p.3 #3Images: 5Ds RAWs with +5.00 Exposure: Noise and DR
Braineack wrote:
wasnt the point of this thread to show the recovery and noise?
The thread was started to share samples that illustrate, well, what the post says they illustrate. Later on, my reply to your post concerned your observations about softness.
There is a lot to say about sharpness, now that you brought it up. My impressions are different than yours.
I wonder if you understand the point that I write about at the link I earlier provided? I mention this because you bring up your impressions of 100% magnification crops from lower MP sensor cameras. Take a moment and see if it makes sense:
Imagine that you put some really fine lens on a 22.1MP camera (e.g. 5DIII — 5760 pixels wide) and you make an exposure using some aperture. You look at it at 100% magnification on your screen, perhaps as a 800 x 800 pixel sample. It looks great. Your 100% magnification crop at 800 pixels wide represents roughly 14% of the image width.
Now put the same really fine lens on a 50.6MP camera (e.g. 5Ds — 8688 pixels wide) and you make an exposure using the same aperture. You look at it at 100% magnification on your screen, again using an section excerpted from the full image that is 800 x 800 pixels. It looks pretty good, but it seems less sharp than your lower photo site density 22.1MP camera! Clearly, the higher MP camera is "softer" than the lower MP camera! What a rip off!
But wait. Your 800 x 800 100% magnification crop represents a much smaller section of the full image. In fact it is only a bit more than 9% of the image width. You are "looking more closely" at the image — the image isn't getting softer. (This is true in the same way that in the film era looking at a negative with a 10X loupe did not mean that the negative was sharper than when you looked at it with a 15x loupe!)
Indeed, if you were to make prints at the same size — preferably a very large size — from the two photographs, in the worst case the higher MP file would not be visibly sharper than the lower MP file. In the better cases the higher MP file has the potential to be sharper, of course depending on accurate focus, choice of aperture, camera stability, and so on. The higher MP file, all else being equal, would never be softer.
Bottom line: If you look at 100% crop of images from camera with different numbers of photo sites, even if you control for all other variables (same lens, exposure, focus, camera stability, etc) at 100% magnification the higher MP source image should and will look softer — even though it isn't in actual photographs.
By the way, the results will be the same when it comes to the apparent sharpness of 100% magnification crops even if you put the lens on a 36MP or 42MP Sony sensor. This isn't a brand or model issue — it is simply an objective characteristic from observing any source files that vary in this way.
(Things will be different if you move to a different sensor format, such as comparing a 50MP Canon full frame nominally 24mm x 36mm image to a Pentax 645z 50MP image from the 33mm x 44mm sensor. That gets into differences in system resolution related to sensor area, a different issue.)
Now, as to noise and dynamic range, those are two related issues that are different that the one you raised about sharpness. As I have written many times, I have little doubt that the new Canon sensor will offer improved noise and DR performance — that's what I'm seeing in the sample files. I also have little doubt that Sony sensors will continue for now to have more DR than Canon sensors. That is one data point that affects my choices, but it is certainly far from being the only one or the most important one. Others may feel differently about that.
p.3 #5Images: 5Ds RAWs with +5.00 Exposure: Noise and DR
We always see these images like these 3 guys in shadow on the boat trotted out as places we want tons and tons of DR.
I still say you lift the shadows on that shot with a Sony sensor and you still have a bad picture. I've seen very few of these that look good after lifting.
You want it to look good:
- Meter on the subject (tourists, if they are indeed the subject), blow some sky or the white boat surfaces. This shot looks like there might be room to expose more anyway without blowing the highlights.
- Convert to B&W if you're bothered by something that gets blown.
- Use fill flash.
- Frame the picture differently to get better lighting or to reduce the shadow area
Actually if you fill flashed this image you'd still have a bad picture...
Same thing with the mother & baby picture. I don't see a DR problem, I see an unhappy baby first and foremost, 2nd a poorly chosen exposure. Use a flash properly and you'll be able to count the baby's eyelashes instead of looking at mush.
I also don't like these lifted shots when they don't accurately reflect what I would see if I was looking at the scene with my own eyes.
p.3 #6Images: 5Ds RAWs with +5.00 Exposure: Noise and DR
bboule wrote:
We always see these images like these 3 guys in shadow on the boat trotted out as places we want tons and tons of DR.
