Last year, <i>New Electronics</i> took a first look at the developing role PXI has to play in the design environment. PXI (PCI eXtensions for Instrumentation) has been around for 17 years and is clearly at a stage in its existence where it has gained recognition from some as a stable, versatile platform, but still viewed as relatively unknown and/or irrelevant by others. <i>New Electronics</i> convened a roundtable to discuss these attitudes and the results of a survey of <i>New Electronics</i> readers.
Use of PXI for manufacturing test is both common and well established. Having a modular system housed in a compact chassis has obvious appeal for test engineers looking for a way of automating production validation tests. By not duplicating processing capability, interfaces, connectivity and more, the cost effectiveness of PXI, along with its flexibility, make it an almost de facto choice for the production environment.
Design engineers, however, have a different approach and different demands from their test bench. While the ability to set up an automated test routine is ideal for the production line, engineers need to use test instruments to probe, fiddle and explore. Does PXI fit this bill?
Our survey threw up the headline statistic that only 10% of respondents – electronics design engineers – currently use PXI and 68% of those that don't said lack of knowledge was a main reason. "It is not really a surprise," said Keith Moore, managing director of Pickering Test. "PXI is a natural fit for manufacturing test and it is difficult to get to other areas – it will be slow progress. Design engineers are not used to having modular test systems on their benches; certainly not from the VXI days – they would not have any bench left!"
On the other hand, Victor Fernandez, European manager for Marvin Test Systems, was surprised. "How can they not know about a standard that is 15 years old? Maybe what they are saying is that they don't know how to use it or know what PXI can do for them. When MTS does PXI seminars and asks the question about how many know or use PXI, I am always amazed at the number of people that either don't know or don't use PXI – although they all seem to know VXI, so card modular architectures are not foreign to them."
There are reasons why PXI based instrumentation works in a design environment, and reasons why it doesn't, but one of the key issues appears to be one of 'hearts and minds'. Gary Clayton, sales director, Mac-Panel, observed: "If the engineer is sat with his bank of instruments because he feels that is the right way to go, how do you tell him there is a better way and that he needs to throw away everything that he has? Engineers never throw away anything –there is a lot of emotion in that."
Ultimately, heads will rule hearts, according to Jeremy Twaits, regional marketing engineer, National Instruments. "If there is enough time saving as a benefit, then people will change. But if someone only wants to make a few measurements here and there, then there is no point in automation and they may as well continue to do it the way they like." The value comes if that engineer is going to take many measurements of a device's many different operating modes. Twaits continued: "If you can prove to them they will save time by going away from their interactive approach and towards an automated approach with PXI, then the change can be made."
PXI is also perceived as too expensive, but is this the case? "It is expensive to set up a new PXI system," admitted Clayton. "But with its flexibility and upgradeability, it is significantly cheaper than having to buy equivalent new benchtop instruments [as the technology demands]. But I suppose the challenge is to communicate that to a market that is well set in one way of doing things."
Along with the lack of knowledge about PXI and its cost, another factor was simply familiarity with existing hardware, something identified by 30% of survey respondents. "As engineers, we like to think of ourselves as rational people," added Twaits, "but you still have that emotional attachment to something; for example, an instrument you like to use or a piece of software."
"Quite frankly," concurred Fernandez, "I think designers probably don't use PXI (or other automated instrumentation platforms) because they don't want to take the time to learn something new or to program an instrument. If designers don't not need to automate a design verification activity, they are probably comfortable with using a 'box' instrument strategy."
Moore believes PXI is now coming of age for the design environment. He said: "When NI came along with PXI, it was the right idea at the right time. People had got fed up with the expensive VXI solutions that were overkill. NI's first PXI might have been underkill, but we thought it would have 'legs' if a few companies got behind it. Those first cards were very simple, but jump forward 16 years and the sophistication today is unbelievable. It has been an amazing transition."
The PXI Systems Alliance now has more than 70 vendors, whose combined portfolio is in the region of 2500 products; far more than comparable platforms. Clayton noted: "The scope of products is growing exponentially in terms of the diversity of product being developed." Moore concurred: "Look at some of the instruments coming out of NI and Agilent – 26G signal analysers and generators, unbelievable technology. Ten years ago, the big defence companies wouldn't touch it with a barge pole – now they embrace it."
There have been a couple of minor additions to the PXI portfolio – the Mac-Panel Scout interface, which combines tidy interconnect management with customisation of the platform, and a small five slot chassis. Both make the platform more useable in a design environment. The five slot chassis, introduced by NI a couple of years ago, opens up the option of deploying it at several sites in identical fashion but at low cost. However, the strength of the platform is in its flexibility and upgradeability and certain sectors have identified this.
"In the semiconductor market, customers are using PXI for design verification in the lab," said Clayton. "They would never dream of using it in production; for them, it's a tool for designers which is configurable for different types of silicon and upgradeable; important, considering the rate of development in semiconductors. It is a requirement to be able to continually upgrade those PXI systems, it gives them that flexibility."
Applications are equally relevant at other timescales – new consumer electronics products appear every six or 12 months, whilst military projects might have a life of 40 or 50 years. The constant change in consumer electronics might necessitate testing new capabilities, such as sensors or software. Twaits said: "A flexible software defined test and verification solution means that, as new products appear, you can modify the system with relative ease to test the new functionality."
Equally, military projects need to be supported over many years and PXI allows new cards to be swapped in as needed over the course of a project. Twaits continued: "With a scalable software architecture, maybe using IVI drivers, it can be possible, if the original instruments become obsolete, to bring in new instrumentation with very little change to code. This makes it possible to continue to use the system."
Clayton claimed that 68% of all PXI units sold went into the military/aerospace sectors and, in a similar vein to the semiconductor industry, PXI's entry point is at the design verification stage. He said: "We then see migration through to production and, almost without exception, a version of that production tester will end up in the repair and maintenance operation. Our strategy is to come up with interface solutions that have been designed for the design stage but can still be used down the line in maintenance."
Simulation is an area in which all panellists saw interest, particularly in safety critical applications. Twaits said: "It is not just aerospace and defence, but also automotive. The ECU is a prime candidate; any ECU is going to have some form of real time test or hardware in the loop tester and it is usually important to have those real time aspects." Such automated testing is ideal on the PXI platform as you can add in any number of variables. Moore added: "You have to test different scenarios to make sure that, when code changes, software engineers haven't fixed one thing, but broken another."
"There is no technical reason why PXI cannot be used successfully for simulation of a system," said Fernandez. "In particular, the use of user programmable FGPA PXI modules provides a high level of flexibility for supporting simulation."
The use of FPGAs has proved a major enabling feature of PXI as it provides truly software defined instrumentation. Twaits said: "The capability for users to really define what they want their instruments to do gives PXI one of its key advantages – it allows you to create your own firmware. Not only does this allow users to create their own instruments, it also protects their IP. Many companies, particularly in the military sector, are understandably reluctant to share their IP with equipment vendors."
As a final observation, while the take up of PXI has been relatively slow in the design verification environment, when asked if those design engineers envisioned using PXI on future projects, the number saying 'yes' rose to more than half.
Participants in this year's roundtable included:
Keith Moore, managing director, Pickering Interfaces
Gary Clayton, sales director, Mac-Panel
Jeremy Twaits, regional marketing engineer, National Instruments
Victor Hernandez, European manager, Marvin Test Solutions