This is the tenth section in the back-to-school series for PCB Designers and those who may want to know more about it.
- Flex Circuits
- Flex Circuit Materials - A World of Their Own
- Beyond Polyimide, Copper, Adhesive, Coverlay, Stiffeners, and Solder Mask
- Button Plating vs. Panel Plating
- Bend Regions - The Main Point of Being Flexible
- Flex To Assemble
- Dynamic Flexing
- Rigid-Flex Designs
This is a whole different thing. A lot of flexible printed circuits (FPCs) are nothing more than a substitute for a ribbon cable with bespoke architecture to get a bus from one board to another. There would be a connector at either end and a number of traces and perhaps a power line between the connectors. Even this simple form has a lot of aspects that are particular to FPCs.
Easily, the biggest difference between rigid boards and flexes is in the material call-outs. Polyimide is the dielectric standard but there will be other dielectric materials in use on the same flex. The stack-up diagrams go into detail on these elements.
Adhesives are used to bond the various laminated materials. Pressure-sensitive adhesive (PSA) is commonly used to attach the coverlay and polyimide or FR4 stiffeners. When stainless steel or aluminum stiffeners are used, a thermoset adhesive (TSA) is usually called out. The adhesive in this case would also be electrically conductive as it forms part of the thermal path. Metal stiffeners have a pullback from the outline edges while polyimide or FR4 stiffeners will be the same size as the rest of the material. The advantage of metal is that it is stiffer per unit of thickness but it adds cost.
Coverlay is the term for the polyimide material that protects the traces from the elements. It helps keep the traces from lifting or getting damaged by abrasion. Black is a popular color for coverlay and solder mask. The solder mask will only be used in the stiffened parts of the flex circuits where parts are soldered down. The area that transitions from coverlay to stiffener and solder mask is a complex region with some carefully controlled overlap of the materials.
Figure 1. Image Credit: Hirose - When is a Two-Layer Not Really Two Layers? When it’s a Flex.
A zero-insertion-force (ZIF) connector is a popular option for connecting a flex to a rigid board. It is a little different from a usual edge connector. The gold fingers are not simple rectangles set side-by-side. It is typical to have two rows that are staggered. The mating connector has a hinged locking mechanism to keep the ZIF connector in place.
The 25 micron (one mil) thick polyimide base film is very common. Two mils is also available. The location where the FPC changes from stiffened area to flexible area is called the transition zone. Pulling the via back 0.4 mm from either side of the transition zone should keep you from getting a call/question from the vendor.
Also, note that the coverlay tucks under the rigid section and comes to an end in the transition zone. The overlap goes on for 0.4 mm. The solder mask overlaps the coverlay by 0.2 mm when a connector or other component is soldered down. Of course, these numbers and material types will vary with different manufacturers. It’s even more important to get out in front of the stack-up when it comes to flexes.
Another glue-down covering can be added over the top of the coverlay for EMI suppression. High speed/high frequency traces will radiate through the ground mesh. When you use the EMI suppression film, the coverlay has to be notched to reveal areas of where the ground mesh is exposed. Putting a patch of solid copper with vias is useful to create a good contact between the EMI shield layer and the ground. There should be at least one exposed area near each connector. A longer stretch of flex would have some openings enroute but not in the bend regions.
Yet another technique uses double-sided tape rather than hardware to keep the flex in place. Adhesive foam is another material that will fill gaps. Both of these materials come with a peelable film over the outward-facing adhesive so that it can be removed when it’s time to stick the flex into its final location.
“ ...a bend table that shows the desired radius and number of times the area will be bent back and forth will be part of the fabrication drawing.”
When it comes to conductors, rigid boards can get by with electro-deposited (ED) copper. The grain structure is vertical so the surface is rough and the plating is too brittle for reliable flexes. Rolled annealed (RA) copper has a smoother horizontal grain which may be good for approximately 10,000 bend cycles. The automotive sector sees a lot of potential vibration which compels them to call out “super flexible” HA copper in the material stack-up section. In any case, a bend table that shows the desired radius and number of times the area will be bent back and forth will be part of the fabrication drawing.
Unlike rigid stack-ups, the flex stack-ups call out the plate-up copper as a separate item. Plate-up is always done as a means of plating the via barrels whether it is a rigid or flex design. This means electro-deposited copper which we may not want on the traces in the flexible areas whether then bend or not.
To get copper into the vias without getting it everywhere, a special fabrication step is included where they mask off everything except the vias. That’s the button plating while doing a plate-up of all of the metal prior to etching is known as panel plating.
Some FPCs remain flat in their installation and usage. Most have to be bent to meet different connector locations. Some have to be flexed during usage. Depending on whether it is a one-time flex to install or a dynamic flex, there are design considerations to follow. The hinge on any laptop will be a very complex part of the solution since the power and data have to get from the graphics processor (GPU) to the display screen.
Not only that, most laptops include a camera module at the top of the screen, maybe a flash or other sensors including a touch sensor for those laptops that convert to a tablet. The various bit streams have to flow through a number of flexes from one side of the case to the other. The repeated bending causes mechanical stress every time the lid is opened or closed. The 360 degree bends are the most stressful. The specific area that has to be actuated over and over is called a dynamic bend region. Special rules are in play.
