$1000 pick place

[fcb]

Before I spend more time working on this side project (just finished a proof-of-concept) I thought I'd gauge reaction/interest.

Base spec for "Magnum":
Eurocard capable (160x100) placement area, facility to build bigger boards in two-halves.
1500+CPH (real-world)
36 location loose tray pick-up (interchangeable)
Integrated pump
Accuracy: 0603 at least, likely much better.
Vision system
Single USB connection to host computer
24V external supply (brick in cord type).
US $1000-1200 / £650+VAT

The software would be fairly comprehensive as it belongs to a more advanced $5K machine ("New York") I've been developing for some-time (which is based loosely on a machine I built for our own in-house assembly).

[mikeselectricstuff]

Tape feeders ? Without them it's a waste of time.
Eurocard is a bit small - would it really be much more expensive to make it bigger?

[EEVblog]

Eurocard capable (160x100) placement area, facility to build bigger boards in two-halves.

I really don't see the practical point unless it can do larger panels, with maybe a dozen feeders as a minimum.

[fcb]

This project is a developmental offshoot from a bigger/faster machine.

Feeders: I solved the conundrum of low cost smart feeders on the "NEW YORK" machine, but without resorting to drag feeders it would be difficult to put them on the baby machine and keep the cost down - however it would be simple enough to lay-strips down.

Board size: Yes, it can be increased a bit and the pick-up area is not rectangular. But much bigger and it starts getting awkward with the technology, so the idea is not to be an all-things-to-all-men type machine.

My main reason for posting is to see if there is any interest in a really low cost/small machine. I'm not convinced that there is a requirement, but I don't want to shelve the tech if it's useful.

[digsys]

I've been looking at manual/assist P+P for ages, and don't have any need for reel feeders.
I like the idea of cut tape feeder, but it wouldn't bother me if it wasn't included. I have HUGE stock variety, so loose bins is exactly
what I need. I'm assuming the theory is - optics > LCD screen, manual guided picker, vacuum pick-up, rotate knob etc
Questions- how would you handle flip/rotate the part? Could it auto-learn a part in future? Will it do the flux application as well.
I've been mulling over designing these type of "heads" for a while, and I'd definitely be interested to see what your plan is.

[fcb]

The loose bin thing relies on not over filling each tray - so long as the vision system can find a enough parts the right way up (for resistors) and physically separate then it can pick them. It can be very quick to setup, but is a compromise if you have large qty of passives on a board.

It's also not a manual or semi-auto machine - it's fully automatic. As for auto-learn, it relies on prepared component definitions - but that's pretty quick to write and we plan on having an extensive library of these to download.

[digsys]

I'm fine with full auto - so IF there's a part it can't place, for eg, it has to be done by hand? Did you ever consider being able to
"take control" for tricky parts? Just a question.
Do you have a pic of how a tray with loose parts would look? Quite interested.

[sleemanj]

My feeling is that it's too pricey for a hobbiest's toy (but then again, people buy 3d printers for about the same, maybe I'm just too much of a tight-ass, or underpaid, or both).

But without feeders it is too, what's the word, underspec'd for a small production tool.

You'll be aware I expect of the 3600 USD ones from China which have cropped up in the last few months - see http://dangerousprototypes.com/forum/viewtopic.php?f=68&t=4903 - a number of purchasers in that thread alone, so there is a market there, you should look into how much it would increase the cost to add feeders.

[marshallh]

$1k? Sign me up. Sick of hand placing crap tons of 0402's

[benemorius]

I'd be quite interested at that price for doing small production runs as a quasi-hobbyist, but not without tape feeders.

Let us see some pictures of your proof of concept and your $5k machine.

[mikeselectricstuff]

I think vision of loose parts in a tray is a potentially viable way to save the cost of feeders if it works well - combined with the ability to automatically shake the tray as necessary. However you still need the ability to use tape  feeders (at extra cost) for stuff like decoupling caps that are used in large quantities.

