Fine pitch SMD soldering... again.

[paulca]

I made myself a breadboard toy partly as a practice board inserted into the queue for a bunch behind it.

It has a 74LVC16T245 in SOP40 package. 0.5mm pitch IIRC.

Again the issue is solder paste application.

I tried two techniques. 

"Pull" a thin line of paste along the pins.  This difficult when the paste just won't 'stick' to the board and just slides on it instead of extruding.

OR

Put a blob on and smear it around the pins.  Leaving only a thin minimal trace of solder.

The first bridged badly.  Right up behind the pins and all.  Took me 10 minutes, 5 applications of flux and 3 goes with the wick to right it.  Simply too much solder.

The second hot plate soldered very clean and neatly.  I still touched it up a bit by hand.

However.  When I wrote a quick 'test harness' on arduino the former (too much solder) passed.  The minimal solder version failed.  I ran out of time debugging, but it shorts and then oscilates on bank 2.  "Fingering" the pins changes the scope trace, so it's likely a poor conductive joint.  Given the behaviour and scope trace my finger points to the DIR pin for channel 2.  A floating or shorted input would have different behaviour when the DIR pin is swapped and this bug is uneffected by dir.  The test fails the moment it activates OE and "B side" is driving 0.  If the DIR pin was no functional that would cause a short and then an oscilation around it's start up voltage, fighting the PSU current limiter.

For large boards I order a stencil.

Is there anything I can do to get the paste to "stick" to the board and not slide around on it?  Would applying flux and letting it dry to tacky hold paste?

The actual soldering, at the moment is done with a hot plate.  (leaded 185C paste).   100*C preheat for a few minutes, then ramp as fast as it will go to 205C, timer over 180C for 2 minutes and then I slide it off the hot plate and let it cool on a soldering mat for a few minutes, then hit it with the fan to get it to handling temp.  Then I touch up with the iron under the scope.

[Psi]

"Pull" a thin line of paste along the pins.  This difficult when the paste just won't 'stick' to the board and just slides on it instead of extruding.

That happens sometimes with some brands of paste which seem to be less 'tacky' than ideal, or with paste where the flux has dried out a bit.
You can try and put some tacky flux on the pcb pads first, then apply your bead of paste and that can help get it to stick.
You can also mix some tacky flux in with your paste to make some extra tacky paste. It works but it's really messy and hard to put back into a syringe, so I really don't recommend doing that.
Ideally just find a brand of paste that works better for you.

Another thing that can cause the paste to not sick is a dispensing needle of the wrong size for what you're trying to do. Experiment with a few sizes.

When you go to use your syringe of paste, if it's been sitting for a while squeeze out a fair bit and throw it away, around the size of a pea.  Paste in the end of the syringe will dry out so you want to get to "the good paste".  (Like how crusty bits form on the end of the toothpaste lid).

When running a bead of paste down a side of pads for a TQFP footprint people always tend to put more paste on than necessary and that leads to bridges.  Also if bridges are an issue try starting the bead on the 2nd pad and finishing 1 pad from the end. For me that helps to prevent it as the start/end always seems to get too much paste otherwise.

Also check your solder paste has an appropriate ball size for what you're doing. (The T number,  T1-T5  (big to small))
Lower T numbers tend to have larger blobs of flux suspended in them because the balls are bigger so the gaps where the flux sits are also bigger and imho that makes them more sticky.
Personally i wouldn't go higher (smaller ball size) than T3 when using a syringe. But that's just been my experience.
Also, Chipquik brand paste quality can be random, sometimes its good sometimes it's bad. It's cheapest for a reason.

In most situations, it will be faster to drag solder a legged SMT part like TQFP/SOP than paste + reflow.  Apply flux to the PCB pads first, place the IC, use 0.3mm solder. Feed and drag it quickly before the flux evaporates and it will flow nicely and you will be able to avoid bridges or clear them by just going over that area again.

[paulca]

The hand drag solder has a bit of a conflict when using a stencil.  Unless I specifically edit the foot prints to modify the mask layer and produce a separate gerber for the stencil.

If I don't populate the ICs, their pads will be "heavily tinned" lets say and while that "could work out", I fear I would be fighting the chip to level it with the iron and would end up with the hot air gun to level it, when I might as well have just populated it on the hot plate.  Shrug.  Donno.

