CCD Chips and their dynamic range - REALLY 16bit ??

[Christoph Puetz]

Dear PI Users,
there might be many Users with a cooled CCD astro camera.
I found an irritating issue with those cameras.

Commercial advertising promise us an "16bit" sampling, that would be outperfoming to those
DSLRs. OK, this seemd logical to me.

But: After I read this article: http://www.ccd.com/ccd111.html and computed the dynamic range
of some Sony or KODAK chips and the readout noise of some electronics by ATIK, SBIG, ..., I realized that
no chip/no camera manufacturer really hit the 16bit limit, that they promise.

Example 1 (low cost camera): ATIK 320E with SONY ICX-274
Readout noise = 3 e
Full-Well-capacity: 14000 e
Dyn. Range -> 14000 : 3 = 4666 or about 12-13 bit sampling depth (2^12 to 2^13).

Example 2: Moravian Instruments G3-11000 with KAI 11002
Readout noise = 14 e
Full-Well-capacity: 60000 e
Dyn. Range -> 60000 : 14 = 4285 or about 12-13 bit sampling depth (2^12 to 2^13).

Although those cameras have a 16bit A/D device, this would be useless as they sample a 12bit range with 16bit device  ?

Although the CCD Cameras provide cooling and a better quantum efficiency than DSLRs:
Do they really produce more differentiated images than the DSLRs with "real" 14bit sampling ?
When will this advantage really come to the picture ? In stacking ? In low areas where there is most noise (I think of
dark nebula) ??

Or am I completely wrong ?

Feedback or explanation would be very helpful,
kind regards !
christoph

[Geoff]

Doesn't a similar argument apply to DSLRs? That is, are they really the claimed 14 bit?
Geogg

[mschuster]

In my experience (very limited) this well depth / read noise formula for dynamic range seems overly pessimistic.  My thinking could be wrong but here it is: My 8300 as a read noise of about 8 e- rms and a dark current of about 0.004 e-/s/px at -20C. So a 1000 second dark frame has about 4 e- more integrated dark current than a bias frame. But 4 e- is less than read noise. The formula would lead you to believe there would be no significant difference in the frames. But in fact there is a noticeable difference, the median of the frames differs by about 8 DN. In my opinion the 16-bit resolution helps even though the formula says it should not.
Having a camera with low read noise and deep wells is great, so the formula is useful for this purpose. But relating it to number of bits is not too useful IMO.
Mike

[astropixel]

I have been reading Berry and Burnell who cover this. As I understand it. bit ratings are based on a single set of test conditions at which the camera can achieve it's design rating, or very closely. Much of the time it doesn't or can't because of a variety of factors.

[Christoph Puetz]

Dear PI Users,

thanks for your replies. In fact this is an issue for *all* camera types, also DSLRs. (you are surely right, Geoff!!)
In PI you can test your camera with the BasicCCDParameter Script that is described and enhanced in this
topic : http://pixinsight.com/forum/index.php?topic=4086.0

Some results would surely be interesting for the community.
 
I was really astonished by the article of http://www.ccd.com/ccd111.html, as APOGEE is designing
cameras for themselves.

If this would really be true, then the dynamic range could be more dependent on stacking techniques and on exposure bracketing / hdr than on technical CCD features ?
See also this article of another famous PI user: http://blog.deepskycolors.com/archivo/2011/01/

This would mean: I could buy a "mean" astronomic ccd camera, use the advantages of cooling and a better snr, but
the dynamic range should be optimized by bracketing/hdr and stacking ?

Am I correct ?
This is surely an issue for a lot of PI users, as CCD-Cameras can be quite expensive.

[Geoff]

I believe stacking can play an important role. As an example, think of the 8 bit web cams stacking thousands of frames. The end result is surely better than 8 bit.
But I'm still confused. I would have thought that a single 1 hour exposure would have a better SNR than 6 ten minute sub exposures because of the fact that we have the read noise of one frame as opposed to the read noise of six frames.
Geoff