Microprocessor advances translate into more CNC machine capability
CNC technology follows mass-market computing trends, so advances in general-purpose computing engines make more capable controllers available for CNC systems.
C.G. Masi, Control Engineering -- Control Engineering, 2/1/2009
|
It is a fact of life that the needs of general-purpose-computing drive microprocessor trends. Simple semiconductor-fabrication economics makes mass-marketable designs very inexpensive, while making designs aimed at smaller market segments, such as CNC, cost prohibitive.
“Advances in technology help all technologies in a general way,” says Paul Nickelsberg, president and chief technical officer at Orchid Technologies Engineering and Consulting Inc. [www.orchid-tech.com]
The underlying “big iron” at the heart of any CNC machine has been under development for hundreds, or even thousands, of years. The basic lathes, mills, drills, etc., were designed in manual form at the dawn of technology.
Starting in the mid-20th Century, and accelerating thereafter, technology was developed to substitute servomotors for geared axis drives (numerical control — NC), with computers calculating intermediate data points for the servos by interpolation from waypoints (computer numerical control — CNC). As computer technology advanced from vacuum tubes to discrete transistors to integrated circuits to, finally, microprocessors, these new developments rapidly found their way into CNC machines.
“Enhancements in microprocessor technology,” says Roger Hart, manager of real-time software R&D with Siemens Energy and Automation, [www.sea.siemens.com] “have allowed more functions to be added to the CNC, and they have allowed CNC performance to be improved.”
“These widely used chips and microprocessors provide a very scalable platform and enough performance at any level,” says Karl Rapp, manager of automation and machine tool branch at Bosch Rexroth. [www.boschrexroth-us.com] “Since the software and firmware is developed for this platform, CNC OEMs can easily provide control solutions that run on the processor of the machine HMI.”
The bad news
The bad news is that sometimes there is a disconnect between what CNC machine technology needs and what general-purpose computing provides. For example, Hart reports that there is a need for large amounts of non-volatile mass storage in CNC machines, “but I don’t see the CNC industry driving that in any way shape or form. I see it following the technology, adopting the low-cost solution driven by the PC industry.”
|
|
Multicore processors promote parallel-processing architecture, which allows multiple program threads to run in parallel with no interference |
Hart says: ”There’s been a lot of hesitation to use mass storage at the machine due to vibration and mechanical resonances of those devices.”
The most effective mass-storage solution from a CNC perspective would be a solid-state drive (SSD), which would be robust against shock, vibration, temperature extremes, and other insults that are part of the industrial environment. It would have to be non-volatile to hold current-state information to avoid having to do a reset in the event of a process interruption. It would have to stand up to many read/write cycles for the same reason.
The mass-market favors Flash memory for non-volatile rewriteable memory. That poses a problem for CNC, however, because Flash memory cells can only be rewritten a limited number of times. OEMs, therefore, have been forced to adopt Flash memory, and use work-arounds to stretch the memory’s longevity.
CNC trends
From a controls viewpoint, CNC is a subset of motion control. As with other motion-control types, torque, velocity, precision, and other parameters depend on application needs. For CNC, torques tend to be high, movements (except for tool rotation speeds, which can be very high) range from moderate to extremely slow, and precision requirements vary from quite high to extremely high.
“Typically the most significant differences between a CNC controller,” says Hart, “and a traditional motion controller are interpolation of simultaneous axes and transformations between kinematic structures of the machine tool and between kinematic structures and the actual path programmed into a CNC part program. These are features that are not available in a standard motion controller, but are unique to CNC motion control technology.”
In general, human programmers provide the computer (now called a “CNC controller”) with sufficient information to specify tool trajectories through the material to be machined. The CNC computer then, through a geometrical interpolation algorithm, calculated servo-control commands on the fly to accomplish the cuts with required precision.
Increasingly, human programmers are being supplanted by automatic translation from computer-aided design (CAD) and computer-aided manufacturing (CAM) data files, eliminating one of the most tedious and error-prone steps in the process. All CNC systems, however, will continue to need a manual control mode for troubleshooting and test during installation, repair, and maintenance activities.
Another trend is a requirement to collect data during the machining process. In the past, machinists were provided with part drawings and inspection gauges for reference during the process. The best machinists could read the drawings and use the gauges to improve their operational control. Comparing how machining operations progress against how they should progress makes the difference between the CNC machine mimicing a clumsy machinist and mimicking an expert. Of course, we all want the CNC to mimic the expert!
