Author Topic: QFN float during flux activation  (Read 4815 times)

0 Members and 1 Guest are viewing this topic.

Offline MarsupilamiTopic starter

  • Frequent Contributor
  • **
  • Posts: 263
  • Country: us
QFN float during flux activation
« on: July 02, 2021, 10:19:43 pm »
Hi Folks,

Thanks to your help I pretty much eliminated the previous issues I was having but now I'm stuck with a moving chip.
It always happens around ~130degC, I'm guessing flux activation. On my last few runs it moved just barely 1 pin worth, but it's enough so that when the solder melts it won't be able to re-align itself.
I significantly reduced the solder amount on the exposed pad. It's around 35% now and I'm using a 4mil stencil.
I tried pressing the chip more or less into the paste but doesn't seem to make a difference. Any expert advice?
The other day I did an experiment with putting a small weight on the chip to make it stay in place and that seemed to help but that's janky. Paste is Chipquick lead free no-clean T5.



Thanks!



 

Offline MarsupilamiTopic starter

  • Frequent Contributor
  • **
  • Posts: 263
  • Country: us
Re: QFN float during flux activation
« Reply #1 on: July 03, 2021, 12:32:22 am »
Is there vibration or strong air current from fans or a tilt to the PCBA or vibration from motors / etc.
which could motivate motion beyond the usual levels for such things in a reflow process?

It's a vapor phase oven with no movement at all.

Are you pretty sure the solder paste you are using is genuine from a reputable authorized dealer vs. counterfeit?
Is the paste expired or stored / used with improper refrigeration, humidity, etc.?

That's a good point. I'm not. It looks genuine but it's from Amazon and I don't have any practical knowledge of the counterfeit solder paste market :D
Markings look legit. Expiration date says November 2021 and I was storing it in the fridge, giving it a few hours to warm up before use.
I'll order something completely different from DigiKey or so.

Maybe you're applying too much paste and need a thinner stencil or smaller apertures. 
Is your stencil / paste design & use compatible with the paste manufacturer and the IC vendor land / paste pattern suggestions?

I read through a number of app notes about QNF package assembly and one thing I gathered that the recommended stencil thickness for 0.5mm pitch is either 100 or 150um (4 or 6 mil). I have the thinner one. The coverage on the center pad is crucial but what I have is very small at around 35% of the pad area. Most documents mention ~60%.
Pads follow the recommendation of the manufacturer.

Did you make the pads extra long outside the periphery of the QFN so you could use IPC least dense design or manually solder them or something?

Everything is nominal size.
I'm wondering if the problem is not that I don't push the part enough in the paste but I do it too much so that I close the nice escape channels in the paste on the bottom. When the flux starts to do it's thing it lifts the chip slightly as it has nowhere else to go but because I pushed it down too much I smeared the pasted on the perimeter pins to much so they are too low tack to the chip.   :-//

 

Offline kylehunter

  • Regular Contributor
  • *
  • Posts: 108
  • Country: us
Re: QFN float during flux activation
« Reply #2 on: July 03, 2021, 12:50:08 am »
Is there vibration or strong air current from fans or a tilt to the PCBA or vibration from motors / etc.
which could motivate motion beyond the usual levels for such things in a reflow process?

It's a vapor phase oven with no movement at all.

Are you pretty sure the solder paste you are using is genuine from a reputable authorized dealer vs. counterfeit?
Is the paste expired or stored / used with improper refrigeration, humidity, etc.?

That's a good point. I'm not. It looks genuine but it's from Amazon and I don't have any practical knowledge of the counterfeit solder paste market :D
Markings look legit. Expiration date says November 2021 and I was storing it in the fridge, giving it a few hours to warm up before use.
I'll order something completely different from DigiKey or so.

Maybe you're applying too much paste and need a thinner stencil or smaller apertures. 
Is your stencil / paste design & use compatible with the paste manufacturer and the IC vendor land / paste pattern suggestions?

I read through a number of app notes about QNF package assembly and one thing I gathered that the recommended stencil thickness for 0.5mm pitch is either 100 or 150um (4 or 6 mil). I have the thinner one. The coverage on the center pad is crucial but what I have is very small at around 35% of the pad area. Most documents mention ~60%.
Pads follow the recommendation of the manufacturer.

