Blog - Ruminating Rigid Flex - Part 5
Applications and examples of rigid flex PCBs design rules.
The more I play with, look at and research rigid-flex PCB technology, the more cool applications I keep finding. Last week I had the pleasure of attending the exhibition at this year’s PCB West conference, in Santa Clara CA (and with that, meeting a few fine folks from the Altium user community and seeing some old friends too!) What I expected to find were a few board fab houses who dabbled in Rigid-Flex and a few who “specialized” and could offer me some tips, hints and tricks.
I have to say, this has come a long way in the last few years, and I was blown away by the plethora of flex circuit fabs now available in sunny California alone, not to mention the quality, precision, and layer counts now available for flex circuit and rigid-flex PCB design and fabrication.
This is awesome.
So in this week’s blog, I want to share with you all a few of the neat tricks and application ideas learned. The best part is, there are plenty of board houses that can do these things and the number is growing every year.
Dynamic Flex Ideas
There are two basic reasons for designing a flex circuit into your product: to build a compact and efficiently assembled device, or to make the circuit dynamically integrated with the mechanical function of the product. You may, of course, lean on both of these reasons for justifying the use of flex circuits. In this category let’s look at a couple of examples and the issues that spring to mind when designing rigid-flex circuits of this dynamic type.
A very typical dynamic flex application - this is the sort of assembly that would be used in a 3D printer or CNC machine head. Naturally, this would be laid along the X oriented gantry, and the z-axis tool head travels along it. This is only showing 2-axes of movement here, as the gantry itself would move in the Y orientation.
The total length of flex is the most extreme end-distance needed in addition to the corners and bends.
The corner that sits behind the z-axis moving tool head would be adhered to the x-axis shuttle that moves along the gantry (probably on sleeve bearings). The ends would have stiffener added for the termination of the flex.
For this type of application it’s best to stick to single layer Rolled Annealed copper and keep the bend radii as large as practically possible, to give a good product lifespan.
In some cases, flex circuit with stainless steel might be considered for longevity.
Figure1: Initial design of gantry flex.
Fab Consideration: Panelizing
The example above raises a good question about fabrication and cost. Using a right-angled L-shaped circuit like this we could - for argument’s sake - fit six identical circuits on a fabrication panel. This results in roughly 50% waste of the panel space, and if components were to be mounted on this particular flex circuit, also add to the tooling cost and time. An example panel made from this particular flex circuit in an embedded board array is shown below in figure 2.
Figure 2: Embedded Board Array panelization of the CNC gantry flex circuit.
The good thing about flex is if we use the right materials and plan the overall assembly right, we can also create low-radius installation folds. This is a good alternative to using curved flex circuits as the one shown above, but only in certain circumstances.
Figure 3 shows the same design done again but using a creased 45° fold to replace the 90° corner of the previous version.
The fold is suitable in this case because that part of the flex circuit is going to be adhered to a larger rigid mechanical body, and therefore it should not suffer excessive fatigue.
By doing it this way, the cost will drop significantly, and the ease of tooling for pick-and-place greatly improved. However you may have this counteracted by having to place components on the opposite side at one end of the assembly, due to the fold.
Figure 4 shows the same panel, with the equivalent length folded flex panelized. This doubled the panel yield!
Figure 3: Gantry flex re-designed for folding.
Figure 4: The Same panel size - the folded flex design doubles the per-panel output yield!
Planning the Layer Stack
For a pure flex circuit (as opposed to Rigid-Flex) the layer stack planning is simplified of course. However, there are still going to be anchor points for the panel, and most flex circuit designs require the placement of stiffeners in areas where there are components mounted or the circuit is terminated. Figure 5 shows the layer stack scheme for the gantry above, where the area with stiffener is modeled by a “rigid” stack which is locked in 3D in the PCB editor.
Figure 5: Layer Stack configuration for the flax circuit gantry example.
Figure 6: PCB Outline of a rotational dynamic flex design.
Take a look at figure 6, and notice the use of horizontal work guides in the PCB editor - this enables accurate design of the board outline based on the curved circumferences of the in-situ flex circuit sections. It also allows for exact placement of flex circuit bending lines in the Board Planning Mode of the PCB editor, which leads to accurate flex circuit bend simulation in 3D mode.
In this example, a stepper motor is to be mounted to an assembly such that the motor and it’s control Printed Circuit Board will be in motion, and the shaft will be stationary. The flex circuits are designed to be terminated at the extreme ends to a fixed base assembly, and folded into a cylinder shape, doubling back to allow bi-directional movement.
The 3D mode view of this design is shown with figures 7 & 8.
Figure 7: 3D View of the Rotating Stepper Control Board. Longer “arms” would allow greater than 360° rotation of the motor and it’s control board.
Figure 8: The fully folded view of the assembly, including the 3D body of the stepper motor.
In figure 8, I’ve annotated the movement arrows and anchored flex-circuit terminators to give you an idea of the movement.
This kind of arrangement makes it relatively easy to achieve greater than 360° of rotation.
This example is hypothetical and shows a stepper motor, though this kind of design would be well suited to rotary sensor applications.
Static Flex Applications
Planar Magnetics (Transformers and Inductors)
The use of flex and rigid-flex circuits for integrated planar magnetics is rising in popularity. In fact, in doing my own repair of my 43” LCD TV about a month ago*, I was examining the backlight inverter board and saw a neat row of step-up DC-DC regulators that used flex-circuit for the transformer windings. The turns were formed by a rolled flex circuit of the form shown in figure 9. These were incredibly compact as a result.
*It turned out to be an SMT fuse that had blown. In retrospect I wish I had taken pictures to show you here...
