Tue, 12/27/2016 - 03:03
How does VPhot calculate SNR? Is it based entirely on the square root of the signal in the measuring aperture? Is the signal from the sky background involved in the calculation? Is the same method used when generating the curve of growth?
SNR is calculated using the "CCD equation" involving ADU, gain etc, ref. the AAVSO CCD observer manual:
"the CCD equation: S/N = Nstar/(Nstar + n(Nsky+Ndark+(Nreadnoise)2 ) )1/2 where N is the number of photons received from each of star, sky, dark current, and the readnoise of the CCD, and n is the number of pixels in your measurement aperture"
Yes, this certainly is the way to do it if you know the dark current and the read noise, but VPhot doesn't have access to the master dark and bias frames used to calibrate the image. It has to determine the SNR of the star using just the calibrated image, knowing only the gain and the ADU values of the pixels in the measuring aperture and sky annulus.
While pondering this question I opened my copy of Howell (Handbook of CCD Astronomy) and out fell two printed pages identified as a "handout" by J.H. Simonetti dated 2004 January 8 and titled "Measuring the SNR of the CCD Image of a Star or Nebula". There was no other identification, URL, or copyright statement. I think this printed handout must have been placed there by a previous owner of the book. Google found a John H. Simonetti, a professor of physics and astronomy at Virginia Tech, so this handout may have come from one of his classes.
I've attached a copy of the 2nd page which, I believe, nicely answers the question. The equation in the box is written in terms of "counts" (ADU's). I think that some version of this equation must be what VPhot uses.
Thank you Dr. Simonetti.
The details of the computations in VPhot are detailed in the attached PDF.
What is the state of the "searching" function?
"Finally, if the image is plate solved, a photometry list of all stars is produced and stored in a file, with
the same name as the image but ‘phot’ as extension. These photometry files are used to enable
variable star searching"
The "VS Search" function on VPhot's Images page has been disabled.
The decision was made a year ago because:
- no one appeared to be using it
- creating the phot files for every image as they were loaded was a significant load on the system.
- it was only a "beta" feature, with still lingering problems.
To tidy up we should remove the item from the menu OR decide if there is interest and programmer power to fix it and implement it so its not a significant load on the system.
VPhot maintenance guy
The speed in which VPHOT handles our data is of the highest priority. Having said that, the value of a constant stream of plates capable of photometric measurements to me has incredible value. The AAVSO database could be considerable enhanced, certainly for long term baseline/variations in field stars otherwise unobserved. I fairly confident there would be other uses as well. If it were a question of consumed memory storage, I for one would be willing to archive my own data and make it available to the AAVSO for other uses.
BTW: your continuing contributions to the AAVSO are appreciated
Happy New Year!
Thanks very much for posting that document on the internals of VPhot. I found a lot of interesting information in there, particularly the method VPhot uses to correct outlying pixels in the background sky annulus. I may have missed it, but didn't find a specific explaination of the SNR algorithm. Presumably, it's in PhotLib.
It does say:
SNR is defined as S / Sqrt(S / (G + Ns * Std^2 * (1 + 1 / Nr)))
S: Total ADU in aperture
Ns: Number of pixels in aperture
Nr: Number of pixels is sky annulus
Std: ADU standard deviation in sky annulus
G: Gain of the CCD detector
I should someday squint at the pixels in apertures and count ADUs to see if it all adds up.
An undocumented feature is how averaging improves SNR or how averaging improves the error. I am on the trail of this.
If you take SNRs or errors that VPHOT provides from averaging 1 to 36 images, you find that after averaging about five to eight images, the improvement flattens out. It is sort of a fourth root (or maybe a log10) improvement rather than a square root improvement as more images are averaged. SNR gets bigger, error gets smaller, so they got the sign right.
Could anyone share information about how VPHOT averages?
Does that equation at the beginning of your comment (#9) come from the Phot.Lib? I have seen exactly that equation quoted before in a comment in one of the photometry courses.
But shouldn't the numerator "S" (total ADU's in aperture) have the sky background signal subtracted? This is done in the equation shown in the attached class hand-out in my post (#3) above. In that equation the numerator is given as
Con - n(xsky)
where Con is the total counts (i.e. ADU's) in the measuring aperture, n is the number of pixels in the sky background, and xsky is average pixel value in the sky aperture. It seems to me you should subtract the sky background signal from the total signal in the measuring aperture to get the signal from the star. Am I missing something here?
Oops! I found the SNR equation you quoted in the VPhot DocumentationA that George attached to his post #6, but I'm still having trouble reconciling that with the Simonetti version from post #4.
Yup. A person should be able to inspect VPHOT star detail and plug numbers into the SNR equation they use and come up with approximately the same SNR. Number of square pixels in a circle approximated. Been to busy to do it here.
It's the VPHOT averaging that has me confused.