PCB X-Ray Services

[48X24X48X]

Hi guys,

Does anyone know any company that would provide PCB x-ray services? Like they take an X-ray image of your assembled board and send a soft copy to you.
I have been prototyping a board with an LGA module package for 3 revision now and it always end up with shorts. I'm on the brink of giving up on the design but I could give it another try providing I know how the joints under the module would look like.

And what are the cost like?

[voltsandjolts]

Maybe mike can help if you ask nicely.
mikeselectricstuff
https://www.eevblog.com/forum/profile/?u=906

[Kean]

Or your local dentist?

[48X24X48X]

Hi guys,

Thanks for the suggestions! Especially on mike! 
After discussing the issue with the module distributor, it proved that I'm not alone, at least I'm not crazy.
Most of the customers have about 30% rate of short circuit with the module. That is scary enough to go into production.
Throw in the towel I guess unfortunately. By the way, it's a Sierra Wireless 3G module.

[Leolabs]

Why dont u try to check with Vitrox for their AXI machine?Since u are from Malaysia.....

[wraper]

Sounds like a wrong amount of solder paste applied, very important for LGA. I hope you don't fill all of the area in the center with a solder paste. Sounds like issues with solder paste stencils you used rather than PCB revisions.

[Kean]

I agree with @wraper - too much paste.  I've seen the same with LCC GPS modules that I've prototyped.
You may need to use smaller apertures on the stencil for these pads, or if that isn't possible you may need to use a thinner stencil material.

[mikeselectricstuff]

Most big assembly houses will have x-ray facilities - they may do it as a service if they're set up for it, or maybe do a freebie if you ask nicely ( or suggest they will get the job for production)!

If this is something for production you should talk to potential assembly houses as they should be able to advise on assembly issues like this.

Dental x-rays probably don't have the penetration or resolution - for BGA type stuff you really need something designed for the job.

[voltsandjolts]

Here's a QFP on PCB courtesy of our local dentist showing vias, tracks and bonds.

[48X24X48X]

Yes, the very first thing I thought was the stencil opening. So, I made them smaller in 2nd revision and even smaller in 3rd revision. Even their recommended stencil opening shape was a weird 3 trapezoid shape for a single square pad. Here's an example of how the pads looks like without the module mounted.

If this is something for production you should talk to potential assembly houses as they should be able to advise on assembly issues like this.

I made them in small quantity that renders sub contracting them to local factory not feasible.

Why dont u try to check with Vitrox for their AXI machine?Since u are from Malaysia.....

I didn't know there's even an option here. Let me check them out. Thanks.

[wraper]

WTF is with that central pad. It should be a solid copper polygon without solder mask.

[mikeselectricstuff]

WTF is with that central pad. It should be a solid copper polygon without solder mask.

Not necessarily - that can put too much paste on. Could be it's a BGA/LGA-style package with multiple pads rather than a single pad.

[james_s]

You might try asking a vet (animal doctor), they usually have xray machines and if they aren't busy one may be willing to xray a board for you. The machine I've dealt with is essentially the same as those made for human medical diagnosis with a few changes to make it easier to use it on animals which tend to not hold still on request.

[wraper]

WTF is with that central pad. It should be a solid copper polygon without solder mask.

Not necessarily - that can put too much paste on. Could be it's a BGA/LGA-style package with multiple pads rather than a single pad.

How is this related? You should never do 100% paste coverage of the pad. In some cases some solder mask may be left on the pad, to divide it in areas but no such nonsense with separate pads. Each with thermal relief, defeating the main purpose of the package.

[wraper]

Same pad size, different amount of solder per same area. Solder will be squeezed out from the pads with higher amount of solder.

[wraper]

I think I know which part he is using. I see the recommended footprint there with separate pads on the center (thermal reliefs are still a nonsense). No solder paste stencil recommendation though.
http://www.eltech.spb.ru/item/HL8548%20&%20HL8548-G.pdf
EDIT, looks like significantly less solder on their photo. What is the stencil thickness, is 100% aperture used?

[amitchell]

It also looks like the solders being used are wetting the pads differently, and have a different surface finish.  I have heard that the LGA packages are fussy, maybe figure out what solder the chips manufacturer recommends. Play with the reflow curve too.

Rounding the corners of the pads may also help prevent bridging.

[48X24X48X]

I think I know which part he is using. I see the recommended footprint there with separate pads on the center (thermal reliefs are still a nonsense). No solder paste stencil recommendation though.
http://www.eltech.spb.ru/item/HL8548%20&%20HL8548-G.pdf
EDIT, looks like significantly less solder on their photo. What is the stencil thickness, is 100% aperture used?

I was about to post that red PCB picture. The different is my GND pads has thermal relief, theirs no. This could be the potential problem. Attached is my stencil opening. For the outer small pads, I followed their recommended stencil opening size but for the inner GND pads I weren't able to generate the complex pattern they recommended (will need to some primitive drawing in KiCad to get it properly done), so I reduced them quite a bit especially on the most outer GND pads. They recommend 0.125 mm stencil thickness and I used 0.12 mm which is even thinner, by a bit. 

It also looks like the solders being used are wetting the pads differently, and have a different surface finish.  I have heard that the LGA packages are fussy, maybe figure out what solder the chips manufacturer recommends. Play with the reflow curve too.

I used ENIG for all my PCB and that is the recommended surface finish. In terms of reflow curve, I followed a generic lead-free curve.

