Printed Circuit Board Assembly & PCB Design SMT Electronics Assembly Manufacturing Forum

Printed Circuit Board Assembly & PCB Design Forum

SMT electronics assembly manufacturing forum.


CCGA

Earl Moon

#15664

CCGA | 29 May, 1998

Got handed an assignment for which I'm not totally qualified. Don't expect this forum to had me all the answers. Some answers are obvious as the device types are fairly self explanatory. As everyone knows, IBM had problems with their ceramic BGA device types because of the interface layer as Kapton and the resultant TCE and Tg issues. It appears, without asking the "bluest one," they have fixed the problem with flip chip. I just want to know how many of you supersonic types are doing this. I know the #2 contract assembler is working with blue to get it done. How about you guys and gals? Earl Moon

reply »

Earl Moon

#15665

Re: CCGA | 30 May, 1998

| Got handed an assignment for which I'm not totally qualified. Don't expect this forum to had me all the answers. Some answers are obvious as the device types are fairly self explanatory. | As everyone knows, IBM had problems with their ceramic BGA device types because of the interface layer as Kapton and the resultant TCE and Tg issues. It appears, without asking the "bluest one," they have fixed the problem with flip chip. | I just want to know how many of you supersonic types are doing this. I know the #2 contract assembler is working with blue to get it done. How about you guys and gals? | Earl Moon Ok! I found some interesting things talking to Solectron that backed up IBM's stuff on CBGA (ceramic ball grid array) and CCGA (ceramic column grid array. Also, IBM (of course to no one's surprise but me) has a huge spread on the stuff. It all comes back to this. IBM says CCGA device types solder the same as normal BGA types - with some tweaking. Does it really? I've been through this, as I am sure some of you have, but don't want to re-invent the wheel with modern reflow equipment and attendant elements. The following is not a paid Clinton political announcement. It is, as IBM see's it, the way to produce ASICS and to allow us to assemble their packages without solder joint defect or "ball dismemberment." IBM's CBGA and CCGA packaging products deliver excellent thermal performance by their ability to exploit heat extraction from the back of a chip as well as through to the card, Junction-to-air and junction-to-case resistances can vary from 1.2 to 20 degrees C/W for a standard capless 32.5mm CBGA with a 10mm die. These resistances drop to <0.1 to 5 degrees C/W when the module is capped and a heatsink is added. Thermal resistance depends on package and die size, airflow, type of heatsink, and whether the module is capped or not. The choice of CBGA vs. CCGA is application dependent, driven mainly by system reliability objectives and use conditions. The CBGA package will usually meet most computer system reliability needs for I/O counts below 624 (or 32.5mm package size). Demanding applications where the on/off cycles are high (eg. 4000), or the I/O count is high (eg. 1088) will require the column interconnect to accommodate the thermal mismatch strain between the ceramic package and organic card. From an interconnect reliability viewpoint, CBGA modules are qualified to carry up to 100 grams of heatsink mass while the CCGA limit is approximately 55 grams. Heatsink attach adhesive may however, dictate a lower mass limit depending upon the type of adhesive system used. For high power systems that may drive large, heavy heatsinks, alternatives to direct heat sink attach may become appropriate, such as supporting the heatsink through the card. Card assembly procedures for CBGA and CCGA modules follow standard BGA surface mount assembly processes. Obviously, allowances should be made for thermal mass differences between ceramic and plastic packages when determining reflow profiles using mixed card layouts. As with all BGA components, attention to card design, process definition, process monitoring, and process control are all key variables that lead to successful assembly. Rework is not trivial and may need more care than initial assembly. However, the need for rework is far less than what is typically encountered in quad flatpack assembly. In fact, recent card assembly data shows the level of CBGA solder defects (per lead) to be at 2-5 ppm, at least 10 times better than typical 0.5mm QFP results. CBGA and CCGA packaging products are gaining wide acceptance throughout the microelectronics industry for a range of device markets covering ASICs, microprocessors, memory, etc., with IBM planning to produce several million modules this year alone. Many system developers are currently designing ASICs from IBM with CBGA and CCGA packages. Again, I thank you. IBM thanks you. Solectron thanks you as they continue to buy more and more IBM real estate in NC, FL, and other geographically desirable locations, and as their stock climbs well past buy recommendations of top Wall Street analysts. Still, do any of you folks have any solderability and reliability data on the stuff. Sincerely, Earl Moon

reply »

