best equipment for SMT prototyping

Recently I designed a robotics “vision system” and ordered prototype PCBs and components (most of which has arrived). However, I haven’t been able to decide what is the best way to assemble prototype PCBs with lots of SMTs, including BGAs, QFNs, QFPs and resistors and capacitors in sizes 0805, 0603, 0402… and 0201 to bypass the GND and PWR pads underneath BGAs. The finest QFN pad-pitch and BGA ball-pitch is 0.50mm (with 0.25mm width pads).

The approaches I’ve considered are:

#1:  http://www.aapcb.com - let someone else with pick-and-place assemble my PCBs.
#2:  http://www.zeph.com/systems.htm - 3LMTZ-7-BGA system ($8000 ~ $10000).
#3:  more-or-less manual assembly system including:
     A:  stereo microscope with dual boom stand - $500
     B:  PCB holder cradle to hold PCB above preheat - $200
     C:  stainless steel stencils for both of my PCBs - $200
     D:  vacuum-pen SMT pick-up tool to place SMT devices - $200
     E:  hot air pencil to solder discretes and parts with pins - $1000
     F:  hot-air heater to pre-heat PCB to 150C~210C for soldering - $750
     G:  hot air gun (plus selection of nozzles) to solder BGAs/QFNs - $200
     H:  high quality reflow oven (if necessary to achieve reliable quality) - $4500
     I:  miscellaneous other related devices, tools and supplies - $2000 plus or minus $1000

My nominal product choices for the above:
     A: http://store.amscope.com/sm-4tz-80s.html
     B: http://www.zeph.com/board.htm
     C: http://www.quickstencil.com/prototype-stencil-kit.aspx
     D: http://www.zeph.com/zt3web.htm
     E: http://www.zeph.com/pencil.html
     F: http://www.zeph.com/smdpreheater.htm
     G: http://www.amazon.com/Aoyue-852A-Digital-Rework-Station/dp/B000HDDXY0
     H: http://www.apsgold.com/reflow-ovens/low-volume/gf-b-ht-reflow-ovens
     I: http://www.zeph.com/directory.htm

I’m looking for advice, especially from people with lots of SMT assembly experience. For sure, I wish I had $40,000 to buy a spiffy automatic pick-and-place machine with 4+ zone conveyor reflow oven… but alas, that’s $10K more than my entire budget for this entire project, including prototype PCBs, tooling for the die-cast aluminum cases, components (on the PCB), a 4-channel 5GSPS PC oscilloscope (picoscope 6403), and “all other expenses” (mostly A~H above). And yes, this is a personal project, but also yes, I’m planning to sell the results as a product when it is complete. So yes, this is “more than a typical hobby project”, but “funded by many years of frugal living and savings… and nothing remotely resembling a real commercial budget”.

The typical response to my questions so far has been “you bit off far more than you can chew”. No problem, that’s probably what they honestly think. But I’ve bitten off vastly more than I could chew on dozens of projects over the past 30 years, and I finished every single one, and sold or licensed most of them to companies, and they became real products. On the downsize, though I designed and prototyped dozens of fancy gizmos with large, sophisticated multi-layer PCBs back in the “dark ages”, I’ve been focused on software the past 15 years, and SMT has taken over everything during my “keyboard years” (for lack of a better term). So I have lots of general context and experience, but pretty much zero up-to-date PCB assembly experience. Hence my request for some considered opinions from those more up-to-date on prototype assembly with fine-pitch SMT.

In case it matters, the details of my PCBs are:

#1: the “eye” - 2.80" x 2.80" x 0.0625" - 6 layers - 50 components - prototype quantity 50

#2: the “quad” - 5.80" x 5.80" x 0.0625" - 8-layers - 300 components - prototype quantity 25

The PCB surfaces are ENIG (immersion gold).

Their minimum trace/space/via-size is 4-mils/4-mils/8-mils.

The PCBs adopt “fill-and-cap” to implement “via-in-pad” to make BGA routing/assembly easier.

All components are SMT except 40-pin ribbon-cable headers and RJ45/power/microphone-jacks.

The ID of every silkscreen box around every BGA and QFN part is exactly equal to the maximum OD of the BGA or QFN package it surrounds (to aid placement).