The guidelines are somewhat relaxed depending on the minimum radius of the bend. A one-time bend that is over a small 180 degree arc is typical so that the flex circuit can double over on itself to save space. That type of bend will take more effort than a mild bend over a large radius.
The fabricator can pre-bend the flex so that it has a natural tendency to bend in the required direction. In any case, there is a ratio of flex thickness to minimum inside bend radius. A safe number is ten-to-one and the most radical bend would be twenty-to-one. This will depend on the specific stack-up. A ballpark estimate for the thickness of a three-layer FPC including top and bottom coverlay will be 200 microns or 0.2 mm. The safe bend radius will be 2 mm measured from the inside of the curve.
The first rule of the bend region is to have no vias in the region or even the immediate vicinity. Vias are like little I-beams through the z-axis that act to stiffen the flex. Flexing them against their will causes them to crack and fail. When an FPC is bent, there is compression of the inner side of the bend and elongation on the outer side. Something has to give. The most reliable place for routing is right in the middle of the stack-up.
Figure 2. Image Credit: Author - A USB Type C connector at the end of a flex.
A really long bend region might require a few “redundant” ground vias for the sake of EMI suppression. Special coatings can be added in lieu of the vias. The thing is that the more material you have, the stiffer the flex becomes. The best compromise isn’t always clear.
The second thing is to consider the direction of the bend with respect to the direction of the traces. The traces should run straight through the bend region. A wide trace should be necked down or divided into multiple thin traces that join back together after the bend. Solid shapes are generally discouraged in flexes but if used, should convert to a mesh through the bend regions.
For the sake of impedance control, it is common to use a mesh with a certain ratio of metal-to-open area. The cross-hatch usually resembles a chain-link fence. Depending on the outline, the direction of travel may put the traces in alignment with the mesh lines. So, if a flex has a branch that launches on a 45-degree angle, it would be wise to consider rotating the flex to a 22-degree angle or whatever is appropriate for the outline geometry.
The same is true for multiple trace layers in a flex. Traces should not run right above other traces. Stagger one layer so that it runs in the gaps of the other layer even if there is a mesh in between the two. High speed or any other controlled impedance on a flex circuit usually involves three layers. The outer layers are ground mesh and the inner layer is for signals. A failure in the ground mesh is less likely to cause the entire flex to fail. The outer layers are the stressed layers while the center one isn’t stretched or compressed as much.
To get higher signal density using a narrow strip of flex material, the traces can be run on the outer layers with a mesh between them. Ultimately, the mesh can be deleted and you would have just the two layers or even one side for traces and nothing on the other side. If the traces are going to be exposed like that, the enclosure of the unit has to account for radiation into and out of the box.
More measures are required for flexes that are in motion. Obviously, you want the largest bend radius that can be managed. Picture the area inside the laptop hinge to know what we’re up against. Go with a wide flex rather than a thick multilayer flex. Adhesiveless coverlay is slightly thinner - and a winner!
The changes in linewidth have a gradual taper and come before and after the flex region, not right at the border. Just say no to vias in a dynamic flex zone. It may help to cut slots in the FPC to reduce the mass where the flex sees a lot of bending.
The edge of the flex material should have a copper line that serves as a tear stop to prevent any cracks from propagating to the signal traces. This is especially critical on any inside bends of the outline. It is common to have a flexible area just after the connector where the outline gets wider to accommodate the connector. That creates an inside bend at exactly the wrong spot. All of the traces are being funnelled into the flex zone.
The border between the stiff area for the components and the flex area is a high stress point. This is true whether there is a bend-region in that location or not. Treat those boundaries as if they were dynamic bend regions. A bead of epoxy along the edge of the stiffener will act as a stress relief.
The above design guidelines still apply, Fewer layers is better. Less copper is better. One way to mimic fewer layers is to not glue all of the layers together. When there are four or more layers in a flex, consider a loose leaf approach to the stack-up in the bend regions. A workaround for a high layer-count is to create flex cores on multiple levels so that two or three pieces of flex are embedded between the rigid materials. Using three stacked flexes with two or three layers each is better than a six-layer flex region when it comes to pliability.
Figure 3. Image Credit: Polar Instruments - A rigid/flex with two stacked flexible subsections.
I would be careful about how close the vias are located with respect to the rigid to flex transition area. Similar to a regular flex, pulling the via back 0.4 mm from either side of the transition zone should keep you from getting a call/question from the vendor. Also note that the coverlay tucks under the rigid section as it did the stiffener.
Figure 4. Image Credit: OrCAD - Subtle distinctions in the transition zone.
When you have a trace running over/under/between a mesh layer in the flex region, the mesh should continue into the rigid section. Not to make the rigid section more flexible but to keep the same impedance geometry along the path of the traces. You could make the mesh global on those layers but it’s really only needed over and under the controlled impedance traces that run across the flex. If you know how to do flex and rigid boards, then rigid-flex is just about combining the two in a stack-up and managing the space where it turns from one to the other.
- IPC-6013 Qualification and Performance Specification for Flexible/Rigid-Flexible Printed Boards
- IPC–2223, Sectional Design Standard for Flexible Printed Boards