Advice on setting up a low-mid volume SMT line for <$130k?

[ar__systems]

Macrofab's prices are "unheard of" and not in the way they mean it. They are super expensive even for North America.

[48X24X48X]

Macrofab's prices are "unheard of" and not in the way they mean it. They are super expensive even for North America.

Thanks for this information, was about to try their online quote system (have to made tonnes of changes to fit into their system just to get a quote, it even has to be in mils!). I think I can just skip them.

[SVFeingold]

If anyone knows of a US based PCBA with a good online quoting system, I'm all ears. They either have a good system but charge stupid prices for it, or ask you to submit an RFQ before they'll dare tell you anything. "Will this cost $500 or $500 million?" "Sorry we really can't say without a formal quote." No I'm not going to "call you" because I have 50 other people that are asking the same thing rather than just saying the damn price.

I hope we see more reasonable prices from these sorts of outfits soon. The reason people pay these crazy prices is the convenience and visibility.

[mairo]

...
 this will be a cramped room. Here is my idealized room plan after getting the machine footprints and measuring the room. I believe I can get everything seen here ...

Make sure all fits via the front door

Also it is very good idea to be able to have quick access all around the machines.

[mikeselectricstuff]

Assembly costs for smaller runs vary a lot, and if comparing costs vs. a big investment in doing it yourself  should be looking at costs from at least 4 or 5 places over a number of boards.
I find that costs can easily vary by a factor of at least 2, and this can be very inconsistent for different jobs quoted by the same two assemblers.
IME in the UK. $0.10 per placement is at the higher end of what I'd expect to be paying, and half that can be findable from the right place for some jobs.

[Jeroen3]

We made a machine once that didn't fit trough the door, because the door wasn't compliant to building standards... We had to reinforce it on the inside and put it on it's side to get in. Fun times. It was a large rittal cabinet.

I am missing a testing station on the floor plan, is testing done somewhere else?
Do you also need a small fridge to store paste in? And an oven for MSL parts?

[mikeselectricstuff]

...
 this will be a cramped room. Here is my idealized room plan after getting the machine footprints and measuring the room. I believe I can get everything seen here ...

Make sure all fits via the front door

Also it is very good idea to be able to have quick access all around the machines.

Yes - check carefully about machine access. Every assembler I've ever visited had access to all sides of all machines. You don't want to have to dismantle all the conveyors to pull a machine out to access something.

You also have a lot of wasted space in the middle. I'd be inclined to have the machines back-to-back in the centre with a walkway between them - that way you can use the wall space for storage shelves etc.

[SVFeingold]

The door can certainly be made bigger.

One area of concern to me is the floor. It is concrete and can support heavy machinery but it is rather rough and not perfectly smooth/flat.

After speaking with someone from SMTNet in greater depth (the founder of Olimex actually!) and getting his detailed advice, I looked for new Samsung machines from Korea/HK. New as in packed for sea shipping in the Samsung factory. I was quoted 70k for a new Samsung SM482. That's pretty crazy considering it's a $200k machine in the US! I'm also looking at the SM451 as it is slower but can place more components.

Has anyone here got opinions on this choice? Choose a new Samsung machine with difficult access to support, or a used Assembleon machine with a service contract + training. The Assembleon MG-8 w/ feeders will cost around $40-50k. The Samsung with feeders will be about 90k (waiting on feeder quote) + freight. On the other hand the Samsung would be a new machine that ought to run for quite a while without issues, while the Assembleon is 10 years old...

The Samsung leaves $40k for the rest which is enough but not much "uh-oh" money. Then again it should be a nicer and more reliable machine, maybe worth skipping all of the fancy stuff and hand loading boards for a while.

Very difficult decisions.

Re: Machine access, I think I can put the PnP in the center of the room, so that the whole line is in almost a square shape. That leaves easy access, maybe not as nice feng shui but, eh...

This order is "only" about 600 panels to process. Maybe worth just getting the basics - stencil, PnP, reflow, and shuttling them by hand for now.

Mike just saw your post, yes that may be a better idea! Plus the floor is a little nicer in the center.

[SVFeingold]

No labels but this seems like a nicer layout from a usability standpoint, no?