While the above PCB assembly techniques (and product choices) are the result of my searching and researching on the internet the past few weeks, maybe you have an entirely different approaches and equipment/devices/supplies/etc: I am happy to hear them. I need to keep my total budget for PCB assembly goodies under $10,000… and hopefully many/several thousand under that.

One specific question: Does anyone know from experience whether it is practical to solder fine-pitch BGAs and QFNs with these hot-air solder guns + nozzles… with the pre-heater below the PCB to pre-heat it to ~170C before soldering from above?

PS: I strongly resist option #1 above, even though that might be considered “the safe way”. Why? Because then I’ve “blown” my PCB assembly budget on a one-time “service” rather than equipment that will serve me well over the next 20 projects and ~50 PCBs I plan to develop over the next 3 to 5 years. Thus I think very much opposite most companies I’ve seen, who are happy to pay infinite sums for salaries, consultants and services… but refuse to spend anything for capital equipment to help make their employees productive!

So… what should I do? Which approach? What equipment/devices/supplies/etc?

I’ve had prototype boards with BGAs assembled by this company:

http://www.asktechnology.co.uk/About%20ASK.htm

Their charges were very reasonable.

I just recently had my first SMT experience, so I’ll share. It was a 1.1" diameter board with 2x QFN (one was 64 pins and 0.4mm pitch),

several 0602 and 0402, and a few other things. No BGAs or 0201s, though.

I used a kapton stencil from ohararp - cheap at $25 (I think) each, good service and fast turnaround. My only complaint was that the fine pitch pads seem a little irregular, although all of the larger pads including the 0402s seem perfect. I never could get an asthetically pleasing solder paste application to those fine pitch pads, it would always smear a bit which on these small pads meant that it pretty much just covered the whole area. Had a few external bridges after reflow, but a couple of minutes with the solder wick took care of that with no problem. The bridges looked like I just had too much solder, possibly due to the HASL board finish adding more solder than what was in the paste.

For reflow, I just took the nozzle off of my hot air rework station gun and sort of bathed the board in the hot air, starting far away and getting closer to somewhat approximate the reflow curve. I did have a few components that blew away, though, so I’m looking at just using a reflow toaster oven for the next batch that I make.

Disclaimer - I’ve never used high quality reflow ovens or other high dollar SMT equipment…

My equipment advice:

Get a decent combo hot air/soldering iron station. I got the Aoyue 908 - you don’t really need a digital temp display, it has a 60W iron, and its cheap!

Try a reflow toaster oven setup. Again, cheap and seems to work well from what I’ve read.

Then, if that doesn’t satisfy your soldering needs you can look at other equipment. The hot air/iron station you’ll need anyway so its not a bad place to start.

Not sure what the difference in your listed hot air pencil vs the Aoyue station other than the size of the handle. As far as I can tell they do the same thing, the zeph just costs 4x the Aoyue.

I used a hand loupe instead of a microscope - but given your budget I’d say the microscope would be very useful. I’m looking at the same one.

I’m not sure if the vacuum pen is worth it or not. Tweezers work pretty well, too, although I did have some problems with the small discretes sticking to the tweezers instead of the board occasionally.

I didn’t use a board preheater, but if you’re going to be doing hand rework I could see it being useful. I swear I’ve seen them in the few hundred dollar range including the cradle to hold the PCB.

Check Madell Corp for cheap SMT stuff http://www.madelltech.com/index.html I got a power supply from them on ebay. Only snafu was a misleading shipping amount, partially because I’m in Hawaii and it costs $$$ to ship to paradise… but he never updated his later auction listings to reflect the proper shipping price so I’ll say that’s a little shady. Ebay has an ok selection of Chinese equipment that’s not too bad.

You’ll have to decide if the Zeph quality is worth it - it looks nice but I’m not sure it’s that much better than the imports from ebay.

So, to summarize, yes, I see no reason why you can’t do this with the hand equipment you listed.

Please tell me that all leadless ICs (BGAs, LGAs, QFNs, DFNs, etc) are on one side of the PCB. If that is the case, a hot plate should be enough to do your prototypes. Getting the BGA to solder properly will require experimentation. I’d recommend getting some scrap PCBs to practice on. If you have leadless parts on both sides, a hot plate will probably not work, especially considering the large size of one of your PCBs. In that case, a pre-heater and a hot air station will do it, but it will be trickier.

