20
TIL the hard way that "it worked on the bench" doesn't mean it's good to fly
Last week I got a tip from an old tech named Pete who told me to always run a vibration test on any repaired LRU before reinstalling, even if it passed bench testing. I figured he was being overly cautious since I've been doing this 8 years and never bothered. So I put a comm radio back in a King Air 350 after replacing a bad capacitor, bench tested perfect, and shipped it. Three days later the pilot called me saying the audio cut out during a hard landing at a gravel strip in Fairbanks. The cold solder joint I missed only showed up under actual aircraft vibration. I had to drive 4 hours to swap it with a loaner while the plane sat on the tarmac. Pete was right all along. Has anyone else found a bench pass fail in the real environment?
3 comments
Log in to join the discussion
Log In3 Comments
faithpatel5d ago
Man that's rough, @hannahj49 nailed it with the cold joints being sneaky.
8
sagejackson6d ago
Bench test gave you perfect specs on the ground but the real tell is how the board handles the shaking during flight. What was the specific fix on that capacitor? Did you use a standard leaded replacement or something like a tantalum that might have different vibration characteristics? Cold joints are sneaky but sometimes the part itself is the issue under high G loads. Also did you scope the output during the shake test or just listen for audio? That matters too.
3
hannahj496d ago
Nah I gotta push back a bit here. Swapping to a tantalum isn't always the magic bullet you think it is, they can be brittle as hell under mechanical stress compared to a good old ceramic cap with some strain relief. I've seen more boards fail from tantalums cracking on the solder joint than from standard leaded parts vibrating loose, especially if you don't have the board potted or something. Cold joints are a pain, sure, but sometimes the real fix is just using a cap with a bigger lead footprint so the solder has more meat to grab onto. Scoping the output is the way to go if you want real data, but honestly if you're in the field and the audio cuts out under load, that tells you enough to know you gotta reflow or rework the part anyway.
2