Top Side BGA Reflowing @ Wave Solder

We've recently encountered this problem where one of our .050" pitch BGA's were reflowing at the wave thus causing shorts... The board was profiled with probes stuck to a QFP solder joint, an 0603, and one of the BGA joints. There was an almost 70 degree difference in peak temperature between the QFP and BGA solder joints!! The BGA joint was actually at liquidus for almost 10 seconds!

What we needed to do was speed the wave conveyor up by 1.5 FPM and turn the chip wave off - this took about 20 degrees off the energy curve and just under the eutectic melting point (361 deg. F). I wish we didn't have to do this since there are small SMD devices on the bottom side....so the question is....

....why would a BGA solder joint get so much hotter than everything else? Any of you solder experts know the answer??

Regards,

C.K.

reply »

| We've recently encountered this problem where one of our .050" pitch BGA's were reflowing at the wave thus causing shorts... The board was profiled with probes stuck to a QFP solder joint, an 0603, and one of the BGA joints. There was an almost 70 degree difference in peak temperature between the QFP and BGA solder joints!! The BGA joint was actually at liquidus for almost 10 seconds! | | What we needed to do was speed the wave conveyor up by 1.5 FPM and turn the chip wave off - this took about 20 degrees off the energy curve and just under the eutectic melting point (361 deg. F). I wish we didn't have to do this since there are small SMD devices on the bottom side....so the question is.... | | ....why would a BGA solder joint get so much hotter than everything else? Any of you solder experts know the answer?? | | Regards, | | C.K. | C.K. -

Welcome to the exiting world of secondary reflow. One reason that BGAs are so susceptible is the grid of vias that they require for routing out of their I/Os. When taken together, the vias can transmit a tremendous amount of heat to the solder balls when they fill with molten solder. The other common "fixes" that you might try are to have plugs added to the vias to stop the solder flow. A more crude (but possibly more effective) approach is to place a square of Kapton tape over the vias that are causing the problem. You will have to use an ionizer when removing the tape to avoid possible ESD damage, or use ESD-safe tape. Also FYI, consider yourself fortunate that you got a full reflow of the solder balls. If they are only partially reflowed, the board can test good but be prone to early failure.

Glenn Robertson

reply »

| We've recently encountered this problem where one of our .050" pitch BGA's were reflowing at the wave thus causing shorts... The board was profiled with probes stuck to a QFP solder joint, an 0603, and one of the BGA joints. There was an almost 70 degree difference in peak temperature between the QFP and BGA solder joints!! The BGA joint was actually at liquidus for almost 10 seconds! | | What we needed to do was speed the wave conveyor up by 1.5 FPM and turn the chip wave off - this took about 20 degrees off the energy curve and just under the eutectic melting point (361 deg. F). I wish we didn't have to do this since there are small SMD devices on the bottom side....so the question is.... | | ....why would a BGA solder joint get so much hotter than everything else? Any of you solder experts know the answer?? | | Regards, | | C.K. | CK: The BGA is hotter because it is receiving more heat. Well duh!!!!

I'll bet virtually every ball on the BGA has a heat pipe (via) connecting it (thermally) directly to the wave solder pot and the QFP lead and the 0603 do not. In fact, the QFP leads and 0603 have a very nice 63 mil thermal buffer between them and the solder pot.

My2�

Dave F

PS CK: Check the archives, you're not plowing new ground.

reply »

| | We've recently encountered this problem where one of our .050" pitch BGA's were reflowing at the wave thus causing shorts... The board was profiled with probes stuck to a QFP solder joint, an 0603, and one of the BGA joints. There was an almost 70 degree difference in peak temperature between the QFP and BGA solder joints!! The BGA joint was actually at liquidus for almost 10 seconds! | | | | What we needed to do was speed the wave conveyor up by 1.5 FPM and turn the chip wave off - this took about 20 degrees off the energy curve and just under the eutectic melting point (361 deg. F). I wish we didn't have to do this since there are small SMD devices on the bottom side....so the question is.... | | | | ....why would a BGA solder joint get so much hotter than everything else? Any of you solder experts know the answer?? | | | | Regards, | | | | C.K. | | | CK: The BGA is hotter because it is receiving more heat. Well duh!!!! | | I'll bet virtually every ball on the BGA has a heat pipe (via) connecting it (thermally) directly to the wave solder pot and the QFP lead and the 0603 do not. In fact, the QFP leads and 0603 have a very nice 63 mil thermal buffer between them and the solder pot. | | My2� | | Dave F | | PS CK: Check the archives, you're not plowing new ground. | CK,

Dave's bang on the money - not an unusuall event as it happen's, your BGA's will have via's on the end on most if not all the ball's, normally you have a pealable mask on the bottomside of the BGA to help protect it from the heat at wave or use selective soldering pallet's, sometime's both. Your 'fix' of speeding up the belt and turning off the chip wave really aint the answer, by speeding up the belt your changing the time in the wave which is a critical one for the flux and for the components. Your peel back effect on the wave exit mightt not be as robust as it could be which is going to lead to short's depending on what's on the bottomside of the board. The chip wave should be helping get the little joit's that your smooth wave will skip if your running too fast through it. best advice.. 1. Mask the underside of the BGA's with the pallet / peelable mask. 2. If there's chip's there then switch to a double sided reflow method and selective solder any PTH lead's via a dedicated machine or proper pallet desige. 3. match the wave profile to the flux that your using and ensure your not cooking the flux too much prior to the wave. 4. Match your conveyor speed to the flux spec to get the right time in the wave, usually it's between 3 5 second's ( chip n smooth combined).

