Electronics > Manufacturing & Assembly

How to solder DFN/QFN/HLGA/etc

(1/6) > >>

I bought some PCBs from JLCPCB with ENIG finish and I'm trying to solder some of these PITA packages with tiny pads that are around the perimeter but under the package.  I got a stainless steel stencil from JLC.  I saw a technique on Youtube in which they mounted spare PCBs (with spray glue) so that the PCB to be soldered is held in by friction and the stencil is held flat over it.  I then taped the stencil in place, and used Chip Quik Bismuth low-temp lead free paste.  I did the best I could to place all the parts on the board with tweezers and then stuck the board in my Controleo 3 oven (I tried with the leaded profile first, and then tried again with the lead-free profile).  All the larger surface mount parts soldered fine, but only 1 of the 4 tiny pitch parts worked after being soldered down.  This is the second time I tried, I had similar problems the first time.

I was never really super happy with how the solder paste laid down.  In some cases, the paste seemed to be thicker/higher than it should be.  Also, I found that slits in the stencil were so narrow it was sometimes hard to get the paste through it.  In fact, before using it a second time, I found many of the smaller slits were plugged up, I tried cleaning it with soap and water and with scraping it, and it wouldn't come off.  I eventually stuck it in my ultrasonic cleaner and that got the stencil clean.

For those unfamiliar with Chip Quik, it actually is lower temperature than most leaded solder, so that's why I used the leaded profile.

Anyone got any suggestions?

I put a little leaded paste on the pads with a toothpick under a 5X microscope.  Then place the parts.  Hot air (Quick 861DW), about 305°C, very low flow rate, and they solder just fine.  I don't bother with stencils, the toothpick method is quick and easy - although it takes a little practice.  The paste is MG Chemicals Sn63/Pb37.

That said, if JLCPCB had the parts I need I'd probably just order boards assembled.

I've thought about using JLC's assembly service, but they have a very limited selection and their reel change fees really run up the price on small quantity jobs.  I'm going to try to go through PCBWay which will allow you to give them a parts kit.

Thanks for the tip about the solder and toothpick. I tried something similar with the tip of tweezers, but found it challenging to lay it down without getting too much on and still getting good coverage.  I then soldered it with my YouYue 858D+, but I'm pretty sure I ended up bridging pads.  If a toothpick and different solder paste help I'm all for it though.

Show picture of the applied paste without components?

I had a little experiment recently. I ordered two sets of identical boards from JLCPCB specifically to have a play with some tiny QFN parts. One set with ENIG, one set with standard leaded HASL. The packages were QFN16 logic gates (74HC595) in 2.5 x 3.5 mm packages. Used a stencil, hot air and three different 63/37 solder pastes, one with some very old paste with a little silver in, one with fresh chipquik straight 63/37 and one fresh mechanic 63/37 - all on ENIG. All went well first time. Of all the solder pastes the cheapo mechanic produced the best finish. I also tried mechanic on the HASL version of the board. Again, no problems, worked well first time.

The thing that sticks out as significantly different is that you're using an odd-ball low temperature paste. From what I've heard the thing that will get you every time with QFN is "too much paste" especially on centre pads. It has never given me problems, but I've heeded others advice to use fairly aggressive paste aperture reduction for body pads and a moderate shrink on the perimeter pads.

Full details of my experiment copied from a previous post -
As a reminder, this is the stupidly small package that I'm working with:

I have stock of three different solder pastes that I've tried out on the same part and board. The first is some old Sn63Pb36Ag2 with type 818 flux that I've had sitting in the fridge for a couple of years; I've lost its pedigree but it's a respectable brand that someone was selling on ebay in economical size syringes that they'd filled themselves with something that normally only comes in giant pots (Multicore, Kesler, something of that ilk). The second is a syringe of fresh Mechanic brand XGZ40 (a common cheap Chinese variety sourced via ebay). The last is a virgin syringe of Chipquik SMD291AX bought from RS. The latter two are straight 63/37 alloys. All are type 3 mesh.

The results:

Old Sn63Pb36Ag2MechanicChipquikWorked fine, a tad drossy, best tackOK, less tacky than the firstLeast tacky
All those results are on boards with an ENiG finish. They all self-aligned equally well, any apparent differences are down to camera angle.

So, my conclusion? The best feel, the easiest to use was the old solder but the results are a bit drossy - it's too old. The Mechanic and Chipquik produced comparable results but the Mechanic was a little bit easier to print and slightly tackier. The mechanic is also about £5 a syringe and the Chipquik £15. So I'm going to stick with the Mechanic for the time being and keep my eyes open for some quality high tack paste with some silver content at a reasonable price. All were good enough to get the job done.

For the sake of experiment I also ordered some HASL(leaded) finish boards at the same time. All the advice I've ever received as "Use ENiG for QFN packages" so I wasn't too hopeful for the results. I was wrong, the HASL boards were as easy to place these tiny QFN packages on and get finished results as good as the ENiG boards. No pics because they're not really helpful, you can't distinguish the results from the ENiG board done with the same paste. Now, this is a one-off, but the takeaway is that fine pitch will work on HASL - your mileage may vary. Why this matters is that the ENiG finish option alone costs significantly more than the boards most of the time from people the likes of JLCPCB (£12.25 extra for ENiG on a £1.44 100x100mm board).
The complete 25 x 50mm test board.

This one was soldered with the old slightly tired paste that was technically well out of date.


[0] Message Index

[#] Next page

There was an error while thanking
Go to full version