PnP Machine for House Garage

[mendez]

I need a bit of advice on a pnp machine purchase.

Due to the Tariff increase, I'm planning on making all my boards in USA (at my garage) (0402, 0603, and 0805 sizes).

My original plan was to get an old Production PnP (ex. Juki, Europlacer) for my garage, but there is a possibility that I need to move to a new house/state in 1 year.

Should I purchase a benchtop (ex. NeoDen 3V) to save some $ and put on hold a more reliable/rigid pnp or stick to my original plan? My biggest concern is moving out a 900kg  (2K Lbs) machine (it will be a nightmare to move it from a garage)

Any advice will be appreciated

[rea5245]

I just bought a small lot (30 pieces) of assembled PCBs from China.

They slipped into the US without being hit with a tariff (shhh!!). However, when the ICs were sent from the US to China for assembly, they incurred a Chinese tariff. In theory, that should be refundable when the assembled boards are exported, but that requires paperwork and the company I used has not done it. (Admittedly, they have little incentive to do so - I'm the one who ultimately paid the Chinese tariff).

So before you take the plunge into buying a PnP and doing the assembly work yourself, have you figured out just how much the tariffs would actually cost you, and whether part of that would be refundable?

- Bob

[mrpackethead]

Just wait for next years election and yoru PNP tarrif problems will go.

[IconicPCB]

And if it does not change.. dont worry.. the whole planet is going to be in an equalised state... another world wide recession is just around the corner courtesy US trade sanctions.

[mrpackethead]

And if it does not change.. dont worry.. the whole planet is going to be in an equalised state... another world wide recession is just around the corner courtesy US trade sanctions.

Are there tarrifs on vacumn cleaners?

[IconicPCB]

to return to the topic.

I'm not familiar with the neoden machine.
The question to ask is what You want to achieve.
Hit the road running ?
How fast do You need to be running that is how many components per hour do you need to place?
How robust does the machine need to be and what performance in term of placement accuracy and repeatability do you need?
Some of the old ironmongery is heavy, bulky and worn out, some of the newer machines have lighter perhaps more flexib;le frame suggesting lower level of performance, Is that good enough?
You can always slow down the particular high resolution placement machine stroke in order to gain placement repeatability.

As in any endeavor there are trade offs to be had.
I await arrival of our new machine , a SmallSMT , to take place within the next couple of weeks.
I too shall add to the commentary on machine performance.

[mrpackethead]

Are you going to be selling those smallSMT machines in OZ iconic?

[mrpackethead]

I need a bit of advice on a pnp machine purchase.

Due to the Tariff increase, I'm planning on making all my boards in USA (at my garage) (0402, 0603, and 0805 sizes).

My original plan was to get an old Production PnP (ex. Juki, Europlacer) for my garage, but there is a possibility that I need to move to a new house/state in 1 year.

Should I purchase a benchtop (ex. NeoDen 3V) to save some $ and put on hold a more reliable/rigid pnp or stick to my original plan? My biggest concern is moving out a 900kg  (2K Lbs) machine (it will be a nightmare to move it from a garage)

Any advice will be appreciated

Its a bit of a depends question.

If you are goign to be doing any more than a few boards, then the desktop machines are just going to end up being a fustration. They requrie constant attention.      The next class of machines, ( SmallSMT, TVM925, etc etc ) are better but slow for anything but small runs, and they need some attention.    You start getting into more professional machines, and you get more reliability and less drama.

However, theres MUCH more to a sucessful SMT line than just the PNP.   DO not underestimate the need for good paste application, and reflow.   

Good 2nd hand is an option, but not one i recommend to anyone who does not have a good workign knowledge of the machinery. Its a steep learing curve, and you might be on your own.

[MagicSmoker]

Generally speaking, there are two broad types of PnP machines: high production rate and rapid prototyping. The former type is absolutely the worst possible choice for what you likely want to do as they are incredibly difficult to set up, the machines and "accessories" (like nozzles, reel holders, etc.) are expensive, and are just miserable to deal with unless you are making runs in the 10^3 to 10^6 range. Their saving grace is that they are *fast*, hence the common sobriquet, "chip shooter."

Rapid prototyping machines are easy to setup but tend to put down parts very slowly, and despite this slower speed may not be terribly accurate unless machine vision is an option (which greatly adds to the cost and setup headaches), and you will need machine vision for 0402 parts (generally, 0603 is the cutoff here). Venerable old standbys made by Quad and Mydata are still popular for rapid prototyping, despite being 20-30 years old at this point. I have no personal experience with "desktop" PnP machines, but all the anecdotal evidence I've seen is that they generally suck and that if one is building a few boards at a time then a manual PnP is typically a faster/better option (what I ended up with after dicking around with a Quad for 2+ years). A manual PnP is basically a vacuum pickup tool on an articulated arm or xy gantry that makes it much easier/quicker to pick up SMT parts and place them by hand than would be possible using tweezers or a handheld vacuum pen (e.g. Hakko 394). The one I have and which I have been very pleased with is: https://abacom-tech.com/product/manual-smt-pick-and-place-machine-ezpick/. For example, one board that I make a few of per year has 188 SMT components on it and it takes about 4 hours to place all the components by hand, but under 2 hours with the ezPick. Granted, the old Quad IV PnP could place all the components on this same board in about 15 minutes, but that was AFTER loading all the reel holders with the components, programming the centroid and Z height for each part (admittedly a one-time job, but it still took several hours), etc. If you need to make more than about 20-30 boards per batch the rapid prototyping type PnP will likely be faster (though still a long way from being economically justified), but if you are just doing a board here or there, especially if they are all different, then the manual PnP is likely a better choice.

