[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
sources of the noise floor
Arne makes some informative comments on TN 76.
> First, with a 1mmag quantization, you have to add that
> to the observed noise.
> Second, if you used ensemble methods for *all* stars
> detected at least 5 times, you weight the zero point
> towards the faint stars (obviously, there are more of
> them). Try with just a subset of stars around 9th mag.
Yes, I did several experiments with different subsets of
stars: the brightest 1 mag, the brightest 1.5 mag, the brightest
2.0 mag from the saturation point. There wasn't a significant
difference that I could see. I am already weighting the
solution towards the brighter stars ....
> Third, at these levels, the accuracy depends on how
> you form the digital aperture and sky annulus, and how
> you subtract sky. Sky is significant for all Mark IV
Very true. As I mentioned in the Tech Note, one of the bugs
I discovered in the course of this experiment was in the
counting of pixels within the aperture. My current method
does include fractional pixels, weighted according to the
fraction which falls within a perfectly circular aperture.
There is an approximation I make in this calculation; I don't
think it's significant for apertures as large as four pixels
in radius, but I could check it.
> [informative notes in the dark and flat frames]
Yes, I should work out the amount of noise one expects
from these sources.
> Bottom line: don't worry about the noise floor until
> the images are taken specifically to get the highest
> photometric quality.
I'm glad to hear Arne say this. I wanted to satisfy my
own curiosity on this point, so I went ahead and analyzed
the Data Disk 16 set -- it contained repeated measurements
of the same stars. But, as Arne says, the mode of observation
wasn't designed for "highest quality photometry."
The silver lining here, I think, is that these frames were
taken in what I would describe as "a typical mode of operation."
Someone who doesn't spend time thinking about signal, noise, etc.,
but simply wants to use the Mark IV in a "general-purpose" fashion,
might very well operate as Tom did on this particular night.
I think that there is some value in learning that images
taken in this "general purpose" fashion may be expected
to have internal uncertainty less than 1 percent (for the
brightest unsaturated stars), but nowhere near 0.1 percent.
Of course, an experienced photometrist like Arne probably
could have told us this off the top of his head, if we'd asked.
I'm just not so experienced, so it did me some good to go through
the exercise of measuring the scatter. Now, when I start to
design an observing program of my own, I can take some of these
effects into account and check to see how the changes improve