PXI has taken the world of automated test by storm. In its 16 years of existence, it has grown to dominate manufacturing test for a host of reasons – cost, flexibility, performance being the principal ones. A Frost & Sullivan report from 2011predicted that PXI will be a $1billion market by 2017.
However, according to a survey conducted by New Electronics, only 14% of design engineers use the PXI platform. Of those who don't, 68% state a fundamental reason why is that they don't know enough about it. At a recent roundtable, a number of key players in the PXI sector gathered to discuss why PXI should be more prominent at the engineering end of the market and why its virtues seem to have, so far, gone unappreciated.
On the second point, the New Electronics survey showed that, amongst the 86% who don't currently use PXI, lack of knowledge was clearly the main reason, with 29% saying they were happy with benchtop instruments and 11% perceiving PXI to be too expensive. Cost is relative, however, as David Owen pointed out: "I think most of us round this table would consider PXI as low cost, but actually many of our customers would not. It is high cost for them because of the cost of the chassis and each of the modules has an infrastructure cost associated with it. There is a whole category of people out there who do not expect to be paying thousands of dollars for a chassis."
Guy Purchon concurred: "I suspect a reason why PXI has not made a huge impression in engineering is that a lot of engineering systems are cobbled together from bits that are lying around, so you only need to invest in the bits of capital that you need, rather than invest in the whole rack." This is the typical experience of the design environment, according to Kyle Voosen: "When you walk round design departments, even NI's R&D facilities, you see benchtop or cart top instruments shared among multiple groups – you check out the cart, go to your desk and make manual measurements as you design. That accounts for 85% of boxed instruments. PXI is not as valuable here because you don't want to automate the measurement, you want to take manual benchtop measurements. In manufacturing, PXI was disruptive – it knocked out rack and stack because it is so appealing in terms of cost, flexibility, performance. But when you are going to the design phase, some of the differentiators for PXI don't hold up. The cost of a PXI system, compared to the cost of box instruments that will be shared around a group, is not so convincing. For design engineers who are automating, whether it regression or characterisation or whatever, that is where PXI comes in."
For all that PXI does to open up the potential for using the same or similar equipment and programs in design and manufacturing test, the uses are different, as Owen pointed out: "An engineer's duty, when they design a product, is to get it to manufacturing without there being any query about the design. All manufacturing has to do is put the thing together in the way the engineer defined. So in test, design engineers are looking for problems that manufacturing engineers are not. If manufacturing is looking for faults, they are covering up for poor design!"
But Voosen also emphasised the potential synergies: "There is definitely an overlap there, the utopian scenario where the design engineer is writing software to automate something, maybe to discover the behaviours they are trying to design in or out. There are layers of that software that you can reuse in manufacturing test. This is a smooth transition when it works – a glorious thing – but it doesn't necessarily go the other way so easily. Qualcomm is an example of a company that has really embraced that – starting with design characterisation and reusing software right and left. And Philips Healthcare organised its design and test teams around the concept of sharing IP that is common to both - it is very elegant when it works well."
The design for test philosophy has been around for years, obviously favoured by manufacturing test engineers, but it is still not being adopted across the board. "It is still only the big groups that have realised the necessity to design for test and have all divisions in the discussion," claimed Victor Fernandez. "It means that when design teams start a new project, they get the test and production people involved from the beginning to make sure the product they are going to design can be tested easily. Then they want to move what they have done to production and then from production to maintenance. So big groups are very sensitive to design for test, but smaller companies just design it and give it to someone else to build and test."
It is the portability and flexibility of the PXI platform that lends itself to the above environment. More than a third of survey respondents who are PXI users, physically transfer the PXI system with the project, with a further 20% only transferring the programs. Interestingly, this left 42% who were using PXI purely in the design environment.
