Unboxing Our New Desktop Pick and Place Machine
sparkfun.com
sparkfun.com
Let me say that the needs or SparkX are slightly odd.
We need to build 10-50 of a design and see how it
sells. As it sells we may need to build 100-500pcs.
If a design needs more than 500 pieces then it
immediately gets moved over to our proper SMD
production lines with much more capable machines.
SparkX needed something small and quick to setup.
After evaluating all the various vendors we decided
on the CHMT36VA. It seemed to be the best fit of
low-cost and most flexible while being able to get
the job done.
... the nail in the coffin for the Neodyn 4 in my eyes
... he says the feeders aren’t that great, they are
challenging to load, and machine makes mispicks quite
often. Why spend ~$10k on a machine when I can have a
cheaper machine with less hassle? The CHMT36VA is far
from perfect but I can work around the problems.
Can you tell me what the basic steps are? Pricing information would be appreciated!
The big issue is really setting up manufacturing and supply chain on a large scale when you do get it working.
We're building prototypes of a complex bit of consumer electronics for consumer testing. We can do full industrial design, mechanical design, electronic design get the components made, assembled and delivered in 6 weeks. The fundamental limiting factor on how fast we can do it is the time it takes for the PCB to be made. From an electronics point of view, schematic capture and layout can be done in a week, and board assembly can be done in a day (if you're in a hurry) but there's nothing we can do about the 3 week lead time on the PCB because it's 8 layers with blind laser drilled vias.
Having an engineer who would otherwise be billed out at $150 - $200/hour hand placing parts and soldering boards is a waste of our in-house resources.
Proximity and flexibility still matters somewhat though, so there might still be a market for a pick and place machine between this type of machine and a professional one. In China there are pcb factories offering assembly for a low fee. They have a multiple pick and place machines with common components after each other.
The reason it seems like a good idea is people don't see things like the time spent tweaking as a cost. It is also to some extent true for laser cutters and 3d printers.
Pick&place, as well as 3D printing, begins to make sense at industrial scale, with huge expensive machines. Sparkfun might be a borderline case, but I suspect even they won't use these cheap pnp machines in the longer term.
If you are a hobbyist or a low-scale manufacturing operation, you are much better off using MacroFab, PCBNG, Small Batch Assembly or AISLER.net for electronics production, and Shapeways for 3D printing (SLS). Alternatively, for quantities of ~10 of electronics devices, it makes sense to order your boards and stencils from OSHpark or AISLER, place components yourself using tweezers, and either use a modified oven or a hot air soldering iron for reflow.
This is speaking from experience (as a hobbyist/maker, electronics design engineer designing proof-of-concept and small-scale production devices, and https://PartsBox.io/ founder).
For 3D printing this is becoming less true every day. 3D Printers are now extremely cheap, where sub-$300 printers make sense at low scale usage. A typical print from an outsourced printer like Shapeways might cost you about $20, and once you factor in the cost of the material when you print yourself, it only takes about 20 prints for the printer to pay for itself, and you have the added benefit of having your print in an hour rather than a week.
Also, it's just lots of fun to work with the printer.
3d printed parts and PCBs can be done in as little as 12 hours. You can then without any modification do small production runs anywhere from ten to a thousand pieces (which isn't unrealistic outside consumer hardware). You could even 3d print moulds and do injection moulding.
I think partly why people say that hardware isn't iterative enough is that they are doing it the wrong way. They spend too much time on prototyping, which means they then have to make everything perfect for a large production runs to average out the costs.
Yes, you can do design prototype revision A, B and C, then do production prototype A, B and C, and finally production version A, B and C. But then you have spent a significant amount of time and money refining prototypes and adding features. That forces you to large production runs with huge risk, because you have already made the investment. What I am saying is that you should just make the production prototype with less features and ship that. Because that is how you go to market quickly, which is usually the larger point of such development.
This is also why Shenzhen wins over hackerspaces (which are also great btw) but that is another story.
For another example, I made this case:
https://www.makerfol.io/project/8ycrZ9t-case-for-the-dp30v5a...
This would also have not been possible without the printer next to me. I constantly had to print segments to see how parts fit. With a week-long turnaround time instead of ten minutes, I wouldn't even have bothered.