Jereon3, I have a paste fridge already that can fit under the table. There is no testing station as test fixtures don't exist yet and I certainly don't think I can fit an automated test system into the budget. The client will be testing each board as it's assembled to the final product so they are not sure they want test fixtures at all. I of course recommend them but it is more time. Their call. As long as my boards are done well. No oven for MSL or dry box. This is high on the list. Worst case I can make one, a dry box is simple enough. An oven for bakeout I don't have yet but that is relatively simple as long as I monitor it so as not to cook $50k of components...

[jmelson]

500 sq ft? Are you sure you have this right?  A stencil printer, Gem-series P&P and a Heller oven?  I don't think the machines can be STORED in 500 sq ft, and you'd go NUTS trying to actually run them in that space.  My CSM84 has a conveyor, and front and back feeder rails, so it needs access on all FOUR sides!    I see you have the P&P against the wall!  So, eliminate the rear feeder, suddenly, you are down to half the number of feeders.  Is that enough?  Also, it may require you to move the machine for maintenance.  So, my CSM is actually only about 35 sq ft, but to use it, you need more like 100 sq ft, and even at that, you'd be quite cramped.  You need space for the oven, which could actually be bigger than the P&P, and it probably needs it's own environment, or it will make the rest of the space unbearably hot.   This could affect the P&P and the stencil printing, too.  (Also note, air compressors, unless water-cooled, also throw off a LOT of heat.)  The stencil printer might be the least of the space consumers.  Finally, you need space to store parts, benches to organize and inspect the work, a station to load and check feeders, and so on.  I'd think you could just do it in 1000 sq ft,  my shop space is 1400 sq ft but has a metal shop in there, too.

Hmm, if you had a double-fold conveyor, you could put the oven behind the printer and P&P, and maybe have a curtain between them to try to isolate the heat.  The finished boards would come out behind the printer.  If you arranged for the air conditioning return to be over the oven, it might keep the rest of the shop barely comfortable.  In that tiny room, you will actually be bothered by the heat radiating off the finished boards!

Zig-zg board?  Oh, you will be using panels.  Your panels will be at LEAST 15 x 21", which is pretty large.  You will need a LOT of push-up pins to keep the whole panel from bouncing the parts all over the place.    Even my old CSM handles panels, with bad marks to indicate which individual board failed electrical test, so the P&P can skip them.

Once you have the solder stencils and programming all worked out, making another batch of boards is a SNAP!  An hour or two to set up all the feeders, align the stencil and off you go.

Digi-Key is great as they have a lot of stock, but in the volumes you are talking about, going direct to the manufacturers is cheaper.  I have a local outfit that carries On-Shore connectors at half the price at Digi-Key.  They carry NO stock, so delivery is 2 - 6 weeks, but for HALF the price, it is WORTH it.  Those darn connectors really ADD UP!

Jon

[SVFeingold]

Once I have the contract and have the SMT line in process I will definitely spend some time looking for cheaper sources. You are right about connectors, going from kinked-pin to SMT connectors makes them 2x as expensive on Digikey. 30 cents per connector, with 6 per board! Unless of course the machine can place through-hole/kinked pin but they do not come in reels...

I will need to add ventilation for sure. Yes it will indeed be cramped and uncomfortable but that's ok for now. As soon as a larger space is available I will move the machines. There is one right down the hall that's 1400 sq. ft. that I really hope will be vacated soon!

I assume that with proper care there is no harm to moving machines after installation. Just a hassle.

[jmelson]

Concrete floor is GOOD!  The P&P bounces around like some kind of crazy kids ride, it even bounces a little while on the foot pads!  Don't worry about level or finish, the machines should all have big foot pads to level them.

I'd turn the new setup the other way, put the oven in back.  The printer and P&P are what you will be dealing with the most, might as well have those closest to the table/bench.  I prefer a tall bench and work mostly standing, as I never get a chance to sit before running back and forth between stencil printing (manual for me) and the P&P and oven.

[SVFeingold]

Makes sense jmelson, that eliminates the need to move the conveyor or have a retractable conveyor too, easier access. Good idea!

[rx8pilot]

I have not read all through this thread - but I will quickly throw out a quick opinion based on my own personal opinion of putting together and operating a small line.

However much money you think it will cost - it will cost more.
You will be short feeders, things will break, there will be a software license, sheared nozzles, fiddly servo driver, etc.

However much time you think it will take to get up to speed and setup jobs - it will take longer.
The learning curve does not really taper off for a very long time.

