Why does it have to be so hard to make a test system? If it's not the PC connectivity, it's coaxing the hardware into actually working. If it's not the hardware, it's the software—the O/S, the application, the languages, the drivers! I've about had it with some vendors—"the givers of pain and delight," to borrow a Star Trek phrase. Can't anybody make test-system development easy?
I've defined test system platforms and have visited many customers who are confronted with difficult test problems. In keeping with the trend of all the best-selling books these days to abuse the number "seven," here are the top seven frustrations I have seen and how I deal with them:
7. Racking and cabling. There's an art to this seemingly simple task. Put heavy items like power supplies at the bottom. Use a good power distribution unit (PDU) that can handle 100-240V ac and that has remote control, emergency off switch input, line-fault detection, and UPS interface. Allow about 20% more EIA units than you need for future expansion. Don't put magnetic fields near CRTs.
6. Grounding. With so many grounds—safety ground, power ground, floating inputs or outputs, high/low remote sense, PC grounds like RS-232C common, quiet analog bus commons, low sides of 2-wire relay matrices—how do you avoid massive ground loops? Use a star ground located as close as possible to the device under test (DUT).
5. Interface Panel. A mass interconnect panel and related fixtures can easily double the cost of a test system, but it's worth it. Invest in a reliable mechanism that mounts to the system and one removable fixture for every DUT or class of DUT.
4. Switching. For functional test applications, the best place to put relay cards is in a dedicated switchbox. Placing simple switches in an expensive card cage that's tuned for high performance instruments is a waste of valuable space. Use fast reed relay matrices with instruments on the rows and DUT pins on the columns to provide easy expansion and the ability to hook any instrument to any pin. Use GP relays to connect power supplies and loads to DUT pins. Plan for more switch cards in the future.
3. Instrumentation selection. Choose the modular and rack-and-stack instruments that have the performance you need at the right price. Modular instruments cost more than their rack-and-stack counterparts, but may have functionality you can't get elsewhere. If you use a card cage, try to use a non-proprietary, fast interface like LAN or Firewire to make PC interconnect simpler.
2. Choice of controller. Embedded PCs in a modular card cage cost a lot more than standalone PCs, and they are often not available with the latest technology.
1. Software—but hey, it's getting better. With no previous knowledge of Microsoft's new Visual Studio.NET, but with existing knowledge of Visual Studio 6.0, I designed and built a brand new test system and wrote a complete test application in one week. Never had to even crack open a manual because of the built-in help. The system had LAN, GPIB (via a USB/GPIB converter), Firewire and RS-232C interfaces, and the software didn't complain a bit. Maybe the givers of pain and delight are finally inching toward the "delight" side.