Sure, if I already know how to design and print, maybe I could have designed something without owning a printer, but even that is not very likely (I generally have to make alterations a lot). For $300 (and a bunch of time spent), it's a no-brainer if you want to get into the hobby.
Given what this device does compared to a 3d printer (xyz axis movement, suction head, part reel incrementer, webcam + identification software), it seems like a sufficiently motivated hobbyist could come up with something passably similar in the same price ballpark.
This is one of those things that sounds trivial but has tons of hidden gotcha's.
EEVBlog usually breaks it down pretty well but I can't find any of the recent ones he's reviewed.
to support your point, from 2 weeks ago: https://news.ycombinator.com/item?id=16184255
There are open source PnP designs though:
https://hackaday.io/project/9319-diy-pick-and-place
it's really just that not many hobbyists need or even want one.
Depending on how complex your board is, building, say, 100 units by hand isn't time consuming. I've soldered literally thousands of through-hole boards and SMT is much faster and easier to do by hand.
But isn't this just begging the question?
If hobbyists can make use of it, then by definition it's a hobbyist venture. Bringing down size, cost, and complexity is exactly the path by which so much has become available to hobbyists these days.
Wasn't that long ago that a laser cutter, heaven forbid a 3d printer, was for pro shops only.
Edit: a good analogy is a homebrewer putting caps on beer bottles. You can do it by hand with a simple tool pretty easily - even capping 100 bottles at once is no sweat, really. It just doesn't make sense to have an automated bottle capping machine even if it was cheap, because it takes up space and ends up taking more effort to set up for a small run than it saves you. Pardon the pun, but the bottleneck lies elsewhere.
Capping them by hand despite the time and effort makes sense until you can afford a significant outlay in cost.
(I work for a machine shop that makes handling parts and retrofits capping/filler machines for the bottling industry. I don't fully know our prices but most of them would be far outside the range of any hobbyist starting out.)
So lets figure out ways to bring down the amount of work required to configure it! ;)
I think your analogy falls down because an automatic bottlecapper does the same thing as you'd normally do by hand.
The difference for me is that a pick-and-place machine is an enabler. It would enable hobbyists to take on large-scale, distributed electronics projects - for which, right now, assembly time is normally the bottleneck.
If all hobbyists were solely interested in producing consumer-style devices, like bad versions of phones, or home-automation systems, then a pick-and-place machine wouldn't be a good idea for anyone. But there are actually lots of hobbyists that do stuff like data logging, where assembly time is a major problem.
> I think your analogy falls down because an automatic bottlecapper does the same thing as you'd normally do by hand.
That was my point. A pick and place does the same thing you'd normally do by hand with a pair of tweezers in about 10 minutes for a board of moderate complexity. Doing surface mount work by hand with tweezers is pretty easy. The only thing a pick and place enables is for you to do things at a large scale, because they have a substantial overhead to use that you only start to recoup around 50-100 units or more. Now, if we had some revolutionary device that you could set up in 10 minutes to feed dozens of parts and cost $500 then hell yes, I'd want one too. But I don't think that's possible at present no matter any economies of scale or design optimization, because a pick and place is actually quite a complex machine by simple necessity.
Not to mention that placing parts is only part of it - you still have to fabricate and drill the boards, load the machine, clean up any mis-places, reflow them in an oven, manually place any through hole or awkwardly shaped components and solder them, cut the boards, test them, program them, mount them, etc. A pick and place would definitely save you some time if you really want to do such a large project, but by itself I don't think it's gonna be enough to really make it accessible to an individual in their free time. I'm not disputing that they would be beneficial, just that they make sense for an individual working in their free time. Especially when you consider that contract manufacturing isn't that expensive, though it does come with its own set of hassles.
I do think that a pick-and-place machine is probably more compatible with the hobbyist world than home CNC - given that CNC inherrently requires robust construction, and robust, accurate components are inherrently expensive. I mean - this machine looks more or less like a 3d printer with some reels, and a scanner. Once 3D printing becomes a cheap technology, things that require accurate repetitive movement become cheaper too.
I wouldn't be entirely surprised if stuff like small robotic arms become available to the hobbyist in the next 50 years or so, simply because being able to do arbitrary tasks with a high degree of accuracy is incredibly useful.