Catherine Wu sees this 'life-cycle' role of PXI as being designer led. "While we see market requests from both sides, those from the design side outnumber those from manufacturing. The reason is that cost per unit is more sensitive in manufacturing. If we only consider the unit price of the PXI platform, it's unlikely to prove less expensive than traditional instrumentation. However, the PXI platform's small footprint, high density, high throughput and flexibility easily satisfy the requirements of long-term investment, in addition to providing the benefits of improved time-to-market and testing system efficiency."
Which bought us to discussing the particular merits of PXI. The New Electronics survey showed flexibility of the PXI platform as its major appeal (69%), with instrument performance (46%), specific instrument functionality (42%), cost (38%) and synchronisation (23%) as other leading attributes.
Voosen believes PXI is a more credible engineering tool than it has ever been: "PXI keeps shattering barriers of performance. They said it couldn't be used for gerber-based oscilloscopes – and we got over that with digitisers. They said it couldn't be used for RF, and now we have reputable companies using PXI cards inside box instruments because PXI is now a viable platform for high end – there are 26GHz instruments out there. In doing this, you can start to overcome the sceptics who know enough to say that they don't want it due to reasons that are perhaps out of date. It really has come a long way in the last 10 years and maybe some of these people should take a second look."
Jeremy Twaits thought the actual application areas would be of importance: "It depends on how many different types of measurement the engineer is trying to take. If they have only taking a small number of measurements then maybe PXI seems expensive compared to a USB plug in card. But as the number of channels and the types of mixed measurements grow, the cost of the instrumentation starts to cancel out the cost of the chassis and the cost of a controller. That is when the cost benefit really comes into play. We need to get the message across that maybe the status quo isn't necessarily the cheapest method of getting your measurements, maybe there are cheaper methods."
Wu agreed: "The PXI platform, as an open architecture, can provide benefits of flexibility and scalability. Engineers can configure the design verification system based on different testing requirements. Reusability of the PXI platform makes upgrades simple, and conserves long-term cost of ownership."
An interesting variation in strategy was pointed out by Fernandez: "Suppliers like us are seeing that the future is PXI – they have a standard instrument and are then releasing it in a PXI format.
So we are seeing customers using boxed instruments at the design stage and PXI format in production. It depends on the company and how much influence the designer has on the test department, but if it is a big company, then it is more likely that the designer will look to see if there is PXI version of an instrument for production before using the box instrument himself – they want to use the same tools."
Owen took this argument one step further: "Defence has a slightly different slant on this because they tend to design test systems in a different way. Whereas most companies would design a system for a product or family of products, the defence industry has a problem. What they don't want is a piece of test equipment for every product deployed in a war zone. What they want is a few test pieces that have mass interconnect and so forth and can adapt to a whole range of products." It is no surprise that more than half of the PXI users in the survey worked in defence and aerospace industries.
Progress with PXI will require further adoption at the engineering end of the electronics sector, but responsibility for that, claims Wu, lies partly with PXI vendors: "PXI platforms still carry certain limitations in replacement of all traditional instrumentation. We believe, however, that by convincing more and more vendors to enter the PXI market, we will attract increasing numbers of users to switch to PXI platforms."
Purchon believed it was more a logical step for engineers, rather than a leap of faith: "There is no reason why design engineers shouldn't do their own automation on their own bench. I would say it is a desirable trend and many engineering teams that I have worked with are going down that path. And PXI is a good tool for that. It may not have reached the market penetration yet, but it is useful."
"Our CEO Dr T [Truchard] always talks about the blurring of design and test and it is a grey area where PXI is helping to foster that blurring," said Voosen. "Hopefully, we will continue to see this merging of abilities between design and test, so you don't need to wait until the end to verify your design, you can build a system to test as you go. And you will have the design engineer doing this, not a separate department down the road."
Twaits encouraged engineers to challenge the way they currently operated and question their efficiency: "I think I would ask design engineers how much time they spend taking measurements and could they be spending more time being innovative and creating new things – which is really what they are paid to do. If they are looking at ways to cut down on the amount of time they spend taking measurements, then automation is a way forward."
So is PXI the answer? Fernandez concluded: "Try it!"