QFN silk screen slipping

[derGoldstein]

I ordered a few boards from OSHpark with QFN16 footprints, and was having as hell of a time trying to figure out why I was getting what seemed to be shorts across some of the leads. I managed to avoid using QFN packages until now, so I thought that I just kept messing up the soldering job. I couldn't see a problem with a loupe magnifier so I took a few photos as close as I could to the board -- I'm attaching them here.
I think that the silk screen shifted up on the Y axis, creating windows which exposed more than one lead. On some boards I don't even have to solder a chip onto it -- just tin the pads slightly and it shorts out the leads (the two pads in the red rectangle get shorted almost always). If you look at the rest of the exposed pads on the board they all have more space exposed on Y+ than on Y-.

My question is -- is this tiny sliver of exposed pad enough to create a short, or would the surface tension of the solder make it "ignore" such a narrow area? I know by testing it that the short exists, but I can't be certain if this is because of the silkscreen being off or me not soldering the device properly.

Is shifting the silkscreen on a particular axis a common problem? Is there any way to check for this before sending boards to fabrication?

I'm not trying to specifically complain about OSHpark here, I've ordered plenty of boards from them in the past which worked just fine, I just haven't encountered this problem before (though I'm going to go back and examine previous boards from them to see if this is common).

EDIT: I checked a bunch of OSHpark boards and this was an exception. Most boards have only a very slight shift, not enough to effect soldering.

[wraper]

First of all this is not silkscreen but solder mask. Silkscreen is for component designators. Secondly, your problem is not in the solder mask and not in the PCB at all. What you need is thin stainless steel stencil with 1.127 mm or 0.1 mm thickness, and hole size reduction so less solder paste will be applied.  No mylar stencils will work even remotely well. Paste needs to be applied exactly on the pads without any overlap, therefore one pass with squeegee is preferable. If you appy paste by hand, then no luck, you will need to rework this anyway.

BTW, center pad need about 70% coverage with paste with 1.127 stencil. And what in the hell is that huge hole in the center?

[T3sl4co1l]

Methinks he's doing the poor man's QFN process... soldering-iron-in-huge-via

Can you take a pic of it in the tinned-and-shorted condition..?

Tim

[wraper]

Methinks he's doing the poor man's QFN process... soldering-iron-in-huge-via
Tim

That crossed my mind but why then complaining about shorts as a consequence of PCB quality?

[suicidaleggroll]

What are you using for the solder mask clearance (aka pad-mask clearance) in your PCB program?  Sometimes this defaults to some huge value like 7 mils, which can allow multiple pads to be exposed without any mask in between on fine pitch SMD parts (or small registration problems in the solder mask application can expose pads or tracks in the wrong places, as you're seeing here).  I generally use 3 mils which works well.

[rs20]

Misalignment of the soldermask layer is common with cheap, crappy PCB suppliers. In fact, it's the main practical difference I see between my cheap supplier and my more expensive one, and is the driving factor (along with price of course!) for choosing one over the other for any given PCB.

[derGoldstein]

I keep saying silkscreen instead of solder mask, I even messed up the title... I know the difference, my brain is just glitching.

Here's an example of one of the boards that just shorts out some of the leads when they're tinned. My camera doesn't do macros well so I had to flood the board with a bright LED. It doesn't actually look like a disaster site, it's just a very bright reflection from flux splatter. The board is in the opposite direction this time, you can see the pin-1 square is on the bottom-left rather than the top-right.

From this distance (I can't get any closer) there's no visible bridging. Yet, pin 13 and ground have relatively little resistance between them (~20k), and pin 9 is practically shorted to ground (~500ohm).

I've been using the tin-and-reflow method. I tin the pads on the PCB, apply flux, place the QFN on the board, and then reflow with a hot air gun (with as low of a temperature as I can get away with, usually around 260c). The hole in the middle is to allow excess solder on the thermal pad an escape path in case I used too much. In this particular picture I didn't bother wicking it off, so there's a lot of it on the thermal pad, but I usually keep it to a minimum so there's only slight curvature on the solder "mound" in the center.

I was hoping to avoid using stencils because at this size my hand isn't stable enough to accurately place the package in one go, it's always off by a bit so the solder paste gets smudged around.

[wraper]

From this distance (I can't get any closer) there's no visible bridging. Yet, pin 13 and ground have relatively little resistance between them (~20k), and pin 9 is practically shorted to ground (~500ohm).

Than is not how solder bridges happen, if there is any, there will be dead short. Are you using some crappy active flux which is incompatible with electronics? BTW there is hell a lot of solder on the middle pad. Only tiny amount actually is needed.

[janoc]

Misalignment of the soldermask layer is common with cheap, crappy PCB suppliers. In fact, it's the main practical difference I see between my cheap supplier and my more expensive one, and is the driving factor (along with price of course!) for choosing one over the other for any given PCB.

OSHPark is certainly not a cheap and crappy supplier. I have used them recently for boards using the MPU9250 motion sensor. That thing is a 4mm wide 24 pin QFN with 0.4mm pin pitch. And no problems - the solder mask was okay. However, I have made sure that I actually have a thin strip of soldermask between the pins.

[Kilrah]

On the pictures in the first post it looks like the entire solder mask layer is shifted, not just the QFN.

It however does look on all pictures like your library part uses esxcessively big solder mask holes, with no separation between the pads.

[derGoldstein]

From this distance (I can't get any closer) there's no visible bridging. Yet, pin 13 and ground have relatively little resistance between them (~20k), and pin 9 is practically shorted to ground (~500ohm).

Than is not how solder bridges happen, if there is any, there will be dead short. Are you using some crappy active flux which is incompatible with electronics?

Damn it I can't believe I didn't check that... It accounts for the higher-resistance "shorts", they're gone after I rubbed it thoroughly enough. I never had this problem with this flux but then I never used a chip where I couldn't access the leads to clean them.
Getting flux in israel is a pain, since western distributors won't ship it. So I can get brand-name solder (I use AIM Solder Sn63/Pb37 GLOWCORE 2%), but only crappy flux. Farnell has a distributor here that will sell BONKOTE flux pens in groups of 5 for $105... All other companies here only deal with 1-liter packages or above. So I'm left with flux from hardware stores that is really only meant for large connectors/wires (or the industrial stuff used to solder copper pipes).

BTW there is hell a lot of solder on the middle pad. Only tiny amount actually is needed.

Yeah I mentioned in the previous post that I didn't bother wicking it after I removed the chip, I usually use the absolute minimum on the thermal pad.

On the pictures in the first post it looks like the entire solder mask layer is shifted, not just the QFN. It however does look on all pictures like your library part uses esxcessively big solder mask holes, with no separation between the pads.

Yeah, the whole thing looks shifted about 0.1mm up the Y axis. If the clearance around the pads were smaller the conductors themselves could be covered by the solder mask, so shrinking the clearance around the pads could make the board unusable.

[wraper]

Damn it I can't believe I didn't check that... It accounts for the higher-resistance "shorts", they're gone after I rubbed it thoroughly enough. I never had this problem with this flux but then I never used a chip where I couldn't access the leads to clean them.
Getting flux in israel is a pain, since western distributors won't ship it. So I can get brand-name solder (I use AIM Solder Sn63/Pb37 GLOWCORE 2%), but only crappy flux. Farnell has a distributor here that will sell BONKOTE flux pens in groups of 5 for $105... All other companies here only deal with 1-liter packages or above. So I'm left with flux from hardware stores that is really only meant for large connectors/wires (or the industrial stuff used to solder copper pipes).

Hopefully there are not many of such boards sold in the wild as they can corrode and start to fail with time. You cannot clean such flux completely unless use ultrasonic bath or kind of dishwasher cleaning with full submerging. That flux is not suitable for wires either, unless you want them to corrode heavily. Bonkote flux pens come without any flux unless someone fills them. 5pcs is standard packaging, they are not sold as singles usually (unless Chinese fakes on ebay). Not good for transporting in filled state anyway. If there is no other possibility, at least get some flux from china, probably gel, it will be much better anyway.

[chicken]

I occasionally get misaligned solder mask with OSHPark. The one attached is about as bad as it gets. The footprint is a .5mm pitch QFN with .25x.7mm pads, so alignment is about .1mm off.

Luckily the alignment is better on the other 2 PCBs of that batch and I only needed 1 to evaluate my prototype.. but still.

With Chinese fabs it's usually the silk screen that's misaligned or pixelated. OSHPark's silk screen is usually superior.

[T3sl4co1l]

Yyyyyeah....

That's kind of your own fault, what with the zero mask expansion -- you can't expect anything but severely reduced solderable area in that case. 

Tim

[derGoldstein]

I took some pictures of the other boards with the QFN16 footprint as well as a couple of additional boards (with TSSOP chips) and they're fine on average, that one board was the exception. I'm attaching some crops.

Apart from buying a DSLR, does anyone know of a camera that's particularly good at taking very close macros (and which doesn't cost a fortune)? I remember one video where Dave talked about the cameras he's used in the past but I can't find it now.

[SeanB]

Try using a magnifier over the phone camera lens when you want close up pics. Works quite well, though you will need a lot of light, as otherwise shake will blur the image, and of course the flash will not work well. Depending on the lens you will have to crop the image to get the circular edges out, but it is a cheap alternative.

[T3sl4co1l]

For the mask2 and mask3 pics, you might consider using narrower pads (and adding trace neck-down and polygon pour thermal relief rules).  It looks like the mask web width is too narrow and should've been removed by the fab, but wasn't.  Which can lead to the webs breaking free and contaminating pads.

Looks like you could maybe stand to use a bit more annular ring too, with this fab.  The vias in these pictures do look okay -- but it's hard to infer the worst-case condition based on that.  So, better to be on the cautious side, perhaps?

Tim

[M4trix]

So OSHPark is using blue soldermask now. Good, using only purple was a bit downside and boring. 

[suicidaleggroll]

Pretty sure those are still purple, it's the lighting/coloring in the picture that's throwing things into the blue spectrum.  I had a couple of boards fabbed at OSHPark a month or two ago and they were the standard purple.

[M4trix]

Hm, it looks like pure blue. Dunno, maybe they will do it if you have a special request. 

[Kilrah]

White balance just seems to be off on his pics.

Can look like this with a little correction:

[M4trix]

So it's purple after all. Heck...