best value for small, slow automatic pick and place Moderators: adafruit_support_bill, adafruit

[bootstrap]

I must be crazy! I am seriously considering buying an automatic pick-and-place machine and multi-zone reflow oven to produce the products I'm developing. Since I need to assemble about 50~100 prototype PCBs for testing and software/firmware development, I might as well buy all this stuff now to assemble my prototypes. The alternative is to have all my PCBs assembled by an assembly house like http://www.aapcb.com then have some chinese company make production quantities (which is probably around 10,000 per year for the gizmo I'm doing first).

So, my primary question is: what is the best value "automatic-pick-and-place machine" these days, given my requirements. My requirements are: high-quality AKA high-precision placement, handles all common SMT components (including 0201, 0402, 0603, 0805, BGA, QFN, QFP, SO8, SOIC), doesn't break every month, and CAN BE VERY SLOW since my volumes will not be large (and it can run all night if it wants to). My current PCBs are 6" square with 300 total components (mostly caps/resistors/diodes) and 3" square with 50 total components (ditto), and the finest pad-pitch (to date) is 0.500mm. Also advise what prices the machines you suggest will cost, and any other considerations or tradeoffs I should know about. Also advise which brands and/or models deserve a "stay away from" rating, and which deserve a "good stuff" rating.

I suppose the secondary question is: am I totally insane to consider doing production myself these days in the USSA - even at these very modest volumes? I suspect so... so let me have the truth right between the eyes. And yes, I can handle the truth! :-)

I suppose the same question applies to a reflow oven. I'm guessing this would be something like the smallest APS multi-zone oven (about $8500) or maybe even their batch oven ($4500) though I suspect that's just too lame for even low-volume production.

[adafruit]

pretty much everyone here uses an MDC 7722
did you read any of http://www.ladyada.net/wiki/mdcpickandplace ?

[bootstrap]

pretty much everyone here uses an MDC 7722
did you read any of http://www.ladyada.net/wiki/mdcpickandplace ?

No, somehow I hadn't run across that site. It was quite helpful, especially
including all the direct and indirect links I followed. I'm still not sure, but
it looks like the total price is more than I want to pay before I start getting
revenue from the products. But I always hate to "waste" money now that
I could have avoided by "going all in" right from the start. In other words,
this is still a difficult decision!

[mikeselectricstuff]

You won't find 'high-tech' (0201's) and 'slow'. high precision means expensive mechanical parts & heavy beds for stability & accuracy, and most P&Ps are used in an environment where they must pay their way, so speed is always a significant consideration.
If you have the space, you could luck out on an old but well-built used machine, but the 'luck' would also need to extend to getting enough info to do maintainace yourself, spares availability, getting the software etc. Unfortunately there is currently almost zero useful information on the net - I've recently been hunting for an old low-end machne and struggled to find even a photo of many models. P&P machines are still very much the territory of industry, with only the occasional old one falling into the hands of hobbyist/low-end folks.

[bootstrap]

The more I look, the more I see you are correct. I'm not brave enough to buy a used machine, because I certainly can't afford to fork over a bunch of cash, then have it not pan out.

My latest guesstimate is $40K ~ $50K for the cheapest pick-and-place that is likely to work reliably. I went to look at a couple in town, and they don't even build PCBs with BGAs, because they have too much trouble, and they need to repair a fair percentage of the parts in QFN packages. And their machine cost well over $50K apparently (not sure that includes all the extra component reels they need). Apparently you can pay more for reels than the pick-and-place machine! Crazy! And they said they have a fair amount of trouble, need to tweak and such fairly often. I don't like the sound of that.

Clearly when it comes time to fork out $50K+ for an automatic pick-and-place machine, I must do lots of checking to see which are most reliable and troublefree. So far I kind of like the products at http://www.apsgold.com, but "what do I know" (answer: not much).

I've had a terrible time getting VERY basic information about important issues like precision, reliability, and so forth. The few places that I've talked to that assemble their own PCBs have no idea whether their troubles are mostly caused when the solder-paste is applied, when the parts are placed, or when the PCBs are in the reflow oven!

Also, I've been unable to learn what percentage of scrap to expect if I have my production done by an assembly house that does nothing but assemble PCBs. Everyone I talk to has a wildly different story! I was originally planning to have my PCBs assembled by some chinese company, but I've heard so many horror stories of so many kinds, I'm getting more and more inclined to handle everything myself, where I can make sure quality control is taken seriously, and careless work is not tolerated.

[mikeselectricstuff]

For short runs, China isn't really worth thinking about. The unit cost of getting it done locally generally isn't too bad, the problem is finding a subcontractor with a good attitude to small jobs, in terms of not charging big setup costs, and turnround time. Most small subcontractors I've encountered have a very few customers who provide most of their work, so your little jobs come a fair way down the priority list & tend to get fitted in between other work.
And larger outfits tend to just not be interested in small jobs at all.

[bootstrap]

Yes, China or elsewhere in asia would only be for production. For my products, even production quantities are not that large - probably batches of 2500 of my larger PCB (5.80" square with ~300 components) and 10,000 of my smaller PCB (2.80" square with ~50 components). Hopefully that's enough to be considered "production quantities". We shall see.

I received a quote of $8200 to assemble my prototypes. I will not pay that much. So I've decided to buy what I need to assemble them myself. What I haven't decided yet is whether to buy "rework equipment" that is designed to install and solder [non-trivial] components one-by-one, or buy low-end cheapskate equipment that mimics production (stencil printer, manual pick-and-place aids, reflow oven) that is designed to install and solder all components at once. When I say "manual pick-and-place aids", I mean things like vacuum-pick-up tools, stereo microscope, etc.