The keyword here is practice. You’re gonna have a few dead boards. Maybe many.

NleahciM:
Please tell me that all leadless ICs (BGAs, LGAs, QFNs, DFNs, etc) are on one side of the PCB. If that is the case, a hot plate should be enough to do your prototypes. Getting the BGA to solder properly will require experimentation. I’d recommend getting some scrap PCBs to practice on. If you have leadless parts on both sides, a hot plate will probably not work, especially considering the large size of one of your PCBs. In that case, a pre-heater and a hot air station will do it, but it will be trickier.

The keyword here is practice. You’re gonna have a few dead boards. Maybe many.

Almost all the non-trivial SMT parts are on the same side, but not quite. All the BGAs and QFNs are on the same side of the PCBs, except for two 28-pad QFN chips (max8717s). And I have oodles of 0603, 0402 and 0201 bypass capacitors and a few less funky packages on the opposite side too (less funky meaning I can see and touch the leads (like SO8, SOIC, TSSOP).

All the “heavy” (BGA) parts that might “fall off” if inverted in a reflow oven are on the same side. Actually, I just received the parts, and I am rather amazed at how tiny and lightweight even the biggest, baddest, heaviest component is. Even the 256-ball BGA is a tiny featherweight little thing. Working on the PCB layout at 3200% to 10,000% obviously distorted my impression of how tiny these things are. I can barely even see the 0201 bypass capacitors unless I take my glasses off so I can look closer (5.50 diopters).

Absent some vibration or jarring, I’d bet solder surface tension would keep even those components on the bottom side of the PCB in a reflow oven, if necessary. However, no need for that, except perhaps for those two teenie weenie little 5mm square QFNs, which weight about 1/10 gram, I’d guess.

Do you think the extensive “practice” and “scrap” is mostly due to “placement precision errors”? Or “overheating and thereby destroying components”? Or what? And what is the best way to avoid those problems (besides a $30,000 pick-and-place machine)?

Part placement can be surprisingly bad, actually, and surface tension will pull it in. I have no experience with BGA, though, so those may be more sensitive.

I hope you didn’t order 50x and 25x of all of the boards before doing a few as a prototype. I can basically guarantee something isn’t quite right on your first one, especially if its the first you’ve ever designed. It took me three versions for my first board, although I now know what steps I should have taken to prevent that… One was a bug in Kicad that ended up with no clearance in the power and ground planes around a few vias, and one was a problem with a bug in the the microprocessor I was using - a Cypress PSOC3 which is still in beta test. Even after that, there are a few things I learned during testing that means I’ll have at least one more version before I go to production.

You’ll fry a few components, have solder bridges, probably move a few components while soldering the other side, and generally screw up a few times, but you’ll get the hang of it quickly. Practice is worth much more than a $30k machine in preventing screwups. I think the big advantage to the machine is speed.

tstalcup:
Part placement can be surprisingly bad, actually, and surface tension will pull it in. I have no experience with BGA, though, so those may be more sensitive.

I hope you didn’t order 50x and 25x of all of the boards before doing a few as a prototype. I can basically guarantee something isn’t quite right on your first one, especially if its the first you’ve ever designed. It took me three versions for my first board, although I now know what steps I should have taken to prevent that… One was a bug in Kicad that ended up with no clearance in the power and ground planes around a few vias, and one was a problem with a bug in the the microprocessor I was using - a Cypress PSOC3 which is still in beta test. Even after that, there are a few things I learned during testing that means I’ll have at least one more version before I go to production.

You’ll fry a few components, have solder bridges, probably move a few components while soldering the other side, and generally screw up a few times, but you’ll get the hang of it quickly. Practice is worth much more than a $30k machine in preventing screwups. I think the big advantage to the machine is speed.