[ebastler]

Due to the Tariff increase, I'm planning on making all my boards in USA (at my garage) (0402, 0603, and 0805 sizes).
[...]
Should I purchase a benchtop (ex. NeoDen 3V) to save some $

If you are worried about tariffs, it seems a bit counter-intuitive to buy a Chinese PNP. 
A 25% tariff does apply to those.

Assuming that the high tariffs are a transient thing, i.e. will last shorter than the amortization life of a PNP machine, you would probably be better off paying the tariffs on populated PCBs for a while. Or look at American-made PNPs and check whether the cost-savings math still works.

[SMTech]

In theory at least, of all the processes you could have done in China its SMT where you stand to save the least as the work is done by machines not people. That appears to be less true for lower volumes and finding an assembly house is not always as easy as it might be (a high quote is code for we have better things to do as much as anything else). However you will notice that large assembly houses with multiple location have them pretty much everywhere right across europe,asia&the americas dodging a China tariff for this one process should be easy, the difference is they might not have a nice online quotation portal.

[SMTech]

Generally speaking, there are two broad types of PnP machines: high production rate and rapid prototyping. The former type is absolutely the worst possible choice for what you likely want to do as they are incredibly difficult to set up, the machines and "accessories" (like nozzles, reel holders, etc.) are expensive, and are just miserable to deal with unless you are making runs in the 10^3 to 10^6 range. Their saving grace is that they are *fast*, hence the common sobriquet, "chip shooter."

Rapid prototyping machines are easy to setup but tend to put down parts very slowly, and despite this slower speed may not be terribly accurate unless machine vision is an option (which greatly adds to the cost and setup headaches), and you will need machine vision for 0402 parts (generally, 0603 is the cutoff here). Venerable old standbys made by Quad and Mydata are still popular for rapid prototyping, despite being 20-30 years old at this point. I have no personal experience with "desktop" PnP machines, but all the anecdotal evidence I've seen is that they generally suck and that if one is building a few boards at a time then a manual PnP is typically a faster/better option (what I ended up with after dicking around with a Quad for 2+ years). A manual PnP is basically a vacuum pickup tool on an articulated arm or xy gantry that makes it much easier/quicker to pick up SMT parts and place them by hand than would be possible using tweezers or a handheld vacuum pen (e.g. Hakko 394). The one I have and which I have been very pleased with is: https://abacom-tech.com/product/manual-smt-pick-and-place-machine-ezpick/. For example, one board that I make a few of per year has 188 SMT components on it and it takes about 4 hours to place all the components by hand, but under 2 hours with the ezPick. Granted, the old Quad IV PnP could place all the components on this same board in about 15 minutes, but that was AFTER loading all the reel holders with the components, programming the centroid and Z height for each part (admittedly a one-time job, but it still took several hours), etc. If you need to make more than about 20-30 boards per batch the rapid prototyping type PnP will likely be faster (though still a long way from being economically justified), but if you are just doing a board here or there, especially if they are all different, then the manual PnP is likely a better choice.

I think you might be surprised by just how easy a lot of "production" machines are these days, there has been much more focus on NPI from all the big players. Some machiens focus on it very heavily, Europlacer & Mydata build their business around low volume high mix as does Essemtec, all perfectly accurate machines (and Essemtecs can be small the other 2 not so much). If you can find one 2nd user that hasn't got into the hands of a broker or back to the manufacturer they can be priced quite nicely. (Essemtec refurb from factory is really not that cheap)

[MagicSmoker]

Generally speaking, there are two broad types of PnP machines: high production rate and rapid prototyping. The former type is absolutely the worst possible choice for what you likely want to do....

I think you might be surprised by just how easy a lot of "production" machines are these days, there has been much more focus on NPI from all the big players. Some machiens focus on it very heavily, Europlacer & Mydata build their business around low volume high mix as does Essemtec, all perfectly accurate machines (and Essemtecs can be small the other 2 not so much). If you can find one 2nd user that hasn't got into the hands of a broker or back to the manufacturer they can be priced quite nicely. (Essemtec refurb from factory is really not that cheap)

Yeah, I had my eye on an Essemtec when I decided enough was enough with the Quad, but I just couldn't justify it now matter how badly I wanted to based on the number of boards I make per year. You really need to be doing annual volumes of at least several hundred boards before any of the automated PnPs start to make sense.

[mendez]

Thanks everyone for all the feedback.

I will be doing min 500 boards per year (production). But, I will be doing some prototype boards as well. I'm trying to get away from China SMT production.

I guess, my biggest concern is moving a 900kg machine to my garage and then moving it out; so spend like 2K (forklift service) each time I decide to move the machine; or buy a desktop pnp for now (1yr) and then sell it.

[TK]

How large is the board and how many components?  500 boards/year, 42 per month, 2 per day... Do you really need PnP machine?  I estimate hand placing will be easier and you will get better results than from a cheap PnP machine.

[ebastler]

How large is the board and how many components?  500 boards/year, 42 per month, 2 per day... Do you really need PnP machine?  I estimate hand placing will be easier and you will get better results than from a chip PnP machine.

That's what I thought when I read about the volume. Look at it as "10 boards once a week", or "20 every other week", and it sounds quite manageable for manual placing. Recruit a kid from the neighborhood -- no tariffs!   