D.Lange

#15666

Re: CCGA | 1 June, 1998

| | Got handed an assignment for which I'm not totally qualified. Don't expect this forum to had me all the answers. Some answers are obvious as the device types are fairly self explanatory. | | As everyone knows, IBM had problems with their ceramic BGA device types because of the interface layer as Kapton and the resultant TCE and Tg issues. It appears, without asking the "bluest one," they have fixed the problem with flip chip. | | I just want to know how many of you supersonic types are doing this. I know the #2 contract assembler is working with blue to get it done. How about you guys and gals? | | Earl Moon | Ok! I found some interesting things talking to Solectron that backed up IBM's stuff on CBGA (ceramic ball grid array) and CCGA (ceramic column grid array. Also, IBM (of course to no one's surprise but me) has a huge spread on the stuff. | It all comes back to this. IBM says CCGA device types solder the same as normal BGA types - with some tweaking. Does it really? I've been through this, as I am sure some of you have, but don't want to re-invent the wheel with modern reflow equipment and attendant elements. The following is not a paid Clinton political announcement. It is, as IBM see's it, the way to produce ASICS and to allow us to assemble their packages without solder joint defect or "ball dismemberment." | IBM's CBGA and CCGA packaging products deliver excellent thermal performance by their ability to exploit heat extraction from the back of a chip as well as through to the card, Junction-to-air and junction-to-case resistances can vary from 1.2 to 20 degrees C/W for a standard capless 32.5mm CBGA with a 10mm die. These resistances drop to <0.1 to 5 degrees C/W when the module is capped and a heatsink is added. Thermal resistance depends on package and die size, airflow, type of heatsink, and whether the module is capped or not. | The choice of CBGA vs. CCGA is application dependent, driven mainly by system reliability objectives and use conditions. The CBGA package will usually meet most computer system reliability needs for I/O counts below 624 (or 32.5mm package size). Demanding applications where the on/off cycles are high (eg. 4000), or the I/O count is high (eg. 1088) will require the column interconnect to accommodate the thermal mismatch strain between the ceramic package and organic card. From an interconnect reliability viewpoint, CBGA modules are qualified to carry up to 100 grams of heatsink mass while the CCGA limit is approximately 55 grams. Heatsink attach adhesive may however, dictate a lower mass limit depending upon the type of adhesive system used. For high power systems that may drive large, heavy heatsinks, alternatives to direct heat sink attach may become appropriate, such as supporting the heatsink through the card. | Card assembly procedures for CBGA and CCGA modules follow standard BGA surface mount assembly processes. Obviously, allowances should be made for thermal mass differences between ceramic and plastic packages when determining reflow profiles using mixed card layouts. As with all BGA components, attention to card design, process definition, process monitoring, and process control are all key variables that lead to successful assembly. Rework is not trivial and may need more care than initial assembly. However, the need for rework is far less than what is typically encountered in quad flatpack assembly. In fact, recent card assembly data shows the level of CBGA solder defects (per lead) to be at 2-5 ppm, at least 10 times better than typical 0.5mm QFP results. | CBGA and CCGA packaging products are gaining wide acceptance throughout the microelectronics industry for a range of device markets covering ASICs, microprocessors, memory, etc., with IBM planning to produce several million modules this year alone. Many system developers are currently designing ASICs from IBM with CBGA and CCGA packages. | Again, I thank you. IBM thanks you. Solectron thanks you as they continue to buy more and more IBM real estate in NC, FL, and other geographically desirable locations, and as their stock climbs well past buy recommendations of top | Wall Street analysts. | Still, do any of you folks have any solderability and reliability data on the stuff. | Sincerely, | Earl Moon Earl, No.

reply »

IPC Certification Training Schedule IPC Questions and answers

Large PCB Dispensing System