Over the last few weeks I have had a number of questions on the use of double-sided mirrored ball grid array (BGA) packages. In reality this technique is being used in the industry due to the pressure for space and speed of  product operation. This may also be contributing to the increasing interest in the use of buried components, also referred to as embedded passives.

In simple terms, mirror imaging of components has been around for many years, if you cast your mind back to early memory modules? Basically the assembly process consists of two printing, placement and reflow operations, just standard double-sided reflow process. Mirror imaging means placing components directly opposite each other on two sides of a printed circuit board. The component footprints on both sides of the board can remain the same size. The latest generation of this concept is stack packaging.

With many large processors there is a need for a range of passive components to be placed as close to the device terminations as possible. Taking up all the usable area under the part is restrictive. In the case of reduced area or peripheral BGA devices it is possible to mount very small passives under the parts, just like in the past when chips were placed under plastic leaded chip carriers (PLCC).

The use of mirror imaging technique forces many design engineers to use micro via and via in pad/blind via technology, which will in turn increase board prices. Straight through via holes can be very restrictive as the pin out count increases. The size of the through hole via may be space limiting; with thicker boards over 3 mm there is increasing strain on the copper platting. Via in pad can be successfully used for BGA and chip scale package (CSP) products provided the design engineers takes the complete process into consideration. Via in pad can introduce inspection concerns, voiding due to outgassing from open vias; however, there are solutions for each of these issues, like pulse plating, via filling or capping.

Double-sided reflow of BGAs is perfectly achievable when correctly processed on an edge conveyor equipped reflow oven or using standoff pallets on a mesh belt system. Engineers should never put double-sided boards upside down on a mesh belt – yes, some people do it, or try it! BGAs placed and reflowed on the second side will not drop off due to weight, even up to 20 grams; there is, however, some noticeable joint elongation if the joints are in a liquid state. Most standard plastic ball grid arrays (PBGAs) without heat sinks are less than 10 grams and see less than 10% joint stretch/elongation. This is based on comparing BGA joints on side one, reflowed twice. It is not common to see a different standoff height between the corner joints and the centre with some designs.

Many BGA and quad flat packages (QFPs) are lost due to their size, not necessarily weight; they simply flex against the board, peeling off rather than falling. If solder joints on the base of the board go into a liquid state, as many obviously do, board flex or warp can be the real production issue. It only takes a small movement of the board to cause pressure on the joints. Misalignment on the reflow oven centre board support is also a potential reason for lost parts. The centre board support positioning is critical; care needs to be taken in checking the position at the entrance, exit and centre of the oven. Convection and now vapour phase reflow has become increasingly used for lead-free manufacture since the introduction of the RoHS legislation.

As with any large mirror image parts, the first side parts previously reflowed can affect the temperature profile on the second side of the board. This normally shows up as increased voiding in joints on the second side due to the reduced time in a liquid state for the paste. This is not significant but often noticeable during routine x-ray inspection or when automated inspection has been introduced. Some designs only mirror image part of the BGA ball array or one line of pins. In the case of a QFP only one set of joints would be affected in terms of reflow duration.

There is a lot of conflicting reliability data on double-sided, mirror-imaged thin small outline packages (TSOPs) and area array parts, and this is probably due to the range of different parts in industry. It has been shown that if you compare a mirror imaged part and then move the parts so there is a reduced overlap, the reliability increases. The reliability results obtained normally revert to what is expected from a BGA without any part on the opposite side. The difference is probably due to the stiffening effect and the changes in thermal expansion. It may also be due to flexing between all three surfaces, two components and the PCB during environmental testing. However, the joints that fail are normally in the same locations as on a single sided BGA. The failure sites are on the corners and directly around the edge of the silicon die. Most of the data reported is related to large BGAs in excess of 1,000 termination points on nickel/gold boards.

Inspection of BGA joints can be conducted with optical inspection; they can also be successfully examined with x-ray. This is provided the board manipulator or x-ray head can be tilted so that you are not just viewing directly down through the substrate. If this is not the case, the view available of the printed board and device interfaces will be lost and difficult to interpret. It is fair to say that even on a high end x-ray system with tilting capability, mirror imaged BGAs can still be a challenge for the inspector. With more complex area array devices like ceramic ball grid array (CBGA), the internal interconnections can add to the challenge.

Rework of any area array component is more demanding when setting up a process for removal and must be correctly profiled. The temperature profiling step is often missed by many engineers through time constraints; however, it should always be done correctly. When profiling a BGA, a combination of thermocouples should be placed on and under the component; temperature rise must also be monitored adjacent to the body.

Close proximity components adjacent or on the underside of the board can tend to reflow again as you cannot just avoid pre-heat. Experience has shown that closely spaced and underfilled parts can be lifted slightly from the surface of the board due to the expansion of the underfill leading to intermittent joints. Obviously you want to reflow and remove the topside parts without reflowing the bottom. Care must be taken to provide the correct pre-heat in this process to avoid flexing the board. Just like reflow any large components on the bottom side of the board are far more likely to be lost due to flex and local vibration.

Consideration should be given to shields for adjacent parts to hold down the temperature on surrounding joints; even if it is only 10°C, it will benefit the process. Simple heat sinks on components, deflecting Kapton tape, thin metal shields or even cooling blocks on high reliability products is currently a reality in the industry today.

This column appeared in Global SMT & Packaging magazine issue 7.1 - January 2007.

Bob Willis is a process engineer providing engineering support in conventional and surface mount assembly processes. He organises production lines and seminars for suppliers at exhibitions and also provides seminar and workshops world wide.  Bob has one of the largest collections of training videos, interactive CD-ROMs and training material in the industry. For further information on how Bob may be able to support your staff or answer your process problem contact him via his web site www.ASKbobwillis.com

Bob Willis will be running the “Process Advice & Defect Clinic” a special feature at APEX 2007 in California sponsored by IPC and Global SMT magazine. Bob will be providing FREE advice on process and production issues including lead-free at APEX 2007, visitors will be able to bring process problems and printed boards for examination and discussion during the show. Engineers intending to visit the show can also email questions to be answered by Bob at APEX by visiting the web site www.goipcshows.org

Keywords : Bob Willis, BGA, ball grid array, BGA design, mirror image design, double-sided mirrored BGA

Average user rating

   (0 vote)