Open source pick and place

[NiHaoMike]

[SMTech]

While you have to appreciate his nerdy enthusiasm, I don't get his pitch, its not a "mid scale production" machine and the role is filled already by countless physically superior Chinese machines not made of of 3D printed crap. Sure they're open source, but what is that worth to someone who doesn't have hours of spare time to tinker and contribute.

There are at least two of us on here who contract in this supposed "niche" of lowish volume as well as the JLC option ( You really cannot and should not use PCB numbers as a volume indicator) competitively with no need for a small design house or startup needing to acquire new skills, equipment and processes for stuff thats surely going to end up in a rusty heap in the corner in short order.

[jduncan]

There's nothing wrong with 3d printed parts for this application, especially in comparison with the "cheap as possible" sheet metal of charmhigh etc.

The advantage of open source is you can run openpnp natively, rather than struggling with badly translated barely functional crapware, and get support directly from the designers through discord etc rather than hoping for a back and forth email exchange with someone in China. 

[tomgat]

There's nothing wrong with 3d printed parts for this application, especially in comparison with the "cheap as possible" sheet metal of charmhigh etc.

The advantage of open source is you can run openpnp natively, rather than struggling with badly translated barely functional crapware, and get support directly from the designers through discord etc rather than hoping for a back and forth email exchange with someone in China.

I think it is safe to say that Stephen doesnt have a clue what "midscale" manufacturing is. He should just call his machine a sub functional DIY hobby PNP machine.  Im sure there are hobby people interested in this, but its certainly not the people doing this work for a living.  Im an extremely small niche "mid-scale" manufacture, and my little PNP machines are still placing 9000 pcs/h.  His business premise is assuming that the capital cost of the PNP machine is a barrier of entry, which simply isnt true.  They are an insignificant cost overall in the business.

Also, the reality is that the low cost chinese PNP machines are actually pretty damn good.  I personally prefer QiHe machines over the Charmhigh machines (I have both), but they are great extremely low cost machines that are very reliable.  In no way are they "barely functional crapware".  Their placing 0603, 0402s, TQFN,and BGAs by the tens of thousands per day while his little machine seems to struggle placing 1 massive LED that is at an angle.

[tomgat]

There's nothing wrong with 3d printed parts for this application, especially in comparison with the "cheap as possible" sheet metal of charmhigh etc.

The advantage of open source is you can run openpnp natively, rather than struggling with badly translated barely functional crapware, and get support directly from the designers through discord etc rather than hoping for a back and forth email exchange with someone in China.

For comparison of what Stephen claims he is trying to compete against... Below is a link to an extremely simple board I just ran a few minutes ago on one of those "cheap chinese barely functional crapware" machines (Qihe TVM-925) which he claims he is trying to compete with.  Keep in mind the machine is only running at 80% speed.

https://www.dropbox.com/s/yt4rtwd7yueowqv/IMG_6040.mov?dl=0

[Mechatrommer]

he's been working on that for years since before pandemic and made youtubes with his enthusiastic pitch... i think he's still figuring out the feeder... you can buy if you want, but everything sold out, so he's pushing the $1.7K of his full LumenPNP kit https://opulo.io/collections/all you'll need the board anyway. otherwise you can try to spin his OSS board yourself.. related threads:

https://www.eevblog.com/forum/manufacture/anyone-any-experience-with-the-lumenpnp-from-opulo-(stephen-hawes)/msg4365838/#msg4365838
https://www.eevblog.com/forum/reviews/opulo-lumenpnp-foss-pick-place/msg4003552/#msg4003552

[tomgat]

he's been working on that for years since before pandemic and made youtubes with his enthusiastic pitch... i think he's still figuring out the feeder... you can buy if you want, but everything sold out, so he's pushing the $1.7K of his full LumenPNP kit

$245 for the mainboard!?!?  Damn... I will gladly build that board all day long for those kinds of prices.  Anybody out there that wants this board?

[Mechatrommer]

$245 for the mainboard!?!?  Damn...

why not? its Made in USA i want to learn how the camera feed (USB?) goes into the ARM mcu and then redirected to its USB to PC. i got an idea of PNP machine, but camera wil go directly to PC, so its many USB connections to PC (2 cameras + and machine will make 3 USB connection to PC)

[tomgat]

$245 for the mainboard!?!?  Damn...

why not? its Made in USA i want to learn how the camera feed (USB?) goes into the ARM mcu and then redirected to its USB to PC. i got an idea of PNP machine, but camera wil go directly to PC, so its many USB connections to PC (2 cameras + and machine will make 3 USB connection to PC)

Sweet!  Ive got one order.... If we can get 10, I will order the components. 

