2009-09-24

Spirit Sol 259

Thisol's TUL is Julie Townsend, and she's a little distracted. She's trying to finalize her (re-)enrollment at Stanford, where she's getting her Ph.D. She's somehow finagled a deal where she'll basically do her regular job here at JPL and get a Ph.D. for it.

Damn, I didn't even know we could do that.

Today is the first time we're trying to plan three sols at once. It's a Friday, and the idea is that we'll plan the entire weekend today -- sols 259, 260, and 261 -- and return Monday. As far as the RPs are concerned, we have two sols to plan: on 259, we do a little IDD work, and on 261, we stow and bump backward uphill.

The uphill bump is the tricky part. We drove down to Conjunction Junction with a specific plan, part of a never-to-be-repeated (well, probably) wind experiment. Before conjunction, we imaged this area, then drove onto it, and the remaining part of the plan is to back out the way we came in and image it again.

As always with these vehicles, the hitch is that distance is easy but precision is hard. And this drive is all about precision. The images will be taken immediately following the drive, which means we have to choose the image pointing now, without knowing exactly where the rover will actually end up. To make life just a little more complicated, we need to get the rover back to the same spot as the pre-conjunction imaging, to make it easier to compare the images. Chris has already done the drive -- it's a simple one, just a straight backup -- but he hasn't allowed enough for slippage.

When coming downhill, we slipped farther than we'd expected -- about 45cm on a 1.5cm drive, giving 30% slip. I tell this to Chris, and he makes a face. "Thirty percent slip -- on this terrain?" he says.

"I was surprised, too," I shrug. "We were expecting about 10%, maybe 20% in the worst case. But both Rich Petras and I came up with the same number." I show him the result in the HAZCAM mesh in RSVP; when you model the rover driving 30% too far, the result in the front HAZCAM mesh looks almost exactly like the actual post-drive mesh.

Chris frowns again. "We shouldn't have slipped that much," he muses. He thinks for a minute. During conjunction, we got NAVCAM images from our new position. It should be possible to compare these with pre-drive NAVCAM images, which would give us a better slip estimate. Chris does this, and what do you know -- according to his results, we actually slipped only about 7%.

Good rover. Good girl.

But what the heck is up with that? Rich Petras isn't around, so we can't ask him how he did his analysis, but we realize that he must have based it on the HAZCAMs, since that's all we had at the time. The RSVP-based analysis I did is also ultimately based on the HAZCAMs, for the same reason. And since the HAZCAMs give less precise ranges, we postulate that the HAZCAMs just misled us.

We might be doing the wrong thing, but the smart gamble is to trust the NAVCAMs. So we do that.

Not for the first time, I marvel over the amount of blood, sweat, and tears that we sometimes shed over a matter of centimeters. Here, we sweated for an hour over a matter of maybe 20cm or so -- less than 8 inches. But when that 8 inches could determine whether we manage to nail a one-of-a-kind scientific experiment, well ....

So Chris's skepticism probably saved the experiment. But I have him beat: he saved the day only once, but I manage to save the day twice. I am on the ball.

The first one is definitely a case of Murphy's Law in action. Early in the mission, I wrote a flight-rule checking script, called fr_check, to help us RPs avoid dumb mistakes. One of the rules it enforces is this: we're supposed to open the MI's dust cover exactly when we're going to take an image. We should open the dust cover, take an image -- or a series of them -- and then close the dust cover again. On rare occasions, we deliberately take images with the dust cover closed.[1] (The dust cover is not opaque but a translucent orangish material; cover-closed images are poorer quality, but are helpful in making false-color products.)

Since fr_check can't tell whether we're taking a cover-closed image on purpose, this rule turned out to be a little "chatty," too eager to warn about cases where in fact everything was OK. Warning too frequently can be just as dangerous as not warning at all, since it can lead you to ignore actual problems, so one of the other RPs recently asked me to dial it down a little.

And sure enough, thisol -- the first MI-using sol after I made the change -- the change caused fr_check to fail to warn about an actual problem. Thisol's MI work is a little unusual -- normally, we take a series of images at different distances from the target, but because we're so downlink-constrained thisol, we just took a single image. And we forgot to open the dust cover. Since there was just one image, the amount of MI work fell under the rule's new threshold, and it decided to keep its mouth shut.

Luckily, I catch this manually and add the appropriate commands to open and close the dust cover. (It ends up returning an absolutely perfect image.)

I save another science observation as well. Originally, it was our practice to call instrument-on and instrument-off sequences in the RP sequences. For instance, we'd place an instrument such as the MB and call the sequence that turned it on; later, we'd turn the first instrument off when switching to another instrument. For various reasons, we don't consistently do this any more, so it's reasonably common to have an IDD sequence that doesn't turn on the instrument it places. Instead, the master sequence -- the day's backbone -- will start or stop the instrument at the appropriate time.[2]

But today, after reviewing the master and IDD sequences, I realize something is missing. "Hey, did I miss it," I ask, "or did we not turn on the Moessbauer?"

Nope. We did not.

"That would have been embarrassing," I say wryly. One trivial change later, and I've saved a twelve-hour science experiment.

So even though planning three sols keeps us here late, I'm in a good mood. I always am, when I know I've earned the paycheck.

[Next post: sol 265, October 1.]




Footnotes:

[1] Sadly, on both vehicles, the dust cover has now accumulated so much dust that it is opaque. Better the dust cover than the lens itself, though.

[2] Now we always do it that way. It's easier to make sure you're getting something right when you do it the same way all the time.

No comments: