OBSERVING PROCEDURES
As described in Section 4 (of the previous web page), a sky-source "flat field" calibration was performed each night, shortly after sunset, before turning on the CCD cooler. Dark frame subtraction was forced upon each exposure (by changing the exposure time slightly). Exposure times were chosen that kept the maximum data number at ~50% of saturation (i.e., 50% of 65,535 is about 30,000 counts). Many flat frames were summed, then rescaled with the maximum data number at 65,535. (I never median combine sky flat field images, for the reasons given in the main web page.)
Blazar observations were started after CCD cooldown. I usually allow ~20 minutes for the CCD properties to stabilize after cooldown has been achieved. During this stabilizing period I do telescope pointing and focus determinations.
A focus sequence using the brightest star in the CCD's FOV was performed several times each night at about hourly intervals. Previous focus data shows a strong correlation of telescope temperature and sine of elevation angle. I use charts prepared from least squares fits to previous focus determinations, and during an observing session I note the offset from the chart for every new focus determination. The night's focus offset from the chart is then used as a guide for when focus checks are likely to be needed, or for resetting focus by estimation.
The CCD's FOV was adjusted until the autogider chip provided a bright star for the AO7. For the observations reported here, a star having Mv = 9.77 was used for the AO7. This required that the CCD's main chip FOV be offset slightly to the east (that's why the images on the main page have the blazar offset to the right). Since I want all images to have the star field in the same approximate pixel locations for all observations (for all dates, even), I maneuvered the telescope to place a specific star (the "121" reference star) at a specific pixel location (950,510) of the main CCD. Based on the brightness of the AO7 guide star, an exposure time for the AO7 is chosen. This is usually in the range of 0.1 to 0.5 second, leading to tilt mirror adjustments being made at ~2 to 8 Hz.
Exposure times were either 5 minutes, or 2 minutes. An observing "cycle" consisted of several light frames followed by 1 or 2 dark frames (of the same exposure). If the number of light frames was 4, then I used one dark frame; if the number of light frames was10, then I used 2 dark frames. I like my dark frame duty cycle to be ~20% (which is close to the signal-to-noise optimum of 25%, a whole new issue beyond the scope of this web page). All cycle choices led to a total light frame exposure of 20 minutes. In all cases the AO7 recovered its guide star after the dark frames, when the camera shutter for both the main chip and autoguider chip was closed.
For most observations I monitored FWHM as the light frame data were downloaded from the CCD camera and displayed on my monitor. This was possible because during an exposure as long as 2 minutes there's time to perform a dark frame and flat field calibration, followed by a manually perforemd aperture reading of FWHM for the pixel location of a bright star. This information provided feedback on the need for focus adjustments (which can often be estimated from charts for the system's focus setting versus telescope temperature and sine of elevation angle, prepared from previous observations).
At the start of each cycle I log the object being observed, outside air temperature, wind speed and direction, focus setting, AO7 settings (exposure time & aggressiveness), bin setting (if different from 1), exposure per frame and elevation angle. I have a Davis Weather Monitor II, with a data logger connected to a notebook computer on my observing desk. During an observing cycle if I determine a FWHM I log it, and maybe log a specific star's intensity. Monitoring intensity is useful in detecting the presence of clouds or sub-visible cirrus. Sometimes I do a quick aperture photometry on the image that's displayed, and can usually get a quite useful indication of the blazar's magnitude from just one 4-minute exposure, for example.
My observing logs are kept with an ink pen, which prevents erasures. My observing logs are never modified or annotated after observations have terminated for the night. Remarks at later times, such as pointing out a probable mistake, are made with a light pencil. Observing logs are "sacred!"
ANALYSIS PROCEDURES - ONE NIGHT'S OBSERVATIONS
Analysis of the observations, performed the next day, are time-consuming, I must admit. You have to love data anlysis to do this amount of work.
All my analysis logs are made with a pencil, and all printed graphs are labeled with their filename, associated spreadsheet and creation date and time (very important information). Part of my analysis philosophy is to avoid mistakes by being careful. Records of what was done (during observations and analysis) allow for later detection of mistakes, and easier correction.
An observing "cycle" consists of several "light" and 1 or 2 "dark" exposures. Dark frame subtraction requires that at least 3 dark frames be median combined (to prevent dark frame cosmic ray defects from causing dark areas in the dark-frame-corrected light frames). If only one dark frame is taken per cycle, then I need to use dark frames from neighboring cycles to get the required 3 (I prefer 4, to be safe from coincidence cosmic ray defects at the same location on 2 frames).