I still say you lift the shadows on that shot with a Sony sensor and you still have a bad picture. I've seen very few of these that look good after lifting.
You want it to look good:
- Meter on the subject (tourists, if they are indeed the subject), blow some sky or the white boat surfaces. This shot looks like there might be room to expose more anyway without blowing the highlights.
- Convert to B&W if you're bothered by something that gets blown.
- Use fill flash.
- Frame the picture differently to get better lighting or to reduce the shadow area
Actually if you fill flashed this image you'd still have a bad picture...
Same thing with the mother & baby picture. I don't see a DR problem, I see an unhappy baby first and foremost, 2nd a poorly chosen exposure. Use a flash properly and you'll be able to count the baby's eyelashes instead of looking at mush.
I also don't like these lifted shots when they don't accurately reflect what I would see if I was looking at the scene with my own eyes....Show more →
I actually think the River Thames photograph was well exposed. It goes very close to maximum luminosity and there are luminosity values of 0 in the darkest shadows. You are correct that if it had been necessary the photographer could have pushed the highlights a bit more and used the highlights fader or similar technique in post to recover the overly-bright highlights. But that wasn't necessary in this (mundane but useful) photographs since the deepest shadows only occupy very small areas of the scene.
I did not select this image because I felt that the alterations I was making were necessary or advisable to make the image better. (Seriously. Take a look at the whole frame of the +5 example I posted!) I picked it because it was at f/8 (not too much effect from things like diffraction blur), ISO 100 (where most people who aren't happy with Canon sensors believe they have their biggest DR and noise problems), and because it did contain a few areas of the lowest possible luminosity values.
p.3 #7Images: 5Ds RAWs with +5.00 Exposure: Noise and DR
Looking forward to Sunday's purported (6/14, according to Adorama & B & H) shipping date for the new models.
Given all the sample CR 2 files available for perusal, a number of things seem fairly apparent:
1) The resolution will be tremendous. But for the very sharpest results, the highest quality L series lenses will be absolutely mandatory. Anything less will be brutally revealed.
2) Attention to technique in the field will yield the best results. Highest quality tripods & ballheads, mirror lock-up, timed shutter release, precise focus, proper aperture choice.
3) As lamentable as it is, the 50mp sensor will not be the equal of Nikon or Sony 36mp sensors in dynamic range or higher ISO/noise performance.
That said, the solution to the wide dynamic range problems your original post addresses, Dan, is a fairly simple one that we've all grown accustomed to using for years: that is, careful adjustment of exposure in the field to maximize the entire dynamic range of the sensor by raising exposure levels as much as possible without blowing the highlights (expose to the right, ETTR).
And importantly, when confronted with wider dynamic ranges likely to yield luminosity values close to zero in the shadows, take a second exposure 1-2 f/stops brighter for quick and easy blending in post processing.
The highlight and shadow sliders are important targeted luminosity balancing adjustment tools in ACR or Lightroom. And when a bit more oomph is required, the adjustment brush using both shadows and highlight is considerably stronger than the global slider.
To reiterate, the best lenses, keen attention to focus and exposure technique in the field, stable tripod/ballhead, mirror lock-up, delayed shutter release, and a thorough understanding of when dynamic range requires additional exposures or GND filtering to capture sufficient file information in the darker parts of the histogram.
A little known extra tip: although it's well known that mirror slap, at certain shutter speeds, can severely affect sharpness, a delayed shutter release can also affect overall sharpness, albeit less than a mirror. The sliding of shutter curtain is simultaneous with sensor capture and there can be internal vibration.
It would be a simple firmware update to delay the sensor capture until after the curtain has slid open but to date companies have not given us this.
BUT, using Live View will give you this additional hedge again internal vibration. Mirror lock-up stills internal vibration from physical mirror slap; timed shutter release allows shutter press vibrations to still; Live View slides the shutter curtain open well in advance of capture eliminating the last of the vibration concerns that will, with a sensor of this resolution sensitivity, affect overall sharpness.
There's a reason Canon has redesigned the mirror mechanism and stiffened the camera/release plate interface. They're well aware how critical all these things will be in order to get the very best results from Canon's new upgrade for landscape and studio photographers.
