Astr 680 Class News

Final Exam: Wednesday December 15, 1-3 pm, usual classroom



Dec 8:
A PDS by any other name would still smell as sweet...
Some closing remarks about power spectra:
1. The power spectrum is often called a "power density spectrum" PDS or a "power spectral density" PSD.

2. The Fourier transform definition has factors of 2 pi floating around since there is no universal standard definition. This is especially true if you use angular frequency "little omega" instead of just frequency. The 2 pi is often split between the FT and the inverse FT, so sometimes you'll see 1/SQRT(2 pi) as a coefficient of both.

3. In astronomy, the power spectrum is often normalized in a funny way: the input light curve is divided by the variance of the light curve. Then the white noise level in the PDS should be equal to 2. This has the advantage that the significance of peaks in the power spectrum can be estimated. It has the disadvantage that the units of power are no longer meaningful. Hence, once often sees the term "arbitrary power" on the y-axis. It is important to note that the significance level estimator assumes random, uncorrelated, Gaussian, white noise. If the noise is not white, then the significance estimates are totally wrong. This is the case for most turbulent/chaotic situations or whenever the "noise" has any memory of the past. Accretion disks, AGN, winds, shocks, stellar flares, etc., exhibit non-white "noise". ("Noise" is not really "noise", meaning error or uncertainty. Rather, it is intrinsic, broadband, non-periodic variability. The term "noise" is a bit of a misnomer and confusing, since all the "signal" could be in the form of "noise".) There is no standard way to define the power in a power spectrum! So choose something convenient and be sure to define it somewhere in your papers.

4. If a light curve is unequally sampled, don't despair: there is a procedure that will sort-of work. It is called the Lomb-Scargle power spectrum (or Lomb-Scargle periodogram). Calculating a power spectrum requires some artistic sensibility, since one really computes the PDS of tapered and detrended data - not the observations themselves. And calculating the power spectrum of unequally-sampled data is really something for the Great Masters of Art. But it can be done.

Dec 7:
Homework extension: Homework number 13 will be due on Friday 5pm. See the important email I sent out about problems with "ft1.o" and hints on how to compute the power.

Nov 30:
(1) It is vital that you read Ch 12.0-12.1, the source code "four1", and the figure showing the input/output array structure for FFTs in Numerical Recipes. Also read Ch 13.0, and the first part of 13.4. Both of these chapters provide an outstanding introduction to FTs and power spectra.
(2) In case it wasn't obvious, the motivation for giving the 5-minute talks in class is that being able to give a good talk is crucial for success. It is a required skill for scientists and teachers. Like anything else, the only way to become proficient at giving talks is to practice, practice, practice. The "practice" should be as realistic as possible - you need folks in the audience who know more than you do and will ask hard questions - that's practice! Lecturing to undergrads and the public does not help much - it is not the same.

Nov 17
Dr. Welsh will resume lecturing on Monday Nov 22. You will get a new homework assignment on that day. Remember that your class presentations will be coming up soon - perhaps Wed Nov 24.
In case there was any doubt, there definitely will be questions on the final exam written and graded by Dr. Eztel.

Nov 4
There's a slight conflict in times for next week's Friday class with Dr. Etzel: 2pm is also when the Scholarship award ceremony takes place. So the class and Dr. Etzel will have to come up with a new plan. In any case, there will be no Astr 680 class on Monday Nov 15th.

Nov 3
Note the due date: Homework assigned on Wed Nov 3 is due on Wed Nov 10.
Reminder: The word "data" is plural! Likewise for "spectra". Nothing says "I'm an amateur!" more than misusing these words.
Useful Reading:
For spectroscopy, the "User's Guide to Reducing Slit Spectra with IRAF" will be very helpful. I highly recommend you read it, at least the first few sections. Also, the StarLink guide "Simple Spectroscopic Reductions" is very useful, especially sections 1-4 (the rest of the document is specific to Starlink software). The Starlink documents are hung off our class website. The IRAF documents are available at the usual IRAF site.
For photometry, the IRAF guides to "apphot" and "daophot" are verbose but essential. Apphot is simpler and easier, so you may want to start there. However, DAOphot is much more powerful (for crowded fields, such as globular clusters), and can do everything apphot can do and more. Either way, get familiar with the material in these guides.

