Luminous Landscape Forum
Equipment & Techniques => Cameras, Lenses and Shooting gear => Topic started by: Ellis Vener on March 12, 2012, 09:57:36 pm
-
These were taken in some really ugly available ambient light just to see what the 5D Mark III can do in an extreme situation. First up a full frame view
-
Next: a 100% crop. Noise reduction in Lightroom 4 is turned off
-
This next one has Luminance noise reduction off in Lightroom 4 but noise reduction set to 25/50. Cleans it up a bit don't it?
-
Now Lightroom 4's Luminance noise reduction is in play as well.
-
Finally, I've added a little Clarity (+14) and increased Lightroom 4 sharpening to 60/0.8/35/0
-
In case you were wondering: all versions were exported as full size sRGB PSDs from Lightroom 4, captioned and either cropped or resized down to 1000 x 1500 pixels in Photoshop CS 5 and saved as maximum quality JPEGS.
I look forward to your comments.
-
Well, it should make a passable wallet-size print or you could reduce it to 500x400 pixels for web use.
-
Yes, I can't get particularly excited.
I presume it's better than the mark II ??
-
I look forward to your comments.
It seems excellent to me. Part of this is the 5DIII that seems like a very good low light camera, part of that is LR4 whose new engine shines.
It seems to be a bit ahead of what the D800 can do, but I guess we will need a 1:1 comparison at a given print size to cast a final judgement.
Either way, both cameras are clearly able to tackle any real world scenario without any problem and a bit better than their predecessors. The question for 5DII owners is whether this is worth 2,000+ US$ or not (assuming they sell their 5DII).
Cheers,
Bernard
-
Many thanks for posting these high ISO crops from RAW.
They are the first I've seen.
I've just compared with 5DII shots taken at ISO 3200 and ISO 6400 both put through LR4 using the same settings.
The ISO 6400 results are pretty close in terms of visible noise and retention of fine detail.
Arguably at ISO 6400 the 5DII shots have the edge, but it is close.
The results back up what everyone has been saying....
The 5DIII has getting on for two stop less noise at high ISO.
I know I can produce an acceptable A3 print from the 5DII at ISO 3200 with no luminance noise reduction and default chroma (25/50). Not perfect, but good enough to sell in a gallery.
It's possible that the 5DIII may be able to produce an acceptable A3 print at ISO 12800.
Could post some samples please so I can check :)
-
Well, it should make a passable wallet-size print or you could reduce it to 500x400 pixels for web use.
Actually it makes a pretty decent 8.5 x 11" print @ 300ppi. Would i want to work at anything above ISO 1600 or 3200 (and even then only in available darkness) with this or any comparable camera? Not unless it was an absolute necessity , but it is good to know that the headroom is definitely usable in extreme cases.
-
Actually it makes a pretty decent 8.5 x 11" print @ 300ppi. Would i want to work at anything above ISO 1600 or 3200 (and even then only in available darkness) with this or any comparable camera? Not unless it was an absolute necessity , but it is good to know that the headroom is definitely usable in extreme cases.
Hi Ellis,
It would be an interesting exercise to also compare the results of an ISO 25600, to an ISO 6400 but underexposed by a -2 EV exposure compensation and subsequently pushed 2 EV in postprocessing, or ISO 3200 underexposed an pushed 3 stops, or even ISO 1600 with 4 stops underexposure and pushed 4 stops.
Once the sensor gain is increased beyond "unity gain" by the ISO setting, the benefits are often minimal or even negative for quality. Depending on how the new sensor design (with additional noise reduction transistor per sensel) is implemented, it may benefit the more extreme ISO settings more than the lower ISO settings, unlike the 5D2 or the 1Ds3 which do benefit from underexposing and pushing.
Underexposing and pushing in post also has the benefit of avoiding highlight clipping, a common risk with ambient light only situations.
Cheers,
Bart
-
Once the sensor gain is increased beyond "unity gain" by the ISO setting, the benefits are often minimal or even negative for quality. ...