Did you make the pads extra long outside the periphery of the QFN so you could use IPC least dense design or manually solder them or something?

Everything is nominal size.
I'm wondering if the problem is not that I don't push the part enough in the paste but I do it too much so that I close the nice escape channels in the paste on the bottom. When the flux starts to do it's thing it lifts the chip slightly as it has nowhere else to go but because I pushed it down too much I smeared the pasted on the perimeter pins to much so they are too low tack to the chip.   :-//

Doesn't matter if it's legit paste or not, the fact that you ordered it on Amazon likely means it's toast. They won't store it properly refrigerated.

Buy paste from an authorized distributor, then see if the issue remains. I'd also buy any brand other than chipquik.
 
The following users thanked this post: Marsupilami

Offline MarsupilamiTopic starter

  • Frequent Contributor
  • **
  • Posts: 263
  • Country: us
Re: QFN float during flux activation
« Reply #3 on: July 03, 2021, 01:05:24 am »
Doesn't matter if it's legit paste or not, the fact that you ordered it on Amazon likely means it's toast. They won't store it properly refrigerated.
Buy paste from an authorized distributor, then see if the issue remains. I'd also buy any brand other than chipquik.

Hmm, that makes sense. I have another syringe that's T7 grade, I ordered from DigiKey I remember that came in a cooling bag. I can't rule out the Amazon did too but probably not.
 

Offline 48X24X48X

  • Frequent Contributor
  • **
  • Posts: 509
  • Country: my
    • Rocket Scream
Re: QFN float during flux activation
« Reply #4 on: July 03, 2021, 01:41:10 am »
Is that vias from pins I'm seeing from the inner side of the pads routed next to the middle ground pad? These are Rogers PCB? 0.5 mm pitch QFN with a T4 paste and 0.12 mm thick stencil would be good enough to do QFN consistently.
 
The following users thanked this post: Marsupilami

Offline MarsupilamiTopic starter

  • Frequent Contributor
  • **
  • Posts: 263
  • Country: us
Re: QFN float during flux activation
« Reply #5 on: July 03, 2021, 06:52:05 pm »
1: I wondered if the paste amount was too large.
2: I wondered if the periphery pad sizes were too large or specifically anyway the stencil openings vs. the pad sizes.

The openings on the previous stencil designs on the perimeter pads were slightly smaller. That was probably closer to the recommended, 0.04mm smaller than the pad itself on all sides. I increased the size while decreasing the opening on the center pad for the latest stencil. (Opening is exactly the same as the pad) I thought that if the chip floats on the solder pile in the middle maybe increasing the amount of solder on the perimeter pads would provide more of an anchor. Maybe that was a mistake and I should get a 3rd stencil going that has the smaller openings on both the center and perimeter pads.

I assume the boards are ENIG or OSP or something other than HASL coated when pasted...?

Yep, ENIG
 

Offline MarsupilamiTopic starter

  • Frequent Contributor
  • **
  • Posts: 263
  • Country: us
Re: QFN float during flux activation
« Reply #6 on: July 03, 2021, 07:02:27 pm »
Is that vias from pins I'm seeing from the inner side of the pads routed next to the middle ground pad? These are Rogers PCB? 0.5 mm pitch QFN with a T4 paste and 0.12 mm thick stencil would be good enough to do QFN consistently.

There's continuous ground all over, except where the traces come in from the 3 sides. They go directly to the pad. The vias are all grounds as well. (They are plugged with silicone)

The board is not Rogers but a similar HF laminate.
 

Offline mon2

  • Frequent Contributor
  • **
  • Posts: 462
  • Country: ca
Re: QFN float during flux activation
« Reply #7 on: July 05, 2021, 02:38:02 am »
We are building with TQFP 144 (0.4mm pin pitch) 14x14 body size; 1809EXL Heller oven using Loctite GC-10 solder paste.

This solder paste does not require refrigeration(!!) Give it a try. Mix it well before use.

You can source from Adtool in Montreal.

