ktownsend wrote:It takes me a couple hours, but I bought it because I generally need to make about 50 boards at a time. Outsourcing assembly usually only becomes cost effective for me at 250+ boards, but then there are still a lot of one-time expenses like stencils, etc., and needing to buy 250 of every part ... which can get pricey if you really are only confident you'll sell 50 this quarter. I can easily hand place 10 boards, but it's horrible dull and the amount of time involved destroys any real profitability. Even if you don't place BGA ... there's a strong case to be made for purchasing a PnP if you have regular requirements for boards in the 50-100 quantity range, and support a number of designs with thos quantiites. I'd say even with setup times, I can assemble boards about 8-10x faster than before overall. It's still not fun, but at least it starts to be financially sane if I attach any value to me time.
This raises another question I need to answer to rationally decide whether to buy a pick-and-place machine now (versus when production starts, or perhaps "never" if BANNED assembly is more practical "all things considered"). That is, *roughly* what kind of costs can I expect to have batches of 250 to 1000 large PCBs made (6" square with 300 total components) along with batches of 1000 to 4000 small PCBs (3" square with 50 total components)? What would you guess my costs per PCB would be for those two PCBs?
Also tell me anything you can about the relative advantages and disadvantages of trying to get production in asia versus a western country (of the modest quantities that we need). My back of the envelope calculation implies $3 to $6 each for my larger PCB if done with my own pick-and-place machine (not counting ?my? labor to keep an ear on the machine, assuming we can "hear" sounds or interruptions when human attention is required).
One more question. Even these modest speed machines that claim about 2000 parts per hour tend to have much faster rates to place solder-paste on pads (typically, 10,000 pads per second). This makes total sense, since the average motion required between pads is infintesimal compared to picking up new components. I went to see a small company near me that has two similar size pick-and-place machines, and they claimed they can trace a rather large percentage of problems with assembled PCBs to imperfect solderpaste application (which they do with a rather expensive solderpaste stencil printer). Thus, if it is more reliable to place paste on every pad of PCBs before placing components, that seems attractive to me --- in theory. What do you think about this issue "in practice"?
I bought it having BGA placement in mind (you can still hand-place QFN), particularly because most ARM9 chips are BGA only. But this actually isn't the driving factor for me ... It's all about the small (but not small enough) volume, and finding a way to efficiently assemble those quantities of boards without very heavy investment in expense parts/stock. The 400€ a month the PnP costs me is quite small compared to the €1200 per boards in assembly costs + parts for 250 boards I would otherwise need to fork out. In that sense, a "shared" machine makes sense, but the idea would kind of scare me personally since for the amount of money involved I'd worry about how that investment would work out longer term. I'd really need to trust the other people or have a safe structure setup on the time sharing, and perhaps some sort of insurance, etc...
You say the PaP costs you 400€ per month. Is that because you took a loan to buy it, and that's your monthly payment (or you leased it)? Or you bought the machine, and 400€ is how you calculate amortizing the machine across n years of predicted work?
bootstrap wrote:You have the MDC-7722 also? How much did it cost? And how much more did you pay for other accessories and feeders?
Yes. With the Euro jumping off a cliff, it actually cost me a lot more than it would have a few months ago, but I got a good price on it since I usually go several times a year to Tokyo and know the director of the company that makes them (thanks to an introduction from a Japanese engineer friend). The reels are what adds up, and are what really determines the overall cost since the machine price itself is farily fixed. I currently only have 12x8mm reels, 2x12mm reels, and 15x8mm cut tape and 4x12mm cut tape. Anything 16mm and higher can be placed without feeders using double-sided tape and programming the parts in as trays. I could use at least 7-8 more 8mm feeders, but they add up and my budget was already strained. As I pick up some business, though, I'll try to work up to 20x8mm and 3x12mm which fills up one side of the machine (27 8mm reels on each side), and use cut tape for any oddball discretes/small parts. The cut tape is annoying since the parts jump out if you bump the tray, but it's still the cheapest solution in the short term. 20 8mm feeders should be ideal for my own needs.
Unfortunately, you'll probably have to go through a distributor to buy the machine which adds some margin, though I suppose it also assures you of a local warranty/repair facility as well. I'd try to negotiate a lot on the reels, though, since that's where most of the margins tend to be from most companies (with any PnP).