Replicating that expertise calls for a lot of computer horsepower to do the required calculations in real time. “More and more, we see CNCs wanting to do monitoring and diagnostics in real time,” says Hart, “by simulating a process and then comparing it to what’s happening in the actual process in real time. If the simulation and the real time don’t match, you’ve got a problem!”
“It’s important to the factories in several ways,” he continues, “such as collision detection and tool breakage. By definition, adaptive control has to be done in real time, so you can’t consider things like finite element models and high-order simulation, like you can in an off-line computer system.”
“As we gain ground with multicore, high performance processors at the CNC, our ability to model what’s going on in the process and then adapt the control dynamically is going to be enhanced,” Hart concludes.
Finally, but certainly not least importantly, CNC systems, like all motion-control applications, must deal with safety issues. Where this affects CNC computer hardware most is in the need for non-volatility in its random access memory (RAM). For safety, the CNC computer must constantly track of its current position, as well as speed and direction of motion. In the event of process interruption, especially control-system-power failure, that information must be stored in non-volatile form.
Following consumer trends
Consumer electronics can provide more advantages than just lower prices for parts. “More standard real-time operating systems, development and debugger tools are available for the latest Intel compatible systems, speeding time to market for CNC solutions,” says Rapp. “Standard PC boards in smaller formats, such as ETX, are available for controller hardware. These PC boards offer a long life cycle with a wider range of CPUs, minimizing the time and effort to keep up with the user’s performance and product reliability and life cycle demands. Memory for on-board program storage is no longer an expensive option.”
|
|
Serial processing speeds computer operations by breaking a single repetitive program into pieces, each of which runs on a separate CPU core.. |
Like CNC applications, consumer electronics is moving toward SSDs. Portable devices, such as cellphones and portable media players, have the same — or worse — temperature, shock, and vibration issues as CNC systems. They have moved to Flash technology for non-volatile memory needs. Thus, Flash device prices have been pushed down, while their capacities have gone up — way up!
What’s a CNC OEM to do? Why, they follow the consumer electronics trend by adopting Flash memory for their SSD needs, that’s what. “The use of Flash SSDs provides enough memory for even the most demanding control system applications,” Rapp points out.
The problem is that Flash memory cells can handle a limited number of rewrite cycles. Now, a limitation of 100,000 write-erase cycles may be no problem for your teenage daughter’s iPod, which she’ll replace in a few months with a “new and improved” model, but it’s a different kettle of fish for a production machine shop’s CNC mill, which is expected to run 24/7 for decades. Constant updating of the machines current status in SSD memory can burn through 100,000 cycles in no time, flat!
Yet, alternative SSD technologies, which are not the darlings of the portable media player and cellphone crowd, simply cannot achieve the production volumes to make them cost competitive with Flash. Things like magnetic RAM (MRAM) and battery-backed dynamic RAM (DRAM), and even exotic technologies like ferroelectric RAM (FRAM), though they might be far superior on technical grounds, won’t supplant Flash for CNC controller memories anytime in the near future. In the meantime, OEMs must resort to less-than-optimal strategies, such as wear leveling in vastly oversized Flash memory blocks, to extend CNC non-volatile memories to acceptable levels.
On the microprocessor front, general computing needs are much better aligned with CNC needs. General computing is always looking for more horsepower, and the likes of Intel and AMD have turned from the strategy of simply boosting clock rates by squeezing more transistors into smaller space to boost performance, to parallel processing using multicore devices.
“Having more powerful processors allows having more applications running,” says Milo Grika, industrial PC product manager at Beckhoff Automation. [www.beckhoffautomation.com] Some of our clients have their CAD software running on the same system, so they can design the part that is going to be cut; they have an HMI, so they can control the cutting device; and they have the CNC software that does the cutting.
Multicore
Essentially, multicore chips provide multiple independent processors in a single package. Since they can work in parallel, the four processors of a quad-core chip can achieve nearly the performance running at 1 GHz as one processor running at 4 GHz, but without incurring the power dissipation penalty paid at the higher clock speed (power dissipation increases nonlinearly with clock frequency). “Multicore provides improved performance for CNC applications — for example, only 5 μ-sec/1,000 PLC instructions,” says Rapp.