The light frames of one observing cycle are calibrated using the night's flat field and the set of 3 or 4 closest-neighbor dark frames (I let MaxIm DL do the median combine by checking the appropriate box in the calibration window). All images are set to the same brightness and contrast, and reviewed for cosmic ray defects close to any of the stars that are to be used. If only a small defect is present near a reference star, I edit the affected pixels using a representative nearby one for copying. I don't like doing this for the blazar, so if there's a cosmic ray defect near the blazar that image is rejected. Each image is also checked for sharpness. Any that "stand out" as being too fuzzy are rejected. The set of calibrated light images is then median combined. This, hopefully, removes all cosmic ray defects.
After all cycles are processed they are imported to MaxIm DL for a photometry analysis. The single reference star is chosen, and assigned it's known V-magnitude (from Arne Henden). The blazar and several other (7 for this particular blazar) are specified as "objects." The 8 objects are entered in a specific sequence every time, to facilitate later spreadsheet analysis. I use the "snap to centroid" option. Then the "View Plot" window is selected and the first aperture/gap/annulus (AGA) choice is set. The magnitudes are calculated automatically, as soon as a AGA is set, so I record the results to a text file (comma-delimited CSV-file, containing the magnitudes of the reference star and all objects). All 7 adopted AGAs are set and recorded.
Air mass must be calculated for each observation cycle. This is done using TheSky, with user-specified date and time. Mid-point times for each cycle are read from the FITS header for the middle cycle of the cycle (adding 1/2 the integration time).
The spreadsheet is used to compare average measured magnitude with predicted (true) magnitude for the reference stars. Agreement of <0.1 magnitude is usually achieved, even over a several magnitude range. This assures that linearity has been preserved throughout the observing process, and that an erroneous magnitude for the reference star was not entered.
The spreadsheet is then used to calculate average magnitude for all objects having the same AGA. The average object magnitude for all AGA sets is also calculated. A graph displays the pattern of AGA minus the object average for the 7 AGA values. An unusual pattern is cause for further study, before proceding, as this may indicate an undetected cosmic ray defect in one of the images, or a bad FWHM image.
The air mass values for each cycle are entered in the spreadsheet. A least squares fit is made for all stars (except the one reference star, "121") with air mass as the one independent variable. The fit value at air mass of 1.2 is recorded in the analysis log for each star. The choise of air mass 1.2 is intended to minimize the influence of stochastic uncertainties and systematic errors. This is like a "hinge point" for a typical data set, since about half the data are at lower air mass values and the other half at higher values. Useof the zero air mass intercept is subject to greater uncertainties.
At this point there is usually present a persistent pattern of cycles being high or low with respect to their extinction curve that's shared by all stars. This "image bias" pattern is calculated and removed from all data. Differences are calculated of observed magnitude from model fit magnitude (incorporating an air mass slope and with the image bias removed). The population standard deviation of these differences for each star, including the blazar, is noted in the log. I'll refer to these later as "Population S.E. Graphs of these differences versus time are created and printed. The graph for the blazar is inspected for a pattern that is possibly related to blazar brightness changes.
Finally, on behalf of the search for the presence short term blazar variations on this specific night, a plot of Population S.E. versus star magnitude (at air mass of 1.2). A model trace is adjusted to fit these data. The model consists of a constant plus another constant diivided by intensity (where intensity is calculated from magnitude using the stnadard equation). The blazar's location on this plot is noted. If it is higher than the model fit, then this night is a candidate for blazar variations. If it is consistent with the model fit (i.e., the other reference stars, adjusted for their brightenses), then there is no evidence for blazar variability.
This last analysis, of Population S.E. versus star brightness with a model trace fit, can be used to quantify the blazar's variability (or lack of it) for that night's observing period. For example, if the blazar's variability is B mMag, and the rest of the reference stars predict a value for the blazar's brightness of R, and if B>R, then the blazar's variability, V, can be estimated to be V = (B2 - R2)1/2. By playing with various reasonable alternative values for B and R it is possible to derive an uncertainty for V.
ANALYSIS PROCEDURES - COMPARING DIFFERENT NIGHT'S OBSERVATIONS
Long term variability involves comparing the average magnitude for the
blazar with that from other nights. This sounds straight-forward,
but subtle corrections are called for.
Return to W Com main web page
This site opened: June 17, 2003. Last Update: June 17, 2003