I was aware of the suggestion that you could “bake” the circuit board from a dead Graphics card in an oven to revive it – at least temporarily. Youtube has too many videos to count on this subject – most being posted because the result was success. The commonly held belief is that it works because you have melted the solder and re-flowed the cracked connections. But it appears that there are two opposing schools of thought on this solder “re-flow” idea.
The majority “believe” the heat treatment fix for failed chips works by by re-flowing the BGA solder connections underneath the package – the other much smaller group point out that most reported fixes never reached a temperature that could have melted the lead free solder used on modern boards – but yet the boards came back to life for a period of time (weeks or months, occasionally longer). Trying to prove if cracked solder joints below a BGA chip are the cause of a failure is almost impossible – except by attempting a reflow or reballing of the chip and then observing if the fault has been fixed. But what if the act of heating the chip without melting the solder was the actual cause of the fix?
Until now I had not had cause to try this type of repair – but I recently acquired a failing 47″ LCD television with dead HDMI sockets – which then died completely and would no longer “boot”. This appears to be a common type of failure on large LCD TV’s built around 2010 and is attributed to a failure of the main circuit board – most likely one of the BGA chips.
Nvidia had a massive problem with BGA chips failing from around 2007 to 2009 both on their own graphics cards and in chips supplied to OEM’s such as Apple (used in Mac Books) and many other expensive Graphics cards still have a higher failure rate than customers expect – even now.
So before attempting an aggressive DIY solder reflow or paying for a professional BGA reballing I decided to try a simple heat treatment at a lower temperature (100°C to 150°C). I removed the simple metal plate heatsink from the main chip and I was able to reach a temperature of 100°C on the chip top surface (monitored using a digital temperature probe) and held that temperature for 3 minutes. I was able to do this with a normal hairdryer at point blank range. The TV came back to life with all HDMI sockets working and is still going 2 weeks later.
To be clear there are BGA chip failures caused by cracked solder connections – but these are generally caused by warping or bending of the circuit board. Certain circuit board designs are prone to thermal warping as the chips get excessively hot – which can cause failures of the BGA connections around the edge of the chip. Some gaming consoles are known to suffer this failure mode. The other type of device where BGA chip connections do fail is in mobile phones mainly due to users bending the phone. Several mobile phone designs are prone to excessive bending of the case and the circuit boards inside bend beyond design limits – almost cracking the chips off the board.
The question I needed answered was: What in a failed BGA chip would be changed at temperatures just above 100°C? That is a temperature well below the melting point of ANY solder (183°C to 220°C).
It turns out there is one critical component of BGA chip construction that does fit the profile – the Underfill – and it is also well known to cause chip failures if BGA chips get too hot. The undefill material bridges the gap between the silicon chip and the carrier circuit board in the BGA Flip-Chip package – and is designed to provide non-conductive mechanical support for the electrical connections between these two elements.
Without the solid glass-like underfill the thermal expansion and contraction of the different layers of the Flip-Chip the would quickly cause cracks to form in the connections between the silicon chip and the carrier circuit board. These connections are extremely small and are commonly referred to as micro-balls or bumps and are commonly around 0.20-0.35mm in diameter.
The underfill material is required to stay in a solid glass-like state at all chip operating temperatures.
However if the undefill is under-specified and begins to soften at chip operating temperatures (or the chip just overheats) it causes thermal stress damage to occur inside the BGA Flip-Chip itself. Just do a Google search for “Nvidia Underfill” to see the real life fallout from this issue.
Many “underfill” materials begin to soften at about 100°C and becomes fully liquid by about 150°C (exact temperatures depend on the particular type of underfill used) which is why this temperature range has been found to “fix” many dead BGA chips.
This thermal stress damage has nothing to do with the actual BGA solder connections underneath the package – which is the component that almost all the “re-flow” and “reballer” enthusiasts are focused on. Heating the chip to a temperature hot enough to melt solder will also always melt the underfill and allow the silicon chip to move slightly and de-stress the “bump” connections. But this will also happen if you heat to a lower temperature that only melts the underfill material without affecting the solder.
It will not generally be a long term “repair” as any cracks will still be present – but the cracks will “sit down” and remake the electrical connection. The underfill greatly slows any oxidation of the cracked joints so re-connection is moderately successful. Adding a better heatsink can greatly improve the longevity of the fix by keeping the operating temperatures down.
Whether it is possible to turn this into a real fix by actually re-making the bump connections inside the chip is still unclear.
