Luminous Landscape Forum
Equipment & Techniques => Cameras, Lenses and Shooting gear => Topic started by: Josh-H on April 28, 2016, 11:37:43 pm
-
I have finally had a few minutes spare to shoot some test images with the full production specification (no pre-production hand built) Canon EOS1DX MKII camera (I now have two of these).
For this test I simply photographed a color checker indoors in natural low light (window light on a heavily overcast day). I needed to start in a very low light situation in order to test the full ISO capability of the camera. If I had not, I would have run out of shutter speed! Likewise, I needed to start with a relatively wide open Fstop or the low ISO exposures would have been more than several seconds.
All of the exposures are Evaluative metering, meter as read. Shot in AV mode, F4 (DOF not an issue). Camera was tripod mounted and I used the Canon 300mm F2.8L IS MKII lens. Mirror Lock up. 2 second self timer.
I did two sets. One with the new EOS 1DX MKII and one with the EOS1DX for comparison of the ISO improvements (same settings). Note the 1DX MKII goes to higher ISO's so there are more files and the last three are the push H1, H2, H3. H1, H2 and H3 are to as to be expected... ordinary...
I only did full stop ISO tests from ISO200 upward, since the intermediate ISO settings are either push or pull and I never use them.
The results pretty much speak for themselves. But, if you don't want to download and do the math.. The EOS 1DX MKII is (to my eyes) 2 stops better than the MKI in the meatball range of ISO where it counts - 800 to 12,800. Personally, I would quite happily shoot this new camera at ISO 12,800 if I needed to.. the original 1DX I preferred not to go over 3200 or maybe 6400 at a pinch if optimal results were the goal. Having said all that.. Its very easy to make ISO 51,200 on the EOS 1DX MKII look awfully good with moderate noise reduction in LR.
Here is a pretty compelling screen grab at 100% comparing ISO12,800 on the 1DXMKII to ISO3200 on the original 1DX. 1DX MKII on the left.
You can download the full size original RAWS here: https://www.dropbox.com/sh/dlc6bx7t5slyxy7/AAC5X0WSvpTnDMUfc8Aa7WWna?dl=0
Now that this is done I am returning to my regular programming of shooting out in the field. Taking two of these EOS1DX MKII's to the South Island of New Zealand tomorrow for 3 weeks photography. That will be the real test for me. When its nearly dark and I am hanging out of a helicopter over the southern alps in -15C with the doors off.. Bring it on. ;D
Edit - The files are taking a while to upload to Dropbox..if they are not all there, try a bit later today.
-
Thanks very much for posting this Josh. Have a great and safe trip to the South Island.
For what it is worth, I really can't see much difference between the 1DX and 1DXMKII files, viewing with default color noise reduction, no luminance noise reduction and default sharpening in Lightroom.
Screen grab at 100% comparing 1DXMKII ISO 3200 (left) vs 1DX ISO 3200 (right)
(http://roryhill.zenfolio.com/img/s7/v152/p2067016119.jpg)
Perhaps those with more expertise can evaluate how much improvement there is between the two cameras.
-
Thanks very much for posting this Josh. Have a great and safe trip to the South Island.
For what it is worth, I really can't see much difference between the 1DX and 1DXMKII files, viewing with default color noise reduction, no luminance noise reduction and default sharpening in Lightroom.
Screen grab at 100% comparing 1DXMKII ISO 3200 (left) vs 1DX ISO 3200 (right)
[...]
Perhaps those with more expertise can evaluate how much improvement there is between the two cameras.
Hi Rory,
When comparing ISO 3200 shots, the photon shot noise is the same between models and that noise still dominates a lot of the total noise footprint. When you start pushing such files, the Readnoise will start to show itself more clearly. The same would happen if you use much higher ISO settings, which will get less and less photon signal, and relatively more of the constant Readnoise floor.
I haven't had time yet to compare the files in detail myself, so I don't know what higher ISOs do, but the 1DX Mark II should be better.
Cheers,
Bart
-
Hi Rory,
When comparing ISO 3200 shots, the photon shot noise is the same between models and that noise still dominates a lot of the total noise footprint. When you start pushing such files, the Readnoise will start to show itself more clearly. The same would happen if you use much higher ISO settings, which will get less and less photon signal, and relatively more of the constant Readnoise floor.
