Why are backplane 24-56G cables replacing many PCBA types?

The advent of digital signaling higher frequencies and data rates at 24G SAS, 25G Ethernet, 26G InfinBand and 28G FibreChannel have caused the need to evaluate and develop alternative media types versus traditional FR4 PCB substrates which can no longer provide acceptable transmission performance. The exponential cost of newer improved dielectric PCB substrates, from 5x to 10x and higher, has caused the PCB to cost more than the rest of the server or storage array box! PCB traces now need to be very smooth and symmetrically formed to eliminate disastrous waveforms and provide optimized skin effect performance. Very complex circuit design, high layer counts and signal integrity analysis can add much higher cost and development time for new products.

Leading PCB fab suppliers have also developed HDI (high density interconnect) structures to help mitigate trace routing challenges, but it takes 20-30 fabrication process steps to build these very high layer count backplanes. Such products represent their corporate flagship value-added products and higher margins. Thus it is natural that they have had some inertia about developing other non-PCB solutions.

Primary suppliers of backplane connector products families have been expanding their flagship product lines to have many more form-factor configurations like orthogonal, mirror and now cable plug versions to support internal cable assemblies that transport data signaling rather than the across a backplane PCB. So now leading connector brand products like ExaMax, Impel, Xcede, Paladin and Whisper are terminated with various twin axial ribbon cable types. Twin axial cable dielectrics provide very good transmission performance for these very short, small internal cable assemblies. They minimize the overall amount of higher performance dielectric usage and provide lower cost compared to using Megtron7, Taconic or Durrion brand dielectric PCB substrates.

So this means that additional higher value business goes to the raw cable suppliers as well as the connector suppliers and less valued PCB and PCBA business for those suppliers, unless they also make these types of cable assemblies. Primary PCB fab and assembly companies usually make flexible circuit board and assemblies and have offered those as solutions. But these higher performance FCBAs usually are not as cost competitive as twin-axial cable assemblies.

However, there are downsides to using twin axial cables as they require more labor to build, test and install and tend to be not such elegant designs due to various cable angled routing folds. Many twin-axial cable assemblies impede or block airflow inside the box, thus solving one problem but creating other problems. All of this gets more complicated at 56G data rates going ahead.

Some OEMs and dc customers are also evaluating the use of copper FinRail PCBAs as a rigid, much lower profile, higher speed transmission path versus twin axial cables and connectors. This is another solution set that minimizes the use of higher performance dielectrics in backplane assemblies as well as geometry transition zones within the transmission path. It minimizes the cost of cable backplane connectors as well. Successful implementation of FinRail PCB backplane interconnects can cut the cost of the overall assembly by 50% in many cases. It can provide more value-added business and a better margin for the PCB fabricators and assembly companies. Third angle FinRail interconnects solve problems much like FinFETs do within semiconductor chips.

As data rates and signaling frequency climb higher beyond 56G internal backplanes, midplanes, baseboards and blades will consider using optical fiber fabrics as the optimized media transmission method. There is a lot of process and product development going on that will provide reasonably priced solutions. We’ll follow up with this in the near future.