Interestingly the full-bore BGA+ rework system (3LMZT-7-BGA) at http://www.zeph.com/systems.htm costs about what AAPCB quoted to assemble my PCBs. I'd rather buy equipment that is useful for many years than vaporize the cash on a one-time service. Still, I'd prefer to go the other route (stencil printer, manual pick-and-place aids, and high-quality 1-zone convection reflow oven), but the oven I like has a one-month lead time, which I dislike (so I may try a cheaper one first - $350 versus $4500 for the one I like (and $8500 for their low-end multi-zone conveyor oven)). Unfortunately, the cheapo reflow ovens have IR heaters, which seems totally stupido to me.

[mikeselectricstuff]

For 1-offs the stencilling isn't a problem - polyester stencils and a simple jig made from scrap PCB is fine. For soldering, toaster ovens are no problem as you can sit & watch it. It's the placement that gets tedious - a foot-operated vacuum pen and whatever optical aids your eyes need are all that's necessary here. I'm not convinced that any of the the 'manual pick/place' gear is particularly useful - I've not used one but hand-placement is quick enough that I can't see how any hand-placing process could be made much quicker, unless you're doing a lot of 0402s or fine-pitch parts.
On the optical aids front I'd very highly reccommend the Vision Engineering Mantis - expensive but much nicer to use than a traditional microscope. I'd certaonly pay for that over a stencil printer or manual pick/place workstation
x4 is the ideal lens for most SMT work

[mikeselectricstuff]

This might be worth looking out for

[bootstrap]

That link is quite interesting. Unfortunately, I already have finished PCBs and parts waiting for me to solve the assembly issue (25 of a 6" square PCB with ~250 components and 50 of a 3" square PCB with ~30 components). The vast majority are bypass capacitors, but they also contain ~7 BGAs, ~15 QFNs, ~3 QFPs, and assorted SO8s, SOICs and such. The caps and resistors range from 0201 to 0805.

[blogger]

For 1-offs the stencilling isn't a problem - polyester stencils and a simple jig made from scrap PCB is fine.

tell me more about this, i'd like to make stencils for 0402 parts and 0.5mm pitch tfqp/qfn/bga etc

[blogger]

The vast majority are bypass capacitors, but they also contain ~7 BGAs, ~15 QFNs, ~3 QFPs, and assorted SO8s, SOICs and such. The caps and resistors range from 0201 to 0805.

the $8500 quote sounds pretty reasonable especially if thats the assembly of 250/10k or whatever qty you mentioned before.
good luck getting anything other than high-end factory lines placing 0201s.

I'm still not sold MDC unit can do 0402s reliably, maybe hardware can, but software sure is dodgy.

[bootstrap]

The vast majority are bypass capacitors, but they also contain ~7 BGAs, ~15 QFNs, ~3 QFPs, and assorted SO8s, SOICs and such. The caps and resistors range from 0201 to 0805.

the $8500 quote sounds pretty reasonable especially if thats the assembly of 250/10k or whatever qty you mentioned before. good luck getting anything other than high-end factory lines placing 0201s.

I'm still not sold MDC unit can do 0402s reliably, maybe hardware can, but software sure is dodgy.

The $8500 quote was for 25 PCBs with ~250 components plus 50 PCBs with 1 IC and ~30 caps. Personally, I think that's rather expensive (the second PCB only has 1 component the other PCB doesn't have, so no need to switch over component reels between the two PCBs). In fact, it makes me realize that a brand new pick-and-place machine costs only 5 times that quote, which means, I could pay for the machine by doing 5 assembly jobs like mine. And that's for a fairly nice machine that can handle 0201s and has vision centering --- like the <APS Novastar LE40V> or equivalent.

Hey, does anyone want to give me some assembly work to help pay off such a machine?

I am surprised to hear the MDC software is lame. Most other folks around here seem to prefer the MDC pick-and-place offerings better than similar-priced competition.

Also, it is only the small components (0201 caps) and tight pad/ball spacing (0.05mm ~ 0.67mm BGAs and QFNs) that make me want a pick-and-place machine AT ALL --- to assemble my prototypes. I can assemble everything else quite easily by eye and hand with stereo-microscope + vacuum-pick-up-pencil + hot-air-pencil + hot-air-PCB-pre-heat-device. 0402s are no problem for me, 0201s are a bit clumsy to place, but the biggest problem is not knowing whether QFNs and BGAs are precisely placed.

[mikeselectricstuff]

The $8500 quote was for 25 PCBs with ~250 components plus 50 PCBs with 1 IC and ~30 caps. Personally, I think that's rather expensive

That sounds expensive to me unless maybe it includes x-raying of BGAs etc. - ask some other, smaller assembly houses.

[bootstrap]

The $8500 quote was for 25 PCBs with ~250 components plus 50 PCBs with 1 IC and ~30 caps. Personally, I think that's rather expensive

That sounds expensive to me unless maybe it includes x-raying of BGAs etc. - ask some other, smaller assembly houses.

The was the cheapest of 10 places I tried. That was http://www.aapcb.com, and they seemed cheaper than anyone else. In fact, someone suggested them to me.

This is the problem with "normal companies" and "normal business models"... they have substantial daily expenses (called "salaries"), and they need to pay those expenses even when "times are lean". Plus, their typical customers are mostly "real companies" with substantial cash-flow. What lone-ranger developers and hobbyists and startups need is something much more informal, efficient and cost effective.