I haven't had time yet to compare the files in detail myself, so I don't know what higher ISOs do, but the 1DX Mark II should be better.
Cheers,
Bart
You've got it the wrong way round.
Increasing ISO means fewer photons collected and more amplication, which means increased photon shot noise. If you want less photon shot noise and more contribution from read noise, you need to look at lower ISOs - at ISO 100, there will be minimal photon shot noise, with most noise being read noise.
-
Increasing ISO means fewer photons collected
ISO is not a part of exposure, which is exposure time and aperture...
-
increased photon shot noise.
more exposure = actually more "photot shot noise", but... more S/N ;D
-
ISO is not a part of exposure, which is exposure time and aperture...
Functionally it is, since you can't collect as many photons before blowing out the image. So higher ISO = fewer photons collected for the same final exposure = more photon shot noise.
-
If you want less photon shot noise
you want more S/N, not less "photon shot noise"... and to get more S/N you actually want more "photon shot noise" as the result of greater exposure (exposure time and aperture)
-
Functionally it is
it is not
, since you can't collect as many photons before blowing out the image.
I can collect as much photons as I want... now sensel well has capacity limitations, ADC can clip, firmware can clip, etc = but you need to use proper wording, which you do not
-
it is not
I can collect as much photons as I want... now sensel well has capacity limitations, ADC can clip, firmware can clip, etc = but you need to use proper wording, which you do not
I'm neither a lawyer nor a technocrat, I have no interest in wording or terminology. I only care about what works.
-
The point is, if you want to see the effect of read noise, you need to look at lower-ISO samples, not higher ISOs. Use whatever terminology you like, if you look at ISO 1000000 you won't see any read noise - only photon shot nosie.
-
You've got it the wrong way round.
Increasing ISO means fewer photons collected ...
Assuming we adjust the exposure, yes, I didn't suggest otherwise.
and more amplication,
Maybe, depending on how the camera is designed ('ISO-less' or with pre-ADC gain).
which means increased photon shot noise.
Yes, fewer photons means more shot noise, but I didn't say that wasn't the case. I did say that the noise floor, the Readnoise, will weigh in relatively heavier as we reduce the signal level, fewer photons are added in the mix. Not only the signal is lower, but the S/N is also reduced a lot. The 'cleaner' the Readnoise is, the less obtrusive it will be in the mix, and that's where the Mark II is supposed to make a difference, the signal being the same between models.
The topic is about the difference between the two models, not how to get a better quality image out of any camera.
If you want less photon shot noise and more contribution from read noise, you need to look at lower ISOs - at ISO 100, there will be minimal photon shot noise, with most noise being read noise.
But that's not what makes the Mark II different. That applies universally, more photons is always better (assuming one can expose long enough without camera shake), and is not the topic of this comparison.
Cheers,
Bart
-
The point is, if you want to see the effect of read noise, you need to look at lower-ISO samples, not higher ISOs. Use whatever terminology you like, if you look at ISO 1000000 you won't see any read noise - only photon shot nosie.
You are missing the point, which is how the difference between the camera models looks (given the same level of exposure). The only difference is supposed to be how the Readnoise contributes to the total noise, now that Canon has changed the ADC architecture.
Cheers,
Bart
-
Hi Rory,
When comparing ISO 3200 shots, the photon shot noise is the same between models and that noise still dominates a lot of the total noise footprint. When you start pushing such files, the Readnoise will start to show itself more clearly. The same would happen if you use much higher ISO settings, which will get less and less photon signal, and relatively more of the constant Readnoise floor.
I haven't had time yet to compare the files in detail myself, so I don't know what higher ISOs do, but the 1DX Mark II should be better.
Cheers,
Bart
Hi Bart
If I understand you correctly, you are saying that the 1DXMKII should be better at opening up shadows.
-
Hi Bart
If I understand you correctly, you are saying that the 1DXMKII should be better at opening up shadows.
Correct Rory, that's the bottom line. If and how much better, that's what we'll have to find out. But the performance improvements that the EOS 80D has shown (see e.g. here (http://www.dpreview.com/reviews/canon-eos-80d-review/12), with 5 and 6 stops underexposed and pushed Raws) over its predecessors is promising. The 1DX Mark II is supposed to have a similarly designed new sensor architecture. In those comparisons the different sensors were Photon starved by approx. the same amount, so it's the Read noise improvements that made the difference to the combined shotnoise+readnoise.
