primarily for production runs? I don’t know that I personally ever heard of anyone using the one we had for a quite a while. It very well may have been.

For one offs it seems you could place the items faster than measuring and programing and getting a reel of parts.

It would be an interesting project, fundamentally a CNC with gripper of some sort. Parts are like a tool changer.

A member donated the use of one and it was rarely ever used.

Yes we had one on loan. Only 1or 2 people figured out how to use it. Back in Aug of last year I posted that if no one stepped up to teach on it, it would be returned to the owner. No one did so I asked the owner to take it back.

We used to have one, yes.

From my point of view the main reason it was never used was the difficulty of the software and setup of the machine we had. The pick and place machine we used to have was one of these Chinese machines:
EDIT: Wrong link new link below.
https://www.aliexpress.com/item/SMT-pick-and-place-machine-TM220A-Small-desktop-Automatic-Cheap-Surface-Mounting-PCB-0402-5050-LED/1104214395.html?spm=2114.01010208.3.1.ANUISQ&ws_ab_test=searchweb0_0,searchweb201602_4_10065_10130_10068_433_434_10136_10137_10138_10060_10062_10141_10056_10055_10054_10059_201_10531_10099_10530_129_10103_10102_10096_10052_10144_10053_10050_10107_10142_10051_10106_10143_10526_10529_10084_10083_10080_10082_10081_10110_10111_10112_10113_10114_10078_10079_10073_10070_10122_10123_10124,searchweb201603_2,afswitch_1,ppcSwitch_5,single_sort_0_default&btsid=e833ac40-f71e-4f99-969e-c0c9a4f2d0c2&algo_expid=06392350-5a36-4ad2-9a7d-73df162e7538-0&algo_pvid=06392350-5a36-4ad2-9a7d-73df162e7538

From my understanding the software was quite bad and difficult to learn.

That is specifically why I think we should explore setting up an OpenPnP Pick and Place machine as the software should be much more approachable.

Please, in very big bullet items, what differentiates OpenPnP from the Chinese stuff?

Great question!

  1. Positioning in the X Y plane

    1. OpenPnP uses computer vision to locate the pad the component will be placed on and correct minor issues in location.

    2. The TM240A uses dead reckoning to locate parts thus minor issues in location can lead to completely failed jobs.

  2. User interface

  3. OpenPnP has a GUI that permits user to track the job as it progresses and more effectively understand what is going on in a job making trouble shooting a bit easier.
    https://cloud.githubusercontent.com/assets/1182323/21791153/962a6e0c-d6a6-11e6-82b7-ba252e94cc79.png

  4. As far as I can tell the user communicated with the old machine via an excel doc (page 12 to 15) : https://dallasmakerspace.org/w/images/9/98/Tm220a_manual.pdf

  5. Locating the part feeders

  6. OpenPnP supports multiple different types of feeders and permits the user to set up these feeders within the GUI. https://github.com/openpnp/openpnp/wiki/Setup-and-Calibration%3A-Feeders

  7. The old pick and place machine needed the feeders location and type setup via excel doc (page 10) : https://dallasmakerspace.org/w/images/9/98/Tm220a_manual.pdf

  8. Documentation quality

  9. OpenPnP’s documentation is not perfect in fact it is not even good but the lead developer is extremely responsive and active.

  10. This: https://dallasmakerspace.org/w/images/9/98/Tm220a_manual.pdf just suuuuucks…

  11. Solder Paste

  12. OpenPnP supports having a solder paste extruder to solder paste is deposited as need while parts are being placed. This is a really poorly documented feature but it is present in the code: https://github.com/openpnp/openpnp/search?utf8=âś“&q=Paste+Dispense&type=

  13. The old pick and place machine required the user to manually add solder paste or use a mask.

Edit: added the bit about the solder paste

2 Likes

Pick and places are more than just one machine, it is a system of production. I know this as I happen to own a nice one at my shop.

For me, the machine was the easy purchase, the harder part was getting working pcbs off of it. You have so many variables in the process that if any are done wrong you end up just spending more time trouble shooting and reworking than it would of taken to hand solder in the first place.

Examples, application of solder paste. You can hand apply on individual pads, which is no faster than hand soldering. You can have a machine apply pad by pad, which is faster, but many boards quickly get to more than 100 pads so if the machine is only 99% perfect means you will have rework on every board. Or you invest into a much more complicated and expensive option of having Visual inspection machine that can spot these errors and correct them. But, then you run into the issue of pads that are too small to have a machine apply solder paste, this might sound unlikely, but most chips that require more than 15 pads are too small to have a machine we can afford apply the solder paste. So in that case you have to go with the production method a stencil printing the pads. This adds a cost of $150 plus to any job plus the need for a stencil printer. This is the best way to apply solder paste, the process is much like screen printing, fast and strait forward as long as you have good new solder paste. But, fresh solder paste is expensive, you want to buy it freshly mixed, there is a shop down from the old Dms that will mix batches fresh for you, they tend to be small bulk packs say 1 to 2 lbs each and run more than $100.

