20D hot pixel/noise tests

[pcasciola]

I'm doing some 20D hot pixel and noise tests to get ready for some astrophotography with a couple of friends next week. I'm supplying the 20D, and one of the guys is bringing a 9" Celestron Schmidt-Cassegrain with some sort of fancy german equatorial mount. I have a cheap $400 Celestron so I'm not even going to bother bringing that.

Anyway, I wanted to do some noise tests to see how the 20D will peform photographing some deep space objects.

I did these tests with the lens cap on using the thrifty fifty at f/1.8 to see how much noise would be generated on super long, dark exposures.

The mean levels across the board were as follows:
ISO 100 - 11 minutes - 1.57
ISO 1600 - 2 minutes - 2.76
ISO 1600 - 5 minutes - 5.11

I believe these are really good results but I have no frame of reference.

Here is a 1:1 pixel crop from the worst area of the 11 minute ISO 100 exposure:
http://www.casciola.com/pics/11minutes-1x.jpg

And here's a 1:1 pixel crop of the worst part of the 2 minute ISO 1600 exposure (This is the upper left corner which is the only area that had noticeable noise):
http://www.casciola.com/pics/2minutes-1x-worst.jpg

This is a 1:1 pixel crop from the center of the same 2 minute ISO 1600 image:
http://www.casciola.com/pics/2minutes-1x-center.jpg

After 5 minutes at ISO 1600 there was some obvious noise
http://www.casciola.com/pics/5minutes-1x.jpg

Here are link to the full size images with full EXIF if anyone is interested:
ISO 100 - 11 minutes (http://www.casciola.com/pics/ISO100-11min.JPG)
ISO 1600 - 2 minutes (http://www.casciola.com/pics/ISO1600-2min.JPG)
ISO 1600 - 5 minutes (http://www.casciola.com/pics/ISO1600-5min.JPG)

[drisley]

Yes, those are very impressive.
Curious, did you have Long Exposure Noise Reduction enabled?

[pcasciola]

Curious, did you have Long Exposure Noise Reduction enabled?

I didn't enable Long Exposure Noise Reduction for the hot pixel lens cap on tests. I am experimenting with those settings right now, trying to photograph my 300mm lens in nearly complete darkness. The only light was from a dim LCD clock. It was pitch black. Even after my eyes adjusted to the darkness, I had to fish for the remote release after a few minutes. All of these were 2.5 minute exposures at ISO 1600, f/1.8 with the 50mm. I was amazed to be able to read the text on the lens after the exposure, so I cropped that area to show here.

No noise reduction
http://www.casciola.com/pics/ISO1600-lens-2.5min.jpg

With Long Exposure Noise Reduction Enabled
http://www.casciola.com/pics/ISO1600-lens-2.5min-lenr.jpg

Original after Photoshop dust and scratches noise filter applied
http://www.casciola.com/pics/ISO1600-lens-2.5min-dust.jpg

The Long Exposure Noise Reduction is pretty bad in my opinion, and the camera locks up for over a minute processing the image after the exposure is completed. I'm not 100% sure what is going on there, but it flashes "busy 93" or "busy 94" for 60-90 seconds before it shows me the picture on the LCD.

[drisley]

Hmm, yeah, the long exposure looks greenish, and the test is very soft.
What it does is takes a second exposure with the shutter closed, and subtracts from the first. It's supposed to reduce noise.
However, from everything I've seen, the 20D doesn't really need it.
In some cases (http://photography-on-the.net/forum/showthread.php?t=45862) I've found that the exposures without Noise Reduction actually look cleaner than those with it (especially for exposures under 1 minute).

[Scottes]

Nice tests. Thanks for sharing.

I was annoyed last weekend since I had NR enabled and had to wait and wait and wait. After seeing the green exposure that's getting disabled. I'll do NR with Neat Image if necessary.

[Jon]

Thanks for running the test. Interesting results!

IIRC from when I read the manual (download while awaiting delivery) the long noise reduction takes a second exposure of the same time as the first. Don't know how it can do that on time exposures or Bulb, so maybe it's just taking a wild stab at duration and you're seeing an electronic equivalent of reciprocity failure, subspecies colour shift. The other thing that might come into play would be your white balance setting. That might be an ideal situation for custom WB. I think I'll go upstairs now and turn off the NR function, though.