Cheers,
Bart
-
Interesting how different people see very different things in these samples: ;)
- Josh sees a clear 2 stops improvement,
- Rory doesn't see any difference.
I have checked ISO3200 and ISO12,800 and the 1DXII seems a tiny bit better to my eyes, but I don't see 2 stops, a shy 2/3 stops at best in these samples (they may not be 100% representative of what the camera can do in the real world).
In the end, beyond our dreams for real breakthroughs, both Canon and Nikon are clearly reaching a plateau in terms of high(er) ISO image quality IMHO.
As mentioned many times, the real value is in AF on moving subjects.
Cheers,
Bernard
-
ISO is not a part of exposure, which is exposure time and aperture...
Of course it is. Shutter Speed 1/100th, Aperture f/4 with ISO 200 versus ISO 1600. Do these 2 situations produce the same image? That is will a specific location in the frame be the same tone in both photos? No! They are by definition different exposures...same as film, preciously because the 3 factors are interdependent and have no significance without the others.
I get your point, but this is a real problem when we arbitrarily redefine terms. The sensitivity of the media (film, digital sensor or other) will always be a part of the exposure.
-
In the end, beyond our dreams for real breakthroughs, both Canon and Nikon are clearly reaching a plateau in terms of high(er) ISO image quality IMHO.
According to the cognoscenti many have been trying but nobody has as yet broken the 1e- barrier for input referred read noise in a widespread commercial camera.
Cameras with near 1e- input referred read noise (http://www.photonstophotos.net/Charts/RN_e.htm#Canon%20EOS%201D%20X_14,Nikon%201%20J5_12,Nikon%20D5_14,Nikon%20D500_14,Nikon%20D7100_14,Nikon%20D7200_14,Nikon%20D90_14,Sony%20ILCE-6300_14,Sony%20ILCE-7RM2_14,Sony%20ILCE-7S_14), roughly in chronological order (I chose 1.25e- as the threshold, figures were not normalized for sensor resolution, different cameras achieve their lowest read noise at different ISOs):
Nikon D90
Canon 1DX
Nikon D7100
Nikon 1 V3
Sony 7S
Canon EOS-M3
Nikon D7200
Sony 7SII
Sony 7RMII
Fujifilm X-PROII
Nikon D5
Nikon D500
Sony a6300
It's therefore unlikely one is going to see earth shuttering improvements upon any of the above cameras in their intended usage range in terms of random read noise performance. The Canon 1DX is in the list.
Jack
-
According to the cognoscenti many have been trying but nobody has as yet broken the 1e- barrier for input referred read noise in a widespread commercial camera.
Cameras with near 1e- input referred read noise (http://www.photonstophotos.net/Charts/RN_e.htm#Canon%20EOS%201D%20X_14,Nikon%201%20J5_12,Nikon%20D5_14,Nikon%20D500_14,Nikon%20D7100_14,Nikon%20D7200_14,Nikon%20D90_14,Sony%20ILCE-6300_14,Sony%20ILCE-7RM2_14,Sony%20ILCE-7S_14), roughly in chronological order (I chose 1.25e- as the threshold, figures were not normalized for sensor resolution, different cameras achieve their lowest read noise at different ISOs):
Nikon D90
Canon 1DX
Nikon D7100
Nikon 1 V3
Sony 7S
Canon EOS-M3
Nikon D7200
Sony 7SII
Sony 7RMII
Fujifilm X-PROII
Nikon D5
Nikon D500
Sony a6300
It's therefore unlikely one is going to see earth shuttering improvements upon any of the above cameras in their intended usage range in terms of random read noise performance. The Canon 1DX is in the list.
Jack
But maybe one stop more is within easy reach for the next Canon 1Dx by going to a back-illuminated sensor?
Edmund
-
Of course it is. Shutter Speed 1/100th, Aperture f/4 with ISO 200 versus ISO 1600. Do these 2 situations produce the same image?
these 2 situations have the same exposure, then what happens is that "camera" is processing the same charge (collected photons converted to e) differently... ISO/gain is not a part of exposure... exposure != "image"
-
But maybe one stop more is within easy reach for the next Canon 1Dx by going to a back-illuminated sensor?
