Surface mount prototype work - help!

[Gizmo]

I've started using surface mount components - up 'till now I've stuck to though hole - and would like some tips on how I can make boards which work first time.

I've made a couple of prototype boards, both of which have failed to work when assembled, resulting in overheating chips. I assume this is a short between vcc and an output pin?

The first board I ended up having to remove the chips one by one until I could communicate with the pic chip on the board then re-add them all again one at a time until it worked. The second board I ended up frying the pic chip after trying to readd an overheating chip. So generally my experience hasn't been good!

I've been using a solder stencil cut using a craft cutting machine from acetate, which appears to be pretty good. Although placing the chips with tweezers I will happily screw up the nice little solder islands on the pads.

I have half finished a manual pick and place machine which will help greatly with accuracy of part placement, and I'm thinking about a diy reflow oven as well. At the moment I use a reflow station to reflow all the chips.

Any tips on how I can actually make boards which work would be really appreciated! I would eventually like to make a small batch of these boards, so being able to confidently populate the board would be extremely handy. So far constructing boards has been a nightmare.

[Charles Creations]

My first suggestion is with your component package selection. I made my first boards with 0603 or larger resistor and capacitor packages and kept ICs to soic packages (0.050" pin pitch). The 0.050" pin pitch for ICs makes eliminating any solder bridging very simple with a solder wick and soldering iron. I manually applied solder paste with a syringe and used a 858D+ reflow station. This setup gave me success on my first few boards.

What size are your current packages?
How dense is your board?
What is the quality of your proto boards? Are shorts resulting from poor etching?
Does the board have soldermask?

While debugging a board, I would also test for resistance between your supply and ground rail before your actually apply power. If this is a dead short, then I would start looking around and trying to fix it because applying power and wrecking something else adds even more debugging work later on. Just go through with a multimeter and make sure things seem right. Also, using a current limited supply can help make problems evident without doing as much damage.

[nctnico]

I'm not a fan of solderpaste and hot air. I usually use some flux and a soldering iron with a big (-ish) tip. If the tip is large enough you can put enough heat into a joint to solder it quickly. That way you hardly get any bridging.

[sleemanj]

I've made a couple of prototype boards, both of which have failed to work when assembled, resulting in overheating chips. I assume this is a short between vcc and an output pin?

Are sure it's not a fault in your design to begin with?  Is it a proven schematic which you have converted to an SMD pcb (and verified against the schematic), or is it a new design?

If it's a home-etched board (especially), did you check for hairline shorts (or breaks) before you soldered it up, always a good idea to spend a few minutes with a multimeter and check for continuity and discontinuity as appropriate.  I find it's best to do this on top of a light table the strong backlighting shows up hairline shorts and fractures better.

If you can't see anything obviously shorted on the PCB, start looking for the hidden shorts, namely, those that are under components.  Look at your design, is there any trace close to an SMD pad running under the component that might have bridged, check with your multimeter.

If you're just doing a one-off, or a first prototype, it may be better for you to hand-solder it rather than reflow, then you can assemble your circuit in stages testing as you go, ie, start with assembling the power systems, check they work, move on to the inputs, outputs, micro...

[mjkuwp]

in the USA:

I do:

1. make the pcb at oshpark.com so that I have good quality finish and soldermask.  Makes soldering a relative breeze.
2. Stencils from ohararp - although for many designs these are not necessary.  Otherwise I dispense solder from a syringe through a needle to the pads
3. Reflow toaster oven.  mine is PID controlled but you can do this with manual on/off if you have temperature feedback.

paste solder for stencils is no-clean lead solder, chipquik from Digikey
syringe solder is from zeph.com - zephyrtronics

The only mistakes I have had were due to using a copper pour and having too small of clearance around it.  I ended up with shorts in so many places that I gave up on the board and started over.  After that I started putting solder jumper pads between various parts of the circuit and I test things individually before connecting the solder jumper.  Of course - since I started doing that I have not had problems to troubleshoot.

