The Sound of Progress

Why haven't backplane connectors shrunk like everything else?

The field of electrical connectors has enjoyed constant improvement ever since 1903, when US inventor Harvey Hubbell submitted patent No. 774,250 for the first "separable attachment-plug" (Figure 1, Reference 1). Connectors today are smaller, carry more current, work to higher voltages, and operate at significantly higher frequencies than their distant cousins—crucial to the operation of high-speed digital backplanes.

Hubbell - the first electrical connector.From 2001 to 2011, the maximum operating speed of the connectors in leading-edge backplanes improved by a factor of eight (Table 1). In 1980, users could employ the VMEbus (Versa Module Eurocard bus) DIN 41612 connector at 100 Mbps when the bus was in differential mode with an appropriate number of ground pins to control crosstalk. The numbers in the table imply a doubling of operating speed approximately every four years.

Over the same decade, the number of instructions per second that a leading-edge microprocessor could execute increased from 1000 MIPS (millions of instruction per second) in 2000 to 30,000 MIPS in 2010—a factor of 30. Backplane connectors aren't keeping up. Each year, they fall further behind, and that failure complicates your ability to design fast, reliable systems.

In 1980, when the VME backplane artchitecture reigned supreme, typical digital-system designers did not model backplane connectors. Beyond establishing a signal-to-ground ratio to control crosstalk, the other electrical imperfections in a backplane connector remained less significant than the ribbon cables, unreferenced traces, and IC packages then available. Today, the situation is reversed. A backplane connector and its associated PCB (printed-circuit-board) vias now contribute more signal degradation than almost any other factor. Leading-edge-serial-link designers now routinely attempt a complex co-design of the connector and its attachment along with all other system components.

Unfortunately, the design of any physical system becomes substantially more complex as you push up against the ultimate theoretical limits of its performance. The key to making inexpensive, reliable systems is to identify the primary limit on performance and find a way around it, rather than attempting to scale what may turn out to be a sheer cliff.

Table 1-Pin spacing of high-speed connectors

Year Product Density (pair/mm2) Rated speed (Gbps)
2011 FCI Airmax VSe 0.119 25
2008 Amphenol Xcede 0.101 20
2008 Ernie Ermet Zd Plus 0.089 20
2007 Molex I-Trac 0.073 12.5
2005 FCI Airmax VS 0.119 12.5
2001 Ernie Ermet Zd 0.089 3.1
1980 VMEbus DIN 41612 0.155* 0.1

*single-ended signals

Size is the dominating factor that influences backplane-connector performance: The smaller the connector, the better it works in a high-speed application. Yet, over the last decade, the size of backplane connectors has changed little, almost not at all in comparison with the enormous improvements in onchip geometry. Witness the key metric of connector size in Table 1: the effective connector pin density. It has remained practically constant for 30 years.

Why have backplane connectors not shrunk as everything else in the world of electronics has? Backplane connectors do not suffer limits to performance from lithographic constraints, as silicon does. The connector pins are thousands of times larger than silicon. Backplane connectors do not suffer ergonomic considerations, as do other interconnection systems, such as HDMI (high-definition multimedia interface), whose minimum size depends on the ability of a typical user to see and operate the connection. Backplane connectors find their seats automatically with the help of card guides and alignment pins.

Backplane connectors have failed to shrink because of the physical scale of the vias and breakout patterns on the PCB underneath the connectors. They control the overall size of a backplane connector. If you shrink the vias and the connectors, performance will skyrocket. Unfortunately, the 19th-century electrochemical process we still use to make vias and pads improves at a maddeningly slow pace.

Reference

[1] Hubbell, Harvey, "Separable Attachment-Plug," Patent No. 774,250, United States Patent Office, Nov 8, 1904.