[tomgat]

$245 for the mainboard!?!?  Damn...

why not? its Made in USA i want to learn how the camera feed (USB?) goes into the ARM mcu and then redirected to its USB to PC. i got an idea of PNP machine, but camera wil go directly to PC, so its many USB connections to PC (2 cameras + and machine will make 3 USB connection to PC)

While we wait for other people, Im going to order the boards.  What color solder mask do you want?  Black?

[NorthGuy]

why not? its Made in USA i want to learn how the camera feed (USB?) goes into the ARM mcu and then redirected to its USB to PC. i got an idea of PNP machine, but camera wil go directly to PC, so its many USB connections to PC (2 cameras + and machine will make 3 USB connection to PC)

Why do you want the camera to go through ARM MCU? I have a programming machine which has a buit-in USB hub with 3 USB lines - camera, control board, and programmer.

[Mechatrommer]

Maybe there is misunderstanding with my super broken english... i have in mind to build my own pnp, but what i have in mind is cameras and arduino pnp board will connect directly to PC. this will make many USB connections to PC which is less ideal esp when PC has less USB ports.. LumenPnp fixed this by making only one USB connection to PC, both up and down cameras go to the ARM mcu, or sort of built in USB host? This is i dont know how he did it. I watched his video closely to figure this out. He claimed to have some issue with this hosting topology.. i was interested to get his board, but since its $245 and sold out, maybe i'll wait for made in PRC or do/figure it myself when i have the time..

$245 for the mainboard!?!?  Damn...

why not? its Made in USA i want to learn how the camera feed (USB?) goes into the ARM mcu and then redirected to its USB to PC. i got an idea of PNP machine, but camera wil go directly to PC, so its many USB connections to PC (2 cameras + and machine will make 3 USB connection to PC)

While we wait for other people, Im going to order the boards.  What color solder mask do you want?  Black?

if you can make your board working with proven made machine and cameras, i will be interested to get the board alone if price is affordable.. i can assemble mechanical parts myself. I'm too lazy to read the github source.. the files are organized in less comprehensive way..

But then i want to make the PC software to operate the machine few decades ahead, ie with interactive pcb/components view and placement. Currently used SW looks archaic so... thats the dream and dilema

[tomgat]

But then i want to make the PC software to operate the machine few decades ahead, ie with interactive pcb/components view and placement. Currently used SW looks archaic so... thats the dream and dilema

OpenPNP is a beast, but is extremely functional and will do almost anything you want it to do.  I wouldnt call it archaic.  Its biggest fault is that vision matching is extremely slow by todays standards.  However, the reality is that most people using OpenPNP dont really care much about high pcs/hr counts.   

[tomgat]

i think he's still figuring out the feeder...

I also do not understand why he is so intent on designing his own feeder when there is so many very good (and cheap) ones already out there.  You can buy used Yamaha feeders on ebay for less than $50, and brand new generic electric versions (which are milled aluminum and will reliably pick 0201) are about $200 each with free DHL shipping.  You can get it cheaper from other vendors if you are willing to wait longer.

Reliable placing is all about repeatability which requires frame rigidity.  This is why good PNP machine's frame so heavy.  Even the feeders are all metal.   3D printed parts simply will not be able to reliably place precision parts repeatably with any real speed, and bolting together 2 pcbs to make a feeder will not give the rigidity needed to reliably place and align precision parts.  Meaning, you cant camera align a part that you can not reliably pick each time from the feeder.

[loki42]

$50 is nice.  I just got a quote for $3000 for a used 88mm universal instruments gold feeder. Dual lane 8mm are around $450 used.

[tomgat]

$50 is nice.  I just got a quote for $3000 for a used 88mm universal instruments gold feeder. Dual lane 8mm are around $450 used.

I really like the yamaha (clone) feeders.  Most of the low cost chinese machines use them.  I use the pneumatic version on most of the cheap passives for anything >=0402 that I dont care about wasting.  The electric version is like the universal instruments feeder in that you can tell it to go back and forth.  Its essential when you have expensive parts that you cant afford to waste.

