LGA / MEMS / magnetic sensors - why do I see them fail so much ?

[kellogs]

i know these kind of sensors are sensitive but, man... how do they fail so easily ?!

My procedure:
- bake the MEMS sensors @ 125 deg C for 9 hours (dry out phase)
- apply solder paste over the boards through an electropolished stencil
- place the sensors by hand as straight as I can then put to oven for soldering as per
J-STD-020 - JEDEC. Not to a tee though; the cool down phase for instance is slower, I can not do -6 degC/s in my hobby oven.
- power the boards and test them in a fixture with a magnet; Same distances, same magnet, swipe direction, algorithm, positioning, etc

My success rate varies in the 30% - 70% interval. The failed chips will either not sense the magnet from the same distance or not sense it at all. Tried it with three sensors from three different producers.

I may also not respecting the general recommendations for soldering these chips to a tee, like the ones below, but really, is that why half my chips are failing ?

https://www.nxp.com/docs/en/application-note/AN3484.pdf
https://mcubemems.com/wp-content/up...andling-Assembly-Guideline-for-LGA-AN-001.pdf
https://www.st.com/resource/en/tech...rs--in-an-lga-package--stmicroelectronics.pdf

Thank you!

 

[Harald Kapp]

I suggest you build a test fixture to test the chips before soldering. You can thus narrow down your problem to either the baking process (if they fail after baking but before soldering) or the soldering process (if they fail after soldering.
Worst case scenario: one half fails after baking, one half after soldering.

 

[kellogs]

Worst case scenario: one half fails after baking, one half after soldering.

Aww, come on!

good idea with the test fixture but these are not so easy to find. Only found one on ebay for some $75. And it says QFN16 not LGA16... Got other sources ?

L.E. Oh, build it you say... I immediately thought about those socket adapters...
Ok how do I build these ? I got some bicomponent liquid plastic at hand, hmm...

 

[Harald Kapp]

All you need is some pogo pins arranged in the correct pad pattern and a clamp to hold them down. Could be made as a 3D-printed unit or simply from a piece of plastic (Teflon is commonly used for text fixtures due to good insulation quality) and suitably drilled holes. Gets you going at considerably less than 75$, I should think.

 

[shrtrnd]

Apparently you're following some procedure in baking these sensors for 9 hours at 257 degrees F for 9 hours.
I'm not familiar with what you're doing, but I'd be worried about warping the plastic packages, causing internal mechanical failure.
I assume I'm off base on this, but I'd go with Harald Kapp's advice to test them before and after baking to see if your baking process
is killing them. One other idea, is to ask about the accuracy of the temperature controller on your hobby oven. Is it possible your
oven is exceeding that 125 degree C threshold?

 

[kellogs]

I think it is the die that is susceptible to cracking, not the enclosing plastic, and I don't think going over 125 a bit would be that much of an issue. After all, these can withstand 260 C for a couple of reflows. What I do know is that without baking them the failure rate is higher. Surely testing them before and after is a good idea, but I would not know how to pogo-test such small parts...

 

[shrtrnd]

It's your project, but a solder reflow at 260 C is not as thermally stressful on a component as 125 degrees C for 9 hours.
You said you've tried 3 different sensors from 3 different manufacturers, and you're using JEDEC specs.
How reliable are your vendors?
I believe you're using a standard procedure developed to test authentic OEM components.
Where are your parts coming from? Are you buying them from an approved OEM vendor with traceability?
There are a lot of counterfeit parts out there, allegedly from many genuine manufacturers, that aren't.
If you're doing your baking and testing correctly, the problem isn't with you.
I'd be suspicious of the origin of the parts you're using.
If this were me, I'd reverify all of my test parameters to be sure I wasn't overlooking a critical step, and
then take a close look at where I was getting my test components from, just in case they don't meet spec.
I know this has to be frustrating, but you sound like you're skilled at what you're doing.
Don't ASSUME anything when you reverify your test steps and parameters, VERIFY what you have.
Good luck.

 

[kellogs]

Yeah, I sound like I am skilled, but reality contradicts...
-
The chips I am buying from a well known distributor and are produced by reputed or ok vendors. I took note on the 125 deg limit and reprogrammed my oven to do 90 deg for 23h. I managed getting it to operate in the 87 - 94 deg interval. Of course, my oven is just a home built unit, not a professional one, so temperature readings may be off by some 10% from temp of actual chips being baked. This should be fine as the standard also allows for 33h @ 90 deg.

-
After much more frustration I have (likely) discovered some more subtleties of reworking; have not used the oven but only hot air

When I only do 185 - 190 deg C just so the solder melts I am getting non-contacting pads. As soon as I go 200+ no more opens, but the sensitivity is badly impacted; also it self-centers much better at higher temps.
I am, how should I say, in despair... Since i am unable to find any folks with my problem on the internet I am pretty sure it is me botching it.

 

[shrtrnd]

I don't know the manufacturer of your chips, but my experience is that the original manufacturer usually has their test procedure data available. Sometimes you have to dig through their website for the technical specs, and if you can't find them on-line, use their 'contact us' email option to ask for them. Manufacturers specify their specs, and they have to back those specs up with the data.
I'd go after the manufacturer's website and get a copy of their procedures for testing and installing their component.
There may be something in their process that you're missing in just reading their list of specs.
Also, I'm going to mention something you may consider.
Humidity plays an important part in testing. If you're in a high humidity area, baking the components and then exposing them to
high humidity immediately afterward might be causing them to suck moisture out of the air while the components are cooling.
May not be a problem for you, but wanted to mention it.
When I experience problems like you're seeing, as annoying as it sounds, I go back over what I want to do step-by-step, and reverify everything I did. I've learned that when I get to step #3, or whatever, during my re-evaluation of my procedure, I inadvertently 'assumed' something, without actually checking that detail thoroughly. It might not help if you have a temperature control problem or some other basic technical issue you can't verify. But for me, I find out what's wrong with the result I got, by re-checking my test procedure and finding something specific I made a wrong assumption about.
I know you're convinced, but I am suspicious of your baking oven temperature regulation. (even though you've eliminated it)
Still offering you good luck in getting to the bottom of this quandary.
Do you know anyone else in your area with your skill level, who can take a look at what you've got, and see if they can find a
fault or omission in the procedure you're using?