I use 0805 for the most part and the toughest part I have done is a 0.5mm pitch quad flat no-lead package.  I think I can switch to 0603 soon.

[T3sl4co1l]

For ISP (in system programming), either make sure the programming pins are accessible (not being driven by other logic), and provided with a socket or header or something like that, or, add small resistors (usually 100 ohms to 1k) so that the programmer can overdrive the MCU end of the signals.

If you're building for automated test, you can install vias or pads on useful nets (hopefully all of them!) so that components can be tested and programmed.  You would have a bed-of-nails fixture made, or make one yourself (pogo pins are available from Digikey) so you can clamp it on and have instant access to everything.  Which for just programming, might be little more than a fanout to a header, but you can go wherever you want from there.

Tim

[rx8pilot]

Over the past few months I have manually placed many 10's of thousands of SMT components in my quest to build up to a complete line. For me, manually placing the components has been a fantastic way to learn a LOT about the process which is generally unforgiving. I have been successful with 0402's and .4mm QFN's, so it's obviously possible. With lots of practice, I have been able to create production quality PCB's with better than 99% success. This post needs to stay short, because solder paste is in the printer as I write this. I hope I can post some pics later.

The most critical tools for me (not in order of importance)
Real vacuum pick-up tool with a variety of tips. The angled tips are important. Tweezers are MUCH slower. It takes a number of picks to get the hang of it, but I not even bother making boards if my vacuum broke. If you want to make batches, this is critical. I built a manual machine and never use it. I get more components on the PCB free hand. I have a 5 axis CNC shop and built a manual machine that was very precise. It was just too clumsy overall. Some of the well thought out manual machines that are commercially available look useful, but not anything that is in the DIY category that I have seen.

Stereo zoom  microscope with wide angle lens on a boom. My eyes are very good, but the microscope is a huge reason I can get such great yields. Before I got a microscope, I was using various magnifiers to check placements after it was all said and done. The microscope is perfectly perpendicular to the PCB and you can see exactly what is happening as it's happening without any eye strain at all. This is another tool that would totally stop me if it broke.

Solder paste is the way to go hands down. I am a very accomplished and fast solder iron slinger, but SMT is a solder paste world. Flux and solder is messy, slow, and totally unpredictable. Ceramic caps do not do well with the sudden heat and fast cooling. For consistent and reliable PCB's solder paste and some sort of re-flow is very important. I have done tons of research and have tons of practical experience. I have been building amateur PCB's since the mid-80's, professionally about 7 years ago. In the middle, I repaired PCB's on high value imaging equipment.

Stencil printer. depends on how many and how complicated your PCB's are. I have done the prototype stencils with pretty decent success, just not with anything fine pitch. I have a pneumatic dispenser that is pretty fantastic unless you have fine pitch. I did a bunch of prototype PCB's with the dispenser placing .4mm QFN's which always bridged but I was able to easily repair with an iron using a flat tip. Great for getting through a few revisions of PCB's that will never have a stencil. I eventually got a cheap-ish printer that can hold stencil frames. I jumped for joy the first time I used it. I have not had a solder bridge since. No re-work, all PCB's work. It's awesome.

Component organization is super important as well. The time I have spent creating an organized way of getting components has really paid off. I have some pics attached. I use trays with little bins for loose passives, diodes, etc. I keep sensitive components in a labeled tape dispenser retrieve them one at a time. The PCB itself sits on a small aluminum plate that slides on a bigger plate beneath. This allows me to move the PCB under the microscope very fast. The scope and my head stays still the whole time. I numbered and color coded the parts bins, the BOM, and the paper print of the PCB so I always know what is next.


Re-Flow.For super low-cost DIY, the hot air re-work stations are great albeit a little slow. I take the tip off and use high air temp from far away to slowly heat up the board evenly. Slow spiral down until you get flow, then slowly back off to cool down. It's very easy to get the flow without any danger of over heating anything. The bonus is that all the crooked components will align themselves when the solder is melted making a good looking PCB as well.  I skipped the toaster oven completely , I don't believe that reliable PCB's can come out of them regardless of the DIY controller. IMO, too crude. I eventually got a convection batch oven JEM-310 that is  great but maybe a bit expensive for DIY. If I had to go super low budget, I would use my hot air re-work station for all the boards.

