SMT assembly issues - continued

It was fun to meet everyone who attended the robotics meeting at pololu last night. Since I ageed to report the [predicted] “demise” or “crash and burn” of my project step by step, the following describes what happened today. Hopefully this information will be helpful someday to someone [trying to be] as stupid as me.

#1: I received a sample of the stick-on BGA alignment product that I mentioned at the meeting, which includes:
___ A: a tiny PCB with BGA pads (and nothing else)
___ B: a matching BGA component
___ C: the alignment stick-on

My first impression is: this will work quite reliably for attaching BGA parts to PCBs, whether the BGAs are soldered with a reflow-oven or a hot-air gun.

It is easy to attach the stick-on gizmo accurately and quickly over the BGA pads. And when you slide the BGA part over this gizmo you can very definitely feel the part fall into the holes — no question about it… the part is seated where it is supposed to be.

You guys should get a sample, because you can make these too (!!!much cheaper!!!), and because of that you can start to make gizmos with BGA components! That should open up a few new possibilities, perhaps. Click the various links under “learn more” on this page: … and order yourself a free sample and draw your own conclusions!

#2: I called to ask why their quote to assemble my PCBs is more than one week late. At the meeting, some folks seemed to think they want to avoid dealing with my BGAs and/or tiny-pitch QFNs. The reason they gave me for the delay was “a major surge in requests for quotes”. Who knows if that’s true, but at least I got to ask the question that seemed to worry many folks at the meeting - whether most or all the PCBs they assemble would have shorts, opens, overheated and degraded/trashed components, etc. In other words, there was worry that perhaps not even ONE of my 25 larger (5.80" square, 300 component) PCBs would be soldered/assembled correctly. So I asked the question, “Do you know what percentage of BGAs or QNFs you solder will have shorts, opens, flaky connections or damaged components?”. Her answer was, “we experience less than ONE improperly soldered BGA or QFN component per month” (for all customers combined).

Sure, maybe I can’t believe them, but that’s astronomically different from “most BGAs (or PCBs) will have soldering defects”, which seemed the consensus of the folks who had an opinion about this. Or maybe I should substitute “worry” for “consensus”. Bottom line: Now I certainly have this issue bracketed - 0% success to 99.99999% success. Surely the truth falls somewhere in that range!

Woops, I just received the quotes by email. Remember:

ice-eye == 2.80" square PCB with image sensor and ~50 capacitors.
ice-quad == 5.80" square PCB with FPGA, CPU, PHY (300 components total).

25 ice-eye == $2600 @ 10-days : $3000 @ 5-days (with stencils, programming, etc)
50 ice-eye == $3000 @ 10-days : $3200 @ 5-days (with stencils, programming, etc)
25 ice-quad = $5200 @ 10-days : $5600 @ 5-days (with stencils, programming, etc)
50 ice-quad = $7800 @ 10-days : $8200 @ 5-days (with stencils, programming, etc)

The stencils, programming and other “bogus charges” costs about $1000 for each PCB, and is included in the above prices.

Since I am willing to wait to save $$$, the bottom line is $3000 + $5200 = $8200.

For $8200 I can buy an entire full-bore zephertronics system like the “3LMZT-7-BG” system described on tiis page:, including nozzles and such. Then I have all this equipment to build and repair other stuff. Also, as Jan correctly mentioned, zyphertronics tends to overcharge for their equipment, and cheaper options also exist. The following links show a few alternatives from “Madell”, though clearly they are missing several items included in the zephertronics system.

$ 520:
$ 625:
$ 755:

The hot-air gun stand alone is here:

$ 300:

I mention this stand/press gizmo separately because I prefer other hot-air guns I’ve seen on, but I haven’t found another stand yet. I think $300 for a trivial little stand like this is preposterous, so I will keep looking. I can buy an entire benchtop drill-press for $300, build a bracket to hold the hot-air gun, and end up with something much sturdier and more flexible. Also note how the gizmo is designed so the hot-air gun cannot be positioned very far from the vertical post, so components more than about 4" from the edge of the PCB cannot be reached. Not smart!

The systems mentioned above are essentially “rework stations”, though obviously they can also assemble a new PCB from scratch — one component at a time.

The alternative to “one component at a time” rework systems like the above is some kind of “one PCB at a time” system. Essentially, a “one PCB at a time” system partly or fully implements (or fakes) a nominal production syste albeit on a smaller scale. The basic elements of this approach are:

#1: solder-paste stencil printer
#2: pick-and-place machine
#3: multi-zone reflow oven
#4: cleaning system
#5: inspection system (stereo microscope [and xray imaging])
#6: rework system (to fix defects == essentially the alternative system above)

At least, the above is what I have come to infer is a “production system”. Here are the various ways I have found to implement and/or fake each of the above.

