Thursday, December 17, 2009

Debugging

My advisor has a paper that he is writing for a conference that is due on Monday. He requires data from me for this paper - and I've known this for a while. We have a meeting on Friday, and I have a list of things I must have ready for figures in the paper.

Having data and figures requires me to do experiments, also something I am aware of.

So after my classes ended last week, I breathed a large sigh and went home and took a nap. However I find that despite my fervent wishes, the experiments have still not done themselves.

I tried to do the experiments on Tuesday, and discovered that the Google calendar had the equipment I needed already reserved. I tried again on Wednesday, and made ten parts. I measured these parts to get the data I needed, and found that the parts were all duds. Bugger.

So today is my last day to run the experiments, measure them to collect the data, and then analyze the data and make pretty graphs. This is possible to do all in one day (though not pleasant), so I got an early start this morning to knock it out.

And then, the vagaries of research struck.

My machine can't seem to make good parts.

Oh, wait. I got one good part - the one that came out of the cycle when my equipment had an error. Because THAT makes sense.

I only have three major variables I can change on the machine - A, B, and C. I have been making these parts for two years - I am an EXPERT at making these parts. I know what values A, B, and C should be to make good parts.

So when I started getting bad parts, I followed my usual debugging routine. In chronological order:

1) Make sure all mechanical pieces are fine (nothing on the machine is broken)
2) Change variable A (the easiest to change)
3) Change variable B (fairly easy to change, though tedious)
4) Change variable C (difficult to change)

Nada. Still bad parts. Then I went into my secondary routine, one at a time trying:

1) Change A a LOT
2) Change B a LOT
3) Change C a RIDICULOUS amount because it's a bear to change and I don't want to do this again.

Still nothing, in fact the parts are maybe even worse. At this point, I go home for dinner.

Now I am back in lab, pondering the problem once more. I adjusted minor mechanical issues, thinking perhaps magically a loose screw was the root of all this evil. Unsurprisingly, it was not.

I thought perhaps I should come at it another angle, so I adjusted A in the OPPOSITE direction that I thought would help. Now the part was worse.

Okay, I can do something with that - if something I did made the part worse, I can go the other way and make it better. Except that my results are not conforming to that very logical argument.

WHY must this be happening? Why is it not working like ANY OTHER DAY?

I've worked all night on this, trying every combination possible. I don't want to go to this meeting tomorrow with nothing to show, but I don't know what else I can do. Look sir, here's three petri dishes full of parts. None of them are any good, but here are the seventeen things I tried to make them better.

Does saying "This is SUPPOSED to work, and I DON'T KNOW why it's not" make me a bad researcher? Does "I have a bus ticket to go home for Christmas on Sunday" make any difference when the paper is due Monday and I have no parts to give him?

I hope my advisor can accept the "I'm having trouble here" speech, otherwise I have to dig into my next debugging routine: 

1) Wiggle my nose
2) Tap my heels three times...

No comments:

Post a Comment