Yeah, I did order 25 and 50 prototypes. OTOH, I designed dozens of products many years ago, including 300-IC CPUs, and by the time I stopped designing hardware 15 years ago, I was so experienced at designing digital systems that I stopped wire-wrapping prototypes, and went straight to PCB... with ZERO "patch wires". I don't want to sound arrogant though, because it has been 15 years, and "total screwups" are always possible. The other factor is, the prototype house charges about twice as much for 50 PCBs as 1 PCB, so "what the hell". However, your point about bugs in the software tools is perhaps more a possibility. All my old time schematics were manually drawn, and my PCB layouts were manually created with stick-on patterns and tape on huge mylar sheets. These are my first devices with schematic and PCB software - in my case the diptrace software. At least I don't have any brand new components; probably cyclone3 FPGA, silicon labs C8051F120 CPU and aptina image sensors are the most advanced (not every).

Yes, I’ve heard about chips self-centering on the pads when the solder melts, but with 0.50mm pitch, anything over 0.20mm error is likely to self-center the ball or pad on the wrong pad! That’s the only advantage I see in the $30,000 pick-and-place machines… they’ll put the components down within 0.001" of their correct position. When I go in to production, I’ll care about speed. Right now, on the prototypes, I just care about getting them assembled correctly. Otherwise I might wonder why my FPGA firmware or C8051F120 assembly language isn’t working right… and the answer is a solder-short or open in my assembly. I don’t want that.

maxreason:

NleahciM:
Please tell me that all leadless ICs (BGAs, LGAs, QFNs, DFNs, etc) are on one side of the PCB. If that is the case, a hot plate should be enough to do your prototypes. Getting the BGA to solder properly will require experimentation. I’d recommend getting some scrap PCBs to practice on. If you have leadless parts on both sides, a hot plate will probably not work, especially considering the large size of one of your PCBs. In that case, a pre-heater and a hot air station will do it, but it will be trickier.

The keyword here is practice. You’re gonna have a few dead boards. Maybe many.

Almost all the non-trivial SMT parts are on the same side, but not quite. All the BGAs and QFNs are on the same side of the PCBs, except for two 28-pad QFN chips (max8717s). And I have oodles of 0603, 0402 and 0201 bypass capacitors and a few less funky packages on the opposite side too (less funky meaning I can see and touch the leads (like SO8, SOIC, TSSOP).

All the “heavy” (BGA) parts that might “fall off” if inverted in a reflow oven are on the same side. Actually, I just received the parts, and I am rather amazed at how tiny and lightweight even the biggest, baddest, heaviest component is. Even the 256-ball BGA is a tiny featherweight little thing. Working on the PCB layout at 3200% to 10,000% obviously distorted my impression of how tiny these things are. I can barely even see the 0201 bypass capacitors unless I take my glasses off so I can look closer (5.50 diopters).

Absent some vibration or jarring, I’d bet solder surface tension would keep even those components on the bottom side of the PCB in a reflow oven, if necessary. However, no need for that, except perhaps for those two teenie weenie little 5mm square QFNs, which weight about 1/10 gram, I’d guess.

Do you think the extensive “practice” and “scrap” is mostly due to “placement precision errors”? Or “overheating and thereby destroying components”? Or what? And what is the best way to avoid those problems (besides a $30,000 pick-and-place machine)?

That is good enough to make this doable. The failures I had while developing a BGA soldering process were due to using the wrong amount of solder paste or not putting it on evenly enough. I also have cooked a PCB once (causing delamination), though that was a more complicated situation.

It is funny how looking at zoomed in PCBs can really distort your sense of size. It can be tricky when making really dense PCBs to keep your components placed far enough away from each other to be solderable.

My advice would be to concentrate on the BGAs before anything else. If your process works for BGAs, most likely everything else is fine. I’d suggest even destructively testing a successfully soldered PCB (or sending it out for X-ray) to be absolutely sure everything is working. I pried up the BGAs to examine them. This would rip pads off the PCB and balls off of the BGA. This made examination a breeze.

I used a timed, air powered solder paste dispenser to put on paste - but a stencil should work too. Probably faster as well (dropping 276 balls of solder paste, one by one, is actually a lot of work!)

As for self centering - yes self centering works OK for most things. One time a colleague messed up his LGA footprint so we couldn’t tell how to center it. But we were soldering it on a hotplate, and when the solder went moltent you could tap the PCB a teeny bit and it’d snap into place. What was even more interesting was taking some tweezers and pushing the component. It felt like you were pushing a ball over a sinusoid. Pretty cool.

