Choosing When to Use VXI Or LXI

LXI instrumentation is making inroads in the test and measurement industry, especially in small- to medium-channel-count applications. Yet, despite the growing popularity of LXI, demand continues for the combination of density and performance that VXI instrumentation provides. In addition, VXI has a significant installed base of systems that will retain their usefulness for the foreseeable future. By combining the strengths of the LXI and VXI platforms, engineers can build systems for acquiring data over long distances, such as from strain gauges distributed along an aircraft runway. Similarly, engineers also can combine the technologies to build a high-throughput system, such as a 24-channel system that can acquire data at up to 64,000 samples.

Merging VXI and LXI

Both VXI and Ethernet-the bus used in LXI instruments-have long histories of development while maintaining compatibility with earlier versions. The VXIbus is based on the VME architecture introduced in 1981, while Ethernet has been evolving for more than 25 years. In addition, Ethernet has inherent benefits that have made it the interface of choice for test system designers and instrumentation suppliers:

It is an established, high-speed bus that continues to evolve

It offers a stable architecture

It is computer-platform and operating-system independent

It is a low-cost interface that works with low-cost cabling and accessories and

It supports up to 10-km separation between host and target device

For these reasons, the LXI (LAN eXtensions for Instrumentation) Consortium selected Ethernet as its communications backbone. The initial LXI specification was released in September 2005 and was quickly adopted by more than 40 suppliers. Despite this movement, many system engineers want to maintain their investment in VXI for higher channel-count requirements, although they are willing to integrate LXI-based products as they become available. This means that LXI and VXI products must work together seamlessly in hybrid systems.

By incorporating an LXI/VXI slot 0 Class A compliant bridge device, VXI instruments can be tightly integrated with instruments residing on an LXI instrument network. Merging LXI instruments and VXI devices in this way allows designers of test systems to take advantage of the strengths of each platform. It also enables them to build powerful instrumentation networks capable of addressing nearly every functional-test and data acquisition need.

LXI Class B compliance requires that an instrument include the IEEE 1588 Precision Time Protocol, which defines a precision clock-synchronization method for distributed networked devices. A VXI system can become time synchronous with LXI devices on the test LAN if the slot 0 interface incorporates IEEE 1588.

For more deterministic synchronization and handshaking between networked devices in close proximity, the hardware triggering required by LXI Class A-compliant devices should be used. The LXI specification incorporates a separate eight-line LVDS trigger bus that can map directly to the VXIbus TTL trigger bus.

To implement the bridge between trigger buses, the slot 0 interface must integrate an LXI-compliant trigger bus connector and level translators and drivers to accommodate the different logic levels of each bus.

Besides combining VXI and LXI trigger buses, the LVDS trigger extension also provides a means to bridge the VXI trigger bus across multiple mainframes. This was not possible with previous serial slot 0 interfaces.

The LXI specification also states that LXI-compliant instruments must include an embedded Web page to allow direct communication to instruments through a browser. The LXI/VXI slot 0 interface now makes this tool available to VXI system users. The Web page provides a GUI that allows users to control devices through common browser applications such as Internet Explorer or Firefox.

Many VXI slot 0 interfaces do not embed the VXI resource manager on the module. Rather, it is hosted on the remote PC and must be executed prior to communicating with any devices on the backplane. In contrast, the LXI/VXI slot 0 interface embeds the resource manager on the module and executes it automatically at power up. This eliminates the need for a separate software utility on the host.

More importantly, an HTML-driven interface can be designed to act as an interactive control panel. The control panel resides on the host controller and permits register or message-based communication to be done through the browser. This capability simplifies software installation and setup.

Distributed applications

VXI was initially designed to handle high-channel count or high-mix functional- test and data-acquisition systems. Applications requiring the distribution of measurement channels across long distances were typically left to small, low- performance modules linked together with proprietary buses. When higher channel counts or measurement quality demanded the performance provided by VXI devices, the options were limited to costly embedded controllers or bus extenders that were not always reliable.