Not trying to talk you out of this, only to be realistic. I am very happy to have the in-house capability - but cost of entry has been far more than the cash transaction.

[SVFeingold]

I hear you rx8. I'm trying to leave at least $40k in reserve to cover things like that. Now if I got a new machine that might go down to $10k - $20k. Do you think it's smarter to get an old (but apparently reliable) machine or a new machine but with less margin? Also I don't think I'd have access to onsite training with the new Samsung machine as I would with the used Assembleon. NOT having this makes me nervous. For feeders the Assembleon I'm looking comes with several hundred. So if at least 50 work I'll be in good shape.

My assumption is that as long as everything works the K&S guys would help me set up and program some parts, one board, I'll be able to speed the process along provided nothing catastrophic happens. I.e. my singular goal is to get this batch of boards finished. Once that's set up there is no big rush. I think I can do this in 2 months. We'll see...

[mikeselectricstuff]

for your scale it seems an expensive route to get a new machine - a bit like buying a new car, which depreciates rapidly as soon as you drive it home. Unfortunately unlike cars you're not going to find a 6 month old ex-demo at a knock-down price.

If a company is offering a service package on older models at a sensible price it means they have a fair idea of its likely reliability.

Some valid points about the heat from the oven - this probably varies a bit between models - just looked at a random Heller spec an it says 7.5-9.5kw continuous. All of this will be coming out of either the extraction duct or  the top/sides.  You're certainly going to want some decent ventilation.

[jmelson]

However much money you think it will cost - it will cost more.
You will be short feeders, things will break, there will be a software license, sheared nozzles, fiddly servo driver, etc.

However much time you think it will take to get up to speed and setup jobs - it will take longer.
The learning curve does not really taper off for a very long time.

Not trying to talk you out of this, only to be realistic. I am very happy to have the in-house capability - but cost of entry has been far more than the cash transaction.

Well, these high-end professional machines are quite well built.  I've been running my (used) Philips CSM84 for 10 years, now.  The only real thing I've had go bad was a sensor that stops the conveyor when the finished board is at the end of the conveyor.  I got a replacement for ~ $60 or something at PLC center.  I have had to get replacement filters for the vacuum generators, there are guys in China making spares for these machines.  Cost more to ship them than for the parts.  Now, I caused myself some trouble by tinkering with a speed controller on the nozzle lift, and not realizing I had to retime the software, so it caused some crashes where the XY motion started before the nozzle and feeder plungers were clear.  I had to make replacement parts for the air pistons.  I hacked in an additional sensor to keep the machine paused until the nozzle was fully up.

Now, I'm not doing the level of production that Mr. Feingold is talking about, here, but this should give some idea.  I've done over 1000 boards on my machine, and many have several hundred components each.  I started off with about 50 assorted feeders, and managed to trade some extra stuff I got with the machine for some 12 and 16mm feeders.  I made a waffle tray holder to go on the rear feeder rail.  That pretty much has me happy for doing all the boards I make.

As for setting up, I took the placement file from my cad/cam system and hand edited it into the format required, and was in full-scale production in less than 2 weeks.  Realize, I had never even SEEN a P&P machine in real life up to that time!

Yes, I have learned a lot since.  I've learned the tricks of reducing stencil apertures to get the right amount of paste to make reliable joints without bridges.  I've learned how to select what part goes in what feeder slot to optimize machine motion.  (Obvious hint, the most frequently-used parts should be closest to the board where the machine fixtures it.)  I've learned the right reflow profile for various boards (I do some tin-lead and mostly lead-free).  I now know to pre-bake any boards that have sat around for a few months before assembling them.  I've discovered that certain feeders don't work right with CERTAIN parts, and have marked them.

So, I have learned a lot, and figured out how to be more productive with the setup.  I do a wide mix, I have at least 15 different boards I do, and so I change over the setup quite often.

Jon

[rx8pilot]

I hear you rx8. I'm trying to leave at least $40k in reserve to cover things like that. Now if I got a new machine that might go down to $10k - $20k. Do you think it's smarter to get an old (but apparently reliable) machine or a new machine but with less margin?

Just to reiterate - I don't want to make any attempt at talking you out of this idea. My very old machine is quite reliable overall. I did a ton of repairs up front which is why it was a low-cost to start. A fixer-upper. That part of it is not much of a problem.