Using the head to pull the tape is a reasonable approach to save cost, but you still need a way to pull the cover tape - the cheap Chinese one partly solves this using a single motor across all feeders. I think the  main hing that machine has got wrong (apart from no vision) is you can't change the mix of tape widths. Having adjustable lane widths wouldn't have been too hard to do.

The thing many people don't understand about pick/place until they've used one is that it needs to be quicker to set up and run than to place manually, so for a single board, the best P&P in the world is still going to be slower than manual. And as you go up in quantity, there always comes a point where it's cheaper to use a subcontractor.
So the market for a low-end machine starting off in a niche of low-to-medium build qtys and people who need fast turnround.
Any further limitation in machine capability makes that niche smaller and smaller.

I have no doubt that there are some opportunities, especially by exploiting cheap processing to do vision, but nobody's going to make a fortune out of it.

I think manual placement aids or "semi-automatic" approaches are a total dead end - their benefit just can't justify the cost of making something that's a good enough improvement over hand placement


 

[fcb]

My feeling is that it's too pricey for a hobbiest's toy (but then again, people buy 3d printers for about the same, maybe I'm just too much of a tight-ass, or underpaid, or both).

But without feeders it is too, what's the word, underspec'd for a small production tool.

You'll be aware I expect of the 3600 USD ones from China which have cropped up in the last few months - see http://dangerousprototypes.com/forum/viewtopic.php?f=68&t=4903 - a number of purchasers in that thread alone, so there is a market there, you should look into how much it would increase the cost to add feeders.

I don't think I would call it a toy, but certainly not something that would be used in-anger for runs of more than say 25 boards.

Feeders. Ahhh, the Achilles Heel of any PnP machine.

Yes, i'm aware of the TM220A/240A units, the fundemental problem with them is the lack of accuracy (no vision). If they solved that then I'd be buying them myself.  I considered modifying one, but at that point your paying a lot of $ for a bunch of rail, a plate and some motors.

I'd be quite interested at that price for doing small production runs as a quasi-hobbyist, but not without tape feeders.

Let us see some pictures of your proof of concept and your $5k machine.

Again, the feeders...

The proof-of-concept is just that, by showing it I just end-up giving my ideas away.  There's no point in showing the $5k machine until it's ready to be sold, I'm not planning on Kickstarting/IGG/pre-selling these. And the smart-feeders are subject to patent application.

I guess I should just duplicate the TM220A, add vision, USB and decent software.

[mikeselectricstuff]

If you really have a viable pick & place at $5000 I'd think it would make more sense to focus on that one, at least initially,  as you should be able  to make much more money on that than an ultra cheap option.

 
 

[Selectech]

I'm interested, but might actually be a candidate for the more fully featured unit.

Most of what I've seen in the low cost end { < 5K } has been lacking things so I've hung on to my $.

I continue to manual place, with CNC paste dispensing, and use some assembly houses for larger qty runs. Need to put down 20K to 50K+ parts per year on about 15 to 20 different assemblies. Mosty 0603 to 1206, with lots of SOT-23. Some QFN & fine pitch stuff. Need about 300 x 300mm working area to cover off singles and small panels. Fairly small # of component types per board { < 30 }, but a panel could well be 700 to 1000 items to place.  Generally building 10 to 50 of each item at a time, 3 to 6 types per week. Raw speed not such a big deal as ease of setup, flexibility and accuracy.

[nctnico]

IMHO it would be nice if the P&P machine could also apply paste. It shouldn't be hard to translate the paste mask into data to 'drive' a syringe.

[mikeselectricstuff]

IMHO it would be nice if the P&P machine could also apply paste. It shouldn't be hard to translate the paste mask into data to 'drive' a syringe.