By the way, once you populate the boards, how do you solder them? Do you have a reflow oven; what failure rate do you get? Maybe better stick with an external service, in China or elsewhere, to do both placing and soldering for you?

Combined with your not-entirely-compelling argument about tariffs being your trigger for this decision, I get the feeling that the real reason is that you want a new toy. Which is a perfectly fine reason, of course, as long as you are honest with yourself. 

[mendez]

Thanks guys.... Well I have around 15 different PCB designs and my largest board has around 150 components (603 size); I tried doing them manually and 2 boards takes me around 4-5 hrs to complete (they have also some Through holes). The stencil printer I have is the UNIPRINT PM-LIFT and I have no issues with it; very accurate.

But manually placing the components (specially in the 150pcs board) is just a waste of man power since PCBs assembly is just 10% of my job. I'm most of the time programming, working in CAD and manufacturing in a CNC. So, my plan was to get the PnP so that I can do the boards in-house only for a few days and continue with my other tasks.

[mrpackethead]

Macrofab. They are now very competitive.

[Reckless]

I have 2 Juki 730s I'd let go for $2500/ea.  They need minor repairs, I just don't have the time to fix.  Something with video card I think.  They are pretty light ~1800 lbs and the 730 has 3 nozzle heads which multiplied placement speed.  For garage manufacturing they are perfect.  Seriously the easiest way to get into SMT business.  Spares are readily available in China.

[OwO]

Also worth looking into is JLC SMT: https://www.eevblog.com/forum/manufacture/jlc-pcb-prototype-assembly-incredibly-cheap/

[mendez]

I decided to make this instead; I might blog the build. I know it might be slow (500-100cph) but it will be a fun learning experience.

[ebastler]

I decided to make this instead; I might blog the build. I know it might be slow (500-100cph) but it will be a fun learning experience.

Is that a kit (which one?) or your own design? What are you planning to use for control and software?

[mrpackethead]

Think a lot of us might have thought we could/would build a PNP machine.    IN the mean time you need to get products otu the door.....

[Ribster]

I decided to make this instead; I might blog the build. I know it might be slow (500-100cph) but it will be a fun learning experience.

Is that a kit (which one?) or your own design? What are you planning to use for control and software?

I know at the school in belgium where i was it, they bought a PnP kit from a german guy. With decent servo drives and stuff.
If you want to know, i can request the exact model.
It was well documented, but they spent a lot of time setting it up tho.
Think the school started to develop their own software in Qt also, to communicate with the machine.
In the end, they just bought a decent one :p

[Smallsmt]

I know at the school in belgium where i was it, they bought a PnP kit from a german guy

https://www.vbesmens.de/de/bestueckungsautomat.html

[mrpackethead]

In the end, they just bought a decent one :p

Its what everyone who's serious eventualy does.      First you try to make a screwdriver out of a bit of steel at home,  but its not the right shape.. You then buy one at the $2.. it works a couple of times,  burrs screws and then you slip with it, and hurt yourself.  you then get a $10 one from the hardware sthpp, and it works, but its still a bit meh!   Eventually you go and get a decent tool from an engineering supply and it works really well, and will last you a lifetime.

[Ribster]

I know at the school in belgium where i was it, they bought a PnP kit from a german guy

https://www.vbesmens.de/de/bestueckungsautomat.html

Exactly that one !
It was very nicely designed. But the hand assembly of the main PCB was a nightmare i remember.
It used a few PIC's IIRC. The power stage for the three-phase motors gave them a hard time.
The parts blew up a few times. They wasted a lot of time with that. But the support of the creator was actually pretty good on that part. So no negative things.
It's a part of doing stuff yourself. Issues everywhere. It's not a problem if you account this in, just don't expect 'kits' to work without any failure.

[Ribster]

In the end, they just bought a decent one :p

Its what everyone who's serious eventualy does.      First you try to make a screwdriver out of a bit of steel at home,  but its not the right shape.. You then buy one at the $2.. it works a couple of times,  burrs screws and then you slip with it, and hurt yourself.  you then get a $10 one from the hardware sthpp, and it works, but its still a bit meh!   Eventually you go and get a decent tool from an engineering supply and it works really well, and will last you a lifetime.

If you look back from the point where you bought the decent tool, you think: what did i do.
But knowing you need a decent thing is something that grows. If you are getting into screwdrivers (bad analogy, but you get the point), you might not know if you need the top notch thing and/if you are going to need it daily or once in a blue moon.
The learning process and identifying what you need is also a part. Going for the best thing each time might sets you back more if you just don't use it.

[mendez]

I decided to make this instead; I might blog the build. I know it might be slow (500-100cph) but it will be a fun learning experience.

Is that a kit (which one?) or your own design? What are you planning to use for control and software?

No, this is a OpenPnP CAD file. However, I'm modifying a few files (ex. double shear belt tensioner, exterior dimensions, and head). The software will be OpenPnP with a Smoothie V5.x Board

[mendez]

In the end, they just bought a decent one :p

Its what everyone who's serious eventualy does.      First you try to make a screwdriver out of a bit of steel at home,  but its not the right shape.. You then buy one at the $2.. it works a couple of times,  burrs screws and then you slip with it, and hurt yourself.  you then get a $10 one from the hardware sthpp, and it works, but its still a bit meh!   Eventually you go and get a decent tool from an engineering supply and it works really well, and will last you a lifetime.

  I won't consider the screw driver analogy an Apples to Apples comparison; however, I understand your point. I know where I'm getting my self into; I will purchase a production level PnP as soon I moved; that is why I will try to keep this under budget. I looked into getting a Manual PnP but I decided to give this a try.