I was more generically referring to the process as a whole. The first year was a LOT of learning and fiddling with various things. After the first year, I started doing more variety which required more feeders. Then, to address the higher mix/volume I needed to work out a more suitable work flow - which is not trivial. So, 18 months in I am finally comfortable with the setup, but still shocked and somewhat disappointed at the overall effort that goes into PCB assembly. Keep in mind that outsourcing itself is no picnic either and generally scares me that something bad will happen and I don't learn about it until the boards are delivered with backward diodes.

[jmelson]

I hear you rx8. I'm trying to leave at least $40k in reserve to cover things like that. Now if I got a new machine that might go down to $10k - $20k. Do you think it's smarter to get an old (but apparently reliable) machine or a new machine but with less margin?

Hmmm, that is so hard to answer!  I do have an older machine, but the CSM84 was built like a TANK!  The nozzles are just tool steel tubes with a partial slot milled with a ball-end mill down the side, this is how the keying ball keeps the nozzle from rotating, and also retains the nozzle in the holder.  So totally SIMPLE and nearly foolproof.  (Well, it is a little fiddly to put the nozzle back in.)  I look at some of the nozzles on other machines and think "Man, that is so FRAGILE looking, one wrongly placed part and SNAP!"  Just like that, the whole CSM84 machine is very BASIC, simple and robust.  So, you would do well to look at the construction and the robustness of the basic machine as well as the consumables.  Gecko Drives makes stepper drives and related gear in California, and I think he had a Mydata.  I forget, but he called it "Matilda" so that's why I'm thinking it was a Mydata.  Anyway, at least once a month, maybe more often, the control program would go haywire and the machine would fly off and crash in the middle of a job, sometimes doing minor damage.  The panel would have to be wiped off and started over.  This drove the owner nuts!  He eventually had to replace the machine.  Apparently, although the maker would not acknowledge it, it was a frequent problem.

Also I don't think I'd have access to onsite training with the new Samsung machine as I would with the used Assembleon. NOT having this makes me nervous. For feeders the Assembleon I'm looking comes with several hundred. So if at least 50 work I'll be in good shape.

I'd be WAY more nervous about parts availability than the training issue.  I learned my machine completely out of the manuals, with maybe the slightest bit of help online.  If the manuals I have are good enough, I suspect that the later ones may even be better.

Anyway, before making any moves, I'd certainly talk to somebody that USES the prospective models, and get their take.  Does it break or just use up consumables?  Does it take a lot of maintenance, or just keep chugging along, week after week?  Does the calibration drift over time, or do you just totally ignore that?  Has the software ever crashed, and how bad is it to recover?

I have to say, my CSM84 has been totally astounding.  It sits in a musty basment, and when I turn it on it comes to life and just starts placing parts.  Other than the issue I caused that I mentioned earlier, and a variety of operator/programming errors like calling for the wrong nozzle to pick up a part it can't reach, I've never had a real problem in 10 years!  Maybe I'm just insanely lucky, and I know I am not using this machine at the normal duty cycle, but it seems to be a really good machine.

My assumption is that as long as everything works the K&S guys would help me set up and program some parts, one board, I'll be able to speed the process along provided nothing catastrophic happens. I.e. my singular goal is to get this batch of boards finished. Once that's set up there is no big rush. I think I can do this in 2 months. We'll see...

2 months to SET UP, or 2 months to assemble the 1000 board?  I'm guessing you mean just to set it all up. 

Jon

[SVFeingold]

2 months for both, but it maybe could be pushed to 3 although I'd prefer not.

As for the spare parts, the same center that does training also does service, they have a warehouse full of spare parts and say they can get anything they don't have. There's ALSO a Yamaha service center about 30 minutes away. They basically built these machines and can also support them with parts. So essentially in the case of Assembleon MG-8 I have 2 service centers within a 30 minute drive. With the Samsung, it's new, probably more capable and probably more modern software (though not necessarily better but I have no idea, haven't used both. But if I get stuck - who do I call?


The issue finding folks to talk to is that the ones I'm talking to who own the machine are also the ones selling it! In one case it is running active production on a line but they are upgrading their equipment. So at least I know it works. I am told the maintenance is pretty minimal, the service techs do 1-2 calibrations and head rebuilds a year as preventative maintenance - it's included in the service contract. But at least the machine is working and actively maintained.