That would be a significant selling point, providing more of an all-in-1 solution. as long as it worked well. You may need to go to an auger based system to get sufficient consistency though.
An issue is additional mass, but it doesn't have to move anywhere near as fast as for placing, as you don't need to keep returning to teh feeder area.  Another approach may be some sort of exchangeable head, where you swap the weight of the rotate/lift head with  a dispense + lift head - the lift range doesn't need to be anywhere near as much, and could probably even be a simple 2 position up/down.

[Eclipze]

Thought I'd add my experience on this topic, as there seems to be a lot of people with opinions but not exposure to actually using one of these machines.

I have a TM240A.  My purpose for it is for small board builds.  i.e. 20 to 200.  If were to get more, I'd get them done by a contractor.  But the important aspect is for getting that first run of boards done for field testing and prototypes.  As well as for niche part runs, where the client only wants a small quantity.

PLACEMENT ACCURACY
So far with 0603 and SOT323 I'm not having any issues.  Quite reliable.  I do need to use a better method to load the boards via drilled hole location, as routed board edge just isn't a good alignment reference.  I doubt I'II ever use it for placing larger chip parts, although it would be good to have a manual mode where I could pick, jog/rotate and place by eye.  Usually I go over the board with my bend hook (thin rod with 90 degree bend at the tip) to push/pull a few parts into better alignment.  I can see in the future placing some chips knowing that they won't be perfectly placed, needing a little help, but still quicker than placing manually.   Going to use it for a board design with 0402 soon.

SOLDER PASTE
I stencil my boards.  Use a syringe dispenser only for a single prototype.  Unframed prototype stencils are so cheap and make that much of a difference to reflow yield, I don't understand why people would do otherwise.  If there was an option for an automated dispenser, I wouldn't get it even if it was cheap, but you'd need a good/expensive auger type to get consistency anyway.  Start stenciling and you won't go back.

PLACEMENT SPEED
I don't run at maximum speed.  I'd need to build a sturdier table, as it can move very quick and shakes the table around.  I was surprised to not notice much of a difference in placement accuracy going maximum speed over what I usually run it at.  I did a time run on a board and could save near 25% of the time.  But I find between stenciling, reflowing and doing other work, the speed is fine anyway.  Where i would do 6..8 boards a day, placing with a vacuum pen, I can run through 24 boards in ~3 hours with the pick and place (including reflow).  As for the time for it to index the tape with the head... might be the cheap approach, but it's plenty fast and simplifies the mechanics (one less thing that can break).  It's also quite good at getting the tape indexed/aligned.

DUAL HEAD
Need to set the story straight here, as the comments I've read in a few threads here haven't quite understood the need.  There are two heads, which provides the option to load two different nozzle sizes.  There are 4 different size nozzles.  I often run two different nozzles, but for one board that has a large array of LEDs, I load both with the same size for dual component pickup speed.  Otherwise, I target the sizes I need to load all the chip parts.  Big stuff I'II load by vacuum pen anyway.  But there is some cross over where two different sizes can load the same parts.  For example, the small head can be used to pickup 0805 through to a SOT89, but if really needed, you might be able to get away with picking up a larger part if you slow the machine speed down.  So in general, you need at least two heads for a practical board load, without having an automated nozzle changer.  But you can also benefit from using both, where possible, to do dual component pickups.

FIXED TAPE ROWS
You are limited on lanes, which is why I went for the TM240A that has more.  Bigger placement area than I need, but more lanes.  4x10mm, 2x12mm and a heap of 8mm.  So far, hasn't been an issue.  I commonise on chip parts as much as possible anyway.  Larger boards with more parts are usually loaded double sided as well, which helps to minimise on the number of lanes needed.  As for needing more wide lanes... the larger the part the more I'II hand place those anyway.  It's the quantity of passives and small chip stuff that's the time saver having machine placed. 

