Multiple images of one object

Hi Xavier:

Good technique, taking more than one image per run! If I may be pushy, I would recommend that you first analyze each image to see how much random noise you normally get with your system. How well do your individual magnitudes agree? With 3-5 individual magnitudes you can calculate your mean and standard deviation. I think this is a realistic error and much better than what others often report as their error (1/SNR)!

For reporting, IMO it is best to stack the images (this also improves SNR) and report one magnitude. Reporting all 5 magnitudes for a target that should not exhibit any significant variation is just quantity over quality. You could have reported your mean and std but VPhot doesn't do that automatically so stacking is suitable. Of course, this assumes that your target doesn't make any "real" variations during the measurements.

BTW, do you do ensemble or one comp photometry? Do you have multiple filters? You could also run them and learn more about the star. Yea, I know, more than you asked for!  

Ken

My take on reducing equivalent images of same target:

• First choice is to stack dithered images, then reduce the stacked image. This is what we will all do when CMOS cameras with on-camera registered stacking become available (and they will).
• Second choice is to reduce separately, then report weighted average (weighted by each image's uncertainty), reported with weighted uncertainty as well. The math is not trivial. Due to constraints on my workflow, this is what I do now; I am trying to rewrite my software to adopt the first choice above.
• Third choice is to reduce separately, then report equal-weighted average (careful of outliers).
• Fourth choice is to remove outliers if any, and then select one of the equivalent values at random, report that. Not a great choice, but at least fairly unbiased.
• Last choice--probably not really acceptable--is report them all separately. I would never do this on purpose.

After listening to inputs from the experts and experienced users here, my recipe is similar. I am not suggesting that my recipe is "correct".  Just saying that, after some experimantation with blue, red, bright, dim, scarce-data, plentiful-data stars, I have developed my own guide-lines that make me happy. 

I like to reduce the solved images (which usually happen to have a few pixels of dither between them). Then stack them with VPHOT. After many experiments, the number that I stack depends on the magintudes of vars, comps, and checks. For var or comp or check of mag 10 to 15, I stack/average 10 to 20 1-minute images. I sometimes stack 40 to eek out a bit better error bar. For mag 7 to 10, I stack 3 to 7, depending on B, V, or I. For magnitudes 4 to 7, I might stack 10x10-second images to avoid saturation. I often use short exposures for red stars in the I band to stay below 1% nonlinearity.

For my camera, more than 47500 ADUs is 1% nonlinear. I know that because I measured it, plotted and analyzed it with a spreadsheet. I developed a way to do that measurement with repeatable results. If you are an electronics expert, and have some kitchen utensils, I can reveal the secret sauce.

In general, I stack 10 minutes of exposures because the atmosphere is the major  limitation for my location. Ten minutes beats the twinkle and much of the high cirrus and contrails which seem to be ever-present. I often run VPHOT through the images to find the outlyers. Those get deleted, then stacking happens. The variance that you got for the time series of 5 images does (in my opinion) yield a usable error-bar. However, after suggesting that some years ago, I have gone back to letting VPHOT come up with the error-bar after doing my best to clean up what I am submitting to VPHOT.  

I average to achieve an error of less than 0.01 magnitude. If I observed from the Atacama desert, I would likely do less averaging. For being less than 1000 feet ASL I average 10 clean, inspected images for every data point submission.

One does need to be careful with averaging if there are rapid changes of magnitude, a planet occultaion, or peaks of fast variables are examples. Averaging 10 or 20 minutes of data can yield a fuzzy ingress/egress time or lower the magnitude at a fast peak.

Often, I have a bright var that is widely reported and another star or two on the image that have no reporting. In those instances I may have 7 or 10 images. VPHOT may report an error of .005 for the bright star and .05 or even 0.1 for the dim one.  I'll submit the dim one with big errors because my input is all that is in the data-base, and 0.1 mag is my cutoff for as good as I can do visually. 

When I VSTAR my own (secondary) data on those dim ones, I am sometimes rewarded with a nice phase diagram. Sometimes I get a noisey, but definite LPV. Great fun. 

You might plot the error bar size versus number of images stacked vs magnitude. I know those numbers for my equipment. Experiment, find your own recipe. 

Thank you for all your answers. I am a beginner and for the time being I am using Itelescope. Also I only measure using the V filter while learning the basics. I always compare the results I get with other observations close in time to see that I am not reporting (too) wrong data, I also measure using AstroImageJ to compare results.

I've been reading a about photometry but there are still many concepts that escape me, what do you mean by reduce the measurements.

Thank you for your replies.

Xavier

My routine is to take  4 or 5 images of an object.  If the SNR of my target is good (>100), I then perform photometry for each image.  If low SNR, then I stack the image.  After performing photometry of each image, I will toss any outlier then report the average and report the error as the std deviation of the check star of similar magnitude as the target.

Barbara