On a related note, how do you all do your stencil apertures? I typically follow the component manufacturers stencil pattern if they provide one, but otherwise I now add a -0.025mm stencil aperture reduction on all fine pitch/small passive pads, and round all pad corners but 5% - 20% depending on the size of the pad, lower % on larger pads to get similar radii. For exposed pads I typically use a "window frame" type pattern with rounded corners and ~70% paste coverage.

EDIT: The real concern with the used vs. new is the cost of ownership after a year. What are the hidden costs I haven't considered? That sort of thing. Is it cheaper over a year to buy a new machine at a premium of 20%, or to buy a used machine that is guaranteed to work on delivery and has easily available service? I think this is just one of those things that's impossible to know. If I had just another $50k to spend it would be an easy decision but that's not the case. Maybe less money spent is better as it gives me room to add automation (magazine loaders, etc...) and accessories/unexpected costs.

[SVFeingold]

A quick machine update, right now I have 2 options:

Buy a new Samsung SM451 for ~$70k. Add feeders (probably $10k - $20k) and shipping to US + whatever is involved with customs. Still waiting on a quote. Purchase a Heller 1707EXL oven and MPM Accuflex stencil printer separately. Total cost: ~$121k - $131k including assumed $5k cost for overseas shipping + stencil printer shipping. Leaves me with $19k - $29k for additional equipment and buffer.

Get a used Assembleon MG-1 with a few hundred assorted feeders as part of a package that also includes a Heller 1707 EXL Mark III, MPM Accuflex printer w/ 2D inspect. Total cost ~$100,000 shipped to my door and guaranteed working - although I don't know what "guaranteed" means exactly until it's formalized. Add a service contract and training on the MG-1 for another $9.5k, leaving about $40k for additional equipment, room prep, general cushion.

I also need to add some temperature/humidity control, pay for electric service ($1k), purchase a HEPA filter unit ($500-$1500), air compressor ($1k - $2k), dry box (), and run a vent for the reflow oven. I had considered using a filter for the reflow exhaust but after consideration of the comments in this thread, that would be dumping 20+kW of heat into a 500 sq. ft. room. No thank you.

On the topic of reflow, I have 2 big questions:

1) Are the exhaust fumes from an oven considered hazardous? Does any special care need to be taken? Filtration, etc...
2) Will I get anyone in trouble if I use the oven for lead-free products and they sell them into markets that require RoHS/REACH compliance, considering I don't know if the oven has been used for lead? Are you allowed to call a process "lead-free" in an oven that's processed leaded solder? I know for wave solder machines you really can't change them over as you will never clean all of the lead, and lead-free is more corrosive on parts that were intended for lead. But for reflow the only thing that will touch the oven (ideally) is the flux vapor. The lead doesn't go anywhere but on the board.

[jmelson]

Buy a new Samsung SM451 for ~$70k.
Get a used Assembleon MG-1 with a few hundred assorted feeders as part of a package that also includes a Heller 1707 EXL Mark III, MPM Accuflex printer w/ 2D inspect. Total cost ~$100,000 shipped to my door and guaranteed working - although I don't know what "guaranteed" means exactly until it's formalized.

Well, if the machine and the service contract are from the same people, that should really cover all possibilities.  (Not sure if this is the case.)  I think you really have to compare consumables, accuracy, programming, reliability, etc. before making a decision.  Maybe you can find a Samsung manual online, even if for an older model, just to see how readable they are.

On the topic of reflow, I have 2 big questions:

1) Are the exhaust fumes from an oven considered hazardous? Does any special care need to be taken? Filtration, etc...

Yes, for sure.  I suspect you may be able to vent the fumes outside in most places, but you SURE want to get them out of there.  I have a little batch oven and just leave the fumes in the shop, but if it were a big conveyor oven, you'd get sick very soon.[qoute]
2) Will I get anyone in trouble if I use the oven for lead-free products and they sell them into markets that require RoHS/REACH compliance, considering I don't know if the oven has been used for lead?
[/quote]
I hear these scare stories all the time.  "If there is even ONE microgram of lead in your factory, the whole building will have to be knocked down, and all your stock and machinery will have to go in a hazmat dump!"  I think it is crazy.  Yes, you need to take some actions to prevent strict no-lead product from being contaminated by leaded production.  Any paste that goes through the oven will be melted, so no little paste specks to worry about.  Cleaning the printer, paste tools and stencils/frames might be the biggest challenge.  Contamination without your knowledge could happen real easily there.