PICK AND PLACE FILE
I use the Excel macro (slightly modified) to convert the Altium placement data to the Neoden format.  It's pretty straight forward.  I shuffle reel positions around as needed... easy done in Excel.  To optimise, I'II manually interleave the rows to get dual pickup running etc...  It's not uber automated and smart, but it's in Excel and pretty straight forward to get what you want.  You will need to have a little history file, so you know from experience that placing 1206 resistors is better with a 0.15, -0.2 offset and that an SMA footprint diode needs a 180 rotation to face the right way (or change your footprint to match).  Or if you are placing bottom layer, you need to flip the X coordinates.


THE BAD...
- Rotation speed is too fast.  I suspect some part accuracy is lost from this.  This is usually the first issue I see when the head is slightly undersized.  Sometimes I just have to clean the pickup head. 
- Would be good to have adjustable control on the blow back and head dwell. 
- Would be nice if the vacuum pumps only operated when the machine was running a file.
- I'm not that impressed with the front tray loading.  I need to change how parts are positioned to make better use of that for more accurate pickup.
- There are few idiosyncrasies with the firmware.  Particularly with file format.  For example, never try to pickup with both heads on the same tape, if it's indexed by 8mm.  e.g. PLCC LEDs.  Can only do this with 4mm index parts, otherwise the firmware locks up and needs to be power cycled.  You have to use two tapes to do this.  This is the typical software you'd see from China.  I avoid using the interface to make changes to the placement file... easier to use the PC.
- Setting up a placement file is a bit of overhead, but at least it's flexible and logical.

THE GOOD...
It actually works for real!  Useful, small footprint, mechanically simple and hopefully reliable.


OVERALL
Wasted a lot of time looking for a second hand machine.  Often to find it's accuracy was no longer up to speed after 10 years of work, old software and needed maintenance.  Possibly not as many options in Australia for second hand machines than overseas.  Big footprint and heavy too.  The cost of feeders... OMG.  Don't look at new machine prices, as you can spend more on feeders than the machine.  I really don't think it's feasible to get a mainsteam pick and place machine unless you are planning to run enough production work through it. 

Getting this Neoden machine was an option where it was a small footprint, doesn't need a compressor run, didn't take long to learn and get running and not likely to need much in the way of maintenance.  It's a good fit for small batch runs.


Here is a video of the first board I programmed it to load.

Was having trouble getting the PLCC LEDs to load reliably.  Usually they load, but of course the video I took had to mess two of them up.  Likely an example where using a bigger head size would be better, but I'II get away with the smaller one just for efficiency.

I also use an XReflow306 convection reflow oven.  Not exactly a budget machine, but good for batch runs.  Posted a picture of this setup on my Facebook, Oct 28th.  https://www.facebook.com/EcliptechInnovations   There is also a picture Aug 21st of a surface mount component tape holder I use... which is great for placing a lot of small parts with a vacuum pen off tape strips.

[Psi]

I would pay up to NZ$2K for something that actually worked reliably and could place down to 0603 on PCBs 20 x 20cm. (ideally a little bigger in one direction)

-Speed isn't much of an issue for a hobbyist as long as its not crazy slow.
-Machine vision is definitely required as is parts rotation.
-Inspection camera for alignment (fiducials)
-Open source, or at least open protocol is a big plus
-The most important part of all, as mike says, is the feeding system. It needs to take all width reels without custom feeders for each. It also needs to take loose parts in trays or loose tape. It's quite common for a hobbist to be using parts that are just too expensive to buy in whole reel quantities.
- Paste dispensing would definitely be awesome, as already said.

For a hobbyist to be interested it needs to be relatively simple parts/hardware wise so repairs/maintenance/mods are possible. Keep the really custom parts which have to be made/ordered specially to a minimum.

[GeoffS]

After following a few links to the TM240, I came across this open source,  pick and place machine. http://vbesmens.de/en/pick-and-place.html
Quite advanced for a one person, DIY project.

[fcb]

It's still an ongoing project for me, but I can't sink huge amounts of time into it unless I make it a commercially viable machine ($5K at least).