Here is my "PLAN":
1. Budget : Less than $1500 (Already spent $1000)
2. Time to finish the project: Before June 30th
3. Use modular trays dedicated for each project file; this helps me reduce setup time. I don't plan to use Feeders (but I might make 1 or 2 just in case);

Yesterday I purchased 80% of the parts required for the built; including the Aluminum Frames, pre-made head, HIWIN Linear Rails, controller board and steppers; so  23 days to go.

[NorthGuy]

Its what everyone who's serious eventualy does.      First you try to make a screwdriver out of a bit of steel at home,  but its not the right shape.. You then buy one at the $2.. it works a couple of times,  burrs screws and then you slip with it, and hurt yourself.  you then get a $10 one from the hardware sthpp, and it works, but its still a bit meh!   Eventually you go and get a decent tool from an engineering supply and it works really well, and will last you a lifetime.

Or, perhaps, as you go through screwdrivers you hone your driving (or screwing if you wish) skills, and by the time you get to the most expensive screwdriver you finally know how to screw (or drive). You swear by your new super screwdriver, but in fact, with your newly-acquired skills, the initial $0.50 screwdriver would work just as well.

[Kjelt]

Always nice  to see a build so yes please do start a topic.
Two things that trigger me on your picture.
- the base plate covers the entire X-Y area so it looks more like a cnc machine than a pnp machine.
look at real pnp machines for inspiration
- the two X axis are connected and driven by belt. For p&p not really needed since the head gets no counterforces , so if one of the x ax is is driven the other will follow.

[djacobow]

Also worth looking into is JLC SMT: https://www.eevblog.com/forum/manufacture/jlc-pcb-prototype-assembly-incredibly-cheap/

This is only available within China for now. But it looks like it could be very interesting.

I've been getting some boards done by PCBWay. It's not as cheap as I'd like. I mean, the PCBs are cheap, but on a per pin basis, they are charging me what I think is a lot. Plus, there will be the tariff, and the last couple of orders I have made from them have not "slipped by customs." I think any order over $500 will get the noticed.

Anyhow, I have thought about doing my own PCBA for my own projects. Boards for my current product have about 115 components on them, but only 24 unique and only a few smaller than 0805. I can hammer out a few an hour even with a soldering iron, though there will usually be a bit of rework.  But I do want to make a few hundred this year and it just doesn't seem like a great way to spend my time.

I dunno. Cost-effective PCBA for a small business still seems "unsolved", at least unless your boards go into something with a healthy margin.

[1276-2449-1-ND]

Last year I bought a little CHMT32VA for my business and it was the best decision ever. I only need to make a couple hundred or so boards a month (for just over a dozen designs), and standardized my components so just about everything can be done with the machine. The workflow is so tight now that I can do an entire run in a day, instead of a week the first few months. The last run I got so caught up in it that now I'm overflowing with PCBs. Did I just go through an entire reel of TL072s?

I'm still using a hacked toaster-oven and a DIY laser-cut paste spreader with stencils cut on a hobby stencil-cutter. These work great and are fast so there hasn't been much of an incentive to change.

I can produce things for much, much less than getting them done in China (still get PCBs made there though), and the savings get passed on to my customers. Who would have thought that it's cheaper to get things made locally than in China? Less hassle too.

[revenue_controls]

check out https://www.eevblog.com/forum/manufacture/smallsmt-pnp-adventure-learning-and-documenting/50/

I have looked at the $10k to $20k US machines (because I did want to buy US) But decided on a SmallSMT machine. I am very happy with it and I think it was actually less expensive than the US machine (SMTmax) I was looking at, even with the added tariff! Additionally the SmallSMT machine is built very rigid, the US machine does not.

I do about 400 production boards / year and prototyping but mostly I bought it to play with! I moved the machine into my work shop (5 steps up) by shear will power.

are you on the east cost? Contact me if you would like the details.

[mrpackethead]

even with the added tariff!

Was it 25%?

[mrpackethead]

What i did, was buy some good 2nd hand machines.  I make about 300-1000 panels per month.    I have 150 Slots for CL feeders, and a big tray. That has almost been neough to allow me to keep all my parts on the machines full time.

I'm able to switch between items with about 10 minutes effort.  ( changing supports, rail widths, etc ).     This is only possibel becuase i dont' have to strip down the machine.  Its not optimal placement and moving components around would speed it up, but I get abetween 12,000-14,000 pph actual placement rates.

Would i suggest others to do this.  Yes. but theres a proviso.  Only if you either have some solid experience with the machines you are going to buy, Or you have someone who can help you out.   Or be prepared for a long learning curve.

Also spend as much effort on Paste applicaiton and ovens as you do on teh PNP. They are not afterthoughts.

[Smallsmt]

What i did, was buy some good 2nd hand machines.  I make about 300-1000 panels per month.    I have 150 Slots for CL feeders, and a big tray. That has almost been neough to allow me to keep all my parts on the machines full time.

I'm able to switch between items with about 10 minutes effort.  ( changing supports, rail widths, etc ).     This is only possibel becuase i dont' have to strip down the machine.  Its not optimal placement and moving components around would speed it up, but I get abetween 12,000-14,000 pph actual placement rates.

Would i suggest others to do this.  Yes. but theres a proviso.  Only if you either have some solid experience with the machines you are going to buy, Or you have someone who can help you out.   Or be prepared for a long learning curve.

Also spend as much effort on Paste applicaiton and ovens as you do on teh PNP. They are not afterthoughts.

