Read noise pictures

[MartinB]

Noise is a complicated business. I know that there are 2 kinds - shot noise and read noise, the s/n ratio improves with exposure time and that this can be done by stacking for shot noise but that stacking doesn't remove read noise. I've never understood why read noise can't be removed by stacking. Still don't but hopefully I will after the next edition of the online magazine Astrophoto Insight.

Narrow Band Paul discussed this recently and Kevin Nelson (of QSI cameras) has an excellent article in this months magazine, the first of 2. As I understand it there are lots of points in the translation of charge held in the individual pixels of a CCD to a FIT on the hard drive where noise is introduced. Some of this noise is random and can be elimiated by stacking but there is another form which is "Periodic" which isn't removed by stacking. Kevin describes a technique for producing an image which allows you to view this periodic noise.

Take a number of bias frames - enough to get a decent averaged stack. Create an average bias frame. Open a single bias frame and apply and overall addition to each pixel value, enough to ensure the next stage doesn't create negative pixel values. Now subtract your averaged bias from the single bias and you have a fit which is shows the read noise. Except there is nothing much to see.

I don't fully understand the next stage. To properly show the noise it is necessary to group the pixels into an image which shows frequency values (in fact a sinusoidal frequency representation) as opposed to the spatial representation we are used to. This is what a Fast Fourier Transform does (in case you didn't know that already). This then clearly shows periodic read noise generated by your camera. You can download ImageJ for free to do your FFT. It is important that images are cropped to be be squares of 2 ie. 512x512 or 1024x1024 etc I have done this and show the results for a QSI 532 and a QHY 8. Both images have had their white and black points adjusted to the edges of the histogram. Remember, the pixels aren't arranged spatially, there isn't a bright point at the centre of an ordinary bias frame

QSI

(click to enlarge)

QHY8

(click to enlarge)

The result for the QSI is pretty stunning whereas the QHY8 leaves a bit to be desired. However I think the QHY8 looks a bit odd. The camera was in quite a dim room and the bias frames are zero exposure but I think it might have been affected by extraneous light. The QSI was in the dark for the capture. The normal patterns for read noise are lines running across the image.

As I said, Kevin is part of QSI. In his article he compares results from 2 cameras with the same chip (read noise is camera design more than chip dependent). A quick trip to the QSI web site makes it clear that the cameras he is comparing are the QS! 532 and an SBIG ST10 both with the same Kodak chip.

So what does all this mean? Well my take on it is that the read noise on the QSI is extremely low meaning that I don't have to get too uptight about long exposures to overcome the read noise. The QHY8 is probably a different matter.

I would be very interested seeing the results of other rather sad people who find this sort of thing interesting!