Robert Jackson (Industrial Measurements and Control Product Manager/Energy, Oil and Gas Marketing Manager), Gerardo Garcia (Industrial and Embedded Software Product Manager) and Rick Kuhlman (Product Marketing Engineer, LabVIEW FPGA) explain why National Instruments is like a Mitre 10 for engineers.
One of your goals is to optimise automation for the masses, but how?
Robert Jackson: There are three main ways to do it: multicore, high speed FPGAs, and ease-of-use for control applications with our Scan Engine, the last of which takes I/O data from the control system and puts it into a memory table so users can access it easily. It’s a unique architecture that gives our customers a 1 kHz update rate by simply dragging and dropping into their LabVIEW project. Version 8.6 gives control capabilities to people who aren’t ready to go into a deeper level programming, but there is a hybrid mode if you need to combine that high-end ease-of-use with very high-end performance and processing capabilities with FPGA.
With the Scan Engine you can see all the I/O appearing in the project and coming through an FPGA, and it’s all been abstracted for the customer.
Is abstraction another one of your main goals?
RJ: Yes, and LabVIEW has been providing abstraction layers for years. In the control segment it’s a way to abstract the programming of the I/O and bringing it to programming in the LabVIEW environment.
Robert, you recently wrote an article on Ethernet adoption in industrial automation applications. What’s your take on that?
RJ: It’s been driven by IT, and all the major players in industrial are moving towards an Ethernet-based bus, whether it’s Ethernet IP, Profinet or EtherCAT. Most of them are using layers 1 and 2 of the Ethernet OSI layers, the actual physical hardware; some are using layers 3 and 4, which is UDP and TCP/IP, and employing their own application protocols on top of that. Some are actually modifying UDP and TCP/IP to get the performance they think is necessary in those protocols.
We’d like to say there is one simple Ethernet-based bus, but there are still separate buses with separate target areas, and separate vendors who are pushing them. We try to support all of them as much as we can.
The movement towards Ethernet is positive because it’s a lot harder to support four or five different physical layers; at least now there’s one physical layer. Of course there’s still software work for each of those buses; for years we’ve used Modbus TCP/IP, with a shared variable that abstracts TCP/IP programming and allows customers to share data between all of our programming real time targets, fairly deterministically, but without changing the Ethernet physical layers. There will be more from us in the industrial communications area in the next year.
What are the major requirements from the market?
RJ: About 20 per cent of our market is very high-speed motion control, very high-speed deterministic data transfer. I see many customers taking our control platform and capabilities and using Modbus TCP/IP to get them to other protocols, to share data and have a hybrid system. Going forward, we’re looking at ways to add more deterministic motion control into our platform itself. The answer is not standardising to one protocol thinking that will be a permanent solution, but instead to develop a software platform where our customers can add everything else they need, and then have within it a well-integrated, high-speed deterministic bus.
What is NI’s major contribution to industrial automation?
RJ: It has to be the FPGA. The concept that we can perform PID algorithms at a hardware layer with very high speed deterministic control is something that nobody else is really investigating and investing in. I had a customer using the FPGA and our control system in oil and gas, and the level of control they had on their drilling platform is a step above what you can do with a traditional system. Our PACs combined with the FPGA really differentiate us in the marketplace, from a hardware standpoint.
From a software standpoint, the common approach is ‘here’s software to serve this market’, however LabVIEW is a graphical system design tool that can serve multiple markets. The advantage for customers in the process and control markets is that we’re bringing them things we’ve learned elsewhere. We provide connectivity to web servers, to third party databases, to legacy devices.
Gerardo Garcia: Many industrial technologies are based on PC architectures, so multicore is making its way to industrial. If we look at applications from embedded machine control to advanced process control, it’s clear that customers have to do things in parallel in order to optimise their systems. They need more advanced software and hardware capabilities, and that’s what we’ve focused on: I/O, analysis and control; you are able to do these in parallel. FPGAs, multicore processors and wireless are the areas we focus on, and we’ve simplified the programming of our PACs to bring them to a broader set of customers. There is also a recently added interface to control systems through standard web technologies.
What are the main LabVIEW capabilities for industrial applications that clients might not know about?
GG: It can do real time control. Some might think it’s just a test and data acquisition tool, but we have rugged industrial hardware platforms where LabVIEW can go. It can program FPGAs, which offer high performance for industrial. And it’s also open to ‘speak’ to other control systems and used for SCADA and HMI programming. Many customers use LabVIEW and a PAC to augment their logistic control system, and LabVIEW to tie those together.
LabVIEW 8.6 is a very significant version for industrial. We’ve been moving into industrial for a while now—over the last ten years. We’ve innovated with our PAC platform and added functionality to our software, which has made a huge impact on the industrial control markets and HMI/SCADA programming. We’ve always had a presence there due to LabVIEW’s flexibility, but these recent efforts have solidified it.
Are there potential clients that fail to see NI’s capabilities in automation and control?