You went on the big way and it's nice to see you got it working.
Some customer are on tight space and need smaller solutions.
I agree your machines are more reliable working and maybe more fast too.

[jmelson]

You went on the big way and it's nice to see you got it working.
Some customer are on tight space and need smaller solutions.
I agree your machines are more reliable working and maybe more fast too.

I bought a Philips CSM84 in 2007.  Yes, it is BIG, and heavy.  Those are actually not bad things in highly dynamic machines.
Mine has 3 nozzles, which I have set up for 3 different part sizes.  My machine has no vision, which is a definite drawback for use with fine-pitch ICs,
but uses centering jaws for the smaller parts.  It has had a few breakdowns over the years, but I have been able to fix all of them.

It definitely is faster than some of the machines I've seen videos of, it can do about 3000 parts/hour.  If you set up the feeder placements and placement order carefully, it can mount 2 or 3 parts each time the head moves over to the feeders.  I don't bother with that optimization.

Jon

[Kjelt]

I bought a Philips CSM84 in 2007.  Yes, it is BIG, and heavy.  Those are actually not bad things in highly dynamic machines.
Mine has 3 nozzles, which I have set up for 3 different part sizes.  My machine has no vision, which is a definite drawback for use with fine-pitch ICs,
but uses centering jaws for the smaller parts.  It has had a few breakdowns over the years, but I have been able to fix all of them.

It definitely is faster than some of the machines I've seen videos of, it can do about 3000 parts/hour.

Use it for the majority of jellybean parts and use a second cheap machine with vision for the few delicate parts. Perhaps build the second one yourself with open pnp

[jmelson]

Use it for the majority of jellybean parts and use a second cheap machine with vision for the few delicate parts. Perhaps build the second one yourself with open pnp

Hmmm, an interesting thought!  But, what I do now is have the CSM84 mount all the parts, then I inspect placement and nudge the fine-pitch parts to better alignment.  Kind of tedious, and if I forget to nudge one that needs it, then I have some rework to do.

Thanks for the idea,

Jon

[Kean]

Use it for the majority of jellybean parts and use a second cheap machine with vision for the few delicate parts. Perhaps build the second one yourself with open pnp

Hmmm, an interesting thought!  But, what I do now is have the CSM84 mount all the parts, then I inspect placement and nudge the fine-pitch parts to better alignment.  Kind of tedious, and if I forget to nudge one that needs it, then I have some rework to do.

Thanks for the idea,

Jon

This is what I do with my TM240A (no vision, drag feeders).  I mostly do 0603 and 0.65mm pitch or larger, but sometimes I do 0402 and 0.5mm pitch.  Placement accuracy isn't great but I just need to do quick manual inspection and a little careful nudging.

Using the PnP saves me time over hand placement, but I'm typically only doing batches of 20 to 50 boards, and only about once a month.  A significant reel swap (which is typical as my designs are for different clients and have few common parts) can lead to almost no time improvement in small batches.  I do have a reasonably decent "3040" manual stencil printer.

If building 10 or less boards I'll usually do hand placement.  Larger runs get outsourced, as well as the occasional non-rush smaller runs of a proven design.

[mendez]

I started the pnp build; the only parts missing are the Head Unit Assy that will be available by the end of the month. I got Juky and CP40 nozzles for testing. I will start a new topic soon.

[Ribster]

Where did you buy the kit from? Or are they individually sourced components ?
Looks like a lot of fun! For the hywin (or knockoff) lineair rails, look it up on youtube how to clean them.
I had a delta 3d printer with those rails and they clogged up fairly quickly.
De assembling them and cleaning them will make them more robust.
Something to consider doing!

[Kjelt]

Where did you buy the kit from? Or are they individually sourced components ?
Looks like a lot of fun! For the hywin (or knockoff) lineair rails, look it up on youtube how to clean them.
I had a delta 3d printer with those rails and they clogged up fairly quickly.
De assembling them and cleaning them will make them more robust.
Something to consider doing!

If they are genuine HiWin do NOT clean them, they are pre greased and they should not run as an oily bobsled, they are made to have zero tolerance on 3 axis so they should be sturdy, mistake many beginners make. They are made for cnc machining to withstand big powers on all axis during milling, they are not really made for low powered 3D printers or p&p machines, then you need to beef up you're motors.

[Harvs]

I've started building basically the same thing, using those same plans as a baseline.

For me, it's more about the interest in toying around with some different CNC stuff. I've worked with and built routers, lots of steel and heavy components. I'm very interested to see how well the small rails, belts and ally extrusion works for positioning something without significant load. If I can make boards with it at the end, then all the better.

I don't have high hopes given the amount of effort that normally goes into the mounting surface precision rails normally bolt to, but I think it's a worthwhile exercise.

Building one of these is more or less "building by shopping" from what I can see so far.

[Kjelt]

If you build yourself you can always upgrade later. For instance replace the belts with spindles , replace the steppers with closed loop steppers or servos, perhaps even add a glassrail sensor to get a few um position accuracy.

[mendez]

Where did you buy the kit from? Or are they individually sourced components ?
Looks like a lot of fun! For the hywin (or knockoff) lineair rails, look it up on youtube how to clean them.
I had a delta 3d printer with those rails and they clogged up fairly quickly.
De assembling them and cleaning them will make them more robust.
Something to consider doing!

I individually purchase the parts. Thanks for the feedback; I purchase HYWIN Brand and I will look at the videos; thanks!