After seven plus years with the always reliable 1Ds III, this new camera is exciting despite it not being all we would have liked.
p.3 #8Images: 5Ds RAWs with +5.00 Exposure: Noise and DR
D. von Briesen wrote:
That said, the solution to the wide dynamic range problems your original post addresses, Dan...
Actually, I don't see a "problem" with the DR of the sensor in the samples that I've looked at so far. (Sony appears to still produce larger DR, but that doesn't mean that the Canon sensor has a DR "problem" any more than, say, the current Sony sensor has a "problem" if a newer sensor provides more DR.)
I understand the risks of saying "never," but while sensors with increasing dynamic range will chip away at the number of situations requiring HDR, exposure blending, recovery in post, etc., I don't think we are likely to encounter a sensor for a long time that can handle every possible DR challenge. There will still be situations in which all of the usual techniques are necessary to maximize the quality of the image.
For my part, I rarely need to use exposure blends any more, given the more robust tools for dealing with bright highlights, dark shadows, shifting the midrange, and so forth. Even with some challenging scenes I can almost always get what I need from a single exposure. There are some cases where I cannot, but not very many.
From what I see in the 5Ds files, it looks like there will be good image data well down into very dark tones, and that it will be possible to push them back to useful luminosity levels with little problem. Even without applying any addition NR, I could use the +3 originally-black tones in a print without issues, and with NR I think I'll be able to push further.
Again, I'm not challenging the claim that the Sony sensor may handle even more DR than the Canon. I just don't find that quite as compelling a thing as some do. I guess we'll have to differ on that one.
I'm with you on much of what you write about getting a meaningful advantage from the 50MP sensor. I suppose that anyone will perhaps get smoother gradients and smaller noise "grain" for the smaller photo sites. However, those who want to take advantage of the potential for higher image resolution from their 50MP cameras will have to make some effort to get that. Like you, I always shoot in live view when doing landscape and similar work from the tripod. I use a large, solid tripod/head setup. I let the camera settle before triggering the shutter release. I'm pretty careful about focus. I know how to handle sharpening and related issues in post. I use very good lenses. Folks who work this way are going to see some improvement in final image resolution, and those who already push the upper boundaries of print size may see the greatest potential improvement.
p.3 #10Images: 5Ds RAWs with +5.00 Exposure: Noise and DR
dolina wrote:
Doesnt larger DR entail a larger RAW file?
That would depend. The file might actually be the about the same if the bit depth isn't changed and the greater DR is the result of cleaner data. If the file adds bit depth to accommodate a larger DR, then the file could be larger. I could be wrong, but I think the former is more typical.
p.3 #11Images: 5Ds RAWs with +5.00 Exposure: Noise and DR
dolina wrote:
Doesnt larger DR entail a larger RAW file?
Depends on whether the Raw is compressed or not, on whether the compression is lossless or lossy like Sony's, and depends on the extent of the high DR areas in the image.
Canon's Raw files have been 14 bit for a long time now.
But in any case, the sensor and the associated processing circuitry do not have to produce a linear or direct readout of light levels. The sensor can capture a dynamic range of X stops and produce a Raw that has fewer bits. After all, a JPEG image (for example) has a nominal 8 bits of dynamic range, but the scene dynamic range is made to fit in it.
An image that captures a large portion of the dynamic range of a scene does not have to have an enormous number of levels of tones. Similarly, an image can have a very large range of tones with subtle gradations but be portraying a low contrast scene with small dynamic range.
p.3 #12Images: 5Ds RAWs with +5.00 Exposure: Noise and DR
[Like you, I always shoot in live view when doing landscape and similar work from the tripod. I use a large, solid tripod/head setup. I let the camera settle before triggering the shutter release. I'm pretty careful about focus. I know how to handle sharpening and related issues in post. I use very good lenses.]
Good for everyone else to know, Dan, but I've been familiar with the quality of your portfolio for years now. Your work has always inspired along with your wise words on all things photography related.
I'm guessing you're in line for one of the new Canons (5DS or 5DS R?). It'll be fun to compare some notes. Can't wait for monsoon season at Grand Canyon with a new high res monster!!
p.3 #13Images: 5Ds RAWs with +5.00 Exposure: Noise and DR
D. von Briesen wrote: [Like you, I always shoot in live view when doing landscape and similar work from the tripod. I use a large, solid tripod/head setup. I let the camera settle before triggering the shutter release. I'm pretty careful about focus. I know how to handle sharpening and related issues in post. I use very good lenses.]