I did toy with the idea of open-sourcing it, but I still don't get the whole 'give it away for free thing' yet.

I've looked at the TM220A and the TM240A, and there is a part of me that says "if you can't beat 'em join 'em".

[Eclipze]

Having one at a $2k price point with vision, tape feeders and options similar to a $15k..30k commercial machine is quite unrealistic.  I don't wish to offend anyone with such a comment, but it doesn't take long to exceed that in parts cost for a serious machine.   But if you do go down that route and spend all the time and money to build one that is fast and accurate, you will still need a descent reflow oven and stencil printer.  Which aren't cheap.  Otherwise, you have a neat PnP bottlenecked by a toaster oven and messy stenciling station.

I started this DIY build back in 2006. 
http://www.cnczone.com/forums/rc_robotics_autonomous_robots/97551-eclipzes_smd_picknplace_build.html
That thread had.... Replies: 724   Views: 125,228  before it was closed.  WOW

Went through two builds myself.  It's a lot of work, but it came down to software being the limiting factor.  I was going from a serious hobby through to a business at the time.  The time investment became impractical, unless I planned to commercially build that PnP as a product.  There is a point where having a serious machine with all the vision assist and tape feeding relates to using it for a reasonable volume of boards.  You are then in a grey area, where all that time spent on the PnP is far more than just hand populating with descent vacuum pen off tape in batches.  Even if you were only doing 10 or 20 boards and had a pick and place machine, it's likely faster to hand build them anyway.

I found that grey area lasted 1-2 years, where having a home built/cheap hobby PnP would have been really nice.  On the flip side... now days, any volume of board loading, I have to get done externally.  I just don't have anywhere near the time to handle that as well as everything else.  But... the cost of the TM240A to help build some small batches is rather convenient, quick and affordable at this end of business, even for the cost.  Means I can test the market with a small batch before committing to a real production run.  Or if really needed, I'm sure I could loose 3-4 days just board loading to do a small production run.  I may not last that long before getting too grumpy and snapping though

If you want to build a pick and place cheaply, consider the end goal.  It's a lot of fun, but for business use, it's a time sink.  Ideally, you'd snap up a cheap 2nd hand machine in good working order, with all the peripheral feeders and parts.  Otherwise, an off-the-shelf no-vision option might be the middle ground.  For now, my expectation is to see a machine with vision, nozzle changer and descent software at the $5k mark in a few years.  But not with feeder cassettes.

[Psi]

Having one at a $2k price point with vision, tape feeders and options similar to a $15k..30k commercial machine is quite unrealistic

The TM220A/TM240A is pretty close to what a hobbiest wants, it just needs vision and perhaps some extra feeding options. Maybe a little smaller in size to save costs, 400x360 pcb/panel is pretty big.

There cant be $5000 worth of materials in that thing. Should be quite doable to get it's cost down to 2K

[nctnico]

Projects like these need a lot of software. Most of the cost is time spend by programming not the materials.

[fcb]

Projects like these need a lot of software. Most of the cost is time spend by programming not the materials.

Actually, the software side isn't the biggest headache and probably not the most time consuming bit in making a low-cost machine.  I suspect that given a couple of days, most of us could program a TM220A style machine, it is basically driving motors to a simple list of co-ordinates.

In my experience, the two challenges are 1. vision centering of a component and 2. accurate feedback of present head position.  The vision centering is complex and people often underestimate how important a good image is.  Accuracy of present position is usually taken care-of by building a very rigid structure, which ends up weighing alot and you get into this sort of performance death spiral.

As far as the vision goes, I rolled my own over several months, rather than using available libraries.  My approach to accuracy is to use light-weight parts and a novel realtime 2D triangulation system for the head. You get speed and instant out-of-the-box calibration, from a light-weight low cost machine.

[Lizerd]

Nice idea!!.
sounds like you already have done the major parts for the 5k machine, care to show us some pictures of your software and hardware ?