[tautech]

@ rx8pilot
Thanks for your input, nothing like hearing all this from an experienced expert.
Good stuff. 
Feel free to add more of your wisdom for others to appreciate too.

[T3sl4co1l]

Cool!

Question: is that 99% success per board, or per pad?  (I would hope not per pad.. )

I've seen results from other "single person" assemblers, which came out under 1 in 1000 failed pads, and those were mostly on 300 pin BGAs anyway.

And is that verified by functional test, or do any of the boards you've done see component level testing?  Be interesting to see quality metrics for assembly techniques over time (full on pro automated systems, manual PnP, manual paste, hand soldered, etc...).

Tim

[calexanian]

I do all of our short run SMA work at our shop. Try to use 0805 passives or larger if you can. The tack method works well if you don't want to use paste. Use the thinnest solder possible and for the love of god use nice fluxy leaded solder. flow one side of the passive, or a corner pin of an IC. Once you like the registration do the other side, or drag solder the IC. There are lots of videos out there about how to do that. As for paste and stencil. I use Stencils unlimited. They make good stainless ones for proto use, or cheap polymide ones. They both work well. Again use leaded solder. Its so much easier for hand use. No clean is best. lay out all you parts to place in little piles on the bench and place them one at a time. Its ok if the registration is off a little. The capillary action will center in when you flow the board. Use a hot air station with a low flow but good heat to prevent tombstoning, and have a little stick or probe to nudge stubborn parts back onto their pads while flowing.

[rx8pilot]

Cool!

Question: is that 99% success per board, or per pad?  (I would hope not per pad.. )

I've seen results from other "single person" assemblers, which came out under 1 in 1000 failed pads, and those were mostly on 300 pin BGAs anyway.

And is that verified by functional test, or do any of the boards you've done see component level testing?  Be interesting to see quality metrics for assembly techniques over time (full on pro automated systems, manual PnP, manual paste, hand soldered, etc...).

Tim

I have been measuring how many good boards I get that do not need any re-work. Out of about 100 PCB's, I had 1-2 pads that needed a fix. Each board has 50-100 components that range from 0805 to QFN, but no BGA. That is many thousands of pins between failures. I never did any real metrics because I am really trying to get a proper P&P line. In bursts, on a PCB that I have memorized, I can do about 250 CPH. The problem is that rate is not sustainable. If I measured my CPH over the entire day, it would be closer to 75-100.  I would consider myself an average skill assembler. I do not do it every day, but maybe a few bursts per week. I have, however, focused very hard on the process and the various sciences behind the processes. When I hire someone to take over, they will get very thorough training.

In the end, all of my boards are used in high-reliability field environments. They see a lot of temperature range, vibration, and other bad things. The best inspection they get is a microscope with human eyes on the other end. No automated tests.

As others said earlier, the PCB quality and the pad designs are CRITICAL to assembly in general, but even more so for manual. My early boards felt impossible to assemble until I optimized the pads. On things like QFN's, the paste deposition on the center is very important, too much and the package will float too high for the perimeter pins to make contact. I may estimate that 50% of my success is due to having good PCB and design. Going through the learning curve of manual assembly has made me a much better PCB designer. At this point, anyone can assemble my boards.

[Gizmo]

Thanks for all the responses, it's given me a good idea of where to go moving forwards, and that's incremental building of the boards for prototype then try and increase the accuracy of my chip placement with a manual pick n place. My main issue being my hands are not steady enough to manage the accuracy of placement even with a vacuum tool. Thanks again, all very useful and makes me think. 

Charles Creations
Package Selection
I'm limited to board size, so am unable to use 1.27mm pitch packages. I've been using 0.65mm packages as these are the largest pitch I can physically fit on the board. These are also the ones I am having issues with. Both memory and PIC IC's are 0.8mm AFAIK, and I don't believe I have had problems with either of these. All caps and resistors are 0603 and I havent had any issues with them either. So I dont believe I can do much here.

