Recently I got a 54855A oscilloscope sent back to me for service under the 1 year warranty I underwrote for most of the unit I’ve sold direct. The unit would not turn on at all after sitting for a long time.

I looked up the forum and it turns out other people had this problem with a certain generation of Infiniiums and sometimes changing power supply or motherboard would disturb the setup a little bit and the unit might power on again. When I tried to do that, the unit does boot a quite few times but the problem randomly came back again. This is frustrating as it’s a heisenbug. I almost thought I was done when the unit worked consistently for a week then it comes back. It just looked like something component was acting borderline and disturbing the setup and got it to click.

In the past I had 54830s that’s ‘fixed’ by changing either one of the motherboard or power supply, but turns out that those were flukes, but I didn’t get lucky with 54850s this time.

The customer gave me time to troubleshoot deeper instead of just getting something to work as a fluke by just blindly changing modules which might break down again at a random time if it’s a gravitating aging problem (i.e. if you figured out a problem, it tends to be a wave of manufactured gears that’ll trip on the the same issue as they age). So I spent a whole month troubleshooting through reverse engineering the circuit and nailed down the cause and the fix.

There’s a lot of mixed info going on in the forum with different modes of failure, but those might not be the real cause, as replacing components disturbs a set up that wasn’t supposed to be in that state in the first place and the unit is prone to get trapped into a bad state when the unit ages. The real fix is to plug the path to the bad state in the first place instead of rocking the boat hoping the new combination doesn’t trigger the bad state.