40 years of test and measurement
By James Truchard, PhD, president, CEO, and cofounder of National Instruments reflects on 40 years of test and measurement and what lies ahead.
February 3, 2017
By Lee Kok Leong
As I approach the end of my 40-year career as CEO of National Instruments (NI), I am reminded of the great progress and innovations the test and measurement industry has witnessed since 1976. We have gone from an industry driven by vacuum tube technology in the era of General Radio to a time when the transistor ruled with Hewlett Packard to today, when software truly is the instrument—a transition that NI helped shepherd.
Moore’s law has taken us for a wild, fast ride to say the least, and just when you think it’s run its course, process innovations extend into new dimensions (literally) and push performance even further.
Just like the transistor, NI started from humble beginnings, but it has relentlessly focused on engineering great products and empowering world- changing innovation through our customers and platform technology. Allow me to reminisce on what the past 40 years have taught me and where I see this market heading as I shift into the next phase of my career.
When Jeff Kodosky, Bill Nowlin, and I started NI in 1976, we saw tremendous room for innovation in how engineers and scientists interacted with and built test and measurement equipment. We founded the company on the premise that there had to be a better way to serve the test and measurement needs that we, engineers and scientists, faced. We couldn’t buy it off the shelf but at least we wouldn’t have to build it from scratch.
The general purpose interface bus (GPIB, IEEE 488) was our gateway. Our vision, as stated in 1983, was to “do for test and measurement what the spreadsheet did for financial analysis.” Stated today, the sentence loses some of its power, but think about the early ’80s. At the time, the tools for financial analysis were “locked up” and too expensive for anyone without a big budget to access them. The early incarnations of spreadsheets turned this situation on its head, and that is exactly what we wanted to do. We wanted to make it so that any engineer or scientist could access the same tools or platform used by the R&D teams of the leading technology companies. It was a radically empowering view at the time and, in many ways, it still is.
The software is the instrument
While others might have seen GPIB as a hardware play, we recognized it for what it enabled in terms of software. As the PC industry evolved (as well as Apple’s Mac, which we had a special affinity for given its graphical user interface), that GPIB cable made it easy to analyze and present data in a customized way for our customers’ needs. They were no longer confined to the front panel of an instrument and their pencils and notepads for data acquisition. The opportunity to innovate then shifted to the software world, where programming languages needed instrument drivers for the connected boxes. Our strategy of writing and supporting those drivers offered a critical service that continues today as NI supports more than 10,000 drivers on the company’s Instrument Driver Network.
But that world still left engineers and scientists with the burden of using tools designed for computer science to perform engineering, test, and measurement tasks. Our answer was twofold: LabWindows/CVI, to offer engineering-specific tools in ANSI C programming, and LabVIEW, a graphical programming paradigm that took the way we think about solving a problem (in flowcharts and pictures) and turned it into compiled code. The story was simple: acquire, analyze, and present. Do it in software tools designed for a customer’s use case that were easy to learn yet extremely powerful. We coined the phrase ”The software is the instrument” to describe this approach, and seeing engineers and scientists save valuable time and get to results faster was all the market validation we ever needed.
Evolving With Moore’s Law
People talk about Moore’s law like it’s about hardware, but computational hardware exists only to run software (and maybe firmware). Once we made test and measurement all about software, we had effectively enlisted Intel, Xilinx, and many other billion dollar companies in our R&D staff. With customers and partners building proficiency with our software tools, we just had to follow the chips to deliver increasing value to test and embedded systems. This has happened, so far, along two key dimensions: multicore processors and FPGAs.
Because LabVIEW is graphical and, therefore, not obviously sequential, it is tailor-made for parallel processing. LabVIEW users were among the first programmers to easily migrate from single-core processors to multiple threads and multiple cores and see almost instant speed improvements. Obviously, it’s possible to take advantage of these trends with other languages, just like it’s still possible to write highly efficient code in machine or assembly language, but why would you? The pace of change in modern electronics means you can’t waste time doing by hand what a tool can easily do for you, and we hear that over and over from LabVIEW users.
