Designing the Stiffness Out of a Flex or Rigid/Flex Circuit

John Burkhert

Flex circuits are floppy by nature but there are times when the circuit calls for a higher current rating along with the typical connections. More current requires more copper so meeting the specification with any margin to spare is likely to stiffen up the flex. Oftentimes, a bit of backbone is just fine. A lot of flexes are little more than a bespoke flat cable. Perhaps a stacking connector at one end and ZIF (Zero Insertion Force) gold fingers at the other end. If the flex is used because it has a low profile to fit in a confined space, it will be fine.

Figure 1. Image Credit: Hyper Edge - Diagramming the potential issue with bending a flexible printed circuit

On the other hand, if the flex circuit is formed around contours there will be areas under stress of compression on the inside of the bend and tension along the outer layer. Flexes are thin so it is not a huge difference but still, the layer that sees the smaller bend radius on the inside of the bend wants to delaminate and buckle rather than actually shrink or stretch its length. Bending a flex is a test of the adhesives that bind the whole thing together.

Less Copper Equals More Flexibility

In a typical case, three metal layers provide a happy place in the center where dainty traces can be routed between a pair of reference planes. The center layer is not as affected by bending, it has what is known as a neutral bend axis.

When impedance is not a concern, one or two layers simplify things greatly. A single layer can be somewhat balanced by layers of coverlay and/or EMI shielding materials. Economizing on copper is the oldest trick in the book for pliable flexes.

For a single layer flex the best case is a bend radius that is no less than 6X the overall thickness of the flex. Add a second layer and not only is the flex going to be thicker in general, the radius to thickness factor increases to 12X as suitable ballpark bend radii. Cut back on layers and go for a wider flex if that makes sense. Dynamic flex situations call for single layer flexes as a rule, maybe two sided but not in my experience. The minimum allowable bend is an order of magnitude greater than with flex-to-assemble technology.

Vias Add Stiffness and Can Become a Reliability Concern

The second oldest trick is to economize on vias. Leave them off of the flexible portions using them exclusively in the stiff parts. There may be specific bend zones defined and you would not want them there for the sake of reliability. It really depends on the use case.

One time we made an exception to the “bend zone is a via-free zone” rule was when the radius was based on the average human forehead. Flex to install means just that. Dynamic flexing is when the flex circuit will be articulated throughout its service life. The two cameras that were used for eye tracking for augmented reality had a minor one-time bend when it went into the headset. All of the vias were connected to the ground net and noted as “redundant”. The vendor still raised a technical question about it.

Vias are basically I-beams that strengthen the flex against forces that want it to bend. Forcing them to comply raises stress that leads to via-failure. A simple rule is to get all of the signals sorted out in the rigid areas so that vias are not required except possibly as ground vias in the case of multi-layer flexes. One of the main features that distinguish high reliability is the size of the capture pad for the via. Flexes will benefit from larger pads if you must go that way.

Story Time: What About High Layer Count Flexes?

An example I keep running into is the USB type C connector. They are popular on laptops, phones, AR/VR headsets and yes, the action cameras on my plate these days. The first time I implemented a USB-C connector was on the Pixel laptop circa 2015. It didn’t go well. To start with, it was a six-layer flex which would be a challenge as far as flexes go. Multiple lanes of “Super-speed” differential pairs plus enough power to meet the charging requirements add up, especially if you’re aiming for good signal integrity as well.

Figure 2. Image Credit: Author - Implementing USB-C on a flex circuit is a challenge for the mechanical tolerances as well as the stiffness of the FPCA

The six layer flex was not meant to bend at any point in time. It was designed as a link between the main logic board and the edge of the laptop case. Everything fit together nicely and the computer worked fine. Fine lasted until we subjected the laptop to the drop test. Yes, there is such a thing! Hold the product three feet above a concrete surface, let go and literally let the chips fall where they may.

The gravity games were unkind to the stacking connector end of the flex assembly. The stiffness of the flex added to the misery as it was determined to be the reason the connector would not stay connected. We fixed the problem by rerouting a mess of traces to make room for a bracket to prevent the connector from popping off. It could have been worse.

We also replaced the full ground planes with a mesh. Power was run on multiple layers with a different lane for each trace so that they were not stacked. Thinner adhesiveless materials and rolled annealed copper were explored and found to be more costly and harder to get - even in 2015. The mechanical revision along with the other remediation efforts were deemed sufficient. Special note: I do not recommend initiating a drop test on your electronic devices; just understand that we allow for the possibility that you might drop it by accident.

Flexible Circuit Stack Up Techniques For Greater Flexibility

You may be familiar with the typical stack-up of a flex circuit. One of the unique items in a flex stack-up is the call-out of additional copper plating. That extra copper can be selectively applied leaving the bend areas with only the copper that was originally clad to the polyimide. As with any selective plating process, this is a cost driver with potential lead-time implications.

While we’re on the subject, the glue that binds the layers together is responsible for a lot of the stress as the layers compress or elongate through a bend. A section without the glue becomes more like loose leaf paper than a laminated flex circuit. An advanced method for rigid flexes builds on this property by using multiple flex cores sandwiched within the rigid sections. I have not done this yet but I can see it on the horizon.

Figure 3. Image Credit: EPECtec - A rigid/flex example where layer pairs are not bound together.

The area where the flex extends beyond the rigid section is a natural location for bad things to happen. The flex goes from being completely supported to hanging free. That ‘knife-edge’ transition zone should be given a nice bead of epoxy as a stress reliever. If a bend is going to start right there, it may be advisable to have the flex pre-bent at the factory. Shipping and handling are more difficult when the flex is not flat. It usually takes a disaster before this kind of thing is implemented but there you go.

In Summary:

Reduce layers to reduce overall thickness
Rolled Annealed (RA) copper flexes better than ElectroDeposited (ED) copper.
Eliminate vias or use Class 3 geometry if vias must be placed in the flex zone.
Use mesh rather than solid copper for reference planes. Slots in the planes help if mesh is not viable.
Be mindful of stress risers, the material will eventually fall apart under constant or repeated tension.
Use destructive testing to find weak points and bolster as necessary.

Looking at flexes from a system level, they open up opportunities to reduce the footprint. Getting there without creating failure prone devices takes solid engineering. Creative packaging will require a compromise that balances cost, performance and scheduling goals. I’d be the one saying “I can take space but I can’t make it. Figure out what you can reliably bend before committing to that smaller overall package. At best, we can endeavor to optimize space. It has to be easier than trying to bring space into existence! Happy flexing.