Reliability HotWire Hot Seat Interview with Dr. Robert Herrick

Dr. Herrick is a senior quality and reliability engineer for fiber optics at Agilent Technologies, in San Jose, California. His primary responsibility is for overseeing vertical-cavity surface-emitting laser (VCSEL) reliability for the sources used in Agilent's fiber optic transceivers. His secondary responsibilities are in assisting other engineers in transceiver qualification, and reliability improvement programs. Dr. Herrick did his Ph.D. work at the University of California Santa Barbara, studying the causes of VCSEL failure. Prior to that, he did research in high-power semiconductor lasers at McDonnell-Douglas in St. Louis, Missouri.

Agilent Technologies is a spin-off from the Hewlett-Packard Corporation, specializing in instrumentation, communication, and life sciences. Agilent's fiber-optic transceiver division is the dominant supplier of fiber-optic transceivers in the enterprise networking market, supplying gigabit Ethernet transceivers to equipment vendors such as Cisco, Nortel, and Alcatel.

Which ReliaSoft applications do you currently use, and how long have you been using them?

I've used Weibull++ for most of the time I've been a reliability engineer: over three years. I use it for a few hours at a stretch, every two months or so. I was a beta tester for ALTA, but I use it only a few times a year.

What, if any, reliability analysis software packages did you use prior to using ReliaSoft applications?

Our organization (then Hewlett-Packard) formed a committee to study all available reliability software in the market, and concluded ReliaSoft had the best package. They made it the standard, so it's the only one I've used.

Have you been able to reduce your test and/or analysis time due to the use of ReliaSoft applications?

I would say the primary benefit for us is in reducing analysis time. We have greatly reduced our test time, through proving out higher acceleration factors, but this was independent of the ReliaSoft software.

Have you been able to implement any other improvements to your testing process due to the use of ReliaSoft applications?

ReliaSoft products have saved a great deal of time in analyzing and visualizing our reliability data. Many of the most laborious analyses now require only one click in ALTA. Many of the graphs are ones you wouldn't have even thought to have created on your own before we had the program.

Have the use of ReliaSoft applications made it easier to spot potential problems with products under test?

Yes, although the problems in question could have been easily detected with less sophisticated tools, or by looking at the raw data.

Have you been able to increase the reliability of your products due to the use of ReliaSoft applications?

To understand the failure distribution, and analyze large data sets, specialized products such as Weibull++ and ALTA are needed.

I guess I would mention two things: first, the technical support is top-notch. I've always received quick, thorough answers to every problem I sent in. Second, the training they offer is outstanding, not only educating users on many of the finer points of reliability analysis, but also letting them know the programs inside and out. I learned about many useful features this way. I've strongly recommended ReliaSoft training classes as the starting point for new employees of our department, as well as for our suppliers.

What would you like to see ReliaSoft do in the future to improve current applications, or add to the current family of tools?

The main thing I'd like to see is a feature allowing you to fix some parameters (e.g., sigma, or activation energy), and then have the program calculate median time to failure with that user-adjusted value. Why? Sometimes, a few stray points can radically change the line fit, even though the overwhelming bulk of points fit a different slope better. The "management results" requires the user to then adjust the other values by hand, or set up a least-squares fit routine themselves in Excel. The other thing I'd like to see is the ability to easily censor certain points from the fitting routines, while still plotting them. I want to see the mavericks, but don't necessarily want them skewing the rest of my results.