Reduce test costs with careful PXI design
By Richard A. Quinnell, Contributing Technical Editor -- Test & Measurement World, 5/1/2009 2:00:00 AM
As the economic downturn increases pressure on companies to reduce test costs, engineers may find that the configurable nature of PXI test systems offers a cost advantage over stand-alone instruments. Developers can trim costs further by fully assessing their test needs, carefully planning the system architecture, and selecting versatile, low-power components.
 Fig. 1 Modules that support simplified upgrade and repair, such as those that employ mezzanine boards (indicated by arrow), can reduce long-term system costs. Courtesy of Agilent Technologies. |
One of the first steps in cost optimization is evaluating how a system will address both present and future needs. Mike Dewey, senior product manager at Geotest—Marvin Test Systems, said “Don’t over-design your system by putting in too many features for possible future use. Narrow it down to just what you need now. Future requirements may be totally different from what you expected, and with PXI, you can always swap modules later.”
This does not mean you should completely neglect future needs; in fact, many vendors recommend leaving expansion room in your PXI chassis. “In chassis selection, the number of slots gives a tradeoff between expandability and cost,” said Matthew Friedman, PXI platform manager at National Instruments and PXI Systems Alliance marketing co-chair. “You will want to leave some space for upgrades.”
Aeroflex’s PXI product manager Tim Carey agreed. “Expansion is a good thing to plan for. It’s false economy to minimize the number of slots in a system. The cost of the chassis is only a small fraction of the system cost, especially in RF, but the rework effort that would be needed to change the chassis can be substantial.”
Vendors also suggest caution when assessing current needs. “Developers sometimes focus on minimizing test time,” said Geotest’s Dewey, “but they need to look at that issue closely to avoid paying too big a penalty. Run the numbers to compare asset cost versus test time cost. A cheaper configuration may be a good tradeoff.”
Dewey also pointed out that requirements may have declined. “With sales dropping off a cliff, for instance, a semiconductor manufacturer may not have enough production volume now to justify the cost of the highest test throughput.”
Placing the processing power
With requirements in hand, you can plan an architecture that keeps costs down while permitting easy and inexpensive upgrades. One of your main decisions will be what type of system controller to use. The controller can be an external PC that bridges into the PXI chassis system or an embedded processor card residing in the chassis. This controller can be responsible for all of the data processing, or it can simply provide a data display and user interface while other modules handle data processing, or it can operate somewhere in between those extremes.
 Fig. 2 By reducing development effort, off-the-shelf chassis, such as this enclosure with a built-in cable tray, may provide more in savings than they add in acquisition costs. Courtesy of Geotest—Marvin Test Systems |
Cost and flexibility vary significantly among the options. Carey said that using an embedded controller yields several advantages, including a more compact system and nearly 40% better data throughput than you get by bridging to a PC. On the other hand, he noted, an external controller PC can be less expensive for a given performance level, can allow you a wider choice of suppliers, and can be more readily upgraded than an embedded module. Using PCI Express instead of PCI as the bridge would help recover the lost data throughput. Thus, the optimum choice may be application specific.
The same sort of ambiguity applies in deciding how much of the system’s data processing the controller must provide. Richard Soden, product marketing manager for the signal networks division at Agilent Technologies, said that having individual modules provide substantial onboard data processing lessens the requirements on the controller’s performance, allowing designers to use a less-expensive unit. But he added that this choice may increase the cost and complexity of the other modules, so you should evaluate total system cost not just individual boards.
A rough rule of thumb calls for the system controller to handle processing when the design has general-purpose applicability and modest performance needs, according to Pat Cupo, president of test system developer Instrumentation Engineering. Highly targeted applications, those requiring the highest performance, or those dealing with complex waveforms, said Cupo, are better handled with dedicated on-module processing. “A single controller just can’t keep up,” he said.
