Jonathan Wong wrote:
Did you do the levels step that we did on the other tests though Daan?
Also hidden on chip noise reduction is another possibility.
Yeah, I pushed the levels to 2 (instead of 25 or 10).
I think Nikon is applying some kind of chip NR indeed. But they do a very good job at it, because even in real world situations ISO 3200 is still pretty clean considering we are talking about a 12MP 1.5x crop cam. Up to ISO1600 noise isn't an issue with the D90. Still excellent colors and details. Amazing value
body cap on,
Manual settings,
1/60s exposure
raw image
iso 200 (D90 doesn't have ISO 100)
in lightroom 2.3
open raw, exposure +4
all tone curve settings +60
sharpening: 0
NR: 0
Saved as 900x600px jpeg
Pitch black
My photos were also pitch black when processed in DPP to the max brightness (2). I had to push them again even further in PS to get the noise to come out.
Will Patterson wrote:
My photos were also pitch black when processed in DPP to the max brightness (2). I had to push them again even further in PS to get the noise to come out.
How did you push in PS CS4 (I mean what method: levels or exposure or something else)?
But besides that, the D90 file stays black even after the maximum amount of pushing in PS (levels 2).
I wasn't able to replicate the problem pushing exposure and levels in Aperture. Can someone else double check , please ?
Also, anyone was able to clarify if there is hardware noise reduction happening at low ISO, with the 5d2, if High ISO noise reduction is turned on, let's say, "standard " ? Of course, shooting RAW and conversion done in anything else but DPP ?
danmitchell wrote:
"Betterness" is composed of more than one parameter. They did not write that "the noise produced when you push the exposure by three stops is better on the 5D II than on the 1Ds3" or whatever we are comparing to. Let's say, just for argument, that the 5D II has a larger dynamic range (I have no idea if it does), higher resolution, and "better color rendition" (whatever that might mean), but that it shows a bit more noise banding when pushed than the other camera. Which is "better?"
Dan
The noise produced when you push exposure does in fact affect the dynamic range directly. You can say this it is one side of the dynamic range itself.
I did a rough calculation on brainiac's and Daan's files, and found that brainiac's 5D2 has about 1 full stop less DR at ISO 200 than the 1Ds3.
For sensors of the same size and pixel count, I would need to see some vast improvement on other parameters to compensate for 1 stop loss of DR at low ISO, and still be able to call it "better" IQ.
alundeb wrote:
The noise produced when you push exposure does in fact affect the dynamic range directly. You can say this it is one side of the dynamic range itself.
I did a rough calculation on brainiac's and Daan's files, and found that brainiac's 5D2 has about 1 full stop less DR at ISO 200 than the 1Ds3.
For sensors of the same size and pixel count, I would need to see some vast improvement on other parameters to compensate for 1 stop loss of DR at low ISO, and still be able to call it "better" IQ.
I completely agree with alundeb. What magical benefit makes up for the loss of a whole stop of dynamic range? Clearly neither resolution nor colour accuracy are a whole stop superior on the 5D2, so why were we told it was the best EOS ever? I would really like to ask Chuck where this significant shortfall in DR/tone comes from, and specifically whether it's due to (false) economies, or the need to process video, or a deliberate handicapping of the camera so as not to upset 1Ds users. The last one would be the most inexcusable of reasons bearing in mind Canon's statements about IQ.
p.10 #10 · Are you seeing 5D2 Criss-Cross Pattern?
alundeb wrote:
I did a rough calculation on brainiac's and Daan's files, and found that brainiac's 5D2 has about 1 full stop less DR at ISO 200 than the 1Ds3.
Wow
I knew (based on my own shooting experiences with both cameras) that there was in fact a difference at lower ISO's with the advantage in favor of the 1Ds3, but one full stop... that is actually quite big...
I take it we are done investigating... I mean, the 5D2 criss cross pattern and reduced DR (as copared to the 1Ds3) are considered "features" now... and not QC flaws anymore?
I was hoping something good would come out of this... But as it is now, the 5D2 has lost its attraction as a studio cam for me
p.10 #11 · Are you seeing 5D2 Criss-Cross Pattern?