Suggest for you to phone up Dave (Adtool) and/or Loctite technical support to discuss with them as well on your issue.

Henkel LOCTITE introduces GC 10 - this Game Changer is the first-ever temperature stable solder paste

·         No refrigeration.

·         Improved stability.

·         Improved printing.

·         Improved reflow.


Dave McAuley <dmcauley[at]adtool.ca>

Dave McAuley
1004 Rue Des Fabricants
Terrebonne, QC J6Y 2A6, Canada
www.adtool.ca
Toll Free: 1-800-361-6727
Cell: 514-943-4857
Tel: 514-482-2548
 

Offline SMTech

  • Frequent Contributor
  • **
  • Posts: 845
  • Country: gb
Re: QFN float during flux activation
« Reply #8 on: July 05, 2021, 07:32:29 am »
I have never seen anything like this but there is one clear difference between the side the QFN is pulling away from and the other 3. It's possible what you are seeing is a water issue, and steam is outgassing from old paste as it boils off and moving the chip.
Track to pin count seems ultra contrary to me.....
 
The following users thanked this post: Marsupilami

Online Berni

  • Super Contributor
  • ***
  • Posts: 4922
  • Country: si
Re: QFN float during flux activation
« Reply #9 on: July 05, 2021, 07:48:00 am »
It could be the copper ground plane that is the issue.

The ground planes touching some pins can cause them to heat up a a different rate in the oven and can make the molten solder pull the chip over before the other pins had time to reflow too. Tho vapor phase soldering is supposed to fix this, id expect this to happen in a regular oven.

The solder mask tends to pretty loosely define the pad shape, so i typically remove copper between pins to help make a more defined pad that is the exact dimensions as the footprint. Just filling the copper enlarges the pad to the size of the opening. Tho this being RF by the looks of it id imagine you don't want to do that, so perhaps one solution is to reduce the size of the soldermask openings on those filled in pins.
 
The following users thanked this post: Marsupilami

Offline MarsupilamiTopic starter

  • Frequent Contributor
  • **
  • Posts: 263
  • Country: us
Re: QFN float during flux activation
« Reply #10 on: July 05, 2021, 06:40:52 pm »
I have never seen anything like this but there is one clear difference between the side the QFN is pulling away from and the other 3. It's possible what you are seeing is a water issue, and steam is outgassing from old paste as it boils off and moving the chip.
Track to pin count seems ultra contrary to me.....
It's really only 3 pins.
But the water thing or something similar is what I've been thinking of. I'll see when my paste order arrives. I ordered some Loctite GC 50 which is not meant for screen printing but I guess it will be ok for a trial to see if I get any different results. Also ordered some SRA brand which I have no idea if it's any good.
 

Offline MarsupilamiTopic starter

  • Frequent Contributor
  • **
  • Posts: 263
  • Country: us
Re: QFN float during flux activation
« Reply #11 on: July 05, 2021, 06:46:31 pm »
It could be the copper ground plane that is the issue.
The ground planes touching some pins can cause them to heat up a a different rate in the oven and can make the molten solder pull the chip over before the other pins had time to reflow too. Tho vapor phase soldering is supposed to fix this, id expect this to happen in a regular oven.
The solder mask tends to pretty loosely define the pad shape, so i typically remove copper between pins to help make a more defined pad that is the exact dimensions as the footprint. Just filling the copper enlarges the pad to the size of the opening. Tho this being RF by the looks of it id imagine you don't want to do that, so perhaps one solution is to reduce the size of the soldermask openings on those filled in pins.

Vapor phase should make it even. Also the board is really small and thin, moreover the shift happens during flux activation (as far as I can tell around 130degC) the chip stays totally put when the solder melts. I'm not even sure if I can call it being pulled, it's more like it's floating on top of the liquid flux and there's nothing to anchor it.
See the pic I linked in my previous post, the manufacturer recommends solder mask defined pads, I suspect because the solid ground plane is more important for RF performance.
I'm really really hoping it's the crappy solder paste I've been using because I don't have any more ideas :D
 

Online wraper

  • Supporter
  • ****
  • Posts: 16792
  • Country: lv
Re: QFN float during flux activation
« Reply #12 on: July 05, 2021, 07:22:02 pm »
Likely the problem is that you have open vias in the center pad and only small amount of paste on it, so all of it gets sucked in, and capillary action pulls the chip from its place. Vapor phase does not help either as unless some form of preheating is realized, it has very steep temperature increase curve. It basically goes from cold to solder already melted.
 
