Need help : Amp glow or light leak Sony nex-3n modified

Thx for you answer, I will try this.

Actually, the darks were taken with the camera cap, no lens attached. So it is not an internal reflexion of the lens.

Because the pattern is the same for both your lights and darks it does look like amp glow. Most cameras have amp glow. It is more of a problem for cameras without cooling, and those built to be compact in camera size like general purpose cameras.  It is also more visible on warm nights. For my cameras that have this problem I try to have at least 30 each light and dark subframes. This is a stochastic process, so enough samples are required to get a true mean of the fixed pattern noise, and the rule of thumb from statistics is 30-40 samples are needed. It is important to match light and dark subframes. The same ISO (aka ASA, gain). The same exposure times. The same camera. The same sensor age within a year (sensors change with time). The same ambient temperature within +/-10F. The same binning. The same camera settings (e.g. automatic noise reduction settings, white balance, etc.). The same image file format (e.g. jpeg, RAW, etc.) The same image format ( e.g. small, medium, large, etc.). Of course at least 30 samples (subframes). This takes a lot of time and disk space, so it is best to try and standardize on a few combinations so you don't have to take darks for every session.

BTW, Goldstein is correct in pointing out that light leaks are a problem, and that a lens or body cap does not guarantee good darks. The camera needs to be in a similar environment for both lights and darks. You cannot use a towel around your camera (except for testing) for your darks, unless you are using a  towel for your lights.

No matter how many lights, you should have 30-40 darks. However, if you also have at least 30 lights, then the darks will be the most effective. My experience is as long as the darks are well matched with the lights, you will have all trace of amp glow removed. If you are using flats the same rules apply. Flat darks must be matched with flats, and there should be 30-40 of each. People with expensive equipment tend to use fewer control subframes because expensive cameras have fewer problems, but everyone could benefit with 30-40.

While flat darks must be matched with flats, flat + flat dark sets do not need to be matched with light + dark sets except for the camera's ISO.

I find that for many of my scopes/lenses I don't need flats+flat darks, but darks always make a big positive difference.

I make a set of darks with my set of standard settings for each of my cameras once a year and use that set with any lights I make that year. I have found 32 subframes each for darks, flats, and flat darks to be a convenient number.

To help your choose a standard setting for ISO, use this camera testing site to help. You want as much dynamic range as possible, but also you want the highest ISO to get the shortest exposures. That is where the dynamic range curve starts being non-linear; for most Sony cameras it is at ISO 100. However, depending on your tracking or guiding you may need higher ISO to get shorter exposure times; in that case choose your ISO based on your desired exposure time.

I'm trained in econometrics so this is an explanation based on the probability and statistics mathematics, and is likely not the best way to explain this.

Every subframe includes both information you want called the "signal", and information you don't want called the "noise". Noise can be either fixed pattern or completely random. Fixed pattern noise has central tendency plus error. Random noise can appear anywhere with equal frequency.

All subframes include random noise. When frames are stacked the central tendency of fixed pattern noise including the signal is exaggerated relative to the random noise. Stacking only removes random noise, but the fixed pattern noise from other sources is still there. To get a better estimate of the central tendency of the fixed pattern noise, one needs more samples. This is why all subframe types including lights, darks, flats, flat darks and biases all need to be stacked.

What is confusing for most people is the the combination of both the signal and the noise is called "shot noise". This is because the signal itself is a stochastic process with central tendency. Shot noise has both fixed pattern noise and random noise (aka non-fixed pattern noise). The fact that the image signal itself is a kind of fixed pattern noise is why increasing integration time (i.e. including more light subframes) improves the quality of the final photograph. In theory, one could continue to get improvements for the number of subframes equal to 2 raised to the power of the number of bits reported by the ADC (analog digital converter). For example, for a 12 bit ADC 2^12 = 4096, however, the benefit is decreasing exponentially, and using that many subframes is not worth the effort; hence the 30-40 rule of thumb.

The image signal is part of the fixed pattern noise that is part of the shot noise of the lights, however, camera imperfections and optical train imperfections are information signals you don't want in the final photograph. The signal in darks (aka light darks) is the fixed pattern noise of the camera. The camera fixed pattern noise signal of the darks is subtracted from the fixed pattern shot noise of the lights to end up with only the image signal plus the optical train signal and other sources of fixed pattern noise.

Flats provide data on the imperfections of the optical train in terms of the evenness of image transmission through the optical train. In particular, flats provide information about problems with vignetting, internal reflections, dust, and dirt. Flats will introduce a different fixed pattern noise from the camera than the lights because they are taken with a different set of exposure settings and in a different environment, thus flats need matching darks called flat darks (aka dark flats) to remove the fixed pattern noise caused by the camera in the flats. Dark flats are subtracted from the flats.

The camera imperfections tend to be added to the image signal. Optical train imperfections tend to be image data lost in the transmission through the optical train. The loss is proportional relative to the image signal. This is why rather than subtracting flats from lights, lights are divided by flats.

Each pixel actually reports zero as a sightly different ADU (analog digital unit). This uneven level of zero is another noise signal, and the one that is measured by the bias frames. Bias frames are darks taken with the shortest possible exposure time in order to display the only the fixed pattern of sensor read noise. Darks include both dark current noise and read noise from the sensor so most of the time you don't need a bias set, however, for some cameras the averaging process can produce negative ADU values, and this can create problems with some kinds of computational algorithms, so sometimes bias frames are better than the alternative fixes of ignoring negatives using absolute values, or adding an offset to raise the negatives above zero.

I agree with Marcus that the noise is highly unusual for amp glow because amp glow usually starts a the corner. Most image sensor chips have the electronics for reading at one corner, and that makes that area hotter producing more thermal noise. However, because the problem appears to be related to the camera, the use of darks should get rid of most of the problem regardless of the actual camera problem.

Hi,

I think it's definitely a light leak. Good point Marcus.

I've put a little piece of lather inside the right side of the sensor.



I've made some darks, no light leak anymore.
I 'll continue the tests to confirm this.
And thank you all for your help.

David

Hey David, you asked a bit of time ago: "Need help : Amp glow or light leak Sony nex-3n modified". I self modified a Sony NEX-3N. Afterwards I was taking "Darks" for DSS and noticed light leak. I was using a Novoflex adapter with a Nikon lens. I got rid of the light leak by wrapping the adapter and the E-mount fitting with black electrical tape on the  camera and adapter. It seems the Sony body E-mount on this NEX-3N fitting allows a tiny bit of light in. Also note I had some external red light sources that also may have contributed to the problem. You may have solved this problem but I thought I would add if it was of help.