Instrumentation and Test Expert

Centralized Switching Architecture for Automatic Test Equipment (page 2)

II. CENTRALIZED SWITCHING ARCHITECTURE
Most ATE designers are familiar with a centralized switching architecture including PXI and VXI mainframes as shown in Figure 4. below.

A number of manufacturers make switch, matrix and multiplexer cards in the PXI and VXI form factors. A single PXI chassis takes less space in the rack than a VXI chassis, but the VXI switch cards are typically higher-density than the PXI switch cards. A single, fully populated VXI mainframe can hold approximately 3,000 discrete relays, but beware the deception of switch density. Even if the cables between the mainframe and the tester interface are only six feet long the total length of wire is almost 36,000 feet (6 feet * 3000 relays * 2 wires per relay), or 6.8 miles! This much wire will make the inside of the tester look like a jungle. Even worse are the number of contacts. If you consider that each wire equates to four contacts (two on each end of the wire plus the two mating contacts); a fully populated VXI mainframe contributes 24,000 (3000 relays * 2 wires per relay * 4 contacts per wire) contacts. By definition these must be in high-density connectors and can make for a reliability and maintenance nightmare. If a VXI chassis fully populated with relays is in your design plan, you should seriously reconsider your choice of switching architecture.

When the connectivity is relatively simple, cables can be also be used as the mapping layer in a centralized switching architecture. More commonly the connectivity involves multiple splices, discrete components and the like such as shown in Figure 5. In these cases the mapping layer will almost certainly involve custom interface chassis.

Another item to note about the interconnectedness of the relays in Figure. 5 is the opportunity to create a custom relay circuit board to eliminate external wires and simplify the cabling inside the tester console. This will be discussed more in the section on a distributed switching architecture, but every time a relay pole only connects to another relay pole it is possible to eliminate one external wire. In an overall system, custom relay cards can eliminate hundreds or even thousands of discrete wires.

Within the custom interface chassis the designer can locate terminal blocks, splices, resistors, capacitors and the like. This approach provides the designer with very much flexibility and has the added benefit that connectivity can be changed relatively easily if there are design flaws or new requirements discovered during integration. The downside is that these chassis can easily become “boxes full of wires” that are subject to manufacturing defects. Figure 7. shows a mapping layer created by wires inside of a custom interface chassis.