Can't account for excessive noise in my FLAT frames

[saublestarman]

Dr. Robin Glover
I have been studying shot noise in my astrophotos and have been using FLAT frames to study this noise. I used a constant LED light source and constant camera gain (QHY294C GAIN=1600) with the cooler off (16 Celsius). I then took FLAT frames at various exposures. I thought of the sensor analysis in sharpcap so I extracted statistics of a rectangle of pixels located at the brightest region of the FLAT frame. I determine the mean values for each color of the raw RGGB pixels. I also calculated the standard deviations for these four collections (R,G,G,B). I did all of my analysis using my raw FITS FLAT frames

I plotted the standard deviations vs the mean values and initially thought that there was a square root relationship as is understood for Poisson noise or shot noise. What I soon discovered was that the standard deviation values are significantly higher than expected. It’s as if there is an additional noise source that is very strong. I use an OFFSET value of zero for my QHY294C. Just considering the red pixels…The 0.01 second FLAT has an average pixel value of 46 (16b) with a standard deviation of 13. My 4 second FLAT has an average pixel value of 10033 with a standard deviation of 238. If the entire pixel value of 10033 been attributable to shot noise I would have expected a standard deviation of sqrt(10033) = 100. If there was an additional noise source it would have to be huge.

A couple of years ago this camera passed the Sharpcap sensor analysis. I have closely studied my BIAS and DARK frames and they are well behaved and are in keeping with the charts provided by QHY. The camera seems to take very good photos but now I suspect there is too much noise in my FITS files and that if I could identify the source of this noise I might be able to remedy it or, at least, understand it. I want to be able to study my LIGHT frames and be able to assess it's noise content by being able to attribute the noise to various sources (Read noise, Dark Current noise, Sky Glow noise etc.)

I would appreciate your comments and directions I could take to learn more about this issue.

Peter Wolsley

[admin]

Hi Peter,

I think that what you are missing is that the Poisson noise (noise is square root of value) only applies when the brightnesses are measured in electrons. You are looking at the ADU values of the pixel (and the ADU value of the noise). ADU is linearly related to the electron count, ADU = gain * e + offset. If we ignore the offset for now and assume the gain is perhaps something like 4 ADU per electron, we would calculate about 2500e from your value of 10000, Sqrt(2500) is 50e noise, which we then convert back to ADU and get about 200, which is close to your observed value.

If you look at the sensor analysis data for your camera (easiest done in the 'Sensor Data' tab of the smart histogram brain window), you will be able to read off the e/ADU values for your camera at different gain settings. Note that these e/ADU values are the reciprocal of the value we used above (which was ADU per electron).

Hope this helps,

Robin

[saublestarman]

Robin,
My brain is hurting. Obviously I missed the math class where the properties of square roots were discussed. I've worked out a few examples and you are , of course, correct. Thanks so much for your wisdom.

Peter