At the point where CNC, not as in cutting, but as in general computer-controlled movement, is an ordinary part of the workshop, I wouldn't be surprised if things like this ended up in hobbbyist spaces.
https://www.sparkfun.com/products/14342
I had actually just discovered those recently and with 0.2mm repeatability my first thought was a pick and place. But the complexity ends up being with the part feed and with configuration. It's really hard to build a machine that can reliably feed arbitrary parts automatically, and then programming it for each job is a pain so you'd also have to revolutionize the tooling so that a person could program one in a few minutes (and don't forget loading it ugh). The kinematics do share a lot with 3D printers, but it's the open-endedness in the feed system that makes them difficult. It's probably not impossible, just really hard, so there needs to be a considerable market to recoup the cost of designing such a machine that is accessible to a hobbyist.
I mostly agree with CNC, though it depends on how ambitious you are. I have a very cheap generic CNC engraver ($100 + shipping) that I use to engrave PCBs and it actually works quite well for that as well as it can do wood and soft materials remarkably nicely. Ditto for nicer machines like the Sienci, especially if you're doing woodworking. But a 5 axis machine capable of milling metal for <$500? It's not gonna happen. And even if it did, would you really want one? Milling metal is always gonna be complex and full of hassles.
I feel the same way about laser cutters. They're actually affordable now and I definitely want one, but I don't wanna deal with the considerable inherent hassle. Fires, toxic fumes, permanent blindness, focusing etc, just not worth it to me even for free.
Also this has been an interesting discussion, thanks.
You're almost certainly right about good metal-milling CNC machines never being cheap. I think the only possibility to make such a thing economical would be by re-purposing car parts (I think if you could work out a way to use common car parts and clever software, you could step around the fact that things manufactured to high tolerances out of steel are expensive - since car parts are both cheap, and well-made).
That said, the thing that really excites me about CNC is just how much it opens up a whole load of projects that I haven't had the tooling or the mechanical skill to take on. I'm a bad carpenter, so when I make things out of wood, I tend to have to allow massive tolerances in everything I do. Even a CNC that could only cut softwood would be really useful for me.
I've actually been ummming and ahhing about buying one of ali-express - where did you get your engraver from?
The tool isn't the issue. If a hobbyist wants to play with a Pick & Place, go for it.
My point was that the typical hobbyist isn't going to be manufacturing 1,000 units of anything, no matter how cheap is to do. Once you get into those numbers, it's far more likely to be a (small) business venture.
Sorry for going off-topic, but do hobbyists actually solder SMT's manually nowadays? Or does everyone use some kind of off-the-shelf/DIY reflow oven? And if so, are stencils obligatory or is it feasible to just apply paste manually?
I've been thinking of getting back (or well, properly starting if you will) my electronics hobby, but I've ever only done through-hole components. At least based on tutorials on the web, it's possible to hand-solder SMT's with some practice. But given how small e.g. 0603's are I guess that might involve an unhealthy amount of swearing..
[1] http://www.aoyue.eu/aoyue-int998-smd-rework-station-hot-air-... [2] https://www.aliexpress.com/item/220V-AD-982-Semi-Auto-Glue-D...
Even with practice I'm orders of magnitude slower than people with steady hands.
The main thing I like about electronics is it makes me feel at home in a world of technology that's kinda crazy alien at least half of the time. Being able to interact with it in a proactive way makes it something I can get a grip on, even if, when we're frank, my soldering is shit.
I do both: for small fixes I solder by hand. Complete boards I usually use an old toaster oven, but often if the boards are really small, I use a soldering iron. I try not to go below 0805, but 0603 isn't hard.
I typically won't get a stencil for custom one-off jobs since they tend to be small and ordering them introduces delay. When I do, I use oshstencils.com. Had great luck with them and they are very price competitive.