#1: solder-paste stencil printer
___ A: solder-paste dispenser ($20 plunger to $800 with air-pressure)
___ B: plastic sheet stencils ($100 ~ $200 for both PCBs)
___ C: stainless-steel sheet stencils ($200 ~ $400 for both PCBs)
___ D: stainless-steel framed stencils ($300 ~ $600 for both PCBs)
___ E: manual stencil printer system ($500 ~ $2000)
___ F: automatic stencil printer system ($2500 ~ $15,000)

#2: pick and place machine
___ A: human hands + vacuum pick-up tool + stereo microscope ($700)
___ B: manual pick-and-place machine ($1500)
___ C: automatic pick-and-place machine ($50,000 to $250,000)

#3: multi-zone reflow oven
___ A: pre-heater - to heat bottom of PCB ($800 to $3000)
___ B: hot-air station/pencil/gun + nozzles ($300 to $1300)
___ C: crapoid 1-zone reflow oven ($1000~$1500)
___ D: quality 1-zone reflow oven ($4500~$5500)
___ E: multi-zone reflow oven ($8500~$25000)

#4: cleaning system
___ A: no-clean solder-paste
___ B: not sure
___ C: not sure

#5: inspection system
___ A: stereo microscope with dual boom stand ($500 ~ $2500)
___ B: xray system ($no_idea)

#6: rework system
___ A: zeph system + nozzles + templates ($9000 ~ $12000)
___ B: cheaper versions of the above ($1000 ~ $5000)

#1: At the moment, I’m VERY uncertain how important it is to buy a quality manual or automatic stencil printer system… versus simply buying the stainless-steel stencils with folded-up edges. Maybe I’ll try the “cheapskate” folded-edge stainless-steel stencils technique, then buy something more expensive if that’s problematic.

#2: I’m stuck with option A ($700) until the product generates enough earnings to buy a $50K ~ $100K automatic pick-and-place machine.

#3: Probably I’ll buy options A and B for rework, and possibly C too. I prefer the reflow oven solder 1000 joints all at once, unless I find I can assemble prototype PCBs quickly and reliably with options A and B.

#4: Until I learn more, option A is all I know.

#5: I’m definitely buying option A because it is so generally useful for so many purposes. I need to learn more about option B if more than a small percentage of PCBs have assembly problems.

#6: I’m inclined to go with a mix of option A and B, and maybe even build all prototype PCBs with only the “rework system” and no reflow-oven.

Thanks for the update. Did you look at the Beagle Board powerpoint presentation Eric mentioned? You can at least get some competing quotes from the assemblers they mention. I’m skeptical about the meaningfulness of the one BGA soldering problem a month claim, but I’m not saying there aren’t places that can do them well–that’s why my advice was to have someone else do it. But, it sounds like you’re gung-ho to do it yourself, and I’m definitely interested in hearing how it goes.

- Jan

I searched and searched and searched, and found lots and lots of pages about the Beagle Board. But I never could find any PowerPoint (or other) presentation that describes PCB assembly issues. Maybe I’m just unlucky, or incompetent, or dumb.

I am still going around in circles on this decision making process. I wish I could find 2 or 3+ other people willing to combine resources ($10K ~ $15K each) and buy a slow but precise-and-reliable pick-and-place machine. Personally, I’d be willing to cut corners on feeders to keep costs down. You know… have the machine place all of one component, then switch feeders, install all of those components, then rinse and repeat until everything is installed. Insane for production, but fine for startups, prototypes, hobbyists.

In not at all “gung ho” to assemble the PCBs myself. I’m simply very “gung-ho-NOT” to fork over $8200 and have no equipment to show for it.

Did you check out those stick-on BGA stencils? If you want, I can drop by and give you the sample they sent to me, so you guys can evaluate it. I found a couple places who sell thick (2-mil and 5-mil) polyimide/kapton tape with release liner. You guys could laser-cut pieces of that tape with your laser cutter to make unlimited quantities of stick-on masks like this for BGAs (and maybe QFNs). According to my calculation, the tape for each would cost about 5 cents! Plenty cheap even to make inexpensive hobbyist PCBs with BGAs! I’ll buy you the tape if you agree to try it. What do you say?

It’s the first Google result when I search for “beagle board BGA” (it’s a powerpoint file).

I looked at the BGA stencils a bit, and I think I got some samples of that or of a similar product several years ago. We’ve cut some stencils out of kapton tape for others before, but I’m not sure how that worked out for them. We’re not very interested in making these parts as standard products to have in stock, but we can cut some stencils for you if you bring in the tape.

- Jan

Wow, you sure know how to choose search terms better than me! Thanks.

Yeah, I figured you probably wouldn’t want to make products out of them, but maybe if I have good luck with soldering my BGAs given this technique, you might want to consider trying a BGA or two on your products someday (when a good non-BGA version isn’t available… or has too few i/o signals).