Underexposing and pushing in post also has the benefit of avoiding highlight clipping, a common risk with ambient light only situations.
Naively, it would make sense for a high-end DSLR to have an option like this: limit actual analog gain to the useful limit, which with good modern sensors seems to be relatively low (1600 or less?) and then handle more extreme low light situations by choosing aperture/shutter speed as indicated by the higher EI (like 35,600) but with analog gain only up to the lower useful limit, and then (a) with in-camera JPEG, gain-up in the digital domain, with simple bit-shifting, and (b) in raw files, simply indicate the intended digital gain (again, just a recommended bit-shift) as the default for conversion, leaving room for better manual handling of extreme "highlight range": when you have more than the typical three stops or so between metered mid-tones and the brightest highlight regions.
In fact, I believe that something like this has long been done with so-called "HI" ISO settings, beyond the normal range, and is down at most or all EI settings by DMF backs, but it could actually kick-in at much lower EI on modern DSLR's. I suspect that marketing gets in the way: people call this digital shift "fake ISO" and create the impression that amplifying as much as possible in the analog gain is more "real" and therefor better, and this leads to analog gain being pushed beyond its useful limits.
-
Is it me, or is this some thermal amplificator noise in the lower right corner (red+blue)?
-
Hi Ellis,
It would be an interesting exercise to also compare the results of an ISO 25600, to an ISO 6400 but underexposed by a -2 EV exposure compensation and subsequently pushed 2 EV in postprocessing, or ISO 3200 underexposed an pushed 3 stops, or even ISO 1600 with 4 stops underexposure and pushed 4 stops.
Once the sensor gain is increased beyond "unity gain" by the ISO setting, the benefits are often minimal or even negative for quality. Depending on how the new sensor design (with additional noise reduction transistor per sensel) is implemented, it may benefit the more extreme ISO settings more than the lower ISO settings, unlike the 5D2 or the 1Ds3 which do benefit from underexposing and pushing.
Just a side note -- it's not the 'unity gain' which determines when pushing lower ISO competes with 'normal' exposure at higher ISO; rather it's when read noise in electrons stops decreasing. 'Unity gain' is a bogus characteristic; for instance it dropped by a factor of four when Canon went from 12 to 14 bit encoding, while the ISO beyond which it made no sense to go stayed at 1600 -- and that was because it was at that point that read noise leveled off, stopped decreasing with increase of ISO.
-
I look forward to your comments.
It can make recognizable photos.
-
Naively, it would make sense for a high-end DSLR to have an option like this: limit actual analog gain to the quite modest useful limit, which with good modern sensors seems to be relatively low, like maybe 1600 or less?, and then handle more extreme low light situations by choosing aperture/shutter speed as indicated by the higher EI (like 35,600) but with analog gain only up to the lower useful limit, and then (a) with in-camera JPEG, gain-up in the digital domain, with simple bit-shifting, and (b) in raw files, simply indicate the intended digital gain (again, just a recommended bit-shift) as the default for conversion, leaving room for better manual handling of extreme "highlight range": when you have more than the typical three stops or so between metered mid-tones and the brightest highlight regions.
In fact, I believe that something like this has long been done with so-called "HI" ISO settings, beyond the normal range, and is down at most or all EI settings by DMF backs, but it could actually kick-in at much lower EI on modern DSLR's. I suspect that marketing gets in the way: people call this digital shift "fake ISO" and create the impression that amplifying as much as possible in the analog gain is more "real" and therefor better, and this leads to analog gain being pushed beyond its useful limits.
The D800 is halfway there; analog gain ends at ISO 1600. Now all Nikon engineers have to do is realize (doh!) that there is nothing to be gained by bit-shifting the data and throwing away all that highlight info, and put the ISO setting in metadata.
-
Just a side note -- it's not the 'unity gain' which determines when pushing lower ISO competes with 'normal' exposure at higher ISO; rather it's when read noise in electrons stops decreasing. 'Unity gain' is a bogus characteristic; for instance it dropped by a factor of four when Canon went from 12 to 14 bit encoding, while the ISO beyond which it made no sense to go stayed at 1600 -- and that was because it was at that point that read noise leveled off, stopped decreasing with increase of ISO.