Board Denisty
Pretty dense -- it's not caused an issue that I can tell, though. Placing the parts doesn't cause an issue.

Board Quality
I'm very happy actually -- I'm using 8mil tracks with 8mil space and 20/12mil vias, so reasonably small. It's being made by TinySine, so one of the generic Chinese Seeed Studio type places.

You're absolutely right I need to do more checking before powering up than visual checks, I need to get some small probes for my meter as the current ones are just too large to be able to probe anything at this pitch. Although my eventual goal is to make quite a few of these, so figuring out how not to require so much testing would be good! Although I may look into outsourcing production then... I have no experience of this though. The Chinese PCB fabs usually offer a complete board service, but I have no idea how good this would be -- they are pretty cheap. And I've been on the lookout for a reasonable lab supply with current limit -- my hacked PC PSU isnt ideal.

nctnico
Soldering
I have tried soldering small pitch soics. Nightmare. I will never try that again.  On this board it's also too dense to do this.

sleemanj
Board Fault
Yeah, I'm sure the board is fine. This second prototype is an iteration on the first -- the first worked after I'd sorted out the initial chip cooking problems (apart from the actual flaws in the design which lead to the 2nd board!). No home etch, all professionally done.

I am going to take your approach of incremental construction for my next attempt though. I have no problem individually reflowing components, as this is what I ended up doing for the first board in the end more or less. I built it up fully, then had to more or less completely disassemble and re-assemble it!

But I keep coming back in my head to the fact I want to produce a few of these... perhaps my manual pick and place will sort out accuracy issues and make this possible. We'll see...

mjkuwp
Copper Pour Shorts
I'm with you there. I've done a board before where I had that exact issue -- solder mask exposing the edges of the pour and causing shorts. Nasty. I've avoided that this time by having no top pour and routing power and ground on the bottom. It's only a slow board, so broken ground planes are not much of a worry.

T3sl4co1l
ICSP
I've setup ICSP as a 6 pin header. For proto-work it's a soldered header, for production programming it will be more like a pogo setup. MCLR is hooked up appropriately with a resistor and the previous board worked fine once the bugs were ironed out.

Although one thing which I've realised which may be an issue while in the car this morning is floating outputs.

The PIC's normal state will be high impedance on all pins before programming, I would expect (I haven't verified, I will do later!). Several of the chips are capable of driving the same output lines, however their CE/OE will be floating. So while in an unprogrammed state I assume I may / will have multiple bus drivers. I'm not sure if this would explain my chip overheating issue? It's only been one chip overheating both times so far... would two chips driving the bus potentially cause this? Or is only a short to Vcc going to cause this? I wouldnt think these chips could source enough current to cause them to cook on two chips driving the same lines.

But certainly one thing to consider is pulling the CE's high on all chips with resistors if using ICSP. Previously I've programmed out of circuit and dropped the chip in, so ICSP on a fresh board is a bit new to me.

rx8pilot
Vacuum Tool
I've actually made a vacuum pickup tool out of an aquarium pump with a foot pedal to operate. This has been part of my pick and place plan -- I agree tweezers are not great from dropping chips precisely in place. Plus my hands are not too steady (which was the whole reason behind the pick and place). )

Your post has really made me come to the conclusion accurate placement without messing up the solder islands is the way to go for getting thins working properly. As much as I like to think the solder will flow away from the mask and toward pads, with the tiny gaps you get between legs, solder bridges are inevitable if solder is between the legs given the hundreds of legs on the board.

[sleemanj]

Thanks for all the responses...

So, it sounds to me like you are mainly having problems of bridged IC pins, your passives etc are OK?

First thought is maybe your stencil alignment when you are using it isn't as good as it needs to be for the fine-pitch components.

Second thought is to the solder paste - has it been stored properly, is it too old.