Lower power dissipation reduces power supply needs, but more importantly it reduces problems getting rid of heat. If microprocessor A draws 5 A at 5 V, it generates 25 W of heat, which must go somewhere. Getting the same computing power with multicore processor B, with a current draw of, say, 2 A, means dropping the cooling needs to 10 W, which makes the IC package, the circuit board, the enclosure, and cooling fans much smaller. While these are only nice for desktop PCs, they are critical for mobile consumer electronics. They make a difference for production machine shops, where dozens of CNC machines convert hundreds of kilowatts to heat, which then has to be pumped out of the plant by air-conditioning units.
“With the next generation of processors, we’re able to move the power supply that would have been separate onto the motherboard itself,” says Grika. “So, now the same 3.5-in. motherboard will include a power supply, which will mean that our PC will itself be smaller. Also, there’s less need for fans, which means we can more easily create systems without moving parts.”
“We’re just scratching the surface on how to properly utilize [multicore],” says Hart. The easy solution is to be able to run software that ran on three processors in the past, such as the user interface on a Microsoft Windows processor, and the PLC, which might have been in a microcontroller, and the NC kernel running the motion planning running on a third processor.”
Microcontrollers
Another computer-architecture trend, system-on-chip (SOC) technology, comes from the embedded control world of smart thermostats, automotive engine-control modules, and computer-controlled microwave ovens. SOCs can combine microprocessor cores with important peripheral circuits, such as communications drivers, digital signal processors (DSPs), non-volatile memories, and display drivers, into devices called microcontrollers, which are proving useful for CNC applications. These different functions are embodied in modular circuit descriptions called intellectual property (IP) blocks (or simply blocks), which SOC fabricators combine to create their designs. Since IP blocks are re-usable, SOC technology allows creating complex, sophisticated designs for relatively modest non-recurring engineering (NRE) costs.
Orchid Technology’s Nickelsberg points out: “A microprocessor is the computer itself, without bells and whistles, which all by itself is not that useful because it needs a lot of support around it to make it a system. Microcontrollers are single devices with built-in external peripherals, like UARTs, or USB ports, or timers.”
While the market for microcontrollers is only about a quarter of that for microprocessors, and is fragmented by a larger variety of designs, microcontroller production volumes are still large enough to provide economies of scale, especially for designs aimed at applications requiring a range of capabilities that align well with CNC needs.
Microcontroller applications tend to be extremely power-dissipation sensitive, so microcontroller fabricators put a lot of effort into minimizing power needs of their designs. Again, this aligns well with CNC application needs. Altogether, these characteristics make basing CNC controllers on carefully selected microcontroller designs very attractive.
Microcontrollers particularly attractive to OEMs building CNC systems are those with multiple processor cores, advanced communications capabilities (especially instrumentation/control bus blocks, such as Ethernet), and display drivers. Co-processors, such as digital signal processing (DSP) blocks, can be important for CNC applications, as well. Some microcontrollers even have on-board field-programmable gate array (FPGA) blocks, which afford OEMs the opportunity to create field-upgradable CNC controllers.
“Some microcontrollers have middle-performing processors with a lot of peripherals,” says Nickelsberg, “such as timers and pulse-width modulators, that are really targeted toward the motion-control industry as opposed to, say, the telecommunications industry.”
Future: $4 motor controllers?
Leveraging these trends in consumer-electronics technology make it possible for OEMs to design CNC control systems with greater capability, at less cost, and consuming less power than would be possible otherwise. “We get requests for super cheap dc brushless motor controllers,” Nickelsberg reports. “When I say 'super cheap,’ I mean 3-4 bucks at the most! On the other hand, motion control in larger-powered systems has been going toward DSP as opposed to 8-bit microcontrollers for quite some time. If you get over a few Watts, a DSP looks very attractive for controlling a motor. Certainly over 50 Watts, a low-end DSP is a nice way to control a motor.” For those applications, SOC microcontrollers with DSP cores provide an economical choice.
“Even though microprocessors continue to evolve and get higher performance and more parallelism,” Hart points out, “there are always uses out there waiting for more computing power. In the future I see more optimal utilization of multicore technology. I also see continuing evolution in how to use the higher-speed communications that are coming out on a regular basis — both wired and wireless — within the CNC industry.”
Real-time modeling for modeling and diagnostics is still compute bound. “We’re still looking at microprocessor enhancements to solve some of those computational limitations,” says Hart. Today, it’s much more difficult than it could be.”
View All Webcasts
