Ask the Experts
June 11, 2019
Circuit Board Bow and Twist
Board flexing has recently started to occur on one of our circuit board assemblies. This assembly has a 40 mm PoP BGA component. The BGA has an integrated heat sink.
We are using a 3 zone reflow oven. The profile worked without issues for at least 6 months. The problem with the boards bowing, or curving upwards towards the middle of the board, has just started recently.
Yet our process has remained the same and there have been no changes. Any help is appreciated.
Expert Panel Responses
There are factors both in the PCB manufacturing process as well as the assembly process that can have affects on board flatness. Some of these items can be variable (not consistent) depending on the control of a various manufacturing or assembly process.
On the manufacturing side: warpage can be introduced by changes in the material being used, the stack up of the board being changed or variations in process. Most of the process variations that could be responsible are heat related lack of control of temperature (greater fluctuations) during lamination, thermal curing of soldermask, legend, HAL process, or baking for flatness at the finished board level.
These fluctuations in temperature can typically add a stress or strain to the material that can be released when it is heated (during assembly) during subsequent processing.
Assemblers will, on occasion, bake boards prior to assembly to fix other quality problems like out gassing etc. These baking processes can also introduce strain and stress into a PCB. Another possible source is boards that are inadequately supported during a wave soldering step (more of an issue on larger boards).
The final area that can introduce warpage into a PCB is the design itself. If there is an unbalance of copper through the design (layer to layer) or unbalanced dielectric thicknesses throughout the design, these can add stress and strain to a board as well.
Trying to distribute plane layers and signal layers evenly through the build, using consistent dielectric thicknesses (at least symmetrical top to bottom) and trying to have well distributed copper on each layer during layout will help make the design less susceptible to warpage.
Director of Marketing
Matt Stevenson has over 20 years experience in the PCB industry. Serving in roles as a Chemical Lab Technician, Process Engineer, Quality Engineer, Quality Manager, and Marketing Manager. He has proven himself to be an invaluable resource.
A three-zone oven? What? That seems crazy in terms of being able to control a repeatable reflow process. If the board is bending, talk to your PCB manufacturer. They could be doing a second pressing operation, which just hides warping issues until you go through reflow.
Dr. Craig D. Hillman
CEO & Managing Partner
Dr. Hillman's specialties include best practices in Design for Reliability, strategies for transitioning to Pb-free, supplier qualification, passive component technology and printed board failure mechanisms.
Your problem with excessive bow and twist may be addressed by the direction of the warp and weft, on each dielectric layer of the glass fibers in the PWB. The warp is the direction of glass that runs length wise, under tension, during the weaving process.
The weft is the direction in which the glass fibers go around the warp fibers during weaving. Typically the warp is under more tension and the weft is under less tension. This accounts for the difference in coefficient of thermal expansion (CTE) in the x and y direction in FR4 material (14 and 17ppm/C).
What can happen is the fiber glass may be laid up with the warp of every layer in the same direction; this can make the board more prone to bow and twisting. If the fiber glass is laid up in a manor that every layer has the warp laid down at right angles to the layer above and below it then bow and twist may be reduced. You can specify the direction of the warp and weft, or have the warp and weft laid up and right angles with each dielectric layer,with the better fabricators.
PWB Interconnect Solutions
Paul Reid has over 35 years experience in bare board fabrication, quality and reliability. Working for PWB Interconnect Solutions, which does thermal cycle evaluations (IST) of representative coupons, Paul provides failure and root cause analysis of how PWBs fail. His area of expertise includes how circuit board's copper interconnections and material fails in assembly, rework and in the field, as a result of thermal cycling.
There are four main possible causes to look at:
Because the boards are bowing upward, we can rule out (3). Design cannot be ruled out as a factor, however it is probably not the main factor since the warping does not always occur.
- Design (copper imbalance)
- Fabrication (process error)
- Support issues(sagging due to gravity)
- Mechanical stresses applied (oven conveyor)
It is possible that a combination of design issues and fabrication process variation is leading to higher warpage on some lots of boards. You can check this by reviewing the manufacturing date code on the boards that are warping/not warping, and determining if it is correlated to a specific date code.
Also, look at the oven conveyor, and make sure that if it is an edge conveyor that the width is not set too tight. Even if it does not appear so at the entrance, it can be too tight internally due to non-parallel rails and due to growth of the board during heating. If the oven has a center support, make sure that the support is properly positioned and is not causing the center of the board to be elevated.
Fritz's career in electronics manufacturing has included diverse engineering roles including PWB fabrication, thick film print & fire, SMT and wave/selective solder process engineering, and electronics materials development and marketing. Fritz's educational background is in mechanical engineering with an emphasis on materials science. Design of Experiments (DoE) techniques have been an area of independent study. Fritz has published over a dozen papers at various industry conferences.
Assuming the reflow process and zone temperatures have not changed (which you should verify with thermocouples attached to various sections of the board), you must assume the PCB supplier made a change to one or more layer materials, most likely the pre-preg.
