According to IPC-A-610 revision G, you can have a 30% void in your BGA device. I every pin in the STM32F417IGH6 device was missing 30% of its connection, would the device still work per the datasheet?

[GVarn.1988]

IPC-A-610 revision G page 8-93 indicates that you can have a 30% void in your solder ball. We recently had a STM32F417IGH6 device over heat. Upon X-ray, we discovered that about 80% of the solder balls had voids near 30%.

[Danish1]

You say "We recently had a STM32F417IGH6 device over heat."

I must say I am surprised at this. My stm32 devices don't seem to run significantly warm. Were you clocking it flat-out (or even overclocking it)?

I assume you worry that the voids will increase thermal resistance rather than ohmic resistance.

You specify a 176 ball package and ST say this has a thermal resistance of 39 C / W junction to ambient.

They don't mention thermal resistance to board - which is what might be compromised if the solder balls had serious voids, or any requirements on the amount of copper in the tracks to spread heat around your pcb.

If we look at the absolute maximum ratings, the most current you can put in / out of the supply is 240 mA*. And at 3.3 V this corresponds to 792 mW. So assuming 39 C / W, I'd expect the processor core to be 30.9 C above ambient. Any idea what your ambient temperature was? And your peak supply-current drain?

I'd also say that there probably isn't anything internally different between your 85C-maximum and the 105C-maximum parts.

*The processor core should only burn 117 mA worst-case at 168 MHz, but peripherals also add to this as do GPIO.

Do you drive much high-current I/O pins (e.g. LEDs)?

Or fast I/O e.g. external memory?

Final question: Does your board connect all Vdd and Vss pins? I know it's hard to route them all with a BGA but without doing that you will increase heat generation inside the stm32.

Regards,

Danish (nothing to do with ST other than as a customer)

[Uwe Bonnes]

A device overheating and non-functional is often a sign of overvoltage applied at some point in time.

[GVarn.1988]

​Danish,

Yes, I was surprised as well to find that the device was overheating. I didn't have a thermal imager at that time so I cannot tell you the exact temperature of the device. This was not typical of how the device normally operates in the design. It typically runs much cooler. The environmental temperature was 20 degrees C at that time.

Now, it could just be a bad device (one that is broken). My question though is more general in nature. Yes, I did find a lot of voids in the solder for this device and I was quite concerned about it. But I guess what I really want to know is do my concerns have merit?  What effect can the voids have on this chip. In specific can it loose 30% of the connection on each and every pin and the device still function per the datasheet? In essence, do I really need to worry about these voids as long as they meet the IPC requirement? My concern was not so much that I had voids, but the fact that almost every pin had a void.

Yes, electrically speaking the connection resistance would go up but probably (depending on the solder conductiviity ) only 1 or 2 milliohms at most per connection. But, as you indicated, these pins also help sink the heat of this device. Is loosing 30% of your connection to the board going to make this chip dysfunctional?