Using flex-circuits for planar magnetics has some distinct advantages - the PolyImide film comes in thicknesses that allow very high isolation of windings, as well as it’s high temperature stability which makes it suitable for hot enamel potting processes. From a loss standpoint; using etched copper traces requires more width of course, but can easily reduce eddy-current losses due to the thinness of the dielectric traces mitigating skin effect.
Figure 9: The un-rolled solenoid turns of a four winding inductor.
A more elaborate entry and exit scheme could be used to overlap the end of each winding with the beginning of the next. This could be done to increase the number of turns versus simply having multiple separate windings, as shown in figure 10.
Figure 10: The rolled inductor windings.
18 Layers for the price of 2
The natural extension of this concept is to include some flex layers in your converter design with the intent of folding them up over each other. In the example shown in figure 11, a 2-layer flex circuit transformer design is shown, where a single E18 planar ferrite core protrudes through cutouts in the Printed Circuit Board. This idea could be arbitrarily extended (albeit with practical limits of the thickness of the final folded board). In figure 11, the top and bottom copper layers on double-sided flex yield 18 usable layers for the transformer windings.
Around each of the core center-leg cutouts, you can make a single turn for the inductor winding. Snaking the track around a side-leg will give you half a turn, since the magnetic path area of the side legs is effectively half. This way, you may not be able to completely surround a center leg but you can make up a full turn by adding one or two extra half-turns along the way.
Figure 11: Top-down view of the flex-circuit transformer. A single heavy current winding is shown on the top and six lighter current windings are on the bottom, routed using Altium Designer’s multi-route tool.
This could be confusing though, because you have to keep track of the proper winding directions with respect to each folded section’s relation to the ferrite core geometry. Given that this whole flex circuit will fold orthogonally, I’ve added arrows on the Mechanical 1 layer of the design facing opposite to each adjacent winding layer to remind me which way to route the copper. This is shown in figure 12 for clarity.
Figure 12: Mechanical 1 layer showing the board outline as well as winding direction arrows for guidance.
The final core-and-flex assembly is shown below. Note that this could be integrated within a rigid-flex design where the majority of the circuit is on a rigid 2-layer Printed Circuit Board, with the flex part being used to get the additional layers needed for all the core windings. Of course, there is going to be a cost trade-off between using a large flex area versus just adding heaps of layers to a rigid-only design.
Figure 13: The final completely folded-up transformer, with 3D model of the Ferroxcube E18 ferrite magnetic core through the cutouts.
How to maintain flexibility and lifespan with multiple flex layers
For many military, aerospace or similar high-density designs that require compact, reliable assemblies in tight spaces, it’s hard to avoid needing several layers of flex circuits between rigid board areas. Even more so this is necessary with high-speed digital designs, due to the requirements for shielding or plane layers between busses traversing the flex regions.
The problem here is that to maintain a good degree of flexibility, the number of flex circuit layers has to be kept to a minimum - usually two copper layers over a single PI substrate with PI coverlays.
In “normal” designs, the length of the flex-circuit sections is the same for overlapping flex. This means you end up with the situation shown in figure 14, where the folds can produce significant tensions in the flex areas between rigid boards once placed into the final assembly.
Figure 14: Tensions in the outer flex circuit, and compression of the inner, result when multiple flex layers are the same length and overlap. Notice the “squeeze out” of the adhesive bead used in this design, where the flex enters the rigid section.
Most specialist rigid-flex board fabricators at this point would tell you to use “bookbinder” construction. Bookbinder construction is a viable method where the in-situ radius of the flex circuit bends is used to determine the correct length for each flex circuit and substrate combination in the layer stack. An example illustration of the concept is shown in figure 15.
Figure 15: Bookbinder construction (source: IPC-2223B, 2008 p26).
You can tell immediately this method is going to cost money and increase the challenge of design somewhat. Often a better alternative is to use the same length and radius flex-circuits, but separate the different flex circuit layers to not overlap each other. An example of this is shown in figure 16 below.
Figure 16: Alternative Bookbinder Construction.
Getting Ultra-tight bends without sacrificing layer count
Now for the juicy stuff that will blow your mind if you haven’t seen it before! While I was at PCB West, I took a few pictures of sample rigid-flex and flex-circuit boards from a few of the vendors there. Figure 17 shows a tiny board I held between thumb and forefinger which used several S shaped flex lobes to increase the minimum bend radius between the sections that had components mounted on them. It’s not visible in this photo, but the components were mounted on sections that had thin stiffener adhered on the back side of the board.
Figure 17: Getting essentially 180° bend radius with multiple copper layers.
This concept can be extended in multiple directions of course. The PCB design shown below (figure 18) was some sort of ultra-flexible display board. You can see the many LEDs in a matrix on the wider, stiffer sections. The whole assembly was more rigid in those sections only because of the sheer number of copper and PI film layers laminated together. Again, using S-bends between those made this whole thing able to bend into a curved housing. I believe each LED was individually addressable in a matrix, so there are a lot of separate traces in this design.
Figure 18: X-Y S-bend flex array.
Take that concept even further, and you have the insanely cool application in figure 19. This is a VERY compact design, from a product designed to be worn in specialized apparel. The flex-circuit sections alone are 8 layers, according to the PCB manufacturing vendor who showed me this. As they were, such flex circuits would not be flexible enough at all. But using the myriad S-bends (notice the top flexlexible material layers are all solid copper for shielding!) allows this to bend enough to go into the final mechanical housing, even with hundreds of high speed memory and display connections.
Figure 19: 8 Layers of flex, plus 4 additional rigid layers. Notice the top layer of flex is entirely copper pour for shielding. Notice also the protective adhesive around the edges of the rigid-to-flex interfaces.
And for really creative designers, there’s always one more application, shown in Figure 20!
Figure 20: Who said Rigid-Flex meant no more origami?