That goes to an entirely different level with FPGAs. Some problems are just better solved in the highly parallel, deterministic world of silicon. But the toolchains and programming constructs were inaccessible to most mechanical engineers or medical researchers who were experts in their measurements and problems to solve (not digital design). We recognized this in the late 1990s with LabVIEW’s graphical paradigm. We were on a quest to deliver the power of FPGAs to LabVIEW programmers, and we’ve done that. A quick look at our Engineering Impact Awards winners demonstrates the power of this technology: applications ranging from regenerating and restoring organ function damaged by disease or trauma to setting a world record in 5G wireless spectrum efficiency with massive MIMO.
A software-centric approach to hardware design
When you think about software as uniquely as we have, it’s easy to think differently about hardware, too. Modular, PC-based plug-in boards were a natural by-product. Make the hardware as lightweight and cost-effective as possible (no dedicated screens, power supplies, fixed buttons/knobs, and so on) and focus on ADCs, DACs, signal conditioning, and data movement.
I have yet to see a test and measurement vendor design a user interface better than a customer, for any specific task, that makes the customer more productive. Even the best front panels on box instruments are cluttered with unused buttons or menu structures. Many of our hardware products have size constraints dictated by the I/O connector. Can it get more efficient than that?
The reality is our strategy is more than just efficient; it’s right. Take the new Vector Signal Transceiver (VST), which combines an RF analyzer, RF generator, parallel and serial digital interfaces, and high-performance signal processing into a 2-slot PXI module. This product delivers industry-leading bandwidth (1 GHz), amazing RF performance, and scalability for MIMO applications for one reason: software. We moved as many technical problems into the FPGA as we could, and Moore’s law (along with Xilinx) delivered a vehicle capable of handling the computation. We, in turn, passed the keys to that vehicle over to our customers by allowing them to customize that FPGA with LabVIEW. From 5G cellular technology development to automotive radar and driver assist algorithm development to reductions in the cost of Internet of Things (IoT) devices, the VST and LabVIEW are helping customers achieve goals conventional instruments quite simply prevent them from obtaining.
We are seeing glimpses of the future everywhere we look. A modern factory features what we call “cyber- physical systems,” which combine software-centric computing technology with electromechanical systems and human operators to improve safety, efficiency, and cost structures. The acquire, analyze, and present concept is still valid, but we’ve added “sense, compute, and connect” as a parallel flow for IoT devices. Wireless technology in general is pervasive. We’ve been saying this a while, but if you aren’t an RF engineer today, you will be. And the more you connect things, the more you’d be crazy not to take advantage of the data you can collect for billions of sensor nodes. For us, this is Big Analog Data solutions, and it’s the richest set of data in the world. NI customers are acquiring terabytes and terabytes of it every day.
But even as our capabilities become more advanced and the scale of the problems we try to solve grows vaster, the tools we use must be easier to navigate. Just as machine language migrated to assembly and to object-oriented, other paradigms, including graphical dataflow programming, are critical to offer the right level of abstraction. The multirate diagram in our LabVIEW Communications System Design Suite is a great example; no single software tool delivered the productivity needed to prototype 5G algorithms until we were bold enough to tackle multiple models of computation inside a single flow that could deploy directly to hardware.
No great innovation will be done alone. The best platforms we use today are effective because they’ve fostered an ecosystem. Our software-centric approach at NI spawned a partner network of more than 1,000 companies and 300,000 active LabVIEW users. The rise of mobile devices and “apps” is possible only because of a healthy ecosystem built on developer-friendly platforms. Team-based development, code sharing, and community support soon will no longer be novel or best in class. They will be expected.
It would be impossible to have witnessed what I’ve witnessed in our industry for the past 40 years and not be excited about where all of these technologies and trends are leading us. My advice to any new engineer is simple: develop a vision for the future and pursue it with intensity. And, at the end of the day, don’t be afraid to have fun.