A LAN-based slot 0 device allows multiple mainframes to be connected to a single host controller up to 100 m away with standard copper cable and an Ethernet switch. Fiber-optic cables and switches permit distances to 10 km. If the slot 0 device implements LXI Class B synchronization techniques, measurements can be synchronized to a high degree of resolution (tens of nanoseconds) across the instrumentation network.

As an example, consider a system constructed to acquire data from six groups of strain gages distributed across 1 km of concrete pavement, with each group consisting of 192 channels. The strain gages measure the stresses on the pavement, which is subjected to repeated loading to simulate the landing of large commercial aircraft.

The high channel count and integrated signal conditioning required for this installation were best handled by a VXI architecture. The original system incorporated six mainframes with six embedded controllers. Each controller independently managed data acquisition from the channels in its mainframe and passed the data over an Ethernet connection to a remote PC, where a complex program post-processed and synchronized the data.

System designers found, however, that this architecture was susceptible to faults due to the number of independent processors involved. The LXI/VXI slot 0 interface provided an opportunity to consolidate control and data processing on a single host, greatly reducing complexity. In addition, because the LXI/ VXI slot 0 implements LXI class B synchronization, channel-to-channel skew has been reduced considerably, and data samples acquired across mainframes (and distances) have tight time correlation.

Increasing data throughput

Perhaps one of the most significant aspects of adding Ethernet capability to VXI is the ability to increase overall data throughput by using the speed of the bus coupled with the fundamental nature of an Ethernet switch. Large-channel count, high-speed systems can push the limits of even the fastest slot 0 interfaces because the single pipeline back to the host is bandwidth limited. Further, the acquisition device often becomes a bottleneck if cycle times (the speed at which a block of data can be placed on the bus) are constrained by factors such as onboard processing.

This is illustrated by a 16-channel, 16-bit acquisition device capable of sampling data on all channels simultaneously at a maximum rate of 100,000 samples/s. An onboard digital signal processor (DSP) runs the data through intense algorithms prior to putting a block on the bus for retrieval, and a single module is limited to about 12- Mbytes/s throughput to the backplane.

If the slot 0 interface can support transfer rates greater than 12 Mbytes/s from the backplane to the host, then the instrument module can be a bottleneck. If the slot 0 interface cannot support 12-Mbytes/s transfer rates, then the interface itself may become a bottleneck. In either case, the instrument faces the prospect of overflowing its buffer because it cannot be emptied fast enough to keep up with the amount of data acquired.

In a recent application, 240 channels of data were simultaneously and continuously sampled at a rate of 64,000 samples/s. The amount of data to be processed by the entire system can be calculated from

RDI = RSDC

where RDI is the data-in rate, RS is the sample rate, D is the number of bytes per sample, and C is the number of channels. For the 16-channel, 16-bit system described above, RDI = 30.7 Mbytes/s.

Because the data input rate exceeded the 12-Mbytes/s rate at which a module could be emptied, an overflow was certain to occur, and significant amounts of data would be lost. Further, even the fastest slot 0 interfaces could not alleviate this problem, because the data transfers occurred across a single bus connection back to the host. For example, a 1-Gbit LXI/VXI slot 0 interface supports block transfers in excess of 40 Mbytes/s, but this would not increase the aggregate data rate through a single pipeline.

The only acceptable way to address this application was to split the acquisition modules across three six-slot VXI mainframes (five modules per mainframe) and interface back to the host controller using a 1-Gbit slot 0 interface and a multiport 1-Gbit Ethernet switch. While the path from the switch to the host PC is across a single connection, the keys to successfully meeting this challenge revolved around the bandwidth of the bus (1-Gbit Ethernet is roughly 100 Mbytes/s) coupled with the intrinsic ability of the switch to manage and buffer the large amount of data passing through each of the three ports.

These examples illustrate effective deployment of combinations of VXI and LXI technology that take advantage of the promise offered by LXI while ensuring that the installed base of VXI systems continue to pay dividends.

[

Leave a Comment