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

[jmelson]

I pull parts from the tail and add them back to the head since my feeders can't pick the last few inches of tape. That also takes time.

This is a nice feature of the Yamaha feeders, they are usually able to pick parts down to the last pocket with no trailer at all!  Very nice when using cut tapes.

If any process takes longer than zero seconds - I do not categorize it as no big deal.

Well, if you have 100 panels to do, and there's 9 parts/panel, and you have a cut tape of 1000 parts, the loading and fiddling with that feeder is going to be small potatoes.  if a feeder jams every 10 minutes, all day long, it is going to cost you WAY more time.  Correcting all those gremlins that cause the machine to stop is much more important.  Sometimes, it takes a while of observing the issue before you finally figure out WHY something is going wrong.  I have certain parts that just always cause trouble, the worst is the 0.1 uF 0805 caps.  For some crazy reason (maybe because I buy the cheapest parts as these are used a LOT for decoupling) these cause constant sticking of the tapes.  I don't know if it is the paper tape is thicker, the cover tape glue is too strong in spots or whatever.  But, I usually have 1 or two C-clamps putting additional pull on tape, and that makes it work at 97%.

A few other parts cause minor problems, and when I get them all feeding properly, the machine just hums.

Jon
Jon

[jmelson]

I had kind of assumed that if I have a sensitive part (say 24 hours) I could unbag it, run X boards, put the feeder/reel in the dry box, and then as long as it's in the dry box it still has whatever life was remaining at the time I put it into the box. Is that not the case?

I'm pretty sure that IS correct.

I.e. even if the part only spends 12 hours out before going in the dry box, I still need to bake it after it's been in the box for 1 month?

Well, if sitting around for a whole MONTH, it might makes sense to bake, just to be sure.  But, I think a lot of this MSL stuff is some kind of CYA action.  I have severely violated MSL requirements on chips, and never had a failure.
(Yes, I know, they claim there could be delayed damage...)

Jon

[jmelson]

Hey guys, I feel pretty dumb! About the air compressor anyway. The service techs told me that each machine needs about 12CFM of air @ 5+ bar. However the MG-8 pre-installation guide shows the required flow as 350 Nl/min. If you convert that using the liter/min to CFM conversion you indeed get about 12 CFM. However, paying attention to that "normal" part and accounting for pressure differences it's really closer to 2-2.5CFM @ 5.5 bar per machine. Which seems far more reasonable. So for the whole thing, including the stencil printer, 10CFM ought to more than cover it assuming I did my math right.

Looks like I don't need a 10HP compressor after all. Phew! I was starting to get really anxious about that. Thought I was crazy, my sand-blasting cabinet barely uses 12CFM...

Well, I have a 2 Hp, single-stage Quincy compressor and a very small tank.  The CSM84 is just about totally pneumatic.
All the nozzles are driven pneumatically, and the nozzle air cylinders drive the small feeders, and trip air valves to operate the 16mm and larger feeders.  The door locks are pneumatic.  The main stop, push in stops and push up table are all pneumatic, as well as the mechanical alignment unit.  But, the BIG air user is the vacuum generators.  My machine has a separate vacuum generator for each nozzle, and once they start, they run until the board is completed.  They use up about one CFM, so my 2 CFM compressor is cycling about 50%.  (I have a computer-controlled dual mode compressor, so it leaves the motor running and unloads the compressor as needed to maintain pressure.  When the board is done, the compressor will shut down.)

So, not too clear on what your machines use for vacuum.  If it has one big vacuum generator (wish mine did) it might makes sense to get a vacuum pump and remove the venturi units.  They make a lot of noise and use air very wastefully.
This is a fairly common retrofit on machines where this can be done.

Jon

[SVFeingold]

OK, how are you supposed to do this, then?  Take all the parts out of the tapes/trays?  RIGHT, NOBODY is ever going to do that!

Clearly you haven't been to Shenzhen. 