The following users thanked this post: Marsupilami

Offline Styno

  • Regular Contributor
  • *
  • Posts: 154
  • Country: nl
  • TÜV-geprüft
Re: QFN float during flux activation
« Reply #13 on: July 05, 2021, 07:45:32 pm »
Surely that depends on the quality of the VP-oven. Our Asscon follows a pretty good profile, as measured on the pcb.
 
The following users thanked this post: Marsupilami

Offline thm_w

  • Super Contributor
  • ***
  • Posts: 6272
  • Country: ca
  • Non-expert
Re: QFN float during flux activation
« Reply #14 on: July 05, 2021, 09:42:28 pm »
If you think VP is the issue could try one with simple hot air or hot plate reflow.

But yeah those open vias also make me a bit suspicious.. even if you do manage to reflow it centered, how will you guarantee the solder has not mostly migrated from under the IC?
Maybe try one where you pre-fill the vias by hand just to compare.
Profile -> Modify profile -> Look and Layout ->  Don't show users' signatures
 
The following users thanked this post: Marsupilami

Offline MarsupilamiTopic starter

  • Frequent Contributor
  • **
  • Posts: 263
  • Country: us
Re: QFN float during flux activation
« Reply #15 on: July 05, 2021, 10:36:24 pm »
My vias are plugged. I took me a while to figure that out. (The vendor I used couldn't do it on the 10 mil laminates I choose)

It's really obvious on the bottom of the board if the solder seeps through, but I eliminated that completely so that's not it. :(

I also measured the temperature profile on the board and I tweaked it to a steady ~1 - 1.5degC / sec. (On a vaguely related note I totally will be swapping out the factory controller to one that does proper PWM PID but I think for now it will do.)


 
The following users thanked this post: thm_w

Offline MarsupilamiTopic starter

  • Frequent Contributor
  • **
  • Posts: 263
  • Country: us
Re: QFN float during flux activation
« Reply #16 on: July 06, 2021, 04:09:21 am »
I have a new idea. I did two more just now and contrary to what I though the direction of the shift is not random but seems to be away from the pin 1 corner mark.
There's a tiny bit less solder there but I suspect more that the notch in the center pad (it's solder mask defined) is creating some sort of surface tension condition that's different in the other 3 corners and it's enough to pull the part when the flux starts to do its thing.
I scraped the mask in the corner off on one board, I'll reflow it tomorrow. Fingers crossed.
 
The following users thanked this post: ANTALIFE

Online Berni

  • Super Contributor
  • ***
  • Posts: 4922
  • Country: si
Re: QFN float during flux activation
« Reply #17 on: July 06, 2021, 05:45:31 am »
Oh now i see the pin 1 notch in the center power pad. Yeah it is not something i would recommend doing since the center pad has a lot of pulling power due to the large area. Tho the notch is pretty tiny so it seams a bit drastic that it would pull the chip this much.

Here for prototypes we use some Loctite brand leaded solder paste and a cheep infrared reflow oven. Never had issues with QFN chips just sailing off like that.

Most of the soldering issues we had came down to the paste. The consistency and flux content make a big difference in how well it applies trough the mask and how well it solders. Even kept in a fridge the paste settles out and needs a good mix here and there, then as the paste gets past a certain age it starts working worse and worse. In a pinch we "freshen it up" by adding some extra flux eventually we throw it out and buy a new jar of it.

But i would say if the weight on top of the chip trick works so well, then why not keep using it. I seen much more janky soldering tricks such as using a hacked electric grill to solder some difficult to heat up large aluminium core boards. If it works it's not stupid.
 