Developers should also evaluate the impact on system flexibility when deciding where processing should occur. NI’s Friedman pointed out that using the system controller to handle data processing makes it easier for developers to create custom functions and algorithms than when processing occurs on a module. Aeroflex’s Carey noted that local processing on a module can reduce system flexibility: “Onboard processing puts a wrapper around a module, defining what it can do and making it more like a boxed instrument rather than remaining configurable for new or different tasks.”
Another way to reduce costs lies in making modules or even the entire system perform multiple roles. Developers should “map resources to needs in order to get double or triple duty out of their boards,” according to Geotest’s Dewey. “Do you need a DMM module, or does your digitizer offer adequate performance?”
IE’s Cupo gave as an example of such architectural multitasking a custom system that his company designed that tested two different devices concurrently using a single PXI rack. The multiplexer and DMM served both tests while other modules were dedicated to a specific device. The design filled an 18-slot cage but proved highly cost effective. “The customer couldn’t afford two independent test stations,” said Cupo.
Blending systems can save
You may also save on costs by leveraging PXI’s ability to blend with other buses. “Not all of your test functionality needs to be in PXI,” said Carey. “Custom or proprietary elements such as RF switching, for instance, may be more cost effective and yield quicker time to market than trying to reconfigure an off-the-shelf PXI module.” Agilent’s Soden added, “You can also avoid buying new hardware and [can] create a hybrid system using other instruments you already have.”
Once the system architecture is planned, developers can further hold the line on costs through their choice of individual modules. Power requirements are one area to consider, said Soden. “Using lower-power cards can allow the chassis to also be low power. This saves on the power supply as well as on cooling, and reduces direct electrical cost during operation.”
How readily a module can be repaired or upgraded can also be a factor. Agilent, for instance, has introduced a series of PXI digitizer cards on a common base that use a mezzanine card to carry the front-end electronics (Figure 1) that dictate function and performance. This allows the company to develop new variations more quickly and at lower cost, reducing upgrade and repair costs for customers.
Consistency in choosing suppliers can also reduce long-term costs, IE’s Cupo pointed out. “When you expect to be designing multiple platforms,” he said, “it pays to standardize on one source for stimulus generators, one for measurement, and so on. Otherwise, each tester design looks a little different and adds cost downstream in software development and maintenance.”
Other system elements can also represent an opportunity for savings, especially if developers discuss their needs with vendors to learn what’s available. Geotest, for instance, has more than 30 chassis types with built-in features such as mass termination receivers and cable trays (Figure 2). “We offer lots of options so developers can be further up the system-integration curve right out of the gate. Having the cable tray integrated into the chassis, for instance, saves developers from having to design and fabricate their own cable-management solutions,” said Dewey.
Software support significant
As important as evaluating hardware costs, though, is examining the software support available. “A big part of the cost in creating a test system is the implementation and deployment of software,” said Agilent’s Soden, “so you don’t want to spend a lot of time getting the system going.”
NI’s Friedman agreed: “Developer productivity is an important consideration, yet the cost for development and integration of a system is often missed when evaluating design options.”
Vendors recommend that developers verify the availability of drivers for the operating system they are using and look for tools and programs for module configuration, calibration, and maintenance. IE’s Cupo, for instance, looks for modules that have .IVI driver support because he uses direct driver calls from NI’s TestStand test executive to speed software development by eliminating a programming layer. Cupo underscored the importance of calibration routines in the support package: “Most customer metrology groups are not set up yet for handling PXI systems, and calibration can be a large hidden cost if not addressed up front.”
Ultimately, cost optimization for PXI systems depends on your making decisions based on the total cost for acquisition, development, deployment, and lifetime operation. In some cases, this may mean spending more up front to garner savings downstream. “People can get hung up on a few hundred dollars,” said Friedman. “You need to look at the impact on your overall test system and your overall business.”
No related content found.
- 0 rated items found.
Datasheets.com Electronic Parts & Inventory Search
185 million searchable parts
- Part Number
- Description
- Inventory
- Products
- Manufacturers
View All Webcasts
