This will help show the difference. This is the red channel from Rubber Soul's test showing that the 1Ds3 is better when pixel peeped at 100%. If you don't pixel peep, don't worry about it and enjoy your 5D2. Just don't say that the quality of the 5D2 is the same as the 1Ds3.
I am seeing the criss-cross pattern in 5D2 noise at high isos, and especially in sraw1. There really is a noticeable difference between these two cameras, and if you're spending £2000 - £5500 on a camera because image quality matters to you, then do take notice of this. It really is visible at common print sizes when (a) you are pushing the limits of what the camera can do, or (b) you didn't have time to get exposure perfect.
thedigitalbean wrote:
Cool! Someone who has both and has compared both. Can you post a side by side example please?
Argh if Id of caught this minutes sooner Id be able to. Files are on my friends computer locked in his office. His 1Ds Mark III is in there too : /
To the OP: You really shouldnt see noise like that. You may have a defective camera. I can't reproduce anything like what you have even under exposing.
I just finished comparing images between the 1D3 and 5D2 in a controlled ISO 100 setting. When you take both images and push the fill light to 100% in Lightroom you can easily see the 5D2 has more noise in the red channel. This is with no NR in Lightroom.
1. Is there anyone that owns both a 5D and 5D2 who could do a similar comparison. I'd be curious to see the results.
2. Lloyd Chambers has reported that the 5D2 has a lot of hot and stuck pixels in comparison to the D3x. For those that own both the 5D2 and the 1Ds3, I would be curious to know how the two compare for long exposures (multiple minute) at low ISO.
I have a 1d3 and the 5d2. In some early personal tests looking at high iso shots esp shadow noise. I found the 5d2 underexposed @1/2 stop. The greater exposure of the 1d3 came at the expense of over exposing highlights.
I have used three different 5D2's the past weeks. I updated the first two from FW 1.0.6 to 1.0.7. The RAW files shot with FW 1.0.7 which I opened in LR or DPP showed lots of noise at low ISO's and magenta casts in dark areas or low-contrast areas. DPP and LR gave the same results
The third 5D2 I use hasn't been updated to FW 1.0.7, but is still on FW 1.0.6. The RAW files from this cam show way less noise (almost no noise at low ISO's, very similair to my 1Ds3) and no magenta casts until the higher ISO's (ISO3200 and up mainly).
So, there is a difference in performance between these three 5D2's. It could be related to the FW update and how that conflicts with the different RAW converters (LR or DPP) or it could be that some 5D2's have "crooked" sensors or something else is going on... You tell me.
Noise comparisons must be done at equal magnification otherwise they are meaningless. If you didn't adjust for the crop factor by changing focal length, then you need to start again.
brainiac wrote:
OK - I can see I'm not getting through here. First, HTP was switched off.
Got you. I thought I read that HTP was on. As far as I'm concerned, HTP does things to your exposure that you might not want done.
If you made no adjustments in post processing, HTP wasn't used, and this is essentially straight out of the camera, then yeah, those look a bit noisy.
Well, if we trust the Brainiac's claims , then DXOMark must be total garbage. Otherwise.... I'll like to see side-by-side RAW comparison of pics from both cameras (taken under identical conditions).
thw2 wrote:
Well, if we trust the Brainiac's claims , then DXOMark must be total garbage. Otherwise.... I'll like to see side-by-side RAW comparison of pics from both cameras (taken under identical conditions).
So, who do we trust?
I'd trust anyone taking real pictures over a numeric score any day of the week.
I had better describe what I am seeing here: the 5D2 file on the right shows a canvas-like vertical and horizontal criss-cross pattern in deep shadows at most ISOs that I have tested. This is with the updated firmware. The 1Ds3 file on the left is only very slightly better, but because there is less of a cross-hatch pattern and the noise seems more random, it is considerably less intrusive. I am in the habit of underexposing to save highlights with the 1Ds3, as a push of a stop or two in DPP still produces good shadows, and detail emerges from the murk. The 5D2 seems that it will be less good at that technique, since pushing the shadows produces a cross-hatch pattern which is quite ugly.
brainiac wrote:
Your 5D2 did better than your 1D3 for noise in my opinion. The 1D3 (top) has horizontal bands which will be harder to clean up with NR software:
Noise comparisons must be done at equal magnification otherwise they are meaningless. If you didn't adjust for the crop factor by changing focal length, then you need to start again.
True, the bands in the 1D3 image are indeed disturbing, I'll have to do some more testing. You are correct in that the crop factor needs to be accounted for. I did a very handwaving job at it, used a 35L on the 5D2 and a 24L on the 1D3. Not the same equivalent FOV but the idea was to give me a basis for comparing the quality of the noise.
I'll probably do this again but with zooms and try to more thoroughly compensate for the crop difference.
...was taken on each camera with the same lens, within a few seconds. Both cameras had all noise reduction and tone tricks turned off. Exposure was set manually to identical settings on each camera. I was careful when changing the lens not to touch the zoom ring at all. AF was used to focus on the trees outside the lower middle window. Images were processed in DPP with identical settings, and nothing automatic, no sharpening, no noise reduction and no lens or tone tricks. Contrast was set to minimum. With higher contrast and pushing the criss-cross pattern can become a nuisance. It's not something that NR software will find easy to deal with.
thedigitalbean wrote:
True, the bands in the 1D3 image are indeed disturbing, I'll have to do some more testing. You are correct in that the crop factor needs to be accounted for. I did a very handwaving job at it, used a 35L on the 5D2 and a 24L on the 1D3. Not the same equivalent FOV but the idea was to give me a basis for comparing the quality of the noise.
I'll probably do this again but with zooms and try to more thoroughly compensate for the crop difference.
Thanks - I would be very interested to see that. I think people overestimate the prowess of the 1D3 compared to higher resolving cameras on account of the 100% crop trap. I think the 1D3 and 1Ds3 are damn nearly equal for noise, even at very high isos. The 1D3 perhaps has the edge tonally, but it's a hair of difference.
brainiac wrote:
One more thing. This...
...was taken on each camera with the same lens, within a few seconds. Both cameras had all noise reduction and tone tricks turned off. Exposure was set manually to identical settings on each camera. I was careful when changing the lens not to touch the zoom ring at all. AF was used to focus on the trees outside the lower middle window. Images were processed in DPP with identical settings, and nothing automatic, no sharpening, no noise reduction and no lens or tone tricks. Contrast was set to minimum.
Thanks for the shots. I think your test pics show conclusively that you're right.