One more word of warning: your silkscreen markers showing the outline of the component may not be good enough. Many PCB fabs have pretty bad registration between the soldermask and the copper. I prefer to put copper corner marks to show the corners of BGAs, as I’m sure those will be aligned with the pads of the BGA.

NleahciM:
That is good enough to make this doable. The failures I had while developing a BGA soldering process were due to using the wrong amount of solder paste or not putting it on evenly enough. I also have cooked a PCB once (causing delamination), though that was a more complicated situation.

It is funny how looking at zoomed in PCBs can really distort your sense of size. It can be tricky when making really dense PCBs to keep your components placed far enough away from each other to be solderable.

My advice would be to concentrate on the BGAs before anything else. If your process works for BGAs, most likely everything else is fine. I’d suggest even destructively testing a successfully soldered PCB (or sending it out for X-ray) to be absolutely sure everything is working. I pried up the BGAs to examine them. This would rip pads off the PCB and balls off of the BGA. This made examination a breeze.

I used a timed, air powered solder paste dispenser to put on paste - but a stencil should work too. Probably faster as well (dropping 276 balls of solder paste, one by one, is actually a lot of work!)

As for self centering - yes self centering works OK for most things. One time a colleague messed up his LGA footprint so we couldn’t tell how to center it. But we were soldering it on a hotplate, and when the solder went moltent you could tap the PCB a teeny bit and it’d snap into place. What was even more interesting was taking some tweezers and pushing the component. It felt like you were pushing a ball over a sinusoid. Pretty cool.

One more word of warning: your silkscreen markers showing the outline of the component may not be good enough. Many PCB fabs have pretty bad registration between the soldermask and the copper. I prefer to put copper corner marks to show the corners of BGAs, as I’m sure those will be aligned with the pads of the BGA.

Yeah, I am worried about the silkscreen precision too, and wish I had put little corner marks in copper like you said. The PCB makers claim the silkscreen precision is 0.003" to 0.004", but I won't fully believe that until I get the PCBs (early next week, supposedly). I figure any precision better than about 0.008" ought to be sufficient (1/3 the minimum pad width).

This conversation makes me wonder if one of the “crude techniques” might actually work well in practice. That is, only cap the BGA vias on the opposite side of the PCB from the BGA instead of both sides. This might drain some BGA-ball solder into the via, but it means the assembler could be CERTAIN the BGA is properly placed, because he could feel the balls drop into the vias like soft, mild detents. And surely a 0.008" hole only 0.0625" long isn’t going to drain the majority of the BGA ball into the via. I wonder if anyone has tried this? Seems like an ideal solution if it works. As it is, I chose “via in pad” under all BGA pads, with “fill and cap” treatment (fill the via with something conductive, then somehow additively plate copper pads over the existing pads and filled via). This certainly made for nice, easier, neater, roomier PCB layout.

I would simply go with AAPCB. I don’t claim to be much of a n expert with hot ovens but I do have some experience with soldering fine pitch components including QFNs and repair using a hot air gun. In my experience, the time and effort that is expended the first few times in checking your boards (since they never come out correct the first time) is just not worth it. Besides you have a 0.5 mm BGA and handling that is going to be fairly difficult I would say. And, cold solders can happen anytime if your reflow process isn’t proper. If you are planning to sell your boards to several people, you don’t want them coming back to you due to these cold solder issues.

The other thing that a professional shop can do, is X-Ray inspection of BGA assembly which is quite helpful in fine pitch BGA.

I have used AAPCB several times in the past and have pretty much got a 100% yield from them. I would at least give them a call and get my first 2 boards assembled from them so that I am 100% sure that my design is correct and there are no bugs in my circuit. For the remaining boards, it is up to you how you would like to proceed depending on the cost/benefit of the two options.

I second (third?) just sending them out. I’ve been getting offers all the time lately from prototype assembly places, and for 50 boards I’m guessing it will work out to be less than a decent stencil printer and oven. I’ve done plenty of small and leadless stuff (although I’ve never attempted a 0201) and if it were me I’d at least send the first board to one of the quick turn places and validate the design with a machine placed board. For the $100 or so one of your boards will likely cost as a one off (in my shop at least) I can guarantee more than $100 worth of frustration debugging the first board if hand soldered.

Aaron