[Jackster]

i think he's still figuring out the feeder...

I also do not understand why he is so intent on designing his own feeder when there is so many very good (and cheap) ones already out there.  You can buy used Yamaha feeders on ebay for less than $50, and brand new generic electric versions (which are milled aluminum and will reliably pick 0201) are about $200 each with free DHL shipping.  You can get it cheaper from other vendors if you are willing to wait longer.

Reliable placing is all about repeatability which requires frame rigidity.  This is why good PNP machine's frame so heavy.  Even the feeders are all metal.   3D printed parts simply will not be able to reliably place precision parts repeatably with any real speed, and bolting together 2 pcbs to make a feeder will not give the rigidity needed to reliably place and align precision parts.  Meaning, you cant camera align a part that you can not reliably pick each time from the feeder.

You would have to build a mounting plate for real feeders. His method is just some aluminium extrusion which is cheap and easy to ship and easy for the end user.
Those cheap Yamaha CL electrics are then quite a lot more $$$ than their DIY feeders. It is not the aim of the project to make a cheap PnP to use with production-level feeders. If they can make their own, for cheap and be decently reliable they can easily get away with what they are doing.
You can get some 8x8mm feeder blocks for cheap from old PnP machines but those are heavier than their whole PnP machine...

They won't need anything that high-end for the market they are targeting. Said market, makers, are quite happy to build a PnP kit like this and mess around with it for a few days to make a handful of PCBs a month.

Yes, you can use this kit for a "production run". Hell, it is better than what I used to do a few years back which was doing it all by hand. Probably made around 1000 boards by hand before I started using a PnP.
Would have liked a kit like this for the Rs and Cs at this price back then.

[tomgat]

IMHO, feeders are essential even on small board runs because the bulk of the time is spent loading and unloading the machine. 

While there is a place for the NeoDen YY1 (and its now less functional Charmhigh equivalent), it takes a long time to load and unload this machines.   Also, their pin needle puller design are their weakest link.  I probably went through about 6 needle pullers on my old unit.

As far as hand making 1000 boards.  Damn... That sux.  Even for 1 prototype board I will still train the PNP machine simply because its allot faster than hand placing, but also because it wont smear the paste like hand placing will. 

[48X24X48X]

i think he's still figuring out the feeder...

I also do not understand why he is so intent on designing his own feeder when there is so many very good (and cheap) ones already out there.  You can buy used Yamaha feeders on ebay for less than $50, and brand new generic electric versions (which are milled aluminum and will reliably pick 0201) are about $200 each with free DHL shipping.  You can get it cheaper from other vendors if you are willing to wait longer.

Reliable placing is all about repeatability which requires frame rigidity.  This is why good PNP machine's frame so heavy.  Even the feeders are all metal.   3D printed parts simply will not be able to reliably place precision parts repeatably with any real speed, and bolting together 2 pcbs to make a feeder will not give the rigidity needed to reliably place and align precision parts.  Meaning, you cant camera align a part that you can not reliably pick each time from the feeder.

The Yamaha feeders are around $40-50 new. I have bought used ones for $10 before but wouldn't recommend it. The electrics ones can be had around $145. As long as you feed them with clean air, these feeders just work reliably.

[Mechatrommer]

where do you get yamaha feeder for $50? the cheapest i see is (ok i saw $48 yamaha feeder in aliexpress) juki feeder in aliexpress $80, the used other brand in ebay can go $200+. and then its not one, at least we need 10-20 or more pieces of them, or maybe 50 minimum for "midscale" production . even if you can get used unit for $50, they are not infinitely available... for me diy cheap cost feeder is crucial part in diy pnp like this.

[tomgat]

where do you get yamaha feeder for $50? the cheapest i see is juki feeder in aliexpress $80, the used other brand in ebay can go $200+. and then its not one, at least we need 10-20 or more pieces of them, or maybe 50 minimum for "midscale" production . even if you can get used unit for $50, they are not infinitely available... for me diy cheap cost feeder is crucial part in diy pnp like this.