Thanks for the offer, I’ll order some tape. There seem to be 2 variants that look interesting. One is normal amber-gold-color polyimide tape with sticky (silicon or acrylic) and release liner, and the other is similar but coated with a high-temperature white layer so text can be printed on it, then stuck to PCBs that are then subjected to high temperatures (reflow-oven). I’ll probably end up with a bunch of extra, since the rolls are rather long (about 1300 square inches @ 1" wide).

I am ready to order some white polyimide label stock (2" wide x 36 yard roll) and some thick polyimide tape (2" wide x 36 yard roll), but want to make sure what I’m buying will work conveniently in your laser cutter.

The roll of white label stock seems most convenient because the label stock is backed by a release liner, but this material is only about 4 or 5 mils thick (1-mil frosty-white coating, 2-mil polyimide film, 2-mil adhesive). The release liner is another 2-mils or 3-mils thick but obviously is removed before apply to the PCB.

The roll of polyimide tape seems less convenient because it lacks release liner, but this material is thicker and likely easier to handle and position on the PCB (5-mil polyimide film plus 1.5-mil thick adhesive).

My question is this. What if any considerations or issues exist for either of these materials?

I would guess the labels are easier to laser-cut, because the release liner prevents the material from sticking to the base plate in your laser-cutter. BTW, what material IS the base-plate?

To laser-cut the 5-mil polyimide tape, I assume I need to find a place to buy a separate roll (or sheets) of release liner to stick the polyimide tape to before we place it into your laser-cutter. True or false?

Finally, what do you want in the gerber file? Should I create a separate gerber file for each package footprint? That works if your laser-cutter software lets you tell it to cut "an X by Y array of the pattern in the gerber file [with some fixed gap between them]. Or should I create a gerber file that has all the footrpints I need to make stick-on stencils for one PCB - which you then execute N times (where N is the number of PCBs I want to make)?

Or some other strategy entirely?

I’m not sure what other issues there will be; basically, we can stick it in the laser cutter and see how it goes. The laser cutter has a honeycomb grid, though we can also put in a flat plate with a few holes in it. For the tape, we would need to put it on some release liner first. I think we might have just used old printer label backing in the past.

We shouldn’t get too carried away with the details of the files yet. You can send us a gerber file with one or two representative parts, and we can see what the results are before worrying about how to make many of them.

- Jan

One more general question. I’m ordering all sorts of soldering equipment and supplies, and I realized I’m not sure whether I should buy “no clean” or “water soluable” solder paste? Also, should I clean the PCB after soldering, and if so, what kind of cleaner should I buy?

Also, maybe you can suggest the best place to buy solder. I intend to try lead-free solder paste first, then “back off” if I have problems. The following is the link to the web-page with the best prices I found so far. Is this brand (Shenmao) and type (SAC305) a reasonable choice as far as any of you know? Is $61 ~ $64 per 500g a reasonable price?

The link:

I should have thought of that (release liner from 8" x 11" sheets of laser-printer labels).

I found a place that agreed to make me a roll of 5-mil thick polyimide tape ON a release liner in exchange for reporting our experiences laser-cutting the material. I think it will be 6" wide and 33 yards long. Free - a good price for a change! I’m bleeding money lately, as you can imagine. Maybe I’ll wait and try that first, and only order the roll of label stock if this first attempt doesn’t work.

Do you laser-cut the periphery of each hole (then the centers fall out or get removed afterwards)? Or does the laser actually vaporize the entire area inside each hole? Do the laser-cuts go all the way through the release liner too? I don’t think it matters… I’m just curious how the process works.

I think the price for the solder paste is about what I would expect, though I still think you should go with leaded solder. If you have no problems with that, you can move on to the lead-free stuff, and if you have problems at that point, you’ll know that you have to focus on the process; if you try lead free first and things don’t work, it will be difficult to know what the problem is. We use no-clean flux since we don’t have cleaning equipment, but if you’re just making some prototypes, the water soluble stuff might work better. I would check with your parts to see what the manufacturers recommend. For instance, I don’t know if you have to take any extra precautions with the CCD and with leaving the kapton stencils under BGAs.

- Jan

Well, it has taken me awhile, but finally I have some polyimide sheets with release liner and some gerber files of the patterns I need laser cut. I’ll attach the gerber files to this message for your inspection. I’ll bring the sheets of polyimide tape to your place as soon as you agree the gerber file is okay for you to cut some samples from.

A couple notes. I attach the gerber files for the following layers to this message so you can see the relationship between pads, soldermask, pastemask and silkscreen. The layer that should guide the laser cutter is the soldermask layer (“top_mask.gbr”), which has 0.002" larger apertures than the BGA pads.

