Re: Canon Digital Rebel Questions
From: Chris L Peterson (clp_at_alumni.caltech.edu)
Date: 12/29/04
- Next message: Tim Auton: "Re: disease"
- Previous message: MT: "Re: disease"
- In reply to: eawckyegcy_at_yahoo.com: "Re: Canon Digital Rebel Questions"
- Next in thread: laura halliday: "Re: Canon Digital Rebel Questions"
- Messages sorted by: [ date ] [ thread ]
Date: Wed, 29 Dec 2004 20:51:08 GMT
On 29 Dec 2004 11:35:22 -0800, eawckyegcy@yahoo.com wrote:
>Do you have any evidence for this claim?
I shot a whole series of gradients from ISO 100 to ISO 3200, at 1/1000 sec,
1/100 sec, 1/10 sec, 1 sec, 10 sec, 100 sec, 1000 sec. I did this at 20°C
ambient and 0°C ambient. I measured the absolute dynamic range looking at the
gradients, and the S/N averaged over several areas of the sensor. I found that
the noise was constant up to 10 seconds, which suggests that readout noise is
dominant for shorter exposures (as you would expect). At 100 and 1000 seconds
(and a few intermediate speeds where I did some spot checking) the noise
increased proportionally to the square root of the exposure time, just as would
be expected for dark current noise. At all exposure times, the black level
remains at zero (no added pedestal) which shows that the dark current signal is
being subtracted off (I could see this in the raw data, so it is happening
before the RAW import stage). Both the calculated dark current signal (the
square of the noise) and the reduced saturation level of the gradients with long
exposure times worked out to a decrease in dynamic range of 6dB/300 sec at 20°C,
6dB/600 sec at 0°C. The S/N is always better at lower ISO settings, so given the
option, stacking more images at low ISO is better than fewer images at high ISO.
The best dynamic range I got was a 1 second exposure at ISO 100, 68dB (11.3
bits), which is very good indeed.
>I have a 4921 second exposure taken with a Canon 10D: looks basically
>fine to me (it's of course not the best one can get). Your 'logic'
>says this is flatly impossible: 90/5 = 18 bits, which exceeds the
>cameras DR by 6 bits.
Your bicycle image? Sure, it looks good. But you only need 4-5 bits of dynamic
range to look good. What is the S/N in that image? Also, I have no idea what the
actual variation is between cameras and sensors. Perhaps your camera does
better. If it lost a bit of range for every 15 minutes (about twice what I saw
with my camera), you'd still have about 6 bits of range, which would be visually
excellent with that particular image.
My point (made in another thread) is that preservation of linear dynamic range
is absolutely critical to stacking astronomical images. IMO, based on the tests
I've made, a 5-10 minute maximum exposure is consistent with this goal.
Empirically, I also note that the vast majority of good 10D/300D astroimages are
being made with 5-10 minute subexposures.
_________________________________________________
Chris L Peterson
Cloudbait Observatory
http://www.cloudbait.com
- Next message: Tim Auton: "Re: disease"
- Previous message: MT: "Re: disease"
- In reply to: eawckyegcy_at_yahoo.com: "Re: Canon Digital Rebel Questions"
- Next in thread: laura halliday: "Re: Canon Digital Rebel Questions"
- Messages sorted by: [ date ] [ thread ]
Relevant Pages
|