[SVFeingold]

Well, if sitting around for a whole MONTH, it might makes sense to bake, just to be sure.  But, I think a lot of this MSL stuff is some kind of CYA action.  I have severely violated MSL requirements on chips, and never had a failure.

Even a month at 0.5% RH? For this project I really don't have too much to worry about, if I get some project that involves a reel of unobtanium components that's worth $60k...at that point I'll exercise more caution and look into baking/vacuum bagging and whatnot.

For the vacuum, I was told by the service tech that they do use a venturi for vacuum. I'll have to talk to him and actually look at a machine to see how easy it is to replace with a vacuum pump/buffer tank.

[jmelson]

Nope, never been to Shenzhen!

Supposedly, putting chips in a low humidity environment is VERY slow to pull moisture out of the chips, without raising the temperature.  And, I'd love to see you get 0.5 % RH without a vacuum pump.  Dessicant just doesn't pull it that low, and if it does, it won't stay that low for any time without regenerating the dessicant.

As for the vacuum system on my CSM84, there are 3 different vacuum generators on a manifold with a stack of vacuum/pressure valves.  Yes, it certainly could be modified, but not just swapping a hose or two.

Jon

[SVFeingold]

I don't mean that the cabinet will regenerate the chips with low humidity (although this one does heat to 60C), only that it will last for several months without needing to be baked if kept at that humidity.

0.5-1.5 or thereabouts is the spec on the cabinet, it does have a desiccant that it automatically regenerates.

[Corporate666]

I didn't mean to imply that the moment you expose chips to humidity that you need to bake them - my understanding is the same as that of others above - you have a window of time, dictated by the MSL, that is cumulative from the time the original seal is broken until the components must be baked.  You can take and remove the components from a dry box multiple times during that time, so long as you don't exceed the total amount of time exposed to non-dry environment per the MSL.

I just didn't know what MSL the parts you were using were subject to - some of the more strict levels are quite short.  I manufacture my own in-house products, so I don't really need to stick to any sort of documented standards, just thought it may be an issue if you were offering contract manufacturing services.


RE: Air.  IMO, air requirements are like power requirements.  Something "needs" 10 CFM like it "needs" a 20A breaker.  The machine might pull 20A at startup or under the worst possible circumstances, but in real world usage, it won't really use that much.  Same with air.  Vacuum venturis are inefficient, but PnP nozzles are very tiny... 10CFM sounds like a real lot, even with pneumatic feeders.  If you have the luxury of getting the machines delivered and up and running with a cheap Home Depot compressor, you could figure out how much air they actually use before spending $$$ on a silent compressor. 

[janekm]

Ball-park local pricing in China in volume is 0.0015USD per pad.

At what sort of volume ?

I've never understood the reasoning for "per pad" pricing - it takes no more time to place an 8-pin resistor network than a single resistor

For volume, it's more that at smaller volume it's totally eaten up by the mininum project fee, at least that's what I've typically seen.

I think the per-pad pricing is a hold-over from the manual soldering days, though it somewhat represents inspection and rework costs, I suppose. But mostly it's just because everyone copies the pricing model from each other...

[Smallsmt]

I don't mean that the cabinet will regenerate the chips with low humidity (although this one does heat to 60C), only that it will last for several months without needing to be baked if kept at that humidity.

0.5-1.5 or thereabouts is the spec on the cabinet, it does have a desiccant that it automatically regenerates.

Could anyone explain which problems caused by this?
I store my parts in 40% humidity environment and never received any problems for soldering.
We use TQFP, SOT23-X, 0603 chips.
If humidity is very low you increase ESD problems.
 

[mikeselectricstuff]

Could anyone explain which problems caused by this?

Micro-cracking caused by rapid water boiling during reflow, also known as "popcorning"
http://anysilicon.com/msl-and-popcorn-effect/

[Kjelt]

Ok so I just found out I have one component with a msl level 4 (72 hours) so what should I do , baking is there somewhere a guide for that ?