> I do have a side business building custom electronics, but I think I still qualify as hobbyist :-)
Taking this even further off-topic, could you elaborate a little? What kind of stuff are you doing? I imagine there is a niche for one-off or small batch stuff, but OTOH on aliexpress (or whatever) you can find almost every electronic gimmick you can imagine (and plenty you couldn't imagine!) for ridiculously low prices.
https://www.sparkfun.com/tutorials/60
https://www.sparkfun.com/tutorials/59
I've been skilletting my boards (fabbed by OSHpark) for the last few months, and it's really easy. As another commenter mentioned, use a syringe to deposit solder paste. It only takes a few boards to dial in the right amount to lay, and is actually surprisingly quick to place the parts (even 0603) with tweezers and then throw 'em on the skillet.
If you're one of the handful of hobbyists that want to do such complex projects I think contract manufacturing is probably a more realistic option. Not sure if you've ever set up a PnP before but it's a huge pain in the ass. I'd rather stencil and tweezer as long it was remotely practical. They are utterly mesmerizing to watch work once they're set up however...
LitePlacer is almost there. IMHO the key is recognizing that production-oriented machines need to be fast, while hobbyist/small-shop prototyping machines don't. If I need to stuff a complex prototype board with 500 parts, I literally don't care if it takes all day and all night as long as I don't have to do it. Obviously production shops don't have that luxury, but I do.
Really, the only unsolved problem (at least at the 0402 and up level) is loose part pickup and orientation.
Slump is a problem if you're trying to run a six-sigma process in a factory, of course, but that's not what's being discussed here.
My guess is that as prices drop under $2500/machine and $10/PCB and safety issues are resolved PnP will become more popular. There's a lot of need for repeatable, low rate production. A PnP is kind of like Vagrant/Puppet/Chef for hardware. Instead of manually building the board each time, use a configuration managed automated build.
OP said:
> Interesting that for such an inexpensive but specialized hardware device it still has a "phone home on boot" feature.
It's a harder problem than it seems, as I've found out over the past 6 years. While the basics are quite similar to a 3D printer, PnP has some unique challenges. 3D printing can be relatively inaccurate but still produce a good output since you are working with a pretty "oozy" substance to begin with. PnP has to be pretty accurate across it's entire work surface, though. If the machine is 0.1mm off at any point in it's travel that creates a likelihood that a part will be placed incorrectly. Feeders are also, mechanically, a lot more difficult to get right than it seems like they would be.
In any case, we're having good luck and a lot of success, so if you are interested in DIY / Open Source / Hobby pick and place, please come check it out!
https://www.youtube.com/watch?v=q9BGVYnaOs8
(great project by the way!)
Looks interesting (to me anyway) for this type of project. I can’t tell if the precision you need for PnP is there yet, but it looks like the speed is.
PnP seems like a hard problem. I thought the same thing as you awhile ago, looked at some efforts that others were putting together, and was surprised at how many opted for complex things like computer vision to verify part positioning.
And it did not take long for me to decide that, like most problems I take a cursory look at, there was no flippin' way I'd be able to make anything workable on my own in a reasonable cost/time-box.
If I had an affordable plug-and-play design, I might take steps to protect it. Personally I'd think twice about a 'phone home' solution because I hate that shit as a consumer - or just open source it outright - but I can understand the stance.
LitePlacer is designed for one-off jobs. It gets its parts from cut tape, not reels. It uses its vision system to line up on the parts tapes, so the tapes don't have to be precisely positioned. It's quite slow for a pick and place machine, but far faster than doing it by hand.
Doesn't put down solder paste, though.
There are a lot of low-end pick and place machines, but few critical evaluations of them.
Computer vision really is the answer, but from reading the article it sounds like the CV on the CHMT36VA isn't great for parts that aren't 0603 and it's not open/easily hackable to be better.
You put the PCB in a reflow oven afterwards.
Put solder on the pads. This can be done with a solder mask - a thin sheet of metal with holes cut where the pads are. Solder paste is scraped over it and then the mask is lifted off, leaving solder on the pads.
Place components. This can be done manually or with the pick and place machines described here. The solder paste is sticky so there is something to hold the components in place once placed.
Reflow. This is done by putting the now populated board into a special oven. This oven will put the board through a particular temperature curve ('profile') which melts the solder then lets it cool. The surface tension of the solder on the pads helps here - it pulls components that are a fraction off into perfect position (assuming the pick and place got it close enough and that solder is on pads where it is meant to be).
What you’re referring to in the first paragraph is just called a stencil.