These step's work very well but you have to control your process.

JohnW

reply »

Thanks for the acvice, Dave and John. Like I stated above, I didn't feel comfortable with changing wave parameters since I knew that we'd be compromising our wave quality. I'll try the peelable mask thing, reprofile, and let you guys know what i found out.

C.K.

| CK, | | Dave's bang on the money - not an unusuall event as it happen's, your BGA's will have via's on the end on most if not all the ball's, normally you have a pealable mask on the bottomside of the BGA to help protect it from the heat at wave or use selective soldering pallet's, sometime's both. | Your 'fix' of speeding up the belt and turning off the chip wave really aint the answer, by speeding up the belt your changing the time in the wave which is a critical one for the flux and for the components. Your peel back effect on the wave exit mightt not be as robust as it could be which is going to lead to short's depending on what's on the bottomside of the board. The chip wave should be helping get the little joit's that your smooth wave will skip if your running too fast through it. | best advice.. | 1. Mask the underside of the BGA's with the pallet / peelable mask. | 2. If there's chip's there then switch to a double sided reflow method and selective solder any PTH lead's via a dedicated machine or proper pallet desige. | 3. match the wave profile to the flux that your using and ensure your not cooking the flux too much prior to the wave. | 4. Match your conveyor speed to the flux spec to get the right time in the wave, usually it's between 3 5 second's ( chip n smooth combined). | | These step's work very well but you have to control your process. | | JohnW |

reply »

| | | We've recently encountered this problem where one of our .050" pitch BGA's were reflowing at the wave thus causing shorts... The board was profiled with probes stuck to a QFP solder joint, an 0603, and one of the BGA joints. There was an almost 70 degree difference in peak temperature between the QFP and BGA solder joints!! The BGA joint was actually at liquidus for almost 10 seconds! | | | | | | What we needed to do was speed the wave conveyor up by 1.5 FPM and turn the chip wave off - this took about 20 degrees off the energy curve and just under the eutectic melting point (361 deg. F). I wish we didn't have to do this since there are small SMD devices on the bottom side....so the question is.... | | | | | | ....why would a BGA solder joint get so much hotter than everything else? Any of you solder experts know the answer?? | | | | | | Regards, | | | | | | C.K. | | | | | CK: The BGA is hotter because it is receiving more heat. Well duh!!!! | | | | I'll bet virtually every ball on the BGA has a heat pipe (via) connecting it (thermally) directly to the wave solder pot and the QFP lead and the 0603 do not. In fact, the QFP leads and 0603 have a very nice 63 mil thermal buffer between them and the solder pot. | | | | My2� | | | | Dave F | | | | PS CK: Check the archives, you're not plowing new ground. | | | CK, | | Dave's bang on the money - not an unusuall event as it happen's, your BGA's will have via's on the end on most if not all the ball's, normally you have a pealable mask on the bottomside of the BGA to help protect it from the heat at wave or use selective soldering pallet's, sometime's both. | Your 'fix' of speeding up the belt and turning off the chip wave really aint the answer, by speeding up the belt your changing the time in the wave which is a critical one for the flux and for the components. Your peel back effect on the wave exit mightt not be as robust as it could be which is going to lead to short's depending on what's on the bottomside of the board. The chip wave should be helping get the little joit's that your smooth wave will skip if your running too fast through it. | best advice.. | 1. Mask the underside of the BGA's with the pallet / peelable mask. | 2. If there's chip's there then switch to a double sided reflow method and selective solder any PTH lead's via a dedicated machine or proper pallet desige. | 3. match the wave profile to the flux that your using and ensure your not cooking the flux too much prior to the wave. | 4. Match your conveyor speed to the flux spec to get the right time in the wave, usually it's between 3 5 second's ( chip n smooth combined). | | These step's work very well but you have to control your process. | | JohnW |

Hi CK, Above recommendation from both Dave F and John W are correctly. But may I ask your process again before your found the problem solder bridging under BGA at Wave soldering. - Reflow Soldering with Clean process or no clean process? - If no clean process, how long you leave PCBA in the WIP? The reason or my above question are remaind review our process before solving the problem. I used to have very bad experience with the same criteria that you have now. In that time, I running with the water soluble process, then wave soldering and found high defect solder bridging under PBGA, 50 mils. The adjust of wave soldering parameter is not good enough to solve this problem. But masking the i/o via under the BGA are help a lot. But I still found the problem. Finally I found the major problem coming from the humidity under the PBGA & vias holes. I need to baking @ 80 degree C for 4-6 hrs before wave soldering. Two root cause I found in my process were leaving the PCBA in WIP longer than 48 Hrs and didn't plug/mask the vias. Then you try with masking under the BGA and still found problem of bridging, in the middle of the PBGA. Please try to baking again. Regards, Wirat S.

reply »