[Aavso-photometry] CCD 'fainter-than' question

[Richard Huziak]

Bob, et al,

The discussion that supposedly answers Bob's straightforward question 
seems to have become quite theoretical and has not, IMHO, provided a 
very practical answer.  Although the theory is 'fascinating', it doesn't 
necessarily actually get you a measurement in a reasonable amount of 
time, and certainly doesn't keep a new(er)bie interested.

So here is some practical advice.  First, with a faint image that you 
are struggling to see (and even with a bright image), a single 
'detection' really is quite meaningless because it cannot be verified as 
representative.  This is the nature of any single datum anywhere.  To be 
certain, you want to repeat the measurement, and if you get the same 
answer within the 'expected' error (ie - mostly common sense), then you 
can begin to believe that you have a reliable measurement set.  We've 
heard recently about ghost images, noisy pixels, saturation, sigma 
detection, yada, yada, and all that is fine except with a single image 
you really don't know which of these effects has polluted your 
measurements, if any.  So...practical advice. If you want to shoot and 
get a good measurement, shoot a number of frames - 3 - 5 and then 
inspect them all and reduce them all.  If they all turn out to have 
values within a tight range, then you can believe the results. If they 
(or single points) are scattered, you can have ghosting, cosmic rays, 
poor skies, bad flats, hot pixels, etc - a host of things that cause 
errors.  But you can reject oddball points because you have something to 
compare them to.  At this point, you can either report all your 
individual measurements and errors (easiest) or stack your good frames 
and report a single averaged measurement which now has a better SNR but 
is smeared over time.  In this way, you can have a bunch of SNR=5 
high-error points or one SNR=50 low-error time-smeared point.  It all 
depends on what you want the data for.  If you need to measure a 
threshold (for the TOADs) and don't care about micro-variations then 
stack and measure for higher SNR.    However, a single hit point has so 
many possible errors at low SNR that I simply do not believe the results 
myself.  

As a practical example, we have to reject most single-point hits on the 
HMXB program because we cannot 'understand' the errors in the 
measurement, though in a submitted cluster of points, we'd have a better 
chance.  And since we are hoping to see light curves varying 
inter-observer by <0.03 mag, it is not practical to have single hit 
points from multiple observers, since many are clearly /way/ off the 
zeropoint for unknown reasons.  Same as with the recent request to 
monitor RE J1255+266.  I could have submitted a single image showing the 
star at 19.302 +/- 2.315 mags, but I chose to stack 10 or 12 frames for 
a better detection and much smaller error or around 0.150 (ficticious 
numbers - representive from memory). SNR was around 4 for any one image, 
but you can either work harder and produce a measurement, or cheap out 
and provide a less than.  Less than's result because your equipment is 
generally incapable (simply working beyond your practical limit) or 
simply from poor sky conditions that you judge to be giving you 
unreliable data.
 
So - expose a few frames, if data good over a few frames - submit the 
data.  If shaky - stack and submit with stdev as the error,or however 
you want to calculate it.  If you are not confident, then take the 
brightest of your cluster of images, and report that value as the less 
than value (not the mean, since you have a 50% change that the star is 
indeed brighter than what you reported).

You really don't have to know much theory to do this.

rick (HUZ)  :-)


Bob Crumrine wrote:
> For CCD observing, what constitutes a 'fainter-than' observation?  If I 
> can see the variable on the image but only 10:1 S/N (50 minimum needed 
> for observation, below), is that a 'fainter-than'?
>
> Thanks,
> Bob Crumrine (CRR)
> near Rochester, NY
>
>   

-- 
* * * * * * * * * * * * *
Richard Huziak
Manufacturing Engineering
SED Systems
Saskatoon, SK, Canada
tel. (306) 933-1676
<huziak at SEDSystems.ca>
* * * * * * * * * * * * *