[SVFeingold]

It's worth saying too that plenty of SOT and TQFP chips have an MSL of 1 (unlimited). I did an audit on the upcoming project and only one part had an MSL of 3, the rest were either unlimited or a year.

[jmelson]

Could anyone explain which problems caused by this?
I store my parts in 40% humidity environment and never received any problems for soldering.
We use TQFP, SOT23-X, 0603 chips.
If humidity is very low you increase ESD problems.

OK, the scare story is that moisture seeps into the IC packages, and then causes them to explode when exposed to reflow temperatures.  Well, that's pretty extreme!  But, one can imagine stresses caused by moisture escaping might crack bond wires and such inside the chip.  Supposedly, LEDs are really sensitive.

I have LEDs that have been sitting in my basement for years and have had ** ZERO ** defects on them.  I try to keep the FPGAs taped-up in their moisture barrier bags, but when I go to use them again, the humidity detector card shows some level of humidity.  Again, never had a problem.

Jon

[jmelson]

Ok so I just found out I have one component with a msl level 4 (72 hours) so what should I do , baking is there somewhere a guide for that ?

72 hours is 9  8-hour shifts.  Will you consume a complete reel or whatever it is supplied on in 9 shifts?  If so, just remove the feeder and store in dry box when you shut down.  If not, cut the tape in half or whatever and keep only enough out to use in 9 shifts.

I think a lot of this MSL stuff is crazy.  Maybe necessary for spacecraft or life critical gear, but just a bit overboard for ordinary electronics.

Jon

[jmelson]

Perhaps I've been overly pessimistic about the amount of hand holding the P&P needs.  I just did 12 boards with 166 parts each, so 1992 parts, in a bit over 2 hours.  The machine was all set up already, but I had to align the stencil on my hand stencil printer, and then turn it loose.  I had one reel that ran out, so I had to change the feeder.  I saw 3 mis-picks, but other than the reel change, the machine ran without stopping even ONE time!  That might not be a first, but it is the first time in quite a while where it just ran through the whole job without SOMETHING going wrong with a feeder or tape.

(I did have a "feedback error" when I first started up that was worrisome.  Might have been a software glitch, but that could indicate an encoder cable going out.  I powered down and back up, and it ran fine, so will just have to keep an eye on it.)

Jon

[Kjelt]

Its a rather expensive component: a BT module and I did the first small batch for my friend three weeks ago, it has been out of the bag that long, next week I will place the next batch, well just hope it will be ok.
BTW this BT module had its own pcb with gold plated connection pads on the bottom close to eachother. Stencilling did not work, too much paste made shorts. Terrible component to work with luckily my friend made testpoints on each pad so we could measure shorts under the modulepcb, all failed. So next week I will put the paste by hand with a syringe and make sure it is way less than with the stencil.

[jmelson]

Chip Quik paste seems to work great most of the time but it's been less consistent for me than Kester. 2)

Just ran another batch of lead-containing boards with a NEW batch of ChipQuik.  Wow, what a difference!  The soldering is beautiful, and the paste spreads like it is supposed to, REALLY smooth.  OK, so now I know I just have to throw the old stuff out every once in a while.  I am definitely going to try the GC10 for the lead-free, on the next batch.

Jon

[jmelson]

Its a rather expensive component: a BT module and I did the first small batch for my friend three weeks ago, it has been out of the bag that long, next week I will place the next batch, well just hope it will be ok.
BTW this BT module had its own pcb with gold plated connection pads on the bottom close to eachother. Stencilling did not work, too much paste made shorts. Terrible component to work with luckily my friend made testpoints on each pad so we could measure shorts under the modulepcb, all failed. So next week I will put the paste by hand with a syringe and make sure it is way less than with the stencil.