Another possibility is that the laminate layer supplier made a change and did not inform the PCB fabricator. My suggestion would be to first do a thorough calibration verification of the reflow oven temperature and speed, and once that failure mode is ruled out, conduct a detailed spec review with the PCB supplier.
No mention was made about the PCB dimensions (including nominal thickness and tolerances), but if the problem persists, there are a number of ways to maintain planarity of the board including stiffeners, and/or different PCB stack up materials and methods.
Round Rock Consulting
Following a successful 20 year career within world class OEMs and EMS providers including Honeywell, Bull Electronics, IBM, Celestica and Plexus, James founded Round Rock Consulting a business consultancy focused on supporting OEMs and EMS providers with product realization strategies.
Given that the circuit pattern and assembly processing are unchanged, bare board processing can be considered. If the manufacturer fails to consistently orient the grain direction of the laminate, variations in flatness may be realized later-on in their processing.
If so, one ill-advised "fix" might be attempted: To flatten the product under uniform weight, in ovens of a temperature sufficient to soften and relax the laminate. After adequate time-at-temp the oven is then turned OFF, the doors opened and the product allowed to cool, still underweight.
But the cause of the stress remains, now restrained by laminate that is cooled, firm and flat at room temperature. However, when the PCBs next experience high temperatures (such as those produced by your Reflow) the laminate may soften and, unrestrained by a uniform weight, yield once again to the effects of internal stress.
Robert "Bob" Lazzara
Circuit Connect, Inc.
Bob has been in PCB design and fabrication since 1976. He has held elected positions with the SMTA, is a member of the MSD Council, has served as a committee member for various IPC standards and is a Certified IPC Trainer.
Instability of the inner layers or improper layering of the board during manufacturing may be contributors to this issue. IIT manufactures board stiffeners from titanium to remedy issues like this.
They slide onto the edges of the board easily and keep the board parallel during reflow. If the inner layer instability is severe enough, warping will still be a possibility once the stiffeners are removed.
Integrated Ideas & Technologies, Inc.
Stephanie Nash is the Director of Technical Services & Marketing for Integrated Ideas & Technologies, Inc., a premier manufacturer of SMT stencils. She has been instrumental in the stencil design and technical support.
Michael J Gay, Isola Laminate Systems, USA
There are several potential reasons for bow and twist of finished PCB's some of which have been mentioned.
Copper CTE 17 ppm versus a range of 11-24 ppm x-y CTE for laminate. Low resin content dielectrics have low x-y CTE values and high resin content dielectrics have high x-y CTE. This can lead to the following problems:
- Unbalanced copper distribution in the x-y plane
- Unbalanced copper, heavy copper versus light copper distribution in the z-axis
- Unbalanced dielectric construction or stack up
- Use of dis-similar materials - hybrid construction
Even if some of the above are present, processing can either reduce or enhance the propensity of the PCB to bow and twist. The problems can generally be traced back to the lamination process. The following problems can lead to bow and twist:
- Rapid cooling from curing temperature. Especially in the first few minutes of the cooling process in the press. Common when using a cold press and cold winter water. Too rapid of cooling is the most common root cause and is easy to fix by following the laminate manufacturers recommendations. Not the average cooling rate, but the maximum rate at any moment of the cooling process.
- Improper press cycle settings resulting in wide temperature variation in the book and uneven curing.
- Press malfunctions such as failing heating element causing large temperature distribution across the platen of the press.
- Excessive force used in breakdown causing warped panels. Panels tend to get glued together and they forced apart by the operator to separate them.
If a baking process with weight is used to dewarp panels, the panels should be cooled very slowly to retain the relaxation. There are a lot of reasons not to use the method to salvage boards, but sometimes it is necessary. Always consider the potential for damaging the surface finish and reducing solderability.
All of the experts here had excellent and very experienced inputs. Here is one more: I recently had a PWB begin to show warping issues, even though nothing had changed in the process. We double-checked everything in the assembly process and then we went to the fabricator and double checked everything there. Nothing had changed, artwork was the same, copper was from the same vendor, laminate was from the same vendor, etc.
We couldn't find anything that had changed. But the fabricator did offer up one little tidbit of info; he had changed the layout of the laminate "flat", in order to get more individual PWBs out of each FR-4 "flat" and reduce scrap. So the original 12 of the 16 boards produced from each flat had been rotated 90 degrees to create more room for 4 additional PWBs along one side of the flat, and those were positioned 90 degrees from the rest.
That was all it took to give us 12 warped PWBs out of every 16, the direction of the weave. When the fabricator went back to the original 3 x 4 array, the problem immediately disappeared. Usually, however, it is never a single issue causing the warpage. It is typically some perfect storm of two or more of the factors mentioned from both the fabrication AND the assembly processes. If it ain't broke, keep the heck away from it!
Richard D. Stadem
Richard D. Stadem is an advanced engineer/scientist for General Dynamics and is also a consulting engineer for other companies. He has 38 years of engineering experience having worked for Honeywell, ADC, Pemstar (now Benchmark), Analog Technologies, and General Dynamics.