Oct 25
Midterm exam on Wed Nov 3rd, usual class time. Closed-book exam (no notes), but calculators allowed. A suggested exam-taking strategy:
a) answers the questions you know first - these are like money in the bank.
b) do not spend too much time on questions that are not worth many points.
c) don't attempt to write a "perfect" answer. Write down the majority of the information and go for the next big catch of points. Don't spend 10 minutes getting the last 2% of a question correct. Go for the big bucks, then move on. If you have time, you can go back and fill in any gaps.

I presented two plotting packages in class, "PGPLOT" and "grace" (xmgrace). There are many additional packages, e.g., "mongo" and "SM", as well as graphics capability within non-graphics specific software, e.g. IRAF, IDL, Matlab, NCAR, NAg, etc.. The point is that you need to know how to plot/graph things so they look clean, accurate, and professional. Whatever graphics package you choose, be sure you become proficient. Take a look at a recent Astrophysical Journal paper to see what the figures look like. Yours should match their quality from now on.

Oct 12:
Notes on signal-to-noise ratio.
We are halfway through the homework assignments and to date the class average on the first 6 assignments is 43.8/50.0; this is good. A rough threshold for what I consider to be unacceptable quality work is an average below 40.

Oct 11:
1) If there are any objections to postponing the midterm exam by a week, please let me know. Right now, the midterm exam is scheduled for Wednesday October 27.
2) Regarding the AAS "class project" poster paper: This is a voluntary thing and will be treated totally independent of the Astr 680 class in terms of lectures, homework, exams, and grades. So we should not call it a ``class project'' since it is not - it is an extracurricular activity. If folks are interested, we can choose a time to meet and work on TrES-1. Friday afternoon when there is no colloquium is a good possibility. Even weekends are possible.
3) Here's a comment on the first question for homework #6. If you haven't already guessed, this has to do with the famous "Central Limit Theorem". Essentially the central limit theorem says that the distribution of the means of *any* distribution of independent random variables (with finite mean and variance) will asymptotically approach a Gaussian. Think about this - any shape distribution will yield a Gaussian if averages of samples are taken. Pretty amazing, huh?! And that is the reason why the Gaussian or Normal distribution is so ubiquitous. The homework asks you to use a uniform (flat) distribution. But if you have the time, try it with any distribution. Or, do a search on the web - there are many really nice java and javascript-based animations showing the effect. Just about any good mathematics textbook will have some mention of the central limit theorem. It is interesting, but do keep in mind that the convergence to a Gaussian is both approximate and asymptotic. This means that in some cases it will be close to, but not exactly a Gaussian. The homework is an example of this. [WHY?] And even in those cases where it does converge to a Gaussian, you may need an infinite number of samples to actually get there.

Sep 26:
- Required class at MLO: Fri Oct 1, 4:00 pm to about midnight. Hopefully we'll get done earlier, but it depends on weather and how efficient we are.
- For MLO: Be sure to bring a notebook to take notes, both for the lecture and the on-hands work at the telescope. Be absolutely sure to read the MLO Handbook and User's Guide to Direct Imaging and CCD Photometry with the MLO 40-Inch Reflector (2004 edition).
- BBQ plans: We need get buns & fixin's for the BBQ at MLO on Friday. Dr. Etzel is donating the hamburgers. Let's make plans on Wed for what we'll need and who might volunteer to get things.

Sep 22:
It is final: we'll meet at MLO on Friday Oct 1, as planned, rain or shine. Meet at the visitors center at 4pm. We'll have a class for an hour or so, then break for dinner, then meet again at the 40" telescope at about 6:45 pm. We'll definitely be done by midnight. If it is really poor weather, there is still plenty we can do, but we'll have to try again the next night.
Be sure to bring a late night snack and warm clothes: hat, scarf, etc., (but you do not necessarily need wear them the whole time, unless you are Carlos!). Check the MLO weather page to see how cold it has been getting at night.
- No office hours on Wed Oct 6th.

