Hammer it home

Roy Timpe, Engineer -June 09, 2011

Hammer it home imageSeveral years ago, I worked for a company that makes expensive fiber-optic transceivers. The company worked with a vendor that made burn-in racks for hundreds of the laser diodes that these transceivers used. It was my job to check out and debug any problems in the design. The racks had hundreds of relays that would switch the system’s voltmeter to appropriate nodes to measure the voltage and current of any given laser. Just by looking at the schematic, I saw a problem. The rack vendor had no series resistors in the relay connections to the voltmeter. If a bad software command simultaneously energized two relays, they would tie parts of the circuitry together that were not intended to connect and perhaps even blow the lasers.

I believe I mentioned that they were expensive lasers. So I asked the company to add the series resistors, which they begrudgingly did.

Because these lasers were new, initial volume did not allow me to fully load the racks. Small batches ran fine. However, as production ramped up and I finally loaded the racks with lasers, trouble started. As the racks cycled through their voltmeter measurements, bad voltmeter readings would intermittently occur. Often, every reading after the initial bad reading would also be bad.

The vendor’s software would interpret bad readings as a bad laser. The software would then turn off the bad lasers and mark them as bad in the database, so they never experienced the full burn-in. This problem threatened my company’s production goals. As it turned out, relays that should not have been operating occasionally remained on when they were in the rack.

Oddly, this situation was not the result of some software glitch. I gained access to some of the relay coils with a DVM (digital voltmeter) and figured out that the relay was going in at the proper time but was not releasing, even though the coil was no longer energized.

When I looked at the routing of the power and ground to these high-power lasers, I understood. The rack vendor had simplified the routing by separating the high-current power and ground cables to different sides of the rack. Worse yet, the power cables looped up and around the relays. That big loop of high current created a magnetic field—too weak to pull in any of the relays but enough to keep a relay or two in even when the test program had turned off the coil current.

The technician I was working with was skeptical that the field with only one loop could hold in a relay. During the next malfunction, I carefully placed a large ball-peen hammer near the card cage with the relays. The hammer head coaxed enough flux from the offending relay, and good voltmeter readings resumed. The technician was now a believer.

We routed the power and ground cables as close together as we could, but there was no easy way to remove the big loop of power cables that circled the relays. So we fashioned a ¼-in.-thick steel plate to shield the relays, solving the problem. Nevertheless, the test group still had to flip the plate over every six months or so because it was becoming magnetized and not working as well as I had intended.

It was a good thing that the vendor added the series resistors because these shorting events were sure to have blown out some of the lasers—something your burn-in rack should avoid doing.

Roy Timpe is a senior engineer with 27 years of experience in measurement development and automation.

Also see:

Loading comments...

Write a Comment

To comment please Log In