This is why I start projects with "mickey mouse" PCBs.  Bread board utilities, etc.  Especially having been out of practice a year now. 

I had prior been using "Mechanic" 67/30 and it does stick better.  You can do "touch a blob" on passive pads pretty accurately.  I switched to QuickChip because I figured a known US brand would be better, but... not so sure now.  When hand fixing it, it did me think, "It's like snotty plastersene" and then blamed my own ham fisted soldering iron... but I did check the heat setting multiple times and it was still at 350, holding 330-345 on it's own sensor.

For flux I have a "Circuit works" no clean pen.  You can "paint" flux on, or for ICs, I tend to push the market nib back in and that allows the flux to flood down it and create a nice puddle.

Oh... I forgot to mention.  The failed PCB passed tests after a simple drag with the iron and some flux.  Including when proding and poking the IC during the tests.

And there is a tip for beginners behind me.  When you solder something you are not 100% comfortable with you must "PROVE" it works.  I use an ardruino mega for these purposes.  Just rubbish copy pasta Arduino framework code.  Not timing testing, just running through all possibly permatations and testing the outputs match expected. 

[Psi]

I prefer tacky/gel flux over any liquid flux from a pen

[stretchyman]

Just solder all the pins by hand with plenty of solder and wick off the excess using a liquid flux and a good solder braid + a Metcal iron with a wide bit.

Easy.

[Andree Henkel]

"Pull" a thin line of paste along the pins.  This difficult when the paste just won't 'stick' to the board and just slides on it instead of extruding.
...
The first bridged badly.  Right up behind the pins and all.  Took me 10 minutes, 5 applications of flux and 3 goes with the wick to right it.  Simply too much solder.
...
The second hot plate soldered very clean and neatly.  I still touched it up a bit by hand.
...
However.  When I wrote a quick 'test harness' on arduino the former (too much solder) passed.  The minimal solder version failed.

is it a package with leads or without?

first approach - "to much solder" and removing the bridges is in my point of view the way to go.
Missing solder joints are much worse to be located than solder bridges.
with lead packages you can take very fine tweezer or fine pick:
https://www.amazon.de/Sintron-L%C3%B6tbesteck-L%C3%B6twerkzeug-L%C3%B6tarbeiten-RW-LB6/dp/B0038L6GZQ
and check if pin is really soldered by trying to move it slightly, with no lead you cannot even do that, good luck with dmm continuity check
microscope view is sufficient to detect all bridges, but not really in case of little solder to determine if there is a good solder junction

[paulca]

Just a comment....  I remembered what made this particular little board extra difficult and extra frustrating...

It's tiny.  It's basically DIP24 package sized for a breadboard insert.  You run into "not enough fingers" left for jobs like solder wicking excess.

When the board is so small touching it with the iron move it.  So you need to add at least one finger to hold it in place... but as it is so small, one finger is not enough to stop it rotating when you touch it.  All while keeping the bit you are working on within a dime sized focal area of the microscope..

Something to mount board in would help, I think I spotted little PCB holders on AliExpress for tiny boards.  It would still need a clamp to lock it to the microscope base though.  Something I need to think about.  I am not enduring soldering another "mini PCB" that isn't as a minimal glued down to something!

[Andree Henkel]

hm can you use like double sided adhesive tape to keep it in place?
https://www.tesa.com/de-de/buero-und-zuhause/tesa-doppelband.html
you can cut it with scalpel to the actual needed size

[paulca]

Thanks.  I ordered a set of those picks/tools. While I was there I found a two slide jigs, one tiny, one larger and some 20mm clamps. 

The other thing, fairly unrelated, I need to remind myself to "lift the work to yourself" and not "hunch over it".  I had a bad lumber pain from hunching over the board.

EDIT:  At one point I thought it would be better if I turned the microscope base 90* and worked it that way.  Then I realised I had to do the 90* transform in my head while looking at a screen 45* to my right of where the PCB is.  My brain "seg faulted". "Nope!".  First attempt to move the iron "left 2mm", resulted in it moving up "2mm".  I can fly an RC helicopter "nose in" to me, but I can't do that.

Turns out....  all I needed to do was rotate the camera in the scope 90* also.