The way I bought mine for $50 was to go through alibaba and I bought 250 of them at one time.  Yes, that is allot of money, but compared to most PNP machines, that is 1 feeder cost.  Buying them in by the 1s, you will pay about $90, but most of the time that is with DHL shipping.  It might make more sense to do this if you want to slowly invest in your feeders over time.  The thing I use them for is building a component libraries.  For the low cost passives, I have entire racks of the pneumatic ones that I only load them once labeling them with their part number.  Loading feeders is what times the majority of the time, and can be frustrating with needing reel tape, extenders, ect.   I can load and calibrate the machine with 50 components in about 15 mins.  The only real downside of the pneumatic version is that they are "dumb" and only go one way.  They can be wasteful, but I dont really care about that when I buy an entire real of 10,000 for $30.

This is the key advantage of the electrics.  All my ICs (or any of the expensive components) I load on electric feeders.  Its a really great combination, and I highly recommend it for anyone doing a DIY setup.  If you use feeders and use the designer pick and place positioning files, you will never hand assemble a board again. 

[Doctorandus_P]

I see a few misinterpretations (or at least differences in opinion) on what Stephen is trying to achieve.
If you've got two or more of those chinese entry level machines running for several hours a day, then this is clearly not the product for you.
If you've got a few prototypes to make each month, then hand assembly is tedious, and using an assembly service is quite expensive because of the setup costs.

If you want to show off how quick your chinese machine is (such as tomcat), then Stephens machine is also clearly not for you.
Hand assembling a PCB can takes up a lot of time, is tedious and error prone. And if you're just building prototypes, it really does not matter much if the machine runs for 20 seconds or 20 minutes, as long as it does not need constant tinkering to keep it working it's all right for me.

I have also not seen the sort of feeders I would like for my idealized PnP machine.
I don't want those USD 50 second hand yamaha feeders, even if they're reliable. They're too bulky and too difficult to load.
For prototypes you would want a quite different type of feeder then for production.
Most of the setup time is for loading feeders and such, and the feeders I'd want would be optimized for prototypes. For this they need a few properties.

1. It must be small. Something like 100x100x20mm, and hold about a (half ?) meter of tape.
2. when it's so small, you can store unused feeders in a drawer somewhere.
3. It must be cheap. Goal is to buy / make enough feeders so you never have to un-load a tape.
4. The feeder must be "smart", and know what part is in it. Either via some chip (eeprom) or a bar code that is scanned by the machine.
5. The feeder must be optimized for cut tape. So no splicing on of extra leader tape or other finicking stuff.
6. It does not treat parts as "expendable", because of the short pieces of tape the waste would be too high.
7. An option for using bigger feeders that can hold a full spool.
8. Integrated non volatile part count (so it remembers the count if it's taken out of the machine). This could be as simple as a light switch that shows when the feeder ahs <10 parts left.

The goal for me is to have feeders that are both very easy to fill and exchange, and that can also be put in random order into the machine, and the machine knows where the feeders are and what part each of them contains.
I'm thinking about using an N20 motor inside the feeder for reliability, and cutting the cover tape lengthwise with a knife and fold both flaps outward (several commercial systems use this).

Also, with those chinese machines, you just get a magical box. If that box does what you want, then go for it. But if you want something else, then it gets difficult very quickly.
I have never ever seen a PnP machine that's optimized for prototypes in the way I've described. If anyone has got a link, please post...

[tomgat]

The goal for me is to have feeders that are both very easy to fill and exchange, and that can also be put in random order into the machine, and the machine knows where the feeders are and what part each of them contains.
I'm thinking about using an N20 motor inside the feeder for reliability, and cutting the cover tape lengthwise with a knife and fold both flaps outward (several commercial systems use this).

Also, with those chinese machines, you just get a magical box. If that box does what you want, then go for it. But if you want something else, then it gets difficult very quickly.
I have never ever seen a PnP machine that's optimized for prototypes in the way I've described. If anyone has got a link, please post...

You make fair points.  Each of us are looking at different combinations for what works best for us.  I do allot of individual rapid prototypes, but I also do this for a living so my setup will be allot different than a hobby person just doing a couple of designs a year.  For that person, the NeoDen YY1 makes some sense.  It is basically taking over the older Charmhigh 36/48 systems.  Its feeder design also looks like it does well with individual strips. 

The main issue with all of these machines, however, is that the amount of time it takes to load the machine is generally > than the time to just do 1 board by hand.  To your point, no one has solved this problem, but a cheap/compact feeder would be a critical piece to solving it. 

