It all started with one seemingly simple request from renowned textile manufacturer Mileta – to replace the old floppy disk drive on the Stäubli CX860 / JC3 Jacquard looms that the company has been using since 1991.
Jacquard machines are electronically controlled looms for the production of fabrics with a plastic pattern (brocade, damask). They use a system of thousands of electromagnets that precisely adjust the position of individual warp threads at each step of weaving. This system can thus create even very complex patterns, as it allows independent control of each thread according to a programmed design.
At the time they contacted us, they still had all the designs stored on floppy disks that had to be manually replaced. Over time, the floppy disks became the Achilles heel of the whole process:
- difficult to find
- they loaded annoyingly slowly
- their reliability was deteriorating rapidly
All of this hampered production, so the company decided to replace them with our emulator.
Replacing the floppy drive was the first step
The UDM-100 emulator that we use allows us to record up to 100 “floppy disks” on one USB FLASH memory, and it is a replacement for a floppy disk drive a routine matter. We’ve tried it on CNC machines (Siemens), robots (ABB), recording machines and musical instruments (Yamaha), so we knew it wouldn’t be a problem in this case either.
At first it seemed that simply replacing the floppy drive would be enough for the Stäubli machines. But it soon became clear that the problem was deeper. It’s clear to everyone that after years, even machine control electronics has long since passed its prime.
Even if we were to replace the floppy drive, many other weaknesses would remain. A single repair would only delay the inevitable – sooner or later the machine would fail completely, with spare parts no longer available.
The solution to such a situation would of course be to buy brand new Jacquard looms. However, this would mean an investment of tens of millions, which the client did not count on at that moment. In addition, the mechanics of the machine were still fully functional . We therefore proposed a much cheaper and more efficient solution – the development of a new control system.
Several solutions were offered: to replace only the problematic parts of the control system or to produce a new control system. Therefore, we examined the existing control in detail, measured and decoded some data communications, and finally prepared a feasibility study with several solution options: we worked with proposals ranging from simply replacing the floppy disk and hard disk or its controller, to replacing everything. How did it turn out?
We have created a control system for 13,000 threads
The original wiring had to go all out. All that remains are the metal cases and the electronics of the individual cartridges with electromagnets, of which there are hundreds on the machine: they turned out to be in good condition and repairable – they use parts that are and will be available in the long term, there would be no point in redoing them.
We didn’t use the proposed emulator either, because everything was solved by a completely new control system. In addition to its development, we also took care of converting all the patterns from floppy disks to the new format.
And how does it work inside?
The first part of the system consists of a control panel with an LCD touch screen. The display shows the pattern being processed, the status of the device, and allows you to run a range of diagnostic functions. The operator also selects the pattern to be woven from the panel. Patterns are loaded from a network drive or USB flash drive.
For the development of this part, which is de facto only the user interface, the obvious choice was embedded Linux. We used one of our standard panel microcomputer units (with a TI microprocessor) and a Python application. The application, by the way, also runs on a PC, the control panel is just for operator convenience.
The second part of the system is the control unit itself. It takes the required pattern from the control panel via Ethernet and sends status information back to the panel: after the pattern is loaded, the panel is no longer needed and any failure of the panel or communication will not cause a malfunction.
The unit uses sensors to monitor the operation of the state and issues data to the electromagnet cartridges. This is a relatively fast process, we have milliseconds to set the state of thousands of solenoids to the cartridge shift registers using a non-standard historical interface when the state shaft is in the correct position.
Also, from time to time the power may go out and we need to remember correctly where we left off weaving so that we can resume work at the right place: that’s why we have a nonvolatile FRAM with unlimited writes in the unit.
The controller has no user interface except for a small service display for setting the IP address and checking the status of inputs and outputs.
The unit is based on STM32H7 microcontroller with FreeRTOS operating system.
The whole system is a typical combination of a microcontroller for realtime timing-sensitive tasks, where we take advantage of the speed and straightforwardness of a simple operating system, and embedded Linux for (pre-)data processing, visualization, user interface and more complex network communicationwhere we take full advantage of the facilities of an “adult” operating system, allowing us to quickly develop even a relatively complex and secure application.
Why didn’t we use an integrated dual-core processor?
Some applications also offer the use of processors with a powerful Linux kernel and another kernel for realtime tasks in a single casing, such as the STM32MP1, where typically Linux runs on the powerful kernel and FreeRTOS runs on the other kernel.
This solution brings the advantages of easier communication between the two systems, lower cost and smaller size of the final device.
In our application, however, this solution could not be used because the Linux and FreeRTOS parts are physically separated by a distance of several meters. However, we have successfully used this approach with an integrated dual-core processor in other applications, such as controlling robots with an integrated HMI panel, where both systems are placed in close proximity.
After the successful test of the first machine, we finally secured the replacement of the control system for 9 other Jacquard machines.
This project shows that even a seemingly simple requirement to replace an obsolete component can lead to a much more complex solution that ultimately delivers significantly more value.
Instead of simply replacing the floppy disk drive, we created a modern control system that not only extended the life of the machines by many years, but also significantly simplified their operation and maintenance. It has proven that a properly designed upgrade can be a very effective alternative to a complete equipment replacement. For the company, our cooperation has thus brought a double benefit – modernisation of production and savings of many millions of crowns.
