Previous development testbeds have assumed that the developer was physically present in front of the hardware being used. No provision for remote operation of basic functions (power on/off or reset) was made, because the developer/operator was sitting in front of the hardware, and could just push the button manually. In this innovation, a completely remotely accessible testbed has been created, with all diagnostic equipment and tools set up for remote access, and using standardized interfaces so that failed equipment can be quickly replaced. In this testbed, over 95% of the operating hours were used for testing without the developer being physically present.
The testbed includes a pair of personal computers, one running Linux and one running Windows. A variety of peripherals is connected via Ethernet and USB (universal serial bus) interfaces. A private internal Ethernet is used to connect to test instruments and other devices, so that the sole connection to the “outside world” is via the two PCs.
An important design consideration was that all of the instruments and interfaces used stable, long-lived industry standards, such as Ethernet, USB, and GPIB (general purpose interface bus). There are no “plug-in” cards for the two PCs, so there are no problems with finding replacement computers with matching interfaces, device drivers, and installation. The only thing unique to the two PCs is the locally developed software, which is not specific to computer or operating system version. If a device (including one of the computers) were to fail or become unavailable (e.g., a test instrument needed to be recalibrated), replacing it is a straightforward process with a standard, off-the-shelf device.
This strategy has paid off several times over the developmental effort. It made it very easy to construct a “portable” version of the testbed to take to a remote site to test the flight model radio: the two PCs were rented laptops, and copies of the required interface boxes were rented or borrowed. Everything was plugged together, the software was loaded, and a few hours later, testing could commence. Compared to the traditional approach of a rack full of customized interface drawers and customized PCs, it was much simpler and less expensive, as well as immediately responsive to changing project needs.
In fact, the experience of creating an ad hoc test capability at a remote site has resulted in some minor changes to the overall design: for example, rather than individual serial ports, the system now uses USB to serial adapters, all plugged into a USB hub, so there is a single USB connection to the Linux PC. Likewise, in the laptop configuration, the private internal network is implemented with USBEther net adapters, because most laptops come with only one Ethernet interface.
This work was done by James P. Lux, Minh Lang, Kenneth J. Peters, and Gregory H. Taylor of Caltech for NASA’s Jet Propulsion Laboratory. NPO-48013