Doubt it Edmund, BSI shows its mettle with smaller pixels - it does not make that much of a difference for larger ones with decent microlenses. And BSI supposedly offers a sensitivity improvement of no more than 0.5 stop, bringing problems of its own. But perhaps the 1DXII's gapless microlenses will help a bit. And maybe they will have tweaked the CFA recipes ;)
If the 1DXII brought ADCs on board the sensor one is likely to see that improvement at lower ISOs, which would be great news. Like Bernard said, the low hanging fruit has mostly been picked. Some incremental improvements are still the order of the day, just don't expect to see the deep shadows revolutionized. Look at Nikon, in order to get an extra, what, half stop of usable DR at high ISOs on the D5 they had to give up at least twice that at low ISOs :-\
Jack
-
these 2 situations have the same exposure, then what happens is that "camera" is processing the same charge (collected photons converted to e) differently... ISO/gain is not a part of exposure... exposure != "image"
I guarantee you do not get the same same data in the file. You get the same light hitting the sensor, that is for sure, but the data will not be the same precisely because the GAIN will be different. We used to push film too and you don't get the same thing with film developed for proper gain and film pushed 2 or 3 stops. Processing matters because it is at that point you have a usable image. RAW sensor data is not.
Only at the point we have gain independent technology, which we don't have yet, would it be responsible to discuss removing processing from the exposure equation.
-
Doubt it Edmund, BSI shows its mettle with smaller pixels - it does not make that much of a difference for larger ones with decent microlenses. And BSI supposedly offers a sensitivity improvement of no more than 0.5 stop, bringing problems of its own. But perhaps the 1DXII's gapless microlenses will help a bit. And maybe they will have tweaked the CFA recipes ;)
If the 1DXII brought ADCs on board the sensor one is likely to see that improvement at lower ISOs, which would be great news. Like Bernard said, the low hanging fruit has mostly been picked. Some incremental improvements are still the order of the day, just don't expect to see the deep shadows revolutionized. Look at Nikon, in order to get an extra, what, half stop of usable DR at high ISOs on the D5 they had to give up at least twice that at low ISOs :-\
Jack
Jack,
It's a relief to be talking to someone who doesn't have the IQ of a snail :)
I do wonder however what effect the gapless microlens sensor has on MTF - I suspect resolution goes down.
And then there is what one really sees at high ISOs on many cameras with external ADCs namely striping due to ADC mismatch ...I think onboard ADCs are easier to calibrate and keep matched. And *practical* ISO is in effect limited by the stripes becoming visible.
Edmund
-
I guarantee you do not get the same same data in the file.
you are trying to switch the topic _from_ exposure _to_ the content of raw files - 2 different things... again : (1) exposure = exposure time and aperture (assuming given illumination of course), ISO/gain does not define how many photons hit the sensor / that what exposure is / when you have set your exposure time and aperture... and then (2) digital data in raw file = exposure _PLUS_ sensel capacity (might be variable) & processing applied (on various stages : analog and/or digital) by sensor|sensel schematics / ADC(s) / firmware based on many things, ISO/gain included...
what you fail to understand (or pretend for rhetorical purposes) is that photographer or camera's firmware make a decision about exposure (only two parameters = exposure time and aperture) based on ISO/gain, which controls things that happens mostly after exposure ends ( analog and/or digital gain and/or tag in raw file as an instruction for raw converters ) and/or in some cases before exposure starts ( sensel capacity )
-
http://www.dpreview.com/news/8090146652/canon-eos-1d-x-mark-ii-studio-tests/1 (http://www.dpreview.com/news/8090146652/canon-eos-1d-x-mark-ii-studio-tests/1)
-
I spent a lot of time yesterday on this comparison at dpreview.
At 12800 ISO which is by far the fastest ISO I would consider I just don't see much if any differences between the 1DX MkII and the 6D especially in the part of the image that features the color checker.
In fact adding in the nikon D5 and all three look similar. The only cameras I tried that were worse were the Nikon D810 and Sony A7 RII.
6D was and still is a benchmark camera in many ways for Canon.
Paul C