I wanted to minimize the quantity of polyimide tape your laser cutter has to burn away, so the patterns for the QFN packages are perhaps non-obvious. Rather than having huge (5mm and 10mm) square pads, I put 5mm and 10mm empty boxes (that is, 4 lines that form a square). We can simply remove these 5mm and 10mm cutouts from the surrounding tape by hand and save your laser cutter time, and more importantly, emit much less smoke from your laser cutter.

Let me know how they look, and when I should drop by with the polyimide tape. Unfortunately, they only sent me three sample sheets, each about 6" square, instead of the nice long roll they promised. But they said they’d send me a whole roll once we confirmed the process works. I hope they do.

attachment: 5 gerber files in file “”.
Except for these “stickers”, I now have all the tools, equipment and components I need. Finally, it’s time to assemble some PCBs — and find out how much more I bit off than I can chew. :frowning: or :astonished: or :slight_smile: … TBD. (4.29 KB)

I don’t understand how those squares are related to the QFN parts, but we can cut them. The rest of the files seem fine.

- Jan

The technique I’ve worked out for QFNs is somewhat different than BGAs. In both cases, the desire is to get the IC pads or balls located correctly on the PCB pads. With the BGAs, the best way to accomplish that is to place the sticker onto the PCB with the holes where the solder-balls belong, then place the BGA package. I assume I’ll feel the BGA balls drop down into the cutouts in the polyimide tape and thereby correctly align the part. I’ll just align and apply the sticker, spread solderpaste into the depressions, set the BGA package down, then solder with 180C hot-air bath beneath the PCB plus 260C hot-air pencil from above. The sticker cannot be removed from the PCB after the BGA is soldered in.

The QFN package requires a different approach. The QFN has no balls to drop into cutouts to align its contacts. However, all the contacts (except the thermal pad) are located around the periphery of the QFN package. So the alternative I chose was to create a sticker considerably larger than the QFN package with a square or rectangular cutout about 0.002" larger than the outside dimension of the QFN package. I will align and apply the sticker to the PCB while visually checking that the QFN pads on the PCB barely [almost] touch the inside boundary of the cutout. Then I’ll place tiny dots of solderpaste onto each QFN pad on the PCB. Then I’ll carefully align and place the QFN package into the cutout. I think I should feel the QFN drop or detent slightly when it drops into the cutout. Of course, this must be done carefully and gently, and maybe under the stereo microscope, because I don’t want to smear solderpaste on the bottom of the QFN between its contact pads.

I can envision three ways to attach the solderpaste to the QFN pads on the PCB:

1: place sticker on PCB, then manually dispense solderpaste dots on the PCB.
2: place sticker on PCB, then apply solderpaste to QFN pads on PCB with stencil.
3: apply solderpaste to QFN pads on PCB with stencil, then place sticker on PCB.

I prefer method #3 the most, but I’m not sure I can apply the sticker without smearing the solderpaste. My next choice would be method #2, except I doubt it will work very well because the sticker raises the stencil above the PCB by ~6-mils. What I don’t like about method #1 is the certainty that I can’t get exactly the same quantity of solderpaste on each pad. I’m not sure how much difference this will make, but maybe it matters.

Unlike the BGA stickers, after each QFN component is soldered into the PCB, the sticker will be peeled off of the PCB. This is necessary because some QFNs have components or jumper-pads very close to them.

I just thought up one more approach, but I rather doubt it will work. That would be to attach the QFN sticker, then insert a stainless-steel or kapton mini-stencil (that barely fits into the cutout) into the cutout. If this insert has openings where the QFN pads are, that would let me apply solderpaste to the pads, then remove the mini-stencil with a vacuum pick-up pencil. Somehow, I bet its pad-opening “prongs” would get bent too easily, maybe even if it is made of stainless-steel.

Method #3 above has one potential advantage that might make the process more reliable. Presumably the solderpaste applied with a stencil will be 0.004" to 0.005" thick. Since my stickers will be 0.006" ~ 0.007" thick, that should mean that I can place the QFN package over the cutout rotated 45-degrees from its final position, and the solderpaste would not touch the bottom of the QFN package (barely). Then I can rotate the QFN package until it falls into the cutout, and theoretically no solderpaste should touch anywhere on the QFN except the pads. If the distance is too small and the solderpaste does smear on the bottom of the QFN, I suppose we could lay down two layers of the tape in the laser cutter, and end up with QFN stickers 0.012" thick, which should definitely provide sufficient QFN to solderpaste clearance.

Unless you tell me this weekend is okay, I’ll wait until monday to bring the polyimide tape to you at pololu. - Max @ 702-270-6167.

I don’t think you’ll be able to feel the BGA falling into place if the cutouts are filled with solder paste. I’m also skeptical about this QFN approach, but as with your whole project, I’m curious to see how it turns out.

- Jan