new ATIK 314 - problem with darks

[Robin Leadbeater]

I have just replaced my ATIK 16IC-S with a new ATIK 314L+ I am making some measurements on the new camera but I have found a possible problem. I am subtracting pairs of frames to look at the noise (measured at -5deg cooling)

The bias frame difference looks good (less than 4e- noise and all values within +-100 pixels)
The 300 sec dark frame difference shows a large number of bright and dark outliers (outside +-500). This is much worse than my 16IC-S. In areas where there are no bright/dark outliers, the noise is good (The same as the bias)

I have attached the dark frame difference.(300sec -5deg) before I contact ATIK, I would be interested to hear how it compares with other 314 users

Thanks
Robin

edit: correct file attached 12:37 UT 23 nov 2011

[Christian Buil]

Robin,

The aspect of the image difference look very well for me.

The deviant pixels are very probably cosmic rays impact,
no noise or camera defect. The phenomena is normal, and
perhaps more evident because the intrinsic very low noise
of Atil314L model.

Christian

[Robin Leadbeater]

Hello Christian,

I agree that many of the outliers look like cosmic hits. There are often many grouped together. But why is the number so much higher than for the 16IC-S? Attached is a similar measurement for the 16IC-S taken last week. This camera also had low noise but far fewer "cosmics"

Is the ICX285 more sensitive to comics than the ICX415 ?
Do I have some radiation fom the camera eg from the cover glass? (Potassium in BAK7 for example can apparently be a problem in this respect)
Are we in the middle of a cosmic ray storm?

Cheers
Robin

[Robin Leadbeater]

I have now made some comparisons of the number of random outliers in dark frames from various ATIK 314L cameras and my ATIK 16IC-S. All measurements were made on the difference between two 300sec darks. (This removes the effect of any non time dependent features). Teleauto software was used to produce the frame subtraction and measure the variation and number of outliers. The two models of camera have different gain. I used figures of 0.26e-/ADU for 314L and 0.4e-/ADU for 16ICS. They also have different size and numbers of pixels. To correct for this, the number of outliers per 10^6 pixels was calculated

Both the 16IC-S and 314L have low read noise (~4e-) and the very low thermal noise of both cameras in the absence of outliers means that the total noise (except for outliers) is ~ the same for short bias frames and 300s darks. Outliers were defined as any pixel with a count of more than 7 standard deviations as measured in areas free of outliers. For normally distributed noise you would not expect to find any pixels with counts at this level in 1 million pixels. This was confirmed by checking some bias difference frames which showed no outliers. Outliers which were obviously cosmic ray hits (lines or close clusters) were counted as 1 outlier. The number of these was low, (between 0 and 5 /per megapixel) which is very approximately in agreement with the mean sea level cosmic ray flux of ~1/cm2/min

Although the two cameras have similar noise levels, measured in e-, the ADU value considered an outlier is different for the two cameras because of the different gain. A figure of 100 ADU for the 16IC-S and 150 ADU for the 314L in the difference frames was used, corresponding to ~7 sigma for each camera.

The results (difference frame outliers/mega-pixel/300s) were as follows:

314 (various) 49, 105, 45, 38, 89 (single spot checks on different cameras)
314 (mine) 64, 53, 56, 60, 68 (repeats on the same camera)
16IC-S (mine) 22, 37, 21 (repeats on the same camera)

It confirms my suspicion that the frequency of random outliers in darks in my new 314L camera is (~2.5x) higher than for my 16IC-S but it appears that this level is typical for the 314L.

Looking at the nature of the outliers:

They are mostly single pixels with ADU values ranging from 100s to a few 1000s.
They appear randomly distributed from frame to frame. ie they are not due to noisy warm pixels varying with time.
The frequency does not appear to vary significantly with temperature and increases ~linearly with the length of the dark.

The reason for the increased frequency in the 324L compared with the 16IC-S is not known. The read noise are similar and the thermal noise is very low in both cameras though the 314L thermal noise is reported to be lower. The QE of the CCD in the 314 (ICX285) is reported to by ~2x that of the ICX415 used in the 16IC-S. Perhaps the rate of outliers is connected with this?

The effect of these outliers is probably potentially most serious in the master dark where unless a median combine is used, they will accumulate to large numbers in a large stack of frames. They will also appear in the lights. In an hour of total exposure for example one can expect, based on these results, typically ~350 of these outliers in 1mega pixel from the 314L. (half the frequency seen in difference frames) If the binned spectrum height is say 10 pixels, then one would expect ~3-4 of these outliers to be included per 1000 pixels of spectrum length.

Robin