[Andree Henkel]

make bigger PCB next time (like if you have a panel) and assemble just one sub-PCB, separate the single pcb after soldering

there are special small unplated drills for that called "mouse bites"
https://jlcpcb.com/de/blog/mouse-bite-panelization-guide

[nctnico]

IMHO using solder paste is not a good route as applying it properly by hand is difficult. You'll at least need a dispensing system and good paste. And then you'll also need to control the viscosity of the paste (which probably also goes hand in hand with temperature and aging).

What works well for me on fine pitch parts with leads is flux (a good rosin flux) and a big flat tip. That way I solder 3 to 4 pins at the same time. For QFN I use a tip with (sharp) 90 degree edges which can get into any corner. Drag soldering QFN with such a tip works very well. But it takes tips which wet very well (like the ones from Ersa) set at the right temperature. If a tip has bad spots which no longer wet, all bets are off. Some desoldering braid (chemtronix only) can take care of excess solder. The key to all this is to let the flux do it's magic.

[pcprogrammer]

I'm refining my technique for fine pitch parts. For instance for FPC 0.5mm I now use normal solder wire 0.5mm (60/40 or 63/37) and pre tin the pads while using a lot of flux to get it evenly distributed. I then use hot air at 320 degrees and watch the process under my camera microscope. It shows perfectly when the tin starts to melt and the pins drop into it.

With drag soldering with either a horse shoe tip or a chisel tip the solder tends to ride up on the pins and wicking with braid is needed. With the hot air this does not happen.

In the beginning I too tried with paste, but without a stencil it is almost impossible to apply it in the right amount, so I stopped doing that.

About debugging a board, sometimes it is very hard to find the why for instance a clock is not running. On my first rotary encoder module the STM32F103 HSE clock did not start. At first I suspected the decoupling caps since I changed from through hole crystal to a SMD 3225 crystal, but reducing the values did not help. Replaced the crystal with a through hole version, no luck either. The SWD connection worked though and programming the Flash was no problem.

It was until I probed the reset line with a scope, that I noticed it was not going down to ground when the button was pressed. It turned out to be the tiniest amount of solder between the reset pin and the clock pin. Very hard to spot even under high magnification.

Edit: added that the SWD connection worked.

[paulca]

IMHO using solder paste is not a good route as applying it properly by hand is difficult. You'll at least need a dispensing system and good paste. And then you'll also need to control the viscosity of the paste (which probably also goes hand in hand with temperature and aging).

What works well for me on fine pitch parts with leads is flux (a good rosin flux) and a big flat tip. That way I solder 3 to 4 pins at the same time. For QFN I use a tip with (sharp) 90 degree edges which can get into any corner. Drag soldering QFN with such a tip works very well. But it takes tips which wet very well (like the ones from Ersa) set at the right temperature. If a tip has bad spots which no longer wet, all bets are off. Some desoldering braid (chemtronix only) can take care of excess solder. The key to all this is to let the flux do it's magic.

Few things to pick up on there. The iron hadn't be used in quite a while.  Even then maybe a 2 minutes job and some headers.  So i did retin it.  Clean it.  Retin it.  But in my head I was saying, "This should be done while it's cooling, when you put it away, not before you use it!"

In reflection.  The "moves like snot", "won't wet right", all got easier as the session went.

When I did the rework bit the next day the iron was back to it's usualy self.  "Fizzle, shiny".

I use a 2mm 45* "hoof" tip.  As long as you are "level" with that 45* on the board, you get the full 2mm diameter flat face. It has sharp edges to the vertical and so you can "register" it against the edge of the pins and drag it.  Gently or you bend pins mind.

[paulca]

Just soldered up a board with a TQFN48 IC and  I think the "Mechanic" Aliexpress paste wins for ease of use.  Especially if you let it sit on the hot plate at 40C for a while. 

The TQFN didn't self centre because I just didn't have enough solder on the pins.  I was deliberately shy because I knew I was going to touch it up anyway and my suspicious was correct.  "Too little solder = not enough surface tension to pull the chip straight".

So I had to hot air reflow it to nudge one side. a pin over.

This photo is of a monitor, nothing is aligned straight with each other and I'm sure the chip is less crucid than it looks here!

[Psi]

Yes about not enough solder on those end pads.

But you might want to try some flux, even if it's just a thin layer over the pads before you place the chip/paste.