[wilhe_jo]

Why not using a cheap laser engraver for XY?

https://www.aliexpress.com/item/miniUSB-DVP-3830-Laser-A3-Desktop-USB-CNC-Laser-Engraving-Engraver-Cutting-Machine-Without-Laser-Head/33037159213.html

should be plenty accurate for the positioning part.

73

[Harvs]

Why not using a cheap laser engraver for XY?

https://www.aliexpress.com/item/miniUSB-DVP-3830-Laser-A3-Desktop-USB-CNC-Laser-Engraving-Engraver-Cutting-Machine-Without-Laser-Head/33037159213.html

should be plenty accurate for the positioning part.

73

I have my personal doubts about it being plenty accurate, but I've never worked with any CNC hardware using plastic wheels instead of precision rails so I have no evidence to back those doubts up.

But maybe more importantly, it doesn't really gain you anything.  The XY stage is a simple and cheap part of the whole build.

As a basis it looks like we're both working off the design from Anthony Webb over at https://hackaday.io/project/9319/logs

All the aluminium extrusion, brackets and hardware for the base and gantry came to ~$100AU.  Linear rails ~$100AU for the knock offs, or about 3.5x that if you get genuine HIWINs.  Once you add motors, smoothieboard clone, belts, etc you can probably add another $100AU.  So it's maybe ~$100AU more than that laser xy stage, but it's the size and shape that's needed to complete the rest of the design.

There's a lot of parts and details required to have a truly neat finished build for a pnp, of which the base and xy stage is only maybe only 1/4 - 1/3 of the cost.  So to me it doesn't make sense to scrounge for savings in this area.

[wilhe_jo]

Well it seems they are much better than one might expect...

That guy shows off some laser exposure tests of PCB material and some photos from a microscope:


In a different thread I mentioned, that I have in mind very cheap machines built for a specific design which is quite simple.

I was thinking of a PNP head which is not able to rotate.
Instead, I plan to let the machine pick the part and place it on a clear (for the vision of needed) acrylic plate which is rotated.
After rotating the part is repicked and placed.

So the head just needs to be lowered.
My vinyl cutter does this using a flexture and an electromagnet.
Would be very easy to replicate this... or just use the spare part...
No need for springs to accommodate variations in height, just adjust the down-pressure by setting the magnets current.

Should be much easier from the mechanics point of view but slow.

For my purpose that would be fine.
I'd just let the machine populate the next PCB while assembling, testing and packing a unit...

so XY is  cheap.
Pick/Place is simple.
Rotating without the need of rotating fittings, hollow steppers and such is fine.

The real problem are in my opinion the feeders.

I haven't found an economic solution for this yet.
Well, that's not totally true.. over in the openpnp group I found this one:
https://docs.mgrl.de/maschine:pickandplace:feeder:0816feeder

But i didn't had time so far to print and test one of these...

73

[Fire Doger]

In a different thread I mentioned, that I have in mind very cheap machines built for a specific design which is quite simple.

I was thinking of a PNP head which is not able to rotate.
Instead, I plan to let the machine pick the part and place it on a clear (for the vision of needed) acrylic plate which is rotated.
After rotating the part is repicked and placed.

So the head just needs to be lowered.
My vinyl cutter does this using a flexture and an electromagnet.
Would be very easy to replicate this... or just use the spare part...
No need for springs to accommodate variations in height, just adjust the down-pressure by setting the magnets current.

I don't think it will work, plus the changes in software...
Hollow shaft nema8 costs 9$ on ebay, a tmc stepper driver is more expensive....

[Harvs]

By all mean give it a go.  It may not be the most conventional, but building robots is interesting and can be fun if there's no looming deadline. OpenPNP looks pretty extensible and well writing from the small amount of looking I've done, so shouldn't be too hard to add your custom pick sequence of moves.

[spanner888]

hollow shaft nema8 9$ on ebay

....I want want want - been looking for cheap ones for years, but can't find, used several different model numbers and all sorts of search tricks. A search just now shows $A30 and up.

Can you share or PM any links please.

[wilhe_jo]

Well, you can find some...
https://m.de.aliexpress.com/item/32865075882.html

But making a low runout adapter to a nozzle which holds a sufficient vacuum is not too easy.

73

[Koen]

Send an email to 439291063@qq.com

He supplied all the specialized parts of my PNP like the head, hollow motors, nozzles, valves. He has his own 2 to 4 head designs or can make customs. Great guy, found him on Taobao in 2016.

And OpenPNP is great. It just works.

[mendez]

I've been busy with personal stuff but I was able to work on my PNP today. Please see below pictures. All I need now is to add the down camera, tape feeders and start placing parts. Everything seems to work fine (vacuum, rotation, repeatability ).

FYI: building the machine was not hard at all; but the cable management was key....prepare to have a lot of long wires.

[mrpackethead]

thats looking really good!  I'm really tempted to build one just for fun now.

[Harjit]

I've been busy with personal stuff but I was able to work on my PNP today. Please see below pictures. All I need now is to add the down camera, tape feeders and start placing parts. Everything seems to work fine (vacuum, rotation, repeatability ).

FYI: building the machine was not hard at all; but the cable management was key....prepare to have a lot of long wires.

Your machine seems very interesting. Can you elaborate on it, the build, performance, etc.?

[Harvs]

@mendez your machine's looking nice.

Here's where my build is at.  As you can see I've given up on cable management off the head, I think the number of cables going up to the head is a bit needlessly ridiculous.  I'm just going to make a custom control board for the head, so only USB, vac, power need to be transported up to it.  At the moment I have 7 cables and lines going to the head and it's in it's most simplest form. The capability to add extra widgets without trying to get things through the drag chains would be very useful in future.

