Saul isn't here yet.
Talking about the redone data of the first 7 SNe. So far, of 6 redone, 4 got dimmer by about 0.3 magnitudes. In at least one case, the data really is dimer! What's going on? What's the source? Unknown. Of course, for the fits, there is also the effect that the K-corrections being used are different from what they were before.
SN3 is yet another odd case. The stretch used to be... something very low, and now is just over 1. What happened? The data doesn't look enough different. This may be likely due to the K-corrections.
One thing to note: on all of the redone stuff for the first seven, there have been no instrumental corrections applied as of yet. (We can hope that for R-band it will all be small, but there is a question of Mould vs. Harris filters.) Apparently Alex is the one who knows about the instrumental corrections.
Saul just got here.
It would be nice to check to see exactly what the K-corrections are doing. Peter says he has the old Alex K-corrections, so we will try fitting with those to see what happens.
Question of zeropoints: finding which zeropoints we used, where, may be a bit of a challenge.
Yet another gory discussion about K-corrections is impending. Greg and Saul were asking Peter how much the underlying spectrum affected the K-correction, or whether it was just the color that was driving things. I.e., if you throw any SN Ia spectrum, with the right color, at it, do you always get the same K-correction. Peter did a plot, first doing things as right as he can, and then he also did them all fixing them to the color of day 0. They K-corrections in fact weren't flat, but were only so in 0+-10 days. Beyond that, the dispersion goes up, and K-correciton as a function of epoch is no longer flat. So, it depends on more than _just_ the color you get out of it.
It does say that for the first shot of our fitting, we don't need to know the date of maximum to better than 10 days. Another conclusion, based on the dispersion, is that it looks like it's OK to substitute color for stretch (but keeping the right epoch!) when doing your K-corrections. This would suggest also that the affect of extinction is just on the colors as well. To do K-correction, we should be able to spit in asking for the K-correction at a given epoch and color, rather than feeding in epoch, stretch, and extinction estimate. These would go into the templates.
Latest HST reamage. This week they've been telling us over and over again about problems we're going to have with our observations... e.g. that we'll have to tell them exactly where the SNe are two or three months before we find them. (I exaggerate.) Right now, Peter is telling us that if we want to switch filters, we have to budget and use 8 minutes to change the filters, even if we don't move. The reason is that that's the maximum conceivable time to reqcquire guide stars. Even though it will take a lot less than that in our case, the shutter stays shut for the full budgeted time....
For moving, in addition to this 8 minutes of acquisition, the slewing is 2 minutes for a handful of arcminutes, and 4 minutes for a degree. So, getting two SNe in one orbit, you waste >12 minutes in between them. (Sad.)
Greg has S/N numbers. For redshift 0.83, planetary camera, 700s at R gives 5% photometry _at the peak_ of the supernova. 1000s in I at the peak gives about 5% photometry. The 1700s is what we were able to fit into an orbit before, including a split for each filter. Regarding the cosmic ray rates... in 1000s, about 2% of the CCD gets whacked by CRs. (I.e. 2% chance that a given 10 pixel area gets zapped.) Of course, the shorter you go, the nonlinearities and CTE get worse.
At z=0.9, we can't afford to take twice as many candidates. At z=0.5, perhaps we could squeeze two in.
Which straegy to do? Peter argues that we should do one split, and three non-split (for a total of five SNe). (Saul wonders what our original proposal said.) One of these will turn into a snapshot SNe.
What about doing one more? There are several (we listed four) good reasons to put more into March, using our HST orbits on the more optimal run. The one reason not to is the eggs-in-one-basket theory. The distribution is 4 orbits (meaning 5 SNe) now, 8 in March. Are people happy with this? How about 5 and 8 (with again one of the first orbits split for a snapshot thingie), and (to squeeze this in) pitch some of the late NICMOS data on the SNe, which are less likely to be useful.
Data stealing: right now we're planning 5 WFPC and 5 NICMOS points per supernova. Greg says, hey, grab 13 of those NICMOS orbits, only do the first 4 points, and do something else with the NICMOS orbits. One question is if we can turn these extra orbits into WFPC orbits.
Now we're debating again whether or not we're not doing enough orbits on the January suprenovae. It seems we're back to doing 5... meaning 3 full orbit spent on solo targets and 1 split of two bright ones, one of the two of which will only be done once for the (Sebastien rolls eyes) snapshot distance. Full R, I, and NICMOS for the first point. Later, we drop R-band for the highest redshift guys. This means that we're putting of doing 8 or 9 in March and we're counting on the March search panning out.
NOTE: Do people agree with this? Collaborators please comment to deepnews.
Sebastien, CFHT. Many things are bad. We have B in references, V and I in news, seeing is bad on the news (2x the references) according to Susanna. Right now, the APM matching isn't working. There is also a horrible problem of dark currents that it varies from one frame to the next. Matthew tells us it looks almost like some sort of flickering. For the reference runs, the sky background is about 300dn. The dark current level seems to vary by as much as 60-70dn in some of the chips. There may be patterns to it, but it's hard to tell. The other four chips are more consistent. On another night, the CCDs that were consistent and "flickering" all swapped. All of this is worst on the references.
Much sadness. There seems to be some public sentiment for spending no more human resources on these data. The color mismatch, the bad seeing, and the other troubles; it looks like we're not going to get anything out of this data. Greg says, what the hell, just try blinking them, see if anything booms out. Saul says, do that and stop.
Things to worry about for the upcoming run:
Peter introduces a new thought: it's not worth following 0.9 SNe from the ground. Just get enough to confirm our HST calibrations. Other than that, the ground based data is much better spent on the more nearby supernovae.