Same pad size, different amount of solder per same area. Solder will be squeezed out from the pads with higher amount of solder.

Both of that 2 pads have same stencil opening size.

But, here's the catch. There is one and only one net that shorts with GND. Which is the main power supply to the module. See the attached  image. It's weird because if we assume that the solder paste does flow from the center pad (solder mask is there though to prevent this) to the outer pads, why only this particular NET. It would have shorts with other pins too. I have checked those pins that were broken out and use in the design and there's no short to GND.

[amitchell]

I used ENIG for all my PCB and that is the recommended surface finish. In terms of reflow curve, I followed a generic lead-free curve.

Sorry for not being clear, I meant the solder had a different surface "finish" i.e., it looks like a different type of solder. Maybe an avenue to explore, along with the reflow curve. 

[KE5FX]

The central pad looks fine to me.  The best practice is probably lots of blind vias under a continuous pad, but board houses charge big-time for those, and most ICs don't care that much. 

If you've already shrunk your paste apertures, I'd try a thinner stencil. 

[SMdude]

Are you able to remove one and confirm that the module hasn't actually shorted within the module itself?
To me it seems odd that that is the only net experiencing shorts. The solder balls don't look excessive to me.
Are you using a low temp paste?
Have you tried removing the paste from a few of the pads nearest the V_batt pads?
Or a short somewhere else, though I assume you have eliminated everything to end up at the module..

Cheers

[voltsandjolts]

The gap between the centre GND pads and the left/right pads looks smaller than the gap to the top/bottom pads.
So why are the shorts not happening at the sides?
As SMdude says, is the short in the module?

[Spikee]

The gap between the centre GND pads and the left/right pads looks smaller than the gap to the top/bottom pads.
So why are the shorts not happening at the sides?
As SMdude says, is the short in the module?

I have had a 50% failure rate for a RF module which was configured on the 433MHz band , the 915 and 868 MHz modules had no issue at all.
The shorts were within the modules itself not on the board I designed. I have seen this multiple times.

I always do 100% past coverage other than for BGA's *which are a pain in the ass" , 0.1mm thick stainless steel stencil or smaller.
I have heard about 0.07mm thick stencils before at assembly houses.

[48X24X48X]

Sorry for not being clear, I meant the solder had a different surface "finish" i.e., it looks like a different type of solder. Maybe an avenue to explore, along with the reflow curve.

Their reflow curve recommended on their datasheet looks pretty standard. But, they didn't mention anything on the paste itself other than it should be lead-free.

Are you able to remove one and confirm that the module hasn't actually shorted within the module itself?
To me it seems odd that that is the only net experiencing shorts. The solder balls don't look excessive to me.
Are you using a low temp paste?
Have you tried removing the paste from a few of the pads nearest the V_batt pads?
Or a short somewhere else, though I assume you have eliminated everything to end up at the module..

I have tried in the past, by the time I have the amount of heat applied on the board, the FR4 started to melt. The center pad is holding the module to the PCB too strongly I guess. No, I'm not using a low temperature paste. I did tried once removing he outer most ring GND paste and the module does work although not reliably. But, in the sense it doesn't short. This was in the 1st revision of my prototype where the paste opening was 95% of the pad size. In the case where other parts of the circuitry could have shorts, I have also removed them totally before.

If you've already shrunk your paste apertures, I'd try a thinner stencil.

I think a 0.1mm thickness would worth the try.

The gap between the centre GND pads and the left/right pads looks smaller than the gap to the top/bottom pads.
So why are the shorts not happening at the sides?
As SMdude says, is the short in the module?

The module is not symmetrical, more like a rectangle. That's the odd part, why not short the pins to the left and right?

There are 2 variation of the module actually, one the 3G version and the other is 2G version. They are pin compatible. But, the 2G module body is what you would expect from a module (metal can on top, and visible FR4 PCB at the bottom). However, the 3G version doesn't look metal at all (it's some coating I guess). If you do a continuity test around the 3G module except the bottom part without pads, it's all connected to GND even on the sharp edges. The individual non-GND pads obviously is not shorted to GND.

But a 30% mount issue as reported by my distributor really say something is not right somewhere...
And it ain't cheap a piece.

I have had a 50% failure rate for a RF module which was configured on the 433MHz band , the 915 and 868 MHz modules had no issue at all.
The shorts were within the modules itself not on the board I designed. I have seen this multiple times.

RN2483 I guess? I think a 0.10mm looks like a possible solution.

As for now, I'm moving on and ditching the Sierra Wireless series. Even if I managed to figure it out, I believe the yield could still be less than optimal. But, I will reorder a thinner stencil with even smaller aperture to try.

[wraper]

so I reduced them quite a bit especially on the most outer GND pads.

Asking for trouble, when there is a different thickness of solder in different places, thicker solder will be squeezed out.

But, here's the catch. There is one and only one net that shorts with GND.

Have you measured resistance on the new not soldered module? Is it even soldering issue at all? You said you always get shorts, does it mean in 100% cases?
I don't know anything about these modules, and not into RF but if depletion-mode GaAs MOSFETs are used, it will be dead short on the power before voltage is applied to their gates. So dunno, maybe your module does not start properly.

I have tried in the past, by the time I have the amount of heat applied on the board, the FR4 started to melt.

You need a preheating on the bottom for this.