Quote:
Originally Posted by Slawomir
Hi Greg.
Well...in short, you were right :-)
Although improvements do diminish as you go lower in temperature, nevertheless my findings illustrate that there are extra benefits of going cooler.
I took samples of 5 bias frames at various temperatures (in 5C increments), stacked them and measured in ImageJ.
As the Mean ADU values for bias frames go down with lower temperatures, so does even more importantly the standard deviation (a measure of noise). It seems that improvements are still significant down to -10 degrees Celsius, and from then on you get very small gains at higher cooling.
Attached are simple graphs that hopefully will help to illustrate that.
The first graph just shows Mean ADU vs Temperature.
The second graph shows ADU percentage difference when going down in 5 C increments.
[For 20C data point the difference is only 3.3% because it is measured between no cooling (ambient) and CCD set cooling to 20 C. ]
The third one shows StDev vs Temperature
One can at least measure his camera and thus correct his thinking when rain has been pouring for the second day in a row...
|
This is a very interesting post. Thanks for testing like that. I can see what you were saying that once you hit 0C there isn't much significant change.
Usually Kodak chips list their dark current doubling with every 6 degrees C increase. Same in reverse. The bias must have some temperature sensitive component. I wonder if the main component of the bias is the read noise? That would explain the really low read noise of the Kingfisher at -110C or is it more they use a better analogue to digital converter?
Sony CMOS DSLR Exmor sensors are famous for shadow recovery and thus superior dynamic range. I wonder if some of that know how is present in the CCDs. In the CMOS chips the gain comes from every pixel having its own A/D converter as opposed to Canon which is off sensor. CCD don't have on sensor circuitry like CMOS do. But the Sony ADC may be superior to the Kodak one. Its quite likely.
Or perhaps the Sony plant runs at a much smaller resolution size on their chips like their CMOS plant does (Canon 300nm or larger, Sony something like 15!).
Greg.