Hi Emil,
I know that "unity gain" is not the whole story, because the amplifier noise and other design characteristics also have their influence on total noise. However, as far as I can determine from the 5D3 Raw files on Imaging Resource, the unity gain level (where 1 electron changes the DN or ADU by 1) is approx. at ISO 800. This means that multiplying the signal by 2, 4, 8, etc, will not add real (un-combed) signal levels, but signal levels from (read- and other) noise at the intermediate new levels. Some of those noise sources are related to the actual signal levels, but it makes no sense to interpolate with noise when the interpolating noise exceeds the shot noise level that a real signal level would have produced. It then becomes a matter of postprocessing pushing to keep the noise as low as possible.
That is, unless the new sensor design kicks in at elevated ISO settings and does a significantly better job of keeping the noise down, QED.
Cheers,
Bart
-
In terms of the results, I’m wondering how this pans out with the new camera compared to the 5DMII in terms of what Eric Chan at Adobe had said:
When shooting raw, there is no reason to shoot at above ISO 3200 on the 5D II because the noise levels will be the same (possibly worse) compared to if you shot the image at ISO 3200 and just boosted the Exposure slider in Camera Raw.
-
In terms of the results, I’m wondering how this pans out with the new camera compared to the 5DMII in terms of what Eric Chan at Adobe had said:
Andrew, if you want to send me a 5D mark II to directly compare it to, I'll be happy to check that out ;D
-
Andrew, if you want to send me a 5D mark II to directly compare it to, I'll be happy to check that out ;D
I believe you are missing Eric’s point and my query. You don’t need a 5DMII. But it would be interesting to know if the higher ISO settings on the 5DMIII behave as reported by Eric on the 5DII in terms of better noise reduction with lower settings and appropriate edits in LR.
-
In terms of the results, I’m wondering how this pans out with the new camera compared to the 5DMII in terms of what Eric Chan at Adobe had said:
Is there an algorithm to evualuate this ISO value?
I always thought the same, but I try to get my shots to be as ready-made as possible, so high ISO levels on camera and no exposure compensation under LR.
-
Is there an algorithm to evualuate this ISO value?
Perhaps, I’m not aware of one. But one can test this by bracketing and setting differing ISO settings and altering the Exposure slider in ACR to normalize all images, then must inspect the images up close.
-
Perhaps, I’m not aware of one. But one can test this by bracketing and setting differing ISO settings and altering the Exposure slider in ACR to normalize all images, then must inspect the images up close.
G. Galilei wins again. Thanks anyway :)
-
I believe you are missing Eric’s point and my query. You don’t need a 5DMII. But it would be interesting to know if the higher ISO settings on the 5DMIII behave as reported by Eric on the 5DII in terms of better noise reduction with lower settings and appropriate edits in LR.
And you missed my grin. I'll shall repeat it here ;D
But yes I will look at that as well.
-
Is there an algorithm to evualuate this ISO value?
This needs to be determined empirically, and may proof to be an eye-opener ...
I tested this with a Color Checker card for my 1Ds3 and found no benefit for ISO settings beyond 800 and benefits for the highlight headroom. Shooting at ISO 1600 or H was more noisy than underexposed at ISO 800 + push.
You basically shoot at a given ISO setting at or above the unity gain level, and with several bracketed exposures (1/3rd stop steps) shorter than that. Push the 'under-exposed' images to the same output level as the base exposure and measure the standard deviation in the same (uniform) area.
Then repeat with the next higher full ISOs, and compare the output noise figures.
Cheers,
Bart
-
Dear Andrew and Bart,
I get your points but if you are being realistic you know that most photographers - working professionals (photojournalists, videographers, wedding and event shooters) as well as casual amateurs -- are going to take the more direct route and simply set the camera's ISO higher, even if a more complicated shooting and processing routine will yield somewhat technically better results.