RJ: Definitely! It’s partly because other vendors in the automation space have very specific and targeted industries, while we’re a horizontal broad-based tools provider.
Integrators who are specific to vertical areas are finding success with our tools, and so are some end-users. The Scan Engine and the release of LabVIEW 8.6 will help more end-users use our tools in those markets, but we’re relatively ‘new’ if you look at our presence in industrial. Gaining the adoption takes time.
How do you decide how to diversify your offering?
Rick Kuhlman: We started our graphical system design with a charter, to meet very broad needs. We focus on few verticals. What’s the Australian equivalent of Home Depot?
Mitre 10, or Bunnings Warehouse…
RK: Well, we’re like the Mitre 10 of test and engineering. You come in and you’ve got everything at your disposal to build a house, but you can’t go there to buy a house. We have all the broad tools to let you build your system, but we don’t sell turnkey solutions or end-user systems.
We try to design around particular verticals, but for the most part we solve the broad-base applications. Every day our customers come to us with a new application that surprises us because we didn’t take it into account as something we were trying to achieve!
Which segment is easier to educate about your offering: end-users or system integrators?
RJ: Both bring different challenges. With end-users, you have to reach them in the masses. With integrators, you have to find the right ones, with the right applications and the right market segments. The entire industrial marketplace is very large, specifically the ones our tools are well suited for, such as high-speed machine control, open connectivity or high accuracy measurements. We don’t pursue large DCS/SCADA system with thousands of I/Os as much as we do other markets.
RK: There are many ways to skin a cat; one specific problem and a million ways to solve it. You need to break it down into your main metrics. Are you looking for low cost, a certain performance, or a certain form factor?
But the second consideration is about internal politics or even opinions. If you’re a big food and beverage manufacturing plant, you have PLCs and know how reliable they’ve always been, so if you’re looking for a new control solution, you’ll get a PLC even though an SBC might be able to do the job if you can industrialise it, or custom hardware might also work. Internal opinions can drive the decision on a particular platform, and you can’t always narrow your decision down to the best option because there are so many factors that affect the final solution.
Robert, being in charge of Energy, Oil and Gas, how is the company’s Geen Engineering working in real life?
RK: We do a lot of power and nuclear applications. We’ve focused on energy, carbon output and pollution; it’s a worldwide issue and our customers are doing innovative things, like developing fuel cells for trains. There is a huge project in Costa Rica; we don’t know how much carbon the Earth can absorb in a given year or decade, or how much carbon it emits. This is a measurement problem that goes back to our Green Engineering campaign, Measure it, fix it (PACE September 2008). We’re deploying wireless CompactRIO systems in the forest to determine what types of trees absorb carbon, and what types of forest floors carbon escapes from. The goal is to learn how much carbon the forest can consume in a given year, and then use that to do calculations on a worldwide basis. It is a lot of work and ultimately, the answer to how long we can sustain ourselves on this planet.
The business angle is interesting; with carbon credit trading systems coming into existence worldwide, what’s the correct market value for this new commodity? It’s going to require measurements.
GG: Many customers, for various reasons, want to reduce energy and optimise their machine with less mechanical parts but more electrical-driven parts, which reduces cost but also introduces different control challenges such as non-linear control, varying dynamics where the machine changes over time with different loads. They need to optimise for different variables and there are different advanced control algorithms that in the past have been the domain of experts. With our platform, we have the tools to take these algorithms and put them in real rugged hardware, and take them to a machine or to the factory floor to optimise control systems, reduce cost, energy consumption and wear and tear trough model predictive control, model-based control and adaptive control.
Rick, we recently published your article on Custom hardware vs. Off-the-shelf technology (PACE August 2008). What are the main reasons for turning to custom hardware?
RK: The first one is performance, as an off-the-shelf [OTS] product can’t always provide what people need, so they have to make it themselves. Then comes the form-factor, because when you’re creating a product, you have to fit it in a certain place and sometimes OTS simply won’t fit. Finally, it’s about cost; by making your own board you can pick the exact components and applications, and drive costs out of your building materials, so it’s cheap when you build a lot of them.
With OTS, we try to provide things like the FPGA, which offers performance and flexibility, as well as the real time processor and the modular I/O, so you can select what I/O you want to plug into the CompactRIO. It’s about giving you aspects of a PLC, a single board computer and custom hardware.
Custom has many hidden costs; you typically spend a lot on non-recurring engineering costs upfront, and you have to deal with other things that are not built into the component costs that you’re trying to save on.
Some engineers feel LabVIEW is so easy that when they finish, they feel the program doesn’t ‘belong’ to them…
RK: Dr. T [NI Founder James Truchard] always says engineers love difficult challenges. There are plenty of things in LabVIEW that are easy to do, but others require a high level of knowledge and understanding. We want to make things easier, but also more powerful, and it’s a balancing act between those two. LabVIEW must always be a great way to map an engineering problem to a solution. I personally believe we need high level descriptions; being able to see the entire system in one screen to understand the interconnections, and then being able to zoom down into each piece and look at the low level details and tweak them if necessary.