Third thought is maybe the stencil itself is to blame, apertures a bit large on those components perhaps.

[Tandy]

Solder mask is the key thing with SMT. You are putting blobs of wet solder on pads that will spread out on your board. Without solder mask the blobs will easily merge with each other causing a bridge. Home etched boards without solder mask are not well suited to SMT work unless you go for largest sized components and perhaps make the apertures on your stencil just a little smaller than the pad size to reduce solder spreading outside the footprint.

[Gizmo]

My main issue is, it doesn't matter how accurate the solder mask application is (and it's pretty good), as soon as I start placing components and aligning them, it smooshes all the solder paste around to the degree I may as well just have applied a line of solder paste across the pads. So my biggest enemy is accurate part placement on the first drop!

[Tandy]

It usually shouldn't matter as the surface tension of the solder paste when heated should draw the solder on to the pad.

[Gizmo]

Usually true -- although this is the only possible source of solder bridges. You cant get a bridge if there is no solder in-between pins. Even more so with smaller pitch IC's, as the gaps between pads / legs are just tiny!

[Precipice]

This sounds as if there's too much solder paste, maybe?
How thick a stencil are you using, and how much are you shrinking the holes?

[nctnico]

Usually true -- although this is the only possible source of solder bridges. You cant get a bridge if there is no solder in-between pins. Even more so with smaller pitch IC's, as the gaps between pads / legs are just tiny!

For prototypes I think you should consider switching to a soldering iron. See how easy it can be to solder a TQFP:

[22swg]

This is a re post , of my experience with QTFP 0.5mm pin devices. http://digitaldiy.io/forum/general-electronics/2396-a-diy-pic24h-dev-board?limitstart=0

[Gizmo]

I've tried it, I make a COMPLETE mess of it. Not a chance. 

I'll sort out my paste and reflow issues.

[Yago]

"Yeah, I'm sure the board is fine"
Assumptions are bad Ju-Ju, never find your way out of a maze on assumptions.
IMVHO.

[Kjelt]

I've tried it, I make a COMPLETE mess of it. Not a chance. 
I'll sort out my paste and reflow issues.

I tried that multiple times also on TSSOPs with limited success mostly a few bridges left.
What worked for me before I got a reflow oven was to apply the paste, put the chip on the pads and solder with a smt micro pencil solderiron on 270 degrees. Just one pad, check the alignment, skip a few pads another pad etc. till you have the whole chip. Be sure to skip pads and take your time for cooling of the chip inside. The paste has soo much flux that usually will evaporate during the reflow process but manually soldering it will stick on the pcb, so also rinse the pcb with flux remover and IPA afterwards.

[MK]

A tube of pure flux on hand for any reflow of chips is a must, just squirt some across the rows of pins on the chip and run the iron along, helps to pull bridges of the gaps, but you then need to use much more ipa to clean the boards afterwards.

[T3sl4co1l]

The PIC's normal state will be high impedance on all pins before programming, I would expect (I haven't verified, I will do later!). Several of the chips are capable of driving the same output lines, however their CE/OE will be floating. So while in an unprogrammed state I assume I may / will have multiple bus drivers. I'm not sure if this would explain my chip overheating issue? It's only been one chip overheating both times so far... would two chips driving the bus potentially cause this? Or is only a short to Vcc going to cause this? I wouldnt think these chips could source enough current to cause them to cook on two chips driving the same lines.

Yikes!  Definitely add weak pull-ups (10-100k?) so all enables reach a defined state!

It's always good practice to assign pin states, even with a weak pull up/down, because a floating CMOS input will permit erratic behavior and unusual current consumption.  When the pin voltage is in the transition region between V_IH(min) and V_IL(max), the P/N input transistors are both conducting, and drawing excess supply current; the voltage can drift into this range due to input protection diode leakage, or ambient fields (electrostatic, or RFI, including merely switching nearby pins).  Unused MCU pins, if not pulled with physical resistors, should at least be assigned either as weak pull-ups or active outputs in software.

Tim