What's the use ... ?

[wimvb]

Now that may look depressing for a post title, but it isn't that bad.

Since the weather is the way it is (= 1 shade of gray, and snow for the whole weekend), what else can you do but reprocess old data and play around with the software (in my case PixInsight)

I found this script in PI that lets you calculate camera statistics such as gain, read noise, dark current, etc. While it is intended for CCD and not DSLR, some threads I read suggest that the numbers you get should at least be in the ballpark. So I thought to myself, why not, and took a few shots with my old (2009) Pentax. I took advantage of the weather (Yaay!) because conditions were low noise (0 deg C). First I was amazed at the results: readout noise < 2.7e and dark current < 0.035 e/sec at ISO 800. But after some more checking I got my explanation: DSLR's may use internal dark clipping.

That made me think; if the camera alters the image data, then what is the use of taking a lot of bias and dark frames?

Ok, dark frames do help with the amp glow of my camera, somewhat, and they do show I have so many hot pixels that even a dark frame looks like a clear photograph of the milky way. But still, amp glow is taken care of with a pair of scissors (read: cropping) and hot pixels can be removed by cosmetic correction.

My question stands: what is the use of taking a lot of calibration frames if your camera does internal correction?

BTW, the cause of the clipping was also revealed to me on a Pentax site: the ADC is 14 bit but the output is only 12 bit per channel. Two bits down the drain.

[michael.h.f.wilkinson]

Cameras often do all sorts of processing on the data, and even raw may not really be raw. It is hard to be sure.

[wimvb]

Correct, of course.

What started me thinking about all this was my inability to create a superbias in PI. The resulting superbias seemed always just a few bits deep, and the high rejection map showed most of the read pattern. This would be expected if the camera clips the low range of values.

The only way to keep the read pattern in the masterbias was using no rejection at all during integration.

Then I found that the same was true for my darks, even those with an exposure time of several minutes. Darks (from my DSLR) are only usefull as hot pixel maps.

Lessons learned!