[Aavso-photometry] Dark Skies / Thin Air

[arne]

ben at davies.net wrote:
> I had never imaged in dark skies, so last weekend I lugged a 12" SCT up 
> to the top of a 7,000 ft mountain.  I brought back some BVR  images, 
> have been looking them over, and am a little confused.
> 
> The trouble is, in 60 second images there is no sky noise at all.  
> Zip.  The average sky background runs something like 104-105 adu 
> (before dark subtraction) and the dark frames run 105 adu as well.  
> Read noise is ~7 adu.
> 
> This is an SBIG ST-10 camera at -25 degrees and processing is with 
> AIP.
> 
> Both single images and track and stack, with bias, darks, flats and 
> flat darks, leave typical mean PV of the sky background at around minus 
> 2 adu.  I am thinking that it might be read noise fluctuations are 
> causing the negative values.  (I'm pretty sure) there are not negative 
> photons out there.  The only other possibility I can think of is 
> differential noise in the bias frames.  I didn't do them up there 
> because the batteries ran out. Drat.
> 
> 1. What can I do about the negative pixel values?    The signal is 
> already as saturated as I want it to be.  And, the snows have come to 
> the Sierra - so there will be no more images from there until at least 
> next summer.
> 
> 2.  What is the nature of Read Noise?  I had always assumed that it 
> was addative - that excess electrons were generated by the alu.  But 
> maybe it is actually just uncertainty expressed in units of electrons?
> 
Readnoise is a random process, and involves the conversion process
from electrons to volts, so is a little complex to describe.  Basically,
all manufacturers apply a bias to the output signal so that the ADC
always sees a positive signal.  Once you subtract bias and dark, that
subtracts *mean* bias and dark, and the definition of a mean says that
there should be as many negative as there are positive excursions about
that mean.  Therefore, if there is no sky background, you should see
small positive and negative values that average to zero.  I never
worry about positives and negatives much except when converting back
to an integer fits image, as floating-point fits images don't care
and software *shouldn't* care.

What is your image scale in arcsec/pixel?  It takes a really, really
dark sky to result in no sky background - I always have sky in wide-bandpass
filters like Johnson/Cousins.
Arne