Are you allowed to call a process "lead-free" in an oven that's processed leaded solder? I know for wave solder machines you really can't change them over as you will never clean all of the lead, and lead-free is more corrosive on parts that were intended for lead. But for reflow the only thing that will touch the oven (ideally) is the flux vapor. The lead doesn't go anywhere but on the board.

Yes, that is what I've been thinking.  Of course, one of the problems with lead is you can't SEE it.  So, if stuff IS getting contaminated, you would have a hard time telling, without one of the expensive x-ray backscatter devices.

Oh, one other thing, dealing with paste is pretty messy!  I notice your sketch didn't have a sink!  The people in the building may not want to share a bathroom with you when you come in covered in paste, even if it IS lead-free!

Jon

[Kjelt]

Make sure the oven can handle the non leaded reflow profile, temp has to be higher than with leaded paste.

[SVFeingold]

The oven is a Heller 1707EXL MKIII which has a max standard temp of 350C so it shouldn't be an issue.

[jmelson]

2 months from dead start to 1000 boards completed?  OK, now I know you are nuts!  Maybe an experienced guy could pull this off, setting up a new shop and cranking out boards.  Let's list the issues:

1.  getting machines in, powered, air supply, air conditioning MUCH greater than needed for office space, vent duct, etc.

2.  getting software to talk to P&P machine, format required files from CAD/CAM data, etc.

3.  Programming vision system for actual components you will use.  Some will come straight out of library, some may
               need special attention.

4.  Setting up stencil printer to get proper registration and print with the pastes and stencils you will be using.

5.  Dealing with errors on the P&P, fiddling with the feeders to get good pick-up, getting the vision system to be able to
                 detect the leads on all the parts without rejects.

6.  Getting the oven set up to reflow everything without tombstones, bridges or head in pillow defects.  With good,
     clean boards and high quality solder paste, this ought to be a non-problem.  But, if it IS a problem, it will drive you
     to distraction.  I've been down this road, and it is littered with carcasses.  I even spent $170 to import some SAC305
     from England on a recommendation.  it was GOOD paste, but just too much hassle and expense.  I'm still looking
     for the "best" paste, but the stuff I'm using now is just OK.

OK, so, assume that everything you need is in place the first day (Ha ha ha!) and that it all fires up and works properly.
You will still be fiddling with learning the machines, the software, setting up the paste printer, and the oven profile for quite some days.  I can't BELIEVE you can get there in less than 2 weeks.  So, that leaves 6 weeks for production.
Assuming 5-day weeks, that is 30 days.  Now, assuming your 9-up panels actually will fit in the P&P and the oven, 1000 boards is 111 panels.  Assuming each panel takes 20 minutes to place (I think you mentioned, but I can't find the estimated time) that gives you 24 panels per 8-hour day.  Well, it would only take about a week to run these.

Of course, after that week, you will need a 3 month vacation before even wanting to look at your machines again!
I can't IMAGINE the horror of running a batch like that!  Now, with all those automated SMEMA conveyors totally automating the process, it should seem like just sitting back and watching the robots do all your work, but PLEASE, talk to somebody who actually RUNS a line.  I'm sure my setup is total crap, but if you have even 1/10th of the feeder issues I have, with a multi-nozzle high-speed machine, I think you will be pretty busy fiddling with things.  (Remember, every time you open the guards to fiddle with a feeder, the machine stops.)

The 0805 resistors just go wit the rarest attention, but some parts just slowly drive me nuts.  Oh, also note that many machines have big problems with round-top parts like MELF diodes and domed LEDs.  So, I have had to find what works and what doesn't on my machine.  0603 flat-top LEDs are just too small on top, so I had to go up to 0805 LEDs.  Way too many mispicks on the 0603 LEDs, even after careful alignment of the feeders.

As for stencils, I have kept notes of what worked and what didn't.  For 0805 and up, nothing matters, but excess paste makes little beads around the pads after reflow.  As the IC lead pitch gets smaller, the aperture size needs to be reduced more an more to avoid bridges.  If you are going to do leadless, CSP and such, then the solder amount is insanely critical, or bridges will be all over the place.  The chip sucks down and spreads the paste.  I did that once, and it was AWFUL.  I ended up redesigning the board to not use that CSP part.

Jon