Daan B wrote:
Wow
I knew (based on my own shooting experiences with both cameras) that there was in fact a difference at lower ISO's with the advantage in favor of the 1Ds3, but one full stop... that is actually quite big...
I take it we are done investigating... I mean, the 5D2 criss cross pattern and reduced DR (as copared to the 1Ds3) are considered "features" now... and not QC flaws anymore?
I was hoping something good would come out of this... But as it is now, the 5D2 has lost its attraction as a studio cam for me
I'm not quite there.
We have the word of FashionBoy in the other thread, that a 5D2 was finally replaced by a good one.
We have the DXOmark results, that give the 5D2 and 1Ds3 about equal DR at ISO 200.
Personally, I think there are good and bad 5D2's out, but we need more time before we hopefully will know what it is, and can make some general conclusions.
p.10 #15 · Are you seeing 5D2 Criss-Cross Pattern?
Daan B wrote:
Yeah, I pushed the levels to 2 (instead of 25 or 10).
I think Nikon is applying some kind of chip NR indeed. But they do a very good job at it, because even in real world situations ISO 3200 is still pretty clean considering we are talking about a 12MP 1.5x crop cam. Up to ISO1600 noise isn't an issue with the D90. Still excellent colors and details. Amazing value
I have not done any testing of the D90, but I have done on its sibling the D300, which shares the same Sony sensor. First of all, Nikon clips the raw data at the low end, sometimes quite severely, so that an average exposure of zero photons corresponds to a negative raw level, which of course the ADC cannot output. (This is in contrast to Canon, which sensibly applies a positive offset so that zero illumination corresponds to some positive RAW level that varies by model, recent 14-bit Canons have the offset at level 1024 out of 16383 levels).
In the D300, a lens cap test at ISO 200 showed that about 95% of the pixels have RAW level zero, IIRC. So of course the lens cap test will show no banding -- it won't show anything at all! This does not mean that there is no noise, it simply means that Nikon has erased both the noise and the signal too. The corresponding pattern noise test for Nikon would have to use a nonzero but low level exposure of a uniformly illuminated surface.
That said, IMO the latest Nikons have much less pattern noise than the latest Canons. But one cannot conclude that from a comparison of a lens cap test of each of the sort people are doing here.
p.10 #16 · Are you seeing 5D2 Criss-Cross Pattern?
brainiac wrote:
I completely agree with alundeb. What magical benefit makes up for the loss of a whole stop of dynamic range? Clearly neither resolution nor colour accuracy are a whole stop superior on the 5D2, so why were we told it was the best EOS ever? I would really like to ask Chuck where this significant shortfall in DR/tone comes from, and specifically whether it's due to (false) economies, or the need to process video, or a deliberate handicapping of the camera so as not to upset 1Ds users. The last one would be the most inexcusable of reasons bearing in mind Canon's statements about IQ....Show more →
I suspects its neither of the last two (deliberate handicapping or process video) and its most likely due to quality of components used (like Emil suggested a while back), though I don't think anyone has any proof to back this up.
p.10 #19 · Are you seeing 5D2 Criss-Cross Pattern?
ejmartin wrote:
I have not done any testing of the D90, but I have done on its sibling the D300, which shares the same Sony sensor. First of all, Nikon clips the raw data at the low end, sometimes quite severely, so that an average exposure of zero photons corresponds to a negative raw level, which of course the ADC cannot output. (This is in contrast to Canon, which sensibly applies a positive offset so that zero illumination corresponds to some positive RAW level that varies by model, recent 14-bit Canons have the offset at level 1024 out of 16383 levels).
In the D300, a lens cap test at ISO 200 showed that about 95% of the pixels have RAW level zero, IIRC. So of course the lens cap test will show no banding -- it won't show anything at all! This does not mean that there is no noise, it simply means that Nikon has erased both the noise and the signal too. The corresponding pattern noise test for Nikon would have to use a nonzero but low level exposure of a uniformly illuminated surface.
That said, IMO the latest Nikons have much less pattern noise than the latest Canons. But one cannot conclude that from a comparison of a lens cap test of each of the sort people are doing here....Show more →
Thanks for explaining that... Still, in the example provided by me earlier the D90 shows no banding at all in the ISO 200 shot while pushing the exposure by 3 stops and adding some heavy fill light (+50). That is all that matters to me
)!