Will Patterson wrote:
Yea I don't see 2 stops difference. I DO see about 2 stops of difference in the overall exposure which may be swaying your opinion, though.
Whoops - sorry Will - I just changed it to 1.5 stops before you posted. I looks like it's about 3 times as bad to me.
And that is only with the "lowly" 1.5x crop D90... Imagine what a D700, D3 or D3x will give you
Yes - thanks Dan, I saw that. I'm still holding off on the D700. Nikon lent me one for a week, but I didn't fall in love. I am amazed to see how good that D90 is at pushing. I really think Canon needs to sort this out if it wants to stop haemorrhaging users.
brainiac wrote:
Sure, but the 1Ds3 pattern is primarily horizontal, and much smoother than the 5D2 patterns which are about a stop stronger and criss-crossed. Actually the 200 iso noise looks about 1.5 stops stronger in my 5D2 compared to Daan's 1Ds3. Who wants to lose 1.5 stops of dynamic range?
Dpreview had the 5D II at .2 EV less dynamic range than the 1DsIII on the low end and equal at the highlight end at around the base ISO.
In that case, explain why Brianiac's 5D2 is better at ISO 100 than at ISO 200
Sunspots. Who knows? We are talking about minute levels of light here. Maybe there is some live preview involved and the sensor was hot. There are thousands of variables here. This cannot be completely understood based on our relatively random screwing around.
Noise is random by nature. You stick it in something that has clocks and motors and magnets in it and it is no longer random. It works on your television and radio also.
Although I don't hear of many people staring at television static and trying to determine if the set is faulty. Yes, that's extreme, but I have that feeling.
I'd call this whole thing a known issue. If you think this is bad, go pick on the M8 for while! Now that thing has sensor issues. I don't think I'll run out and get a 1DsIII over this.
Since nobody has posted proof that this issue doesn't appear on their 5D mark II, (or any other digital camera for that matter), I will conclude it is the nature of the beast, and then go out and try to tame it.
edrodgers731 wrote:
Sunspots. Who knows? We are talking about minute levels of light here. Maybe there is some live preview involved and the sensor was hot. There are thousands of variables here. This cannot be completely understood based on our relatively random screwing around.
Noise is random by nature. You stick it in something that has clocks and motors and magnets in it and it is no longer random. It works on your television and radio also.
Although I don't hear of many people staring at television static and trying to determine if the set is faulty. Yes, that's extreme, but I have that feeling.
I'd call this whole thing a known issue. If you think this is bad, go pick on the M8 for while! Now that thing has sensor issues. I don't think I'll run out and get a 1DsIII over this.
Since nobody has posted proof that this issue doesn't appear on their 5D mark II, (or any other digital camera for that matter), I will conclude it is the nature of the beast, and then go out and try to tame it....Show more →
I think everybody had got their lenscaps on... Besides that, in case of Jonathans 5D2 it seems consistent that ISO 200 produces better results than ISO 100. I have a feeling this applies to Brainiac's cam as well but then the other way around.
Are you willing to be openminded to the possible explanation that the differences between these 5D2 are caused by inconsistent sensor quality? Being caucious on purpose here...
Daan B wrote:
I think everybody had got their lenscaps on... Besides that, in case of Jonathans 5D2 it seems consistent that ISO 200 produces better results than ISO 100. I have a feeling this applies to Brainiac's cam as well but then the other way around.
Are you willing to be openminded to the possible explanation that the differences between these 5D2 are caused by inconsistent sensor quality? Being caucious on purpose here...
I think inconsistency of other components is just as likely. Sensors are produced under very very controlled circumstances. I don't know if the same can be said for subsequent electronics, which may even be sourced from different suppliers.
Jonathan Wong wrote:
I am still confused why my 5DII appears to have less noise at ISO200 again...
If you're referring to the pattern noise, it's because pattern noise is introduced into the signal processing chain after ISO amplification. Therefore at higher ISO the signal is more amplified while the noise is not further amplified, so the noise relative to signal is lower. See my post in the other thread.
The current stats are that 34 have seen criss-cross banding and 16 haven't. To my mind that is beginning to look like a feature of the camera rather than a problem in a batch of anomalous bodies. That's depressing.
I received the 5DII back from Canon in record time today. I haven't yet tested it, but here are the service details:
"We have exmained the product according to your request, and, it was found that due to an 'other cause' in the optical assembly causing noise to occur. Software adjustments were carried out on the CMOS assembly. Other electrical adjustments like ISO compensation and auto focus were carried out".
So, they did find something that was irregular and did make adjustments. I suppose the proof of the pudding will be in the tasting, though.
When I get some test photos, I'll put them up.
BTW, I requested that they look at the AF as well, as the microadjust seemed to have a wide variance at f/1.2 for me.
AFAIK pattern noise reduction is done in two steps by Canon. First, on-sensor with separate transistors for measuring the black current. This stage deals with the pattern noise from the sensor itself.
Second, in firmware, by using the masked pixels for correction of differences in the readout circuits and amplifiers. The rows on top (or bottom) can be averaged to get a measure of the vertical banding. The columns on the left (or right) can be averaged to get a measure of the horisontal banding. This can then be subtracted from the image. I don't know to what extent this is done, or with what efficiency, but a number of masked rows/colums in the order of magnitude around 10-100 should make significant banding reduction possible.
Also from what I've read, the information from the masked pixels is not included in the Canon RAW files (but in Nikon's).
If all this is correct, and turning in-camera NR completely off disables this stage of correction, doing so is like asking for criss-cross pattern noise.
To be honest, I doubt this is the case, it is just a theory.
alundeb wrote:
If all this is correct, and turning in-camera NR completely off disables this stage of correction, doing so is like asking for criss-cross pattern noise.
Interesting idea. I will try switching on NR. But which NR should I switch on? High ISO NR would not seem to be applicable for 100 or 200 ISO. I'll give it a try...
brainiac wrote:
The current stats are that 34 have seen criss-cross banding and 16 haven't. To my mind that is beginning to look like a feature of the camera rather than a problem in a batch of anomalous bodies. That's depressing.
I would take these results with a grain of salt... If only everybody who voted could show the same "lenscap" samples as we provided...
i get noise when i boost the shadows on iso 400, but nothing nearly as bad as yours brainiac
no patterns show or anything of that nature, just good old fashion noise
yale goodman wrote:
i get noise when i boost the shadows on iso 400, but nothing nearly as bad as yours brainiac
no patterns show or anything of that nature, just good old fashion noise
Could you try the "lenscap" procedure (like described earlier) and show the results, please?
Remorse.