Yeah, sorry to hear about the euro pain. Just a few months earlier and the dollar was poopoo and the euro was king. Chalk up another one to Murphy and his Law. Pretty soon no self-respecting company will accept anything but gold for their goods (if they're smart). I just noticed manncorp sells that machine for $30,000 US, which I guess is roughly €25,000... without feeders and options, presumably. But I guess that model inherently comes with top and bottom (2-camera) vision, at least. This is almost exactly what APS Novastar charges for the LE40V that catches my fancy so far (though the MD2277 is a close second in my book, not even counting this website and forum). I think the MD2277 can handle larger PCBs (the LE40V only handles 340mm x 560mm PCBs), but I also guess the footprint of the LE40V is probably quite a bit smaller (1016mm x 1067mm x 686mm high)... and can thus sit on a desk, countertop (or industrial strength table) instead of the floor. Otherwise, they're probably fairly similar, I'm guessing.
Woops! I just checked, and I have it backwards! And WOW, the videos on the internet make the 7722FV look HUGE, but the specifications at manncorp claim its footprint is only 710mm x 760mm x 660mm (high) - which is quite a big SMALLER than the LE40V unit I'm looking at. Wow is that a surprise! All the videos make it appear HUGE, and they all make it appear it sits on the floor. But obviously not.
Okay now that I have both specs in front of me, they are even more similar than I thought. I'm rather blown away that the 7722FV (at least on manncorp webpage) claims 0402 is the smallest supported component (versus 0201 for the LE40V). The claimed placement accuracy of 7722FV is 0.10mm with flying vision and 0.05mm with bottom camera (which I thought was standard equipment, but I guess not). This compares to 0.025mm for the LE40V. Gads, the LE40V is a BEAST - it weighs twice as much as the 7722FV (113kg versus 60kg). The LE40V web-page and product brief PDF don't even say how many feeders it takes (maximum).
Interestingly, even though the 7722FV says its placement accuracy is 0.05mm (at best), it claims its "X,Y resolution" is 0.015mm. I guess some precision is lost between the X,Y linear encoders and the pick-and-place head --- either flexure, incomplete compensation for thermal expansion, or limitations of the bottom-up (and/or top-down) vision system. Just guessing. Their placement rates are not stated the same way, but seem similar. I was also wrong about maximum PCB sizes, which are 390mm x 310mm for 7722FV versus 560mm x 343mm. At first that seemed totally irrelevant since my largest PCBs are less than 150mm square --- but then I realized panels of PCBs might be a wise move when production time arrives (that can easily cover a panel of 6 of my large PCBs and 28 of my small PCBs... hmmmm). But then I need some way to "de-panelize" them after assembly, which sounds rather unpleasant.
Overall, that makes them seem fairly close. As it turns out, the 7722FV is smaller and lighter, which comes in handy when locating it and moving it... but might be partly responsible for not being as rigid - which shows up as slightly lower placement accuracy... IN THEORY (who knows how honest either set of specifications is). The LE40V comes with a standard "automatic 4 position nozzle changer", but also says the following, which I totally don't understand the meaning of: "automatic tool changer - up to 8 positions". I don't see this topic mentioned in the 7722FV specifications, so I cannot compare this aspect of the two machines. Somehow, I'm guessing the 7722FV manages to handle 0201s, no matter what the manncorp specifications say. Do you know about this? I have a fair number of 0201 capacitors on the opposite side of the PCB from my BGA pads (I have "fill-and-cap" treatment on all BGA pads to support "via-in-pad).
Now all we need (both of us) is to find someone who bought a machine that exploded, and who wants to sell his huge but now-useless set of 256 feeders for $200... or €200. That would make us happy, no? Seems like pick-and-place machines are a little like ink-jet printers - the machines are kept [relatively] inexpensive, but they totally nuke your budget on supplies (feeders and ink cartridges).
A big factor for me deciding on the 7722FV is that I know there are numerous other people here with the same (or similar) machines. For a small scale company like my own, that's a pretty big plus when you can bounce questions off other people, and share your experiences/parts/etc. The other big factor (aside from budget) was just physical size. The 7722FV it big, but not so big that it can't fit in a small office or apartment. It's very 'movable' with two people and fits on a table unlike some massive industrial beasts that require much larger office facilities.
Yup, that's a big plus for the 7722FV. I haven't been able to find a single video of a LE40V... or even one of their other pick-and-place machines on the internet. That makes my analysis more difficult. The only partial compensation is... they're made in the USA and the company is in the USA, so I can call them up easier than a japanese or european company... and much easier than a chinese company (for which the documentation seems to be horrific for these kinds of machines). Probably either of these machines would be fine for me, though I'm still leaning towards the LE40V (mostly due to 0201s and [claimed] placement accuracy).