Good for everyone else to know, Dan, but I've been familiar with the quality of your portfolio for years now. Your work has always inspired along with your wise words on all things photography related.
I'm guessing you're in line for one of the new Canons (5DS or 5DS R?). It'll be fun to compare some notes. Can't wait for monsoon season at Grand Canyon with a new high res monster!! ...Show more →
Thanks!
I'm in line for the "R" model, though I could have almost as easily selected the non-R. And, yes, I'm looking forward to seeing what I can do in large prints with it. Friends who have been using the A7r and the D810 have been very pleased with the potential of the higher resolution sensors for larger work, and I figure it is about time to join the club!
p.3 #14Images: 5Ds RAWs with +5.00 Exposure: Noise and DR
dolina wrote:
Doesnt larger DR entail a larger RAW file?
Here is how it goes within the paradigm of linearly encoded and losslessly compressed raw files, what we get from Canon today:
The theorethical limit for what Dynamic Range that can be represented within 14 bits, is just above 15 stops. Explained in this thread. That would give some posterization around the blacks though, 15 stops DR would look visually better represented by 16 bits, and the theorethical DR would also be slightly higher with a 16 bits file representing 15 stops of dynamic range in the signal. But it is not meaningless to provide 15 stops of DR and stick to 14 bits files.
The increase in file size going from 14 to 16 bits to begin with, is like going from 28 to 32 MB or from 70 MB to 80 MB raw files. A rather modest increase in file size compared to what increased resolution brings.
Now for the goodie: Higher DR means less noise, wich means smaller file sizes with lossless compression. The effect of this is modest, but still large enough to almost compensate for the increased bit depth. If Canon has the technology of 15 stops DR, and give us that in 16 bits raw files, the net increase in raw file size will be negligible at low ISO. At high ISO we will not have that effect. It would make sense to provide the option for 12 bits files at high ISO since the noise floor is so high there.
p.3 #15Images: 5Ds RAWs with +5.00 Exposure: Noise and DR
The whole 14 bits of dynamic range thing is weird.. maybe they are already doing this but it makes more sense to store it in 16 bits because of the way computers work. Maybe they are doing something else where they pad it with 2-bits of something else, in which case they might not increase file size at all by storing it as 16-bits.
Dan thanks for posting your 5D workflow to lift shadows while preserving blacks.. I tried it last night on a few of my files and like the results, although it was funny, I had to look back through quite a few files before I started finding a few candidates. I just don't shoot many keepers that I feel like I need to lift shadows.
p.3 #16Images: 5Ds RAWs with +5.00 Exposure: Noise and DR
D. von Briesen wrote:
A little known extra tip: although it's well known that mirror slap, at certain shutter speeds, can severely affect sharpness, a delayed shutter release can also affect overall sharpness, albeit less than a mirror. The sliding of shutter curtain is simultaneous with sensor capture and there can be internal vibration.
It would be a simple firmware update to delay the sensor capture until after the curtain has slid open but to date companies have not given us this.
BUT, using Live View will give you this additional hedge again internal vibration. Mirror lock-up stills internal vibration from physical mirror slap; timed shutter release allows shutter press vibrations to still; Live View slides the shutter curtain open well in advance of capture eliminating the last of the vibration concerns that will, with a sensor of this resolution sensitivity, affect overall sharpness. ...Show more →
This is implemented in Live View with an electronic first curtain shutter, which performs a rolling electronic reset of the sensor at the start of an exposure to mimic the light-blocking effects of the first mechanical curtain shutter. The D810 actually supports this outside of Live View as well; Canon should be able to implement this also. The A7rii takes this a step further by supporting a fully electronic shutter mode, where both the first and second curtains are performed electronically; it remains to be seen if this affects the read noise in the image as it does in the A7s's implementation.
p.3 #17Images: 5Ds RAWs with +5.00 Exposure: Noise and DR
snapsy wrote: D. von Briesen wroteA little known extra tip: although it's well known that mirror slap, at certain shutter speeds, can severely affect sharpness, a delayed shutter release can also affect overall sharpness, albeit less than a mirror. The sliding of shutter curtain is simultaneous with sensor capture and there can be internal vibration.
It would be a simple firmware update to delay the sensor capture until after the curtain has slid open but to date companies have not given us this.
BUT, using Live View will give you this additional hedge again internal vibration. Mirror lock-up stills internal vibration from physical mirror slap; timed shutter release allows shutter press vibrations to still; Live View slides the shutter curtain open well in advance of capture eliminating the last of the vibration concerns that will, with a sensor of this resolution sensitivity, affect overall sharpness.This is implemented in Live View with an electronic first curtain shutter, which performs a rolling electronic reset of the sensor at the start of an exposure to mimic the light-blocking effects of the first mechanical curtain shutter. The D810 actually supports this outside of Live View as well; Canon should be able to implement this also. The A7rii takes this a step further by supporting a fully electronic shutter mode, where both the first and second curtains are performed electronically; it remains to be seen if this affects the read noise in the image as it does in the A7s's implementation....Show more →
In my view there are perhaps two circumstances in which electronic first-curtain shutter is a good thing, and I've used the feature in both of them.
The most common is when shooting from the tripod and trying to come as close as possible to eliminating an camera vibration. While mirror lockup eliminates "mirror slap," when a mechanical shutter is triggered it can produce some small amount of vibration. The electronic first curtain shutter does not, and this is the most stable way to operate the camera. I use it by default for landscape photography from the tripod.
In terms of image stabilization, these benefits are probably insignificant when shooting handheld, as there are larger magnitude factors that introduce camera motion, so in most cases it isn't really worth it to use this feature for that kind of shooting. However, the fully electronic shutter systems are very quiet. Believe it or not, I've actually used it for handheld photography with a 5DII when minimizing camera sound is important, for example when photographing certain kinds of musical performances. It is awkward, but I've handheld with long lenses while shooting in live view this way. (Canon's implementation is far from ideal for this — for example, AF is complex, and I usually use manual focus.)
p.3 #18Images: 5Ds RAWs with +5.00 Exposure: Noise and DR
gdanmitchell wrote: snapsy wroteD. von Briesen wroteA little known extra tip: although it's well known that mirror slap, at certain shutter speeds, can severely affect sharpness, a delayed shutter release can also affect overall sharpness, albeit less than a mirror. The sliding of shutter curtain is simultaneous with sensor capture and there can be internal vibration.
It would be a simple firmware update to delay the sensor capture until after the curtain has slid open but to date companies have not given us this.
BUT, using Live View will give you this additional hedge again internal vibration. Mirror lock-up stills internal vibration from physical mirror slap; timed shutter release allows shutter press vibrations to still; Live View slides the shutter curtain open well in advance of capture eliminating the last of the vibration concerns that will, with a sensor of this resolution sensitivity, affect overall sharpness.This is implemented in Live View with an electronic first curtain shutter, which performs a rolling electronic reset of the sensor at the start of an exposure to mimic the light-blocking effects of the first mechanical curtain shutter. The D810 actually supports this outside of Live View as well; Canon should be able to implement this also. The A7rii takes this a step further by supporting a fully electronic shutter mode, where both the first and second curtains are performed electronically; it remains to be seen if this affects the read noise in the image as it does in the A7s's implementation.In my view there are perhaps two circumstances in which electronic first-curtain shutter is a good thing, and I've used the feature in both of them.
The most common is when shooting from the tripod and trying to come as close as possible to eliminating an camera vibration. While mirror lockup eliminates "mirror slap," when a mechanical shutter is triggered it can produce some small amount of vibration. The electronic first curtain shutter does not, and this is the most stable way to operate the camera. I use it by default for landscape photography from the tripod.
In terms of image stabilization, these benefits are probably insignificant when shooting handheld, as there are larger magnitude factors that introduce camera motion, so in most cases it isn't really worth it to use this feature for that kind of shooting. However, the fully electronic shutter systems are very quiet. Believe it or not, I've actually used it for handheld photography with a 5DII when minimizing camera sound is important, for example when photographing certain kinds of musical performances. It is awkward, but I've handheld with long lenses while shooting in live view this way. (Canon's implementation is far from ideal for this — for example, AF is complex, and I usually use manual focus.)
I agree, handheld shots have different vibration harmonics; either they wont suffer from mirror slap or will have their acuity reduced for other reasons. Jim Kasson performed a series of careful experiments to demonstrate scenarios where an EFCS produces a noticeable improvement at 36MP, including measuring the effect over a range of shutter speeds. His