Sep 20:
- Note on programming: Be sure to always test your code with simple cases where you know the answer. This is a necessary part of writing code. Get into the habit of doing this.
- Guest lecture by Dr. Bob Leach on Monday Sep 27 on the topic of CCDs. Please review the material in Howell's book as prep for the lecture.
- I have yet to decide on a final date for the MLO class. I'd prefer Saturday Oct 2, because Oct 1 is the transit of the newly discovered extrasolar planet TrES-1. But it may still be Fri Oct 1. In any case, it seems like a BBQ is a good plan for dinner.
- As some of you may have noticed, some versions of xfig do not properly close the PostScript file they produce. For some unknown reason, sometimes the very last line is missing. To fix this, so that you can actually print something, add the command "showpage" to the very end of the PostScript file. So for example, the last lines should look something like this:
$F2psEnd
rs
showpage
 Use the unix "tail" command to quickly look at the end of a file.

Sep 7:
Dr. Welsh will not have office hours on Friday (Sep 10) or next Mon, Wed (Sep 13, 15) because he'll be at MLO.
Reminder: MLO BBQ on Sat.

Sep 3:
Note typo in HW#1 Q 4: The question should read "...uncertainty in magnitude (sigma_m)", instead of the incorrect "(sigma_m / m)". This typo has been corrected and the LaTeX and PostScript files revised. Thanks to Tamara for pointing this out.

Sep 2:
For Questions 2,3 in HW #1, use the propagation of errors formula at the end of Ch 3 in Bevington (see page 48). Because of the upcoming holiday on Monday, the class is a bit off pace and part of the homework is one class ahead of the lectures. My apologies.

Aug 31:
Errata/typos in textbooks (according to wfw):
Bevington & Robson:
- Eq. 1.10, 2nd line: limit in wrong place
- On page 196 (a bad page for the author!) the following errors:
- - - Eq. 11.7 : should be Gamma(n+1), not Gamma(n-1)
- - - Eq. 11.8: exponent should be (n+1/2)
- - - Eq. 11.10: should be lowercase p inside integral
Howell:
- Figure 2.1: In the bucket analogy, realize that each bucket is slightly different, therefore holds slightly different amounts.
- Figure caption 3.6 (page 41): should be "inverse gain".
- Fig 5.2: In general, use many more sky pixels than this.

Aug 30:
Please try to attend the Department BBQ at MLO on Sat Sep 11. We will use this opportunity to tour the facilities at MLO. In particular, I'd like you to:
- see the dormitory room you'll be staying at when observing
- see the 40" telescope and control room
- get acquainted with operating the visitor's 21" telescope (for public viewing) especially if your are a TA.
Remember to bring a flashlight and warm clothing! Be careful of rocks and cows on Sunrise Highway....

We will have class at MLO on Friday Oct 1.

Meet at the Visitor's Center at 4:00 pm, rain or shine. We will have a one hour lecture first, then break for dinner, then meet in the 40" dome/control room at about 7:00pm. We should be finished by midnight so do not plan to sleep at MLO. If the weather is so bad we cannot open the dome, plan to meet again the next night (Sat) at 7pm. Remember to bring a flashlight, warm clothing, and dinner + snack. If you have a camera, that could be useful to bring.

General Comment on Programming: The first few written homework assignments are purposely designed to be easy to give you time to acclimate to SDSU and get up to speed in scientific programming under the unix environment. Don't squander this time! Use it to become fluent in a scientific programming language (fortran is best). Although programming is not astronomy, you can't do the latter without the former, and that's why there is such emphasis on programming in this course. A recent survey of graduate students echoed the sentiment: more programming in Astr 680 was thought to be very important. You need to be proficient in programming to do research in astronomy. So if you aren't already a decent programmer, use this time to become one.