Our goal as photographers should be to make better photographs than we made yesterday and not to become technique jockeys. Well that is my goal but I fully recognize that others might have different goals. For me, great or even good photographs have never been about technique and tools but about communicating my visions of the world. I am foursquare for whatever tools make that possible and will happily explore the possibilities combinations of available tools will offer, but will not sacrifice concentrating on seeing for the distraction of technology.
I will run the tests
-
I get your points but if you are being realistic you know that most photographers - working professionals (photojournalists, videographers, wedding and event shooters) as well as casual amateurs -- are going to take the more direct route and simply set the camera's ISO higher, even if a more complicated shooting and processing routine will yield somewhat technically better results.
That’s dumb. It plays into what could be considered marketing hype in terms of ISO settings. IF you can get better quality and the same results setting at a lower ISO and adjusting the raw, that says a lot. Assuming you are writing a review of this camera, it would be useful to separate the reality of the ISO benefits from the hype.
As you can see in another post here on LuLa, there are conditions where an ISO800 capture can have less noise than one set for ISO100. An eye opener for those so called working and casual photographers who take a more direct (if misleading) route.
Our goal as photographers should be to make better photographs than we made yesterday and not to become technique jockeys.
I agree and hence why I asked that you examine the results of setting the ISO real high versus lower and using the raw converter to possibly produce a better photograph.
-
That’s dumb. It plays into what could be considered marketing hype in terms of ISO settings. IF you can get better quality and the same results setting at a lower ISO and adjusting the raw, that says a lot. Assuming you are writing a review of this camera, it would be useful to separate the reality of the ISO benefits from the hype.
Is it really "dumb" to look at how most people will use a camera? This has nothing to do with hype. It is looking at the camera from the perspective of the actual user experience.
And yes I am reviewing the camera and if the technique we are discussing yields better results I will certainly point it out. But once I do that, I am reviewing the camera + a specific raw processing software's capabilities + an informed user and not just the camera.
-
Is it really "dumb" to look at how most people will use a camera?
IF they are using it incorrectly yes. Look again at Eric’s post about raw capture and high ISO settings. You’ll get less good results if you up the ISO for raw. Just because there is a high setting, doesn’t mean it is a good idea to use it, be it on a camera or your sound system. If you want to write about the new ISO capabilities of the camera, well test it so that you can inform if the settings are useful or not. IF you get better results with a lower ISO setting and a normalization with the raw converter, why on earth would you use the higher setting?
Or to put it another way, if using ISO 1600 produces better results in terms of noise with a raw adjustment than using 3200, how would you describe that 3200 setting? Useful? Or marketing hype? If it produces a better JPEG than a raw, that’s worth knowing and mentioning. Treating a camera system set for raw and JPEG identically isn’t useful in such conditions no?
-
IF they are using it incorrectly yes.
It may be dumb for them to use it that way. It isn't dumb to recognize that most are using it that way.And then I can point out that that there may be a better way - if they have the time and the inclination to explore that route.
That we still have to fight the raw vs. jpeg battle is bad enough.
-
I t may be dumb for them tosue it that way. It isn't dumb to recognize that most areusing it that way.
Then teach them! Isn’t that part of the goal of a review?
And then point out that that there may be a better way - if they have the time and the inclination to explore that route.
They can’t make such a decision of someone hasn’t tested the two processes and informed them that possibly, a higher setting produces worse images than using a lower setting and just moving a darn slider!
That we still have to fight the raw vs. jpeg battle is bad enough.
Yes, and consider where we’d be today if more knowledgeable people didn’t point out the differences.
You asked for comments on your captures. My comments were questions: Does the higher ISO settings bring anything to the party. I’ll let you know if you want to send me that 5DMIII <g> (grin emphasis)
-
Then teach them! Isn’t that part of the goal of a review?
no that is the goal of a tutorial. The goal of a review is to look at the product or performance and point out its good and bad points.
I’ll let you know if you want to send me that 5DMIII <g> (grin emphasis)
In the interest of your prolonged good health I don't want to tax your ability to hold your breath for long enough for that to happen. ;)
-
just moving a darn slider!
for some people just moving a darn slider in a specific piece of software doesn't work in terms of their necessitated by reality or choice workflow. What if they prefer, possibly by training or habit to use a third party piece of noise reduction or sharpening software?