Offline Smallsmt

  • Frequent Contributor
  • **
  • Posts: 598
  • Country: de
Re: QFN float during flux activation
« Reply #18 on: July 06, 2021, 06:48:57 am »
It looks like too much paste on pads.
reduce paste pads opening by 10%.

Try to remove 4 paste dots on center pad manually and check if it is still shifting.

Another point is your QFN little pushed in paste or only laying on surface?
« Last Edit: July 06, 2021, 06:50:46 am by Smallsmt »
 

Offline Matt-Brown

  • Contributor
  • Posts: 41
  • Country: us
Re: QFN float during flux activation
« Reply #19 on: July 06, 2021, 02:16:08 pm »
The vias on the center power pad do not look like VIPPO (VIA in pad plated over). It looks like the epoxy fill in the vias is not level with top of the board surface. So as the flux and solder melt, flux is pushed into the vias, then explodes out and moves the part (my best guess). This is causing the part to shift. It also looks like there is much more than 4 thousandths of paste height (but could just be how good the picture is). The print looks perfect though
 

Offline IconicPCB

  • Super Contributor
  • ***
  • Posts: 1527
  • Country: au
Re: QFN float during flux activation
« Reply #20 on: July 06, 2021, 10:19:35 pm »
I have had similar experience with a wi fi module which would consistently float of into the sunset.
Upon inspection and analysis the drift was consistent ( unidirectional ) and it was due to the design of the footprint.
The central land area was asymmetric designed to accept two different modules.
The surface tension would be imbalanced and the module  would float off its central position.
 
The following users thanked this post: Marsupilami

Offline MarsupilamiTopic starter

  • Frequent Contributor
  • **
  • Posts: 263
  • Country: us
Re: QFN float during flux activation
« Reply #21 on: July 07, 2021, 12:17:39 am »
I though I had it but apparently not I screwed up 2 more, same result. I scraped off the corner notch but it still pulls.
I have a suspicion that this is happening:

The pad on the bottom of the chip is significantly larger than the pcb pad and it seems to align itself by the edges of that. I had a moment of panic that I just grossly screwed up the footprint but double checked and I exactly follow the datasheet. The chip center pad size is 3.6x3.6mm while the PCB pad is only 2.79x2.79mm.  :-//
 

Offline SiliconWizard

  • Super Contributor
  • ***
  • Posts: 14297
  • Country: fr
Re: QFN float during flux activation
« Reply #22 on: July 07, 2021, 02:09:48 am »
Just chiming in, sorry if it has been said. Or if it's irrelevant. But, although it's not necessarily easy to tell from your pictures, to me they look (especially the first one) like the amount of disposed solder paste is too much for a QFN. Both on the center exposed pad and on the peripheral pads.

Not sure if I'm the only one with that impression.
 
The following users thanked this post: Marsupilami

Online wraper

  • Supporter
  • ****
  • Posts: 16792
  • Country: lv
Re: QFN float during flux activation
« Reply #23 on: July 07, 2021, 07:22:17 am »
The pad on the bottom of the chip is significantly larger than the pcb pad and it seems to align itself by the edges of that. I had a moment of panic that I just grossly screwed up the footprint but double checked and I exactly follow the datasheet. The chip center pad size is 3.6x3.6mm while the PCB pad is only 2.79x2.79mm.  :-//
IME don't outright trust footprints in the datasheets. I have a feeling that sometimes they simply don't care and provide it just because. Not to say that for the same package, different manufacturers often provide radically different footprints.
Quote
I scraped off the corner notch but it still pulls.
If the pad on the chip has such a corner as well, there is nothing bad if the pad on the PCB follows its shape as well.
« Last Edit: July 07, 2021, 10:03:15 am by wraper »
 
The following users thanked this post: Marsupilami

Offline MarsupilamiTopic starter

  • Frequent Contributor
  • **
  • Posts: 263
  • Country: us
Re: QFN float during flux activation
« Reply #24 on: July 14, 2021, 03:33:27 am »
Oookay, I got an additional 3 different kinds of solder paste and the 3rd stencil variant with smaller perimeter pins and no reduction at the pin 1 notch. If I can't solder this stupid chip this time I'll give up and move to the Himalayas to be a goatherder.
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf