Pages: [1]   Go Down

Author Topic: 20d noise  (Read 3218 times)

thierrylegros396

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 1947
20d noise
« on: November 16, 2004, 09:48:37 am »

What about noise in RAW ?

Has somebody verified it on a marketed one ?

Some reviews noticed that the pre-production models had less noise !
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
20d noise
« Reply #1 on: November 16, 2004, 07:56:06 pm »

I believe the noise improvement of the 20D over the 10D is significant only at ISO 1600 & 3200. At ISO 800 it's very marginal and below that, probably insignificant or non-existant.

I picked up my 20D just a few days ago. For me it's an upgrade from the D60, so noise improvement at higher ISO's is obvious and much appreciated.

However, I should add that converting the first 20D RAW images I took, I used the Canon software (Digital Photo Professional Ver.1.1). Noise at ISO 800 was fairly obvious. Perhaps worse than the D60 at ISO 400  ???

Then I tried Photoshop's ACR with luminous smoothing and color noise reduction at 50% (which is what I use with my D60) and found that noise is at least as good at ISO 800 (20D) as the D60 at ISO 400.

When the pixel-peeping part of my nature gets the better of me, I'll do a direct, apples for apples comparison  :D .
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
20d noise
« Reply #2 on: November 18, 2004, 01:15:48 pm »

Hi! again, BJL,
I suppose I will eventually get around to doing some dynamic range tests, although I'm probably more interested in comparing the 20D with the D60 rather than arriving at a (perhaps) dubious f stop range.

What appeals to me greatly is the 20D's performance at ISO 800 and 1600. I always considered the top 2 D60 settings (ISO 800 & 1000) to be unuseable, yet I often found a need for these speeds with the 100-400 IS zoom. Despite the IS function, anything less than 1/250th at 400mm I found a bit risky for a hand-held shot.

Not sure about the 10-22mm. I've always attached great importance to the cropping factor. Will this lens be up to scratch? Reports I've read so far almost without exception mention corner softness, and some reviews, softness all over.

I'd like to compare the 10-22 with my Sigma 15-30 (in the 15-22 range) before any decision.
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
20d noise
« Reply #3 on: November 19, 2004, 12:56:26 am »

Quote
.... the noise is tolerable in the brighter areas of the image, but can be much worse in the shadows, as you might expect.
I think this is true at all ISO settings. I was very pleased with the first test shot at ISO 1600, taken indoors, in the evening, with a normal amount of artificial lighting and no flash. With my 28-135 IS, at 28mm and f5.6, I was able to get an acceptably sharp and noise-free result, hand held at just 1/10th sec exposure. I was impressed. However, this was not a high dynamic range scene.

I've photographed waterfalls in a rainforest at ISO 100 (with my D60) and found lots of noise in the shadows as a consequence of exposing for the highlights. I would expect the 20D to be no different in this respect since noise at ISO 100 is about the same as that of the D60, 10D and 300D etc.
Logged

Misirlou

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 711
    • http://
20d noise
« Reply #4 on: November 22, 2004, 12:46:13 am »

I spent some more time looking at noise today, pixel peeping perhaps. I said earlier that I thought at high ISO, there was a bigger spread in the noise between low and high brightness areas. I'm now convinced that color has at least as much impact as brightness.

I'm still not as concerned with the clumpiness as I am the linear features. Is anybody else seeing linear noise?

Wes Baker
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
20d noise
« Reply #5 on: November 22, 2004, 07:37:11 pm »

Once again, when discussing noise in RAW images one should always mention the RAW converter being used. Adobe Camera RAW has a 'luminance smoothing' and 'color noise reduction' control which can be very effective. When shooting above ISO 100 I usually increase these controls above the default setting. At ISO 1600 & 3200 I would recommend a 100% setting for luminance smoothing and color noise reduction.
Logged

wjy

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 51
    • http://www.billyyunck.com
20d noise
« Reply #6 on: November 23, 2004, 11:48:04 pm »

I have gotten linear noise "lines" even at 400.  I notice it in shadow areas that I try to pull out with photoshop's shadow highlight tool.  I never exceed eight on this setting and I never noticed the linear lines with the 10d.  I have been using photoshop cs for my raw conversions, maybe I'll try some with canons software to compare.  It is only at 400 or above in the shadow areas, but I have also seen some strange looking color noise in the properly exposed areas of the photo at these iso's.  I don't use 1600 or 3200 very much, but I was suprised to see these results at 400.
Maybe it is just a few of the 20ds experiencing this and that most of the bodies are ok. I don't know anyone else with one to do tests with.
anyway, I can always use 200 and good old 100.
Thanks.
Billy Y.
Logged

wjy

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 51
    • http://www.billyyunck.com
20d noise
« Reply #7 on: November 15, 2004, 11:49:16 pm »

I know all the reviews are saying the 20d is quite a bit better than the 10d when it comes to noise, but I'm seeming to get what seems like a lot even at iso 200.  I have used both cameras and see no difference in noise between the 2.  Has anyone one else noticed this?  Is it possible it could be a bad sensor?  Anyway, It almost seems worse in some cases than the 10d.  Go figure.
Thanks Billy Y.
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
20d noise
« Reply #8 on: November 16, 2004, 11:43:40 am »

I would love to see tests of not only noise but also resolution at higher ISO sensitivity settings from the 20D, and from other cameras that show low noise at high ISO. The reason is that my rough calculations show that at ISO3200, a 20D pixel with mid-tone illumination will receive less than 100 photons in its colour spectrum (R, G or , so that the S/N ratio at mid-tones must be less than the square root of that, or less than 10:1 which is ten decibels.  This is already in the realm of poor image quality, and when you go to moderate shadows two stops lower, the photon count should be less than 25, and hence S/N at most 5:1; very visibly noisy.

