Northwest passage

As a product engineer at Analog Devices' Northwest Labs in Beaverton, OR, Ron Simonson puts new IC designs through rigorous tests. His tests verify that a device meets engineering specifications, and he also develops application circuits that go into data sheets to help engineers with their designs. Simonson evaluates devices such as log detectors, rms detectors, I/Q modulators and demodulators, log amplifiers for communications circuits, and RF ICs.

T&MW: What is your role in evaluating a new IC?

Simonson: When I receive the first pieces, I measure voltages and currents to make sure that the device draws the proper amount of current and that it drives the specified load. I also test for stability and frequency response. At first, I manually test a part on the lab bench. When I reach a point where I need to repeat a set of measurements many times or I have a time-consuming test, I automate the data collection.

T&MW: What do you use to automate your tests?

Simonson: I use several software tools to automate tests and analyze test data. For instrument control and data collection, I use Agilent Vee running on Windows 98. I often write code modules and DLLs in C to enhance Vee's capabilities. I also get sample programs from the Vee e-mail user group that I can modify to suit my needs. I then move the data to a Linux PC where I use gnuplot to view waveforms and use Octave for data processing.

T&MW: Why do you still use Windows 98?

Simonson: I use it because it lets me write directly to a PC's I/O ports. Many of our devices have registers and I can use the parallel port to write to them. Writing to a parallel port is more difficult with Windows 2000 or XP.

T&MW: Why do you use Linux tools such as gnuplot and Octave?

Simonson: I'm a Linux user, although I use Windows for instrument control. I often test dozens of parts to get a feel for a design's typical performance. With Linux, I can use tools that don't bloat my data files. For example, my tests can easily collect over 1 Mbyte of data. If I move that data into Excel, the file can easily exceed 30 Mbytes when you add math and plots. Although I can use Vee to store data in a Excel files, I don't because I find that writing directly to Excel while a long test runs isn't reliable. If I run a test over a weekend and the PC crashes, I lose a lot of data and time. I prefer to save the data in binary format because it's more reliable and the files are small.

With gnuplot, I can leave the data in its original file, and gnuplot will read and plot it. If I need to manipulate data with FFTs or other functions, I use Octave, an open-source program similar to Matlab. I just write a script and Octave produces the results I need. Because I prefer Linux, I don't use Matlab, even though it comes with Vee.

T&MW: What can test-equipment makers do to make your job easier?

Simonson: Test-equipment makers may use proprietary binary formats to save data. They don't always clearly specify data formats for these proprietary files. I also find that instrument manuals have errors relating to programming. On rare occasions, example programs don't work.

I'd also like oscilloscope makers to get Windows out of their instruments. I want my scope to make measurements and collect data, not run Excel. I have scopes that run Windows 98 and I'm forbidden from connecting them to the corporate network. I had to create a local network in the lab using a PC running Linux that has two network cards. The PC lets me connect instruments to a local network while isolating them from the corporate network.

From a performance perspective, I need instruments with better dynamic range. I'd also like a scope that doesn't change its sample rate when I change the timebase setting.

Every other month, we will publish an interview with an electronics engineer who has test, measurement, or inspection responsibilities. If you'd like to participate in a future column, contact Martin Rowe at mrowe@tmworld.com.