||[Nov. 15th, 2018|09:28 pm]
sprockets, sockets, grommets & gaskets
Work problems, which aren't real problems:|
1. Looking at a huge stream of data and an eagle-eyed coworker pointed out a lot of repeated numbers. The instrument producing these is supposed to provide six significant digits: 1.23456. It's doing that. But between, say, 1.23456 and 1.35000 it's only providing (multiple instances of) 1.23456, 1.24567, 1.26789, and 1.31234. If you have twenty megabytes of data and plot through this and do a linear regression it looks completely uniform: nice, well-behaved data. If you go look at the individual data points after sorting them, you see this stair-step behavior that is for sure not what's actually happening.
The instrument that's producing this is introducing digitization artifacts, even though I've set it up to not do so, and I ended up having to rewrite the driver that interfaces with the instrument to get it to improve (but not entirely solve) the issue.
2. So with that hammered out, now I can look more carefully at the actual data, and I see all these weird dropouts, where a very small value that should be increasing monotonically is instead increasing monotonically with a second monotonic line somewhat below it. That took a huge effort to track down and I still don't understand it: when I agitate an output controlled by a relay, the relay control line stops working. They're not connected that way. This doesn't make any sense. It's like if you unscrew the lightbulb suddenly the wall switch physically snaps down into the off position. That's not supposed to happen.
So all this is going on while my manager's wife is just about to have a baby, and another project that I'm not on is turning into a huge flaming dumpster fire that I can't talk about but is absolutely catastrophic both for how badly it's going, and for how badly it's going to affect our group, and my manager can't do anything to make his wife's situation better (she was due today) and can't do anything about the dumpster fire so instead he was running over to try to see why my setup was acting up, because it's both interesting and potentially easy to fix. Except he doesn't software and when he looks at my code he clutches his head.
(I have no idea how to run my oscilloscope. I know exactly how to set it up to produce the measurements I need by programming it. So he sits down and says "how do you put cursors on the screen?" and that's not something I would ever try to use, so I have no idea, but I can quite literally get the oscilloscope to play a recognizable song by beeping a series of data points I send to it.)
So he's channeling all his frustration with all this other stuff into handling what are really fairly simple problems I'm having, because then he doesn't have to think about what he's not able to do.
Which isn't the worst thing in the world.
My coworker who sometimes drives me bonkers interrupts me about every ten minutes to tell me something he's done, mostly because he wants recognition that he's doing things and our manager no longer gives him that recognition because he's too busy and knows he'll just snap something unkind. So, repeated talking, while I'm trying to teach myself how to use excel to compare linear regressions to see if I can say with confidence that they're statistically similar.
Then he puts on a pair of hearing protectors "so I can concentrate".
This entry was originally posted at https://randomdreams.dreamwidth.org/53250.html. Please comment there using OpenID.