High-Speed Rigid-flex

Rigid-flex can meet the speed you need – with the right design and materials.

I am designing a rigid-flex and was just told it will be higher frequency. What do we have to do differently?

Answer: As the phrase goes, “I feel the need…the need for speed!” We are seeing more and more applications with signal speeds in the GHz ranges. As clock speeds on chips increase, it is important for the circuit boards to keep up the pace.

When it comes to high speed, rigid-flex is very capable of meeting the challenge. It all comes down to good design and material selections.

For the high-speed signals, think in all three dimensions. Map out which layers carry these signals. Rigid and flex layers will have different dielectrics and copper types. If all the high-speed signals reside within the rigid layers only and do not span the flex region, concentrate just on those rigid layers and associated materials. For this discussion, however, let’s assume that one or more of the flex layers will be involved.

On the rigid layers, you may select from a wide range of rigid copper clad laminates as well as various copper profiles just as you would for a rigid PCB. Manage loss by selecting low-profile coppers and low-loss tangent laminate.

On the flex layers, we get to enjoy a couple of advantages right out of the gate. The polyimide film substrate is very low-loss, with a Df in the 0.002-0.003 range. That is coupled with a rolled annealed copper (if 0.5oz. or thicker), the lowest profile copper available.

Two areas require careful material selections in order to meet loss budgets. These are the prepreg and any flexible coverlay or bondply.

For the prepreg, board fabricators will typically use a no-flow prepreg throughout the stack-up, especially if the board is a single lamination construction. This is because not all prepreg materials have the same lamination profile. The good news is that a few material suppliers have developed low-loss no-flow prepreg materials over the past few years, which has helped tremendously.

In some cases, when a construction requires multiple laminations, it opens options after the first lamination to use standard-flow prepreg. That means you could once again choose most anything for those outer layers to satisfy signal integrity. Just keep in mind that your fabricator will not have qualified every material from every supplier, so consider generic callouts to allow a selection of materials they have experience with.

In the flex region, standard flex coverlay and bondply do not have good loss tangent values. To combat this, we can select alternate flexible materials developed specifically for low-loss. These will have Dk values in the 0.002-0.004 range.

Coupling these various material sets together permits a design that is fully capable of carrying high-speed signals throughout the system with minimal insertion loss. Now you can concentrate on the circuit design. All the standard design practices for circuit layout still apply. You will still want smooth flowing conductors without sharp corners or turns and you will need to model trace pairs for the optimal line width and spacing.

If backdrilling is needed for stub-length reduction, that can be accomplished as well. The one caveat here is that if the critical signal is on a flex layer, it will be difficult to cut the adjacent copper layer using backdrilling. This is because the flex laminates run thinner than rigid laminates, so the adjacent layer is much closer. Total stub length will remain the same, but not all layers you want to disconnect may get cut.

Another consideration when it comes to rigid-flex is whether to incorporate cross-hatch patterns on the reference planes. As we have discussed in past columns, cross-hatch planes are sometimes employed to achieve impedance values while keeping the flex as thin as possible. This has proven to work very well in many applications. However, modeling and testing tells us that as frequency rises, insertion and return losses rise rapidly when cross-hatch planes are used. It appears that at frequencies of 1GHz and above, the losses will be too excessive to use cross-hatch planes.

Hopefully you see it is very possible to use rigid-flex in high-clock-speed applications. Also, an intrinsic advantage of rigid-flex is that it eliminates signal disruptions due to the use of connector and solder joints when comparing a rigid PCB and wire harness to a rigid-flex. In the end, signal integrity, speed and skew are improved in a rigid-flex design. All that in a lighter weight, more compact and more reliable package solution. Your boss should give you a raise for all the money you are going to save the program!

Nick Koop is director of flex technology at TTM Technologies (ttm.com), vice chairman of the IPC Flexible Circuits Committee and co-chair of the IPC-6013 Qualification and Performance Specification for Flexible Printed Boards Subcommittee; nick.koop@ttmtech.com. He and co-“Flexpert” Mark Finstad (mark.finstad@flexiblecircuit.com) welcome your suggestions.