That is just dealing with solder paste, not even talking about buying components in quantities needed to be able to feed in the machine, or dealing with placement, bad components reflow ovens, or cooling. Using pick and places are strictly for production environment, they don’t have the over lap into prototyping that other tools like the Haas, Cnc router, plasma cutter, or other tools we have around Dms do.

My two cents spending 2+ years building a pcb production system and spending $40,000 plus on it. Ps I still don’t have it figured out for more than one product.

2 Likes

All in all I think it is clear that a pick and place machine built on OpenPnP would be of much more utility for the DMS membership. I also believe if we had an OpenPnP machine it would be significantly easier for members to learn how to use it than the chinese machine was.

I would agree that production systems are complex to setup.

But the question is if the software and setup for an OpenPnP machine is sufficiently easy to save time on a short (5 to 10 boards) run for our membership. Here is a video of the type of machine I am envisioning building:

You will notice the feeders are short strips minimising the need for bulk purchases.

Frank,

Look into my section on just applying solder paste. This step alone from the running of the machine is a Bitch. Find a video showing the full work flow from empty machine to working pcb and see if you feel the same. At best you are investing 4 to 8 hours just setting up an already proven and working system, making your 5 boards, cleaning up and leaving it all in an acceptable state for the next person. 4 hours honestly on the low end.

Also, that system is effectively no different from the system we have from Falgout. If you are unwilling to put in the time to learn the system we have, which is very easy to program, then you are unwilling to learn the open pcb system.

As I stated in my analysis of the OpenPnP system the solder paste deposition is a poorly documented feature but from reading some of the issue logs it seems to use the same computer vision based position correction.

I think your very comments show the lie in this:

If the old system was “easy” to program and set up good parts would be easy to produce. But of course the old system was not easy to setup and program.

Frank you are hilarious I your ignorance. You claim my expertise is a lie and then claim a system you haven’t used is easier than another system you haven’t used.

I’ve used three pick and place systems, my professional one and the one we have at Dms and a step up from the one at Dms which has spool feeders Visual centering. They work the same, they have a training manual programming where you drive the head above every placement, also they have the ability to import pre programmed designs. This is how it is done. But open pcb which uses the same system would be easier?

The same system would not be easier, they would be equally shitty. But they are not the same.

You are right I have not used either system, however I have read the documentation for both systems. Given the fact that I have read the documentation of both tools I can come to a few conclusion with a high degree of certainty.

I assume you are referring to computer vision. The TM220A does not have any computer vision systems. From all of the reading I have done good and accurate computer vision is fundamental to making an reliable pick and place machine.

Both tools clearly import the same data file in order to populate the board that is not in question. What I am arguing is that the OpenPnP GUI makes it easier to debug jobs and determine where and how the job is failing.

But something tells me you have not bothered to look at any of the documentation provided.

Edit:

I don’t mean to be an asshole but pick and place machines have clearly given you some kind of hell for some time. I am just hoping to find a better system that will make pick and place tech more accessible and useable.

Would it be possible to build a small inexpensive machine to test your hypothesis?

Something that will only place “large” chips like SOICs for example?

(While not economical given Chinese manufacturers propensity for dumping such things, it would be fun to knock out a fistful of Teensy / Micro / Nano compatible boards whenever needed.)

2 Likes

The fastest way to test the hypothesis would be to retrofit @Andrew_Falgout’s TM220A to run OpenPnP.

However if Andrew does not want to risk his rather expensive machine the next best test bed would be the shapeOko. We could set it up as a test bed for about $332 (82+18+24+8+34=$166, $166 * 2 new project fudge factor = $332).
Here are the core parts as far as I can tell:

Edit: the pump is just a guess I need to do some more research to be sure we are specing the correct pump

The Shapeoko is in use for milling and I don’t think you could use it for this.

I’m curious; if we had an OpenPnP machine, what projects would we use it for? Just teaching?

I don’t want to shut down the idea, thanks for bringing it! I just don’t know where we’d put such a thing, who would use it, etc.

I understand I can lead a horse to water, but I can’t force it to drink.

The easiest part is wasting the money buying a machine and not knowing the full process. I’ve highlighted multiple examples of how the items external to the pick and place are much harder than the pick and place process and you ignore these points.

Good luck young prince in your dreams. I’ll keep my eyes open here in reality. You need much more than s pick and place to do what you want.

Sorry guys. I was told the PnP was not of value to the Electronics
committee, so I took it back. It has been sold and no longer available.

./Andrew Falgout
KG5GRX

4 Likes