Design, construction, tuning of spectroscopes
Information and discussion about softwares (telescope remote, autoguiding, acquisition, spectral processing ...)
Given the systems efficiency initiative I was prompted to look at to look again at the possibility of using the mean bias level, rather than the bias frame for subtracting from the Darks , Target frames etc. For a CCD this should be a constant (not for CMOS). Obviously you need to look for structure across the bias frame etc...
However, I came across a different issue. I use IRAF and remove the hot pixels using "fixpix" I had set this up carefully looking at the processed images and manually adding hot pixels.
When I looked at the Dark - mean (of bias) I observed a faint snow of warm pixels. Using ISIS to get the data I plotted the number against the threshold see attached. There is a clear knee in the graph with about 10000 warm pixels.
The camera is an SXVR-H694 at -10 C I will look at deeper cooling and removing the warm pixels.
I also have an SXVR-H694 so was curious to see what I would get from a similar exercise. I used my current 300 sec dark frame generated using the normal ISIS protocol. The calibration frames had been collected with the camera cooled to -20C. I plotted up the number of hot pixels eliminated as I varied the threshold parameter in the ISIS Masters tab. There is a similar knee although not so dramatic. Looking at a histogram of number of pixels vs pixel value in the dark frame using AIP4WIN, this is due to a small excess of hot pixels around 490 ADU. Once these have been eliminated the graph jumps down to a lower level. Comparing our results shows the dramatic difference that cooling to a lower temperature makes to the ADU level of hot pixels.
Regards, David
Hot pixel count vs ADU threshold.png (11.04 KiB) Viewed 8265 times
ATK428_warm_pixels.png (16.05 KiB) Viewed 8233 times
1200s thermal (median dark frame - median bias frame) ATIK ATK428 (ICX674) 2x2 bin -10 deg C
but it does show the knee clearly so confirms a likely sensor effect rather than something camera specific.
I have had the feeling for some time as I have upgraded through cameras using newer Sony CCD (ICX424, ICX415, ICX285, ICX674) that although the measured thermal noise remains low, the "snow" from warm pixels has been increasing, though I have not quantified it.
Interesting result David, I plan to make some new darks tonight at -20 C set on the camera and the MRes spectrograph set at 8 C ( the camera is inside the MRes). My darks are for 600s and 1x1 bin.
Do you manage the SXVR-H694 at -20 C all year as I can't get mine to sustain a delta T of more than 30C ?
Regards Andrew
Last edited by Andrew Smith on Sun Dec 03, 2017 3:34 pm, edited 1 time in total.
Hi Robin an interesting confirming result for the ATIK, I think it shows we need to be aware of warm pixels as well as hot ones.
We should of course normalise the results to electrons e rather than ADU and adjust for the different exposure but nonetheless it show a potential additional source of noise that needs eliminating if at all possible.
I can't quite decide yet whether this warm pixel excess matters or not. They are a long way from being saturated so If they are stable, always in the same place and react to light in a linear fashion, the dark/flat calibration should deal with them. If they react inconsistently and are popping on and off all over the place then clearly they are a problem.
I have now done some flats at -20 bin 1x1 Exp 600s processing them as before the ADU threshold is low as you would expect but the knee is at a similar pixel number ~ 10,000. Looking at Robin's and Davids data it looks like different chips have different numbers of warm Pixels.
My original darks were exposed for 900s not 600 as I stated before.
Not sure if they matter or not Robin it will be a real effort to see if they are linear.
My SXVR-H694 just about managed to hold a temperature of -20C during the summer although there were warm nights when it struggled to get to -19C. I always run it with 1x1 binning. My dark frames are taken with 300 sec exposure and scaled by ISIS. In my experience the hot pixels are stable and ISIS removes then reliably. If I see evidence of one appearing in a spectrum, it is always a new one which has just become "hot" and adding it to the hot pixel file successfully removes it.
I have noticed on the SXVR-H9 I use for photometry, where I take darks with a wide range of exposure times, that the hot pixels scale more or less linearly with exposure time whereas the dark frame level barely rises.