Biggest issue for me at the moment appears to be the 40T pulleys I've used.  The machine in the HaD link used 52T, so I though 40T would be a safe bet, but I probably should have done the math as 40T is 0.4mm per full step! So I'm at 1/4 microsteps just to get 0.1mm, which is holding back it's accuracy (although it can move pretty fast.)  I have some 16T pulleys I might switch over to, but that does mean making new brackets and probably changing my idler guides somewhat.

Some testing last night without the bottom camera setup yet:
https://youtu.be/q1Wxt_wY1CY

[Harjit]

@Harvs, very impressive.

If you are going USB, why not go Bluetooth or Wifi? Then, you only have to send power over?

Can you elaborate on the pulley math? When you say one full step @ 40T is 0.4mm, this is based on some pulley pitch?

Will the head rotate to fix any angle error during component pick up?

Do you step the motors yourself? Or do you issue a high level command the there is a stepper motor controller board that does the motion profiling and stepping? Are you using pieces out of 3D printers? They have a similar need.

Can you share more details around the components and design? I'd like to learn more before jumping in and building one.

[Harvs]

@Harvs, very impressive.

If you are going USB, why not go Bluetooth or Wifi? Then, you only have to send power over?

Can you elaborate on the pulley math? When you say one full step @ 40T is 0.4mm, this is based on some pulley pitch?

Will the head rotate to fix any angle error during component pick up?

Do you step the motors yourself? Or do you issue a high level command the there is a stepper motor controller board that does the motion profiling and stepping? Are you using pieces out of 3D printers? They have a similar need.

Can you share more details around the components and design? I'd like to learn more before jumping in and building one.

Hi Harjit, there's nothing very impressive about it (other than the time investment), it's pretty much following the fairly well trodden route of building an OpenPNP build.

The machine I'm largely copying is this one https://hackaday.io/project/9319/logs

But head over to the OpenPNP project to see other examples.

But to answer your questions:

The USB is mainly for the camera, but may as well put a hub IC up there and run the control signals as well.  The OpenPNP folk don't recommend using a IP camera due to latency issues, never tried it myself but I'm pretty confident they've been down that rabbit hole.

The belts are GT2 size, which has a 2mm pitch, so it's simply 40 tooth*2mm for the circumference / 200 whole steps per rev for standard steppers. So 40*2/200=0.4mm.  I do get reasonably reliable 0.1mm moves, but it's not that great.

The vid I showed wasn't using the bottom camera, so it can only go off the feeder (tape) alignment and the rotation of the part on the board.  I've since got the bottom camera working, so now it takes the part over the bottom camera after pick and corrects for misalignment of both offset and rotation from the pick (in theory, it does seem to work pretty well in practice as well).  There's lots of videos on youtube of this if you search for OpenPNP.

OpenPNP generates G-code outputs which you direct to whatever controller you've got.  The "standard" route is pretty much using a smoothieboard (or compatible), as they have enough stepper drivers and IO to do everything from the one board.  So you can pretty much get one of those and the problems solved.

I really have nothing to add beyond what's on that HaD link above unless you want info on something specific.  There's a google docs spreadsheet in that link that has the BoM, which is very useful as there's a lot of parts to order.

Other than that, there's nothing particularly difficult about it, just it takes quite a time investment (both waiting for the 20odd aliexpress orders you'll make) and the careful build/setup.  There's lots of interesting problems in the setup and tuning, like I never thought to consider that the sensor on the USB "microscope" camera I used is actually a 10 or so degrees off being true to the body which causes all sorts of issues.

But overall it's only worth it really for the interest in doing it.  The difference in cost between the DIY route and the low end Chinese machines doesn't really make it worth it from a time perspective.

[Harjit]

Thanks for gently pointing out that there is USB camera data that needs to be transported. I didn't think about that but it makes perfect sense.

How are you handling the interface between the pick up motor and the vacuum line? Are you doing it the way the HackaDay article has it?

Several people had linked to stepper motors with a hollow shaft and I didn't follow why that was important but now I get it.

Have you come across a block diagram of the entire machine i.e. the controller, the motors, limit switches, wiring, vacuum circuits, etc.?

[GerardG]

Hello,

In some high schools in Europe they use the Neoden4 PnP machine.

http://neodentech.eu/

[Ribster]

Maybe something interesting on topic :

[mendez]

@Harvs; Your machine looks nice too!!

Here is an update of my PnP. I been busy with work but I finally finished. I must say....making this PnP took me a lot of time; because I had to modify a lot of parts.
-Autofeeder (up-to 24)
-Nozzle Tip changer (4 Max)
-Dual Cameras
-Dual Nozzles
~500CPH (this is for prototyping not for mass-production)

[Harjit]

@Mendez - super impressive!

I've got a similar build planned. Where did you get the brackets that connect the different rails and extrusions - essentially, the orange pieces?

Looks like you are driving the Y axis on one side. How has that worked out? I've read that it isn't a good idea and I've read it works fine.

It seems like if you use linear rails, driving on one side is okay but if you use wheels on V-style rails, maybe not?

What width belts are you using? 6mm? 9mm?

[Kjelt]

The orange pieces look 3D printed. If so which material did you use and is it sturdy enough?

Looks like you are driving the Y axis on one side. How has that worked out? I've read that it isn't a good idea and I've read it works fine.

