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.
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.
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.
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.
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
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.
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.
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.
The big issue is really setting up manufacturing and supply chain on a large scale when you do get it working.
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.
You put the PCB in a reflow oven afterwards.
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.
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.