-
The goal of a review is to look at the product or performance and point out its good and bad points.
This will be the last I’ll say on the subject unless you care to answer my initial question: If the performance of the product is inferior using a higher ISO setting than a lower one and adjusting a slider in a raw converter, don’t you think your readers would find that useful? If setting a higher value for ISO produces an inferior quality in this context, would you consider that a bad point?
Yes or no is all needed.
Better still would be an answer to the qualities and limits of the ISO options. Or you could just bracket some images and upload the raws. Some of us will be happy to inspect them.
for some people just moving a darn slider in a specific piece of software doesn't work in terms of their necessitated by reality or choice workflow.
So these raw shooters are expected to simply render the images provided with no alterations what-so-ever in their converters? Sounds more like JPEG shooters. The questions I asked are specifically targeted to the raw data. I fully expect JPEG users to futz with the higher settings and be done. The question wasn’t directed to that audience. Eric’s post make the distinction between JPEG and raw clear to me.
-
This will be the last I’ll say on the subject unless you care to answer my initial question: If the performance of the product is inferior using a higher ISO setting than a lower one and adjusting a slider in a raw converter, don’t you think your readers would find that useful? If setting a higher value for ISO produces an inferior quality in this context, would you consider that a bad point?
Yes or no is all needed.
Fortunately you are not an attorney and I am not in the witness box and so I can give a more a more complicated, reality based, larger than binary answer.
The question your question begs is a simple one: At the end of the day, when a photographer delivers a photograph either as a digital file or as a print to a client how much better or inferior is one way of doing things than another?
I think it is very important that a photographer either get trained or train themselves to deliver the best quality photograph they are capable of delivering but even inside that criteria there are nuances -some assignments will require a lot of processing and post processing to reach that goal and some will not. Elliot Erwitt makes the point eloquently when he said that his goal as a photographic technician was to "Work myself into a position of total versatility, so that I can do anything I want to do at the time I want to do it. Whether I do it or not is another question."
An even deeper look at the question you raise Andrew is to see that what you essentially propose is that for my (or anyone else's) review to be fair I need to look at this every possible combination of a camera's ISO settings put through every possible permutation of raw processing (Lightroom v4 /ACR 7; Capture One, Capture One pro, DPP, etc.) , sharpening, and noise reduction software. And maybe we should throw interpolation software in there as well. I have no doubt that after a few weeks of doing so we can each find the perfect combination for every photographic situation that presents itself - but in the meantime many photographs won't be made in the pursuit of am ever receding grail of absolute best technique.
-
Fortunately you are not an attorney and I am not in the witness box and so I can give a more a more complicated, reality based, larger than binary answer.
You asked for comments. I asked a question. Just tell me you have no desire to answer the question and that the posts you made were for some other purpose and we can all go our merry way.
-
As I said you aren't an attorney.
Hey tell you what : you are obviously a far better man than me Andrew Rodney. Far superior in every possible way. Who can possibly doubt that? I bow deeply and apologize to you and everyone else who is reading this for not following your explicit orders by not giving you the simple answer you ordered me to deliver.
-
But by the way I did point out long ago that i would be testing the under-expose-and-over-develop technique Bart and Eric suggested . If it yields better results, hats off to both of them and I'll share the results.
-
The D800 is halfway there; analog gain ends at ISO 1600. Now all Nikon engineers have to do is realize (doh!) that there is nothing to be gained by bit-shifting the data and throwing away all that highlight info, and put the ISO setting in metadata.
Thanks; that is nice to know. (As an aside, I recall many years ago that with the Olympus E-1, ISO settings beyond 800 had no effect on raw except through meta-data: is this common?)
And the rough figures I have seen suggest that full-well capacity on these sensors is in the range 30,000 to 60,000, or 2^15 to 2^16, so with 14-bit ADC output, there are 2 to 4e- per level at base ISO speed, which is 200 or less, so that by ISO 800, there are already at least as many levels as electrons.
