[Aavso-photometry] When to submit"FainterThan"versusactualnumbers.

[Arne Henden]

Jeff Hopkins wrote:
> Thanks for popping in here.
> 
I can't "pop" in for long, though. My time is limited today. :-)

> My point is that you can have a standard of deviation for CCD photometry 
> if you group three readings. If there are no errors with CCD the three 
> reading should be very close if not exactly the same. The idea implies 
> that the program star would not actually change greater than the 
> measurements error during the the time of the three measurements. What 
> bothers me is it seems some CCDers are giving a standard deviation for a 
> magnitude that is not a real standard deviation (determined by the 
> spread of data) of the program star.
> 
Of course you can obtain a standard deviation with results from
three or more separate frames.  This is seldom done in the CCD world.
If the object is highly variable, such as a CV, the results from 3 frames
can be quite different.  If the object is slowly varying, such as a Mira,
then this technique of multiple combined images can be used for a
better error estimate, but most people just move to the next object
after the first frame to optimize the number of datasets.

> I wonder just how valid using a surrogate star is. So what if it changes 
> over the set of exposures. It may well do just that due to changing sky 
> conditions. The results will show a large standard deviation, but how 
> does that have any meaning on the accuracy of an individual program 
> star's magnitude determination? Since as you say all stars are measured 
> at the same instant (or close to the same), variation of the surrogate 
> star or comparison star from image to image should not matter, only the 
> actual difference between the comparison star and program star. A 
> standard deviation of the surrogate star would provide an indication on 
> sky quality, but have no value indicating the quality of the program 
> star's magnitude determination.
> 
I think you need to do CCD photometry before you can get a feel
for this process.  It actually does work.

> In reality photon arrival and scintillation of the light (which both 
> follow Poisson statistics) from the comparison and program stars is 
> unpredictable and will vary from image to image and star to star. I have 
> spent hundreds of nights observing this and see where one star has a 
> fairly constant set of counts while another may be constant for some 
> sets and jump around a lot (not only decrease, but sometimes increase 
> significantly) for others. In other words, there seems to be no 
> predictable correlation of this scintillation. The best that can be done 
> is to try to smooth it by averaging data. This is where the grouping of 
> the three sets of measurements should help. In my observing I have 
> reduced instrumentation error to near zero. Can CCD instrumentation 
> error be assumed to be zero? Since I am watching counts real-time I can 
> easily spot a drift of the star and other possible errors. The only 
> significant errors are the changing sky and the scintillation of the 
> star light. Large sky changes can be fixed by waiting for the sky to 
> settle (e.g., cloud or contrail to pass) or packing it up for the night. 
> Perhaps the long period CCD exposures sufficiently reduce the 
> scintillation to a non-significant level.
> 
Granted scintillation will be present on each frame, so there are
error sources that must be added in quadrature to the Poisson error.
That is why it is a lower limit to the true error.  Yes, CCD instrumentation
error is zero.  The two types of photometry are really different.
Changing sky is a second order effect for CCD photometry since all objects
are imaged simultaneously.
   PEP is still useful in many areas - bright stars or very rapid time
series come immediately to mind.  CCD photometry has pretty much
replaced PEP for most amateurs and professionals because of the wider
bandpass, better signal/noise on faint objects and multiplexing advantages.

Arne