In Stephen's case, from a business perspective he's stuck because on the one hand he's competing with the $3K PNP market which he has little chance in penetrating because its established, and the DIY build it yourself market.   Paying $1745 (LumenPNP V3) plus shipping/taxes, and your basically back to the same price as the other players which are more functional.  I dont doubt he will sell some machines.  I just dont see any real business problem he has solved.

Also, with those chinese machines, you just get a magical box. If that box does what you want, then go for it. But if you want something else, then it gets difficult very quickly.

I cant speak for the other chinese machines, but I know the Charmhigh and QeHi machines pretty well.  All PNP machines (even the $1M) have their quirks you have to work around.  Once the machine is running and stable, though, they all pretty much get the job done. 

As far as "want something else"... That is what the IC trays are for.     

Nothing stops someone from buying one of these low cost machines and putting OpenPNP on them, but it would just be a step backwards.  Their just a mainboard with stepper amplifiers.  The QeHi machines are ideal for this because their about as simple of a design as it gets.  You could convert them in a day, but the reality is that there is no real reason to do this.  The chinese software is actually pretty damn good, and gets the job done. 

My larger point and I see this with people complaining about the LPKF prototyping systems is that nothing about designing, making, or populating pc boards is "point and click" out of the box.  A person has to fully understand how each of these systems operate as well as learn their individual limitations/quirks/work arounds.  Its not rocket science to do this, however.  I make and populate custom 4 layer boards daily, and can do it in a couple hours.  Its not some very expensive equipment I use either.  Its just allot of hard, methodical, and persistent work. 

[48X24X48X]

There's a guy on YT from Vietnam that built machine used in his production using OpenPnP. The feeder is based on the Yamaha CL feeder.
For me, it is the most cost effective solution and proven. Loading a tape is just 1-2 minutes.

The way I bought mine for $50 was to go through alibaba and I bought 250 of them at one time.

You should have gotten a much better deal at 250 units! Something like $40 is more like it. Try to go to the source of these feeders. Places like KSUNSMT and Shenzhen Wenzhan provide better quality and also lower price. Just avoid any Aliexpress store, inflated price stores for foreigners. Alibaba with proper business background is still okay(ish).

I don't want those USD 50 second hand yamaha feeders, even if they're reliable. They're too bulky and too difficult to load.

They are not 2nd hand to begin with but of course they are (of those original made by Yamaha in the earlier days). These are new copies of the CL feeders. The design is simple but reliable and many if not thousands of factories in China are still using them. Close to 2 years using my machine, there's zero of them that went kaput except 1 where I bought it 2nd hand $10 just to try out. Even these feeders can be simply repaired with it's spare parts readily available.

[tomgat]

I don't want those USD 50 second hand yamaha feeders, even if they're reliable. They're too bulky and too difficult to load.
They are not 2nd hand to begin with but of course they are (of those original made by Yamaha in the earlier days). These are new copies of the CL feeders. The design is simple but reliable and many if not thousands of factories in China are still using them. Close to 2 years using my machine, there's zero of them that went kaput except 1 where I bought it 2nd hand $10 just to try out. Even these feeders can be simply repaired with it's spare parts readily available.

Until someone comes up with a better solution, the CL feeder design currently is the fastest and cheapest way for doing rapidly individual prototypes and also being able to run full production later if you choose to.  For me, the difference between prototyping and production has more to do with needing a conveyor versus a fixed frame design.  If a person has more time than they have money, then the DIY approach is the best approach.  However, if your trying to run a business (which ultimately just means you have the ability to write off the expenses), then a $6K/$8K machine which supports feeders is an easy choice.  I dont know all of the machines out there, but I know that Charmhigh and QIHE make great low cost PNP machines.  QIHE is superior only in the fact that their vision matching and pick/place speed is blazingly fast.

On the $10 feeders.  I have several of these.  The only issue I have found with the used feeders are the advancing gear and pusher is generally worn and occasionally misses its advance. For me, its not a big deal because the PNP machine will just give it a couple tries.   You can also buy replacement gears for these feeders and fix it yourself.  I have had limited success in adjusting out the wear, however. 

The electric feeders are definitely allot easier to load.  My only complaint with the electric feeders are that they all seem to have a software bug so that once they start advancing using their pneumatic sensor, they will at some point stop accepting requests to advance manually through the keypad.  A quick power cycle resolves the issue.  My approach to "fix" this issue was to mount a reset button on the machine feeder panel, but on 0201 components it can be frustrating because you have to recalibrate the feeder anytime it is reset/looses power.