So either my physics is wrong, or Canon (like Fuji) processes high ISO images with some smoothing to reduce noise levels, which must be at the cost of some resolution. Not that there is anything wrong with that; higher speed films also have less resolution too; trade-offs of sensitivity against resolution are demanded by basic physics, even though electronic sensors can potentially achieve a given resolution at a far higher sensitivity level than film due to their far greater efficiency in detecting light.

How do various RAW conversions look? Maybe the smoothing is part of the Bayer interpolation process, since different interpolation algorithms can make different trade-offs between sharpness and noise (like different film developers!)
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
20d noise
« Reply #9 on: November 17, 2004, 11:27:04 am »

Quote
I picked up my 20D just a few days ago.
Welcome back Ray, and congratulations on the 20D!

Now I will have to stop teasing you about the wide angle lens restrictions of the D60/10D: are you planning to get the 10-22 EF-S? I am curious, because this seems set to be the classiest of the EF-S lenses so far. Maybe I could also interest you in doing a dynamic range test, in the style described on this site somewhere, since I care about DR more than any other supposed limitation of DSLRs.
Logged

Misirlou

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 711
    • http://
20d noise
« Reply #10 on: November 19, 2004, 12:12:55 am »

My 20D is my first digital SLR, so I have nothing to say about how it compares to the D10. Also, I only shoot in RAW, so I can't comment on JPG captures either.

I can tell you about my experience with the 20D noise characteristics though. 1600 and 3200 are actually quite usable, with a few caveats. First, I'd give them a good shot of NeatImage as a matter of course. Second, the noise is tolerable in the brighter areas of the image, but can be much worse in the shadows, as you might expect. This spread characteristic at high ISOs requires extra attaention with NeatImage. At lower ISOs, say up to 400, there is much less noise difference between bright and dark areas.

The annoying thing about the noise in dark areas of high ISO images is that there is sometimes a non-random quality to it, especially in the green channel. NeatImage can handle random noise, but faint green lines are harder to deal with. Not impossible, and if they're dark areas detail is less critical anyway, but still annoying.

Wes Baker
Logged

wjy

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 51
    • http://www.billyyunck.com
20d noise
« Reply #11 on: November 20, 2004, 12:01:59 am »

Lets see, after reviewing images from both my 10d and the 20d, I would say the 20d overall is better with noise at high ISO.  What I was seeing in images that made me post this thread in the first place, was a trend for the 20d to have "clumps" of noise in the shadows.  What I also realized is that after reviewing all my older 10d images was that I rarely used 800, and I didn't once try 1600.  So I guess all in all the 20d is better.  I just didn't ever remember the 10d having the underexposed noise "clumps".  Now I realize I never really pressed my 10d to higher ISOs, mostly because when I read reviews prior to buying they all stated that 800 and above on the 10d were unusable.  This was not the case on the 20d.  Even the test on this site said the 20d good at 800 and usable at 1600.  I just had my sights set a little higher than what you actually get.  I do admit that my 20d is better however, now that I have made a closer comparison.
Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
20d noise
« Reply #12 on: November 20, 2004, 06:34:33 pm »

Quote
What I was seeing in images that made me post this thread in the first place, was a trend for the 20d to have "clumps" of noise in the shadows.
I had also seen that in the single 20D ISO3200 example I have found; do other 20D users confirm it? Is this clumpiness limited to the higher ISO speed settings? Is it confined to in-camera JPEG or is it also seen with Canon's RAW conversion software? Is it confined to Canon JPEG conversions (in-camera or later) or also seen with third party RAW conversion software? [I am trying to "reverse engineer" Canon's generally very successful noise handling; the imperfections are the place where the techniques can best be guessed at.]

The only explanation I can think of for such "clumpy" noise, with length scale a lot larger than a single photosite, is that at high ISO and at photosites with low illumination levels, there is a high degree of spatial mixing of information from multiple photosites in producing the output values for each pixel; probably something in the Bayer interpolation/JPEG conversion algorithms. (Bayer interpolation algorithms apparently vary widely in how many nearby photosite values are used.)
Logged

jd1566

  • Full Member
  • ***
  • Offline Offline
  • Posts: 149
20d noise
« Reply #13 on: November 22, 2004, 01:42:48 am »

On the question of the 20D noise characteristics, I find that my unit shows horrizontal banding or stripes at high (between 800 and 3200 ISO)... This generally shows up when the picture is underexposed or there is very little light.  However not every single photo shows it..  As no-one else seems to have picked up this problem I am wondering whether it's just my unit or whether I'm the only user who has managed to find the "right" conditions for this phenomenon to creep up. "normal" photos are great and up to 400ASA the noise is really under control and no horrizontal banding  whatsoever.  Before everyone jumps in and says it's moiré, I don't believe it is because it;s across the whole image and not just a section of it.  I have been shooting RAW+JPG but have only worked with my JPG files as my software can't handle the new CR2 files yet...
Logged
B&W photographer - Still lifes, Portrait

wjy

  • Jr. Member
  • **
  • Offline Offline
  • Posts: 51
    • http://www.billyyunck.com
20d noise
« Reply #14 on: November 23, 2004, 11:44:31 pm »

oops
Logged
Pages: [1]   Go Up