Yes, the trick with anything leadless is to make the paste aperture WAY less than the pad size.  When the leadless part pulls down against the board, it squeezes the solder out, causing shorts.  I think there are articles about how to figure the factor of aperture reduction.

Jon

[Corporate666]

Ok so I just found out I have one component with a msl level 4 (72 hours) so what should I do , baking is there somewhere a guide for that ?

72 hours is 9  8-hour shifts.  Will you consume a complete reel or whatever it is supplied on in 9 shifts?  If so, just remove the feeder and store in dry box when you shut down.  If not, cut the tape in half or whatever and keep only enough out to use in 9 shifts.

I think a lot of this MSL stuff is crazy.  Maybe necessary for spacecraft or life critical gear, but just a bit overboard for ordinary electronics.

Jon

I used to agree... however, I spent a long time and a lot of frustration chasing a problem with ATTINY chips in SOIC-8 packages.  We were getting anywhere from a 1% to 50% failure rate after the reflow process depending on how long the chips had been outside of the package (but they had always been used within 2-3 months and the MSL on those is level 2 = 1 year).  Atmel blamed it on popcorning due to improper storage, which was BS... but even after I sent them a couple of hundred failed chips for decapping, they said "yep, it's popcorning - not our fault". 

I've never had any issues with any other parts, ever.  And it was only ATTINY chips from one specific fab plant... so I always felt it was an Atmel problem, but I can't prove it.  But who knows.  I never got failures on anything in the field.  It was either dead after reflow or it worked long-term.  But since that experience I've been more cautious about moisture.  I don't have a "real" dry storage unit (although that one shown above looks nice) - but I do take care to store parts carefully and use sealed/dry-packed parts quickly after opening.  And I no longer use Atmel

[rx8pilot]

For what it's worth....I have violated MSL severely and not seen any failures in a few years.

[Kjelt]

Yes, the trick with anything leadless is to make the paste aperture WAY less than the pad size.  When the leadless part pulls down against the board, it squeezes the solder out, causing shorts.  I think there are articles about how to figure the factor of aperture reduction.
Jon

Thanks, I hoped that the pro pcb and stencil factory would have taken care of this, but I guess they just take the pad size as the stencil size, so I have to make two pcb layouts one for the pcb and one for the stencil, does not make much sense 

[mikeselectricstuff]

Yes, the trick with anything leadless is to make the paste aperture WAY less than the pad size.  When the leadless part pulls down against the board, it squeezes the solder out, causing shorts.  I think there are articles about how to figure the factor of aperture reduction.
Jon

Thanks, I hoped that the pro pcb and stencil factory would have taken care of this, but I guess they just take the pad size as the stencil size, so I have to make two pcb layouts one for the pcb and one for the stencil, does not make much sense 

You don't need 2 layouts - you need a pad-stack definition with the different pad sizes for the paste layer, so teh correct apertures appear in the gerber files. The stencil manufacturer doesn't know enough about your design to know how to shrink the pads.

[coppice]

Yes, the trick with anything leadless is to make the paste aperture WAY less than the pad size.  When the leadless part pulls down against the board, it squeezes the solder out, causing shorts.  I think there are articles about how to figure the factor of aperture reduction.
Jon

Thanks, I hoped that the pro pcb and stencil factory would have taken care of this, but I guess they just take the pad size as the stencil size, so I have to make two pcb layouts one for the pcb and one for the stencil, does not make much sense 

What would be professional about taking someone's stencil design and arbitrarily changing it? If a manufacturer doesn't exactly follow the CAD data I give them, I'll wouldn't use them again. Its good if they query apparent anomolies, but if they just make changes as a standard practice its a recipe for dismal failure. They have no unambiguous way to know what I want or need.

[Kjelt]

Ok thanks guys learned again something new today 
This was the first time I encountered such a module and something went not as expected during reflow. This was not my own but a friends project, I will have to dig into the datasheet myself to see if there is additional information to be found about the stencil pad format.