I also shared this worry in the past.
It depends on the application. For P&P where there are literally zero counterforces it should be fine.
For a CNC mill it would be a big nono.

Actually what could happen is that the bridge that is fixed on the belt in this case could be less than 90 degrees straight. Since the bridge is a sturdy fixed connection this can only be a small offset.
And could vary with the position of the head.
You can try this by feeling it by hand, if your linear rails are not pretensioned or have slack this could become worse. A solution would be to use double wagons with some space in between on both sides fixed with the bridge, at that point it will be very hard for the bridge to stay out of line.

If you read in the SmallSMT topic their $20k P&P machine also use a single sided ballscrew with servo motor and they can place 0201 parts so it looks like it is feasible.

[Harjit]

@Harvs; Your machine looks nice too!!

Here is an update of my PnP. I been busy with work but I finally finished. I must say....making this PnP took me a lot of time; because I had to modify a lot of parts.
-Autofeeder (up-to 24)
-Nozzle Tip changer (4 Max)
-Dual Cameras
-Dual Nozzles
~500CPH (this is for prototyping not for mass-production)

Any chance you can post your design, STL and bill of material files? Thanks.

[Fire Doger]

Any chance you can post your design, STL and bill of material files? Thanks.

https://docs.mgrl.de/maschine:pickandplace
http://www.betztechnik.ca/store/p34/Pick_and_Place_head_-_dual_nozzle_-_OpenPnP_compatible.html
I identified some parts from these 2
And some parts can be replaced from ebay common parts like motor mount.
I would guess on 800-1k € I don't know if he is selling the head but I don't like it. I would choose something similar to these https://www.aliexpress.com/popular/pick-and-place-heads.html
3D printed nema8 shaft to fitting coupler is a joke and very cheap to skip cost...
*I wouldn't recommend building it if you need a machine.
It would be cool though if more PnP experienced users contributed to make a nice open source machine (designed to work, not all 3D printed craps).
https://www.ebay.com/itm/for-NEMA-17-Stepper-Motor-Mount-Plate-Fixed-Bracket-Holders-For-3D-Printer/352650307456?hash=item521b98a780:g:8wEAAOSwPT5cwtwM

[stoyanoff]

I have question! Can you share what are the tiniest parts which your home made machines can place correctly? For example have you tried with TQFP64? I am talking about machines with top and bottom cameras?
I am building one and I am waiting for my control pcb and drivers and I am not sure if have much bigger expections then really achievabled!
Thanks!

[jmelson]

Looks like you are driving the Y axis on one side. How has that worked out? I've read that it isn't a good idea and I've read it works fine.

My $100K Philips CSM84 works the same way.  It has two parallel linear slides, and a coarse-pitch ballscrew on only one slide.  Works fine.  But, these are pretty big THK linear slideways, and on the driven end, the gantry spans quite a distance between the two slider blocks.

Jon

[Harjit]

@Fire Doger - thank you for the links, especially the eBay ones. I'll check if those will work for me.

@mendez's design seems interesting because he seems to have made the pulley mounts more robust and the 3D printed pieces look really good. I'd love to get the design files, 3D print material information, and info. on how well they are working.

[wilhe_jo]

FYI there's currently an EleksDraw XY deal on bangood... 136€ delivered... couldn't step back from ordering such a toy.

At this price point I'll give it a try for a weekend or so to see if this could actually be a plattform for a single-prototype-board pnp-machine.

Just to repeat it for the new folks: For a single prototype, you're typically dealing with small 5-pcs strips from digikey and such odds. So feeders are not the main problem.

If you could manage the toolschain in such way, that the placement is a 1-click operation, this machine would/could be helpfull.

73

[mendez]

@Harvs; Your machine looks nice too!!

Here is an update of my PnP. I been busy with work but I finally finished. I must say....making this PnP took me a lot of time; because I had to modify a lot of parts.
-Autofeeder (up-to 24)
-Nozzle Tip changer (4 Max)
-Dual Cameras
-Dual Nozzles
~500CPH (this is for prototyping not for mass-production)

Any chance you can post your design, STL and bill of material files? Thanks.

I honestly build it by trial and error; I don't have an assembled 3d model or bom. I do have the 3d parts if you need anything specific. As soon I get some relax time I put a model together with bom to share.

[Harjit]

Thanks.

Please do post the files you used to make the 3D parts so that others and I can use what you have by modifying/adjusting them.

Also, what material did you use? Any concerns about the stability (warping?) of the pieces?

On the X and Y-axis, you used a large pulley on the stepper. I keep thinking it may be better to use a small pulley so that you have more precision on the Y axis. What do you think?

Finally, you only drive the Y-axis on one side. I've seen people do this and also drive on both sides. Is it working well?

<edit: added more questions>

[Fire Doger]

This machine is not ment for build and forget, it always needs adjustments and fiddling around.
If you take a look at any 3D printer community with self assembly printers you will see that even with identical aluminium machined parts everyone is getting a different result. Something that worked well for him doesn't mean that will work on you.

For example in linear rails if they are good quality on the undriven side gantry won't jog noticeably but in your case it might be.

Precision depends mostly on the motors, one step skip and it's gone because they don't have encoder for closed loop control.

3D models are available somewhere in the links that I have posted before.
Building a xy plotter is the easiest part of the process.

Assembling a 3D printer may take a couple of hours for someone unexperienced, finding the best settings may take weeks... (And we are talking for exactly the same hardware with thousands of other users)
Without machined parts settings will be all over the place, drifting, etc...

I would suggest to try to design your own parts and you will see on what parts/faces you need precision