But as you indicate in this comment
it's not the 'unity gain' which determines when pushing lower ISO competes with 'normal' exposure at higher ISO; rather it's when read noise in electrons stops decreasing.
the relevant maximum useful gain level is about "one level per photo-site dark noise electron count", rather than counting every electron, and that noise floor seems to be at least 2e- of dark noise from the well itself, so even one level per 2e- is enough, and that probably happens by ISO 400.
But speaking as one semi-physicist to a real physicist: it is not that increasing ISO speed decreases the number of electrons of read noise, since that converse of output noise levels back to an electron count is at best a convenient fiction, converting measured output noise levels back on the false assumption that all noise comes from noise present in the charge in the electron well, before ISO gain. I would prefer to say that the useful threshold for analog gain is when any additional increase in gain produces a linearly proportional increase in read noise, measured downstream of the amplification. Or more simply,
"further analog gain is only useful if it improves the S/N ratio in the output signal".
-
"further analog gain is only useful if it improves the S/N ratio in the output signal"
yep.
-
Hi Ellis,
Aside from the banter here, I would like to see a test at a smaller aperture. Say f8 with a longer exposure, to see the lens at its maximum potential. But maybe the noise was the main reason for the test, and you're not as concerned with the resolving detail?
Thanks,
John
http://www.johnmaclean.com/
https://www.facebook.com/JohnScottMacLean
-
Hi Ellis,
Aside from the banter here, I would like to see a test at a smaller aperture. Say f8 with a longer exposure, to see the lens at its maximum potential. But maybe the noise was the main reason for the test, and you're not as concerned with the resolving detail?
Thanks,
John
http://www.johnmaclean.com/
https://www.facebook.com/JohnScottMacLean
Thank you John for your contribution. I can definitely do that but of course that might compound long exposure
noise with high sensitivity setting noise. That's a different kind of worst case scenario!
-
An even deeper look at the question you raise Andrew is to see that what you essentially propose is that for my (or anyone else's) review to be fair I need to look at this every possible combination of a camera's ISO settings put through every possible permutation of raw processing (Lightroom v4 /ACR 7; Capture One, Capture One pro, DPP, etc.) , sharpening, and noise reduction software. And maybe we should throw interpolation software in there as well. I have no
No, that would be a more superficial look. Andrew's question is fairly simple to answer once you have the camera in hand (for example with the IRIS freeware). If you can't use IRIS (for example because you have no adequate RAW decoder at this point) changing a slider on a relatively under-exposed image is just a shortcut, an approximation. There's no need to test dozens of approximations. One will probably be enough to get a good idea.
And remember that the interesting camera reviews are the ones that go a bit further than the standard ones. We all know that there will be hundreds of reviews of that camera, most of them written in the very same mold that was used for the D30.
No need to let this discussion evolve into ad-hominems.
-
Thank you John for your contribution. I can definitely do that but of course that might compound long exposure
noise with high sensitivity setting noise. That's a different kind of worst case scenario!
Yes when I did the math I got 6 seconds @ f8. WOW that is a low light room shot!
-
Is it me, or is this some thermal amplificator noise in the lower right corner (red+blue)?
Sorry to not get back to this sooner. I am pretty sure it is flare from light reflecting off the glass door on the bookcase , not thermal noise. If the camera is producing that much thermal noise at only 1/6th second that would be a really bad problem. I will check for that tonight.
Edit 11:21PM 03/13/2012
I cannot get it to replicate, so I am tentatively going with my hypothesis that what we are seeing is flare or glare.
Edit 3:26PM 03/14/2012
I absolutely cannot get the effect to replicate after trying several times at even longer exposures. As this is a pre-production camera body it could be a transient effect but most likely is just flare.
-
I absolutely cannot get the effect to replicate after trying several times at even longer exposures. As this is a pre-production camera body it could be a transient effect but most likely is just flare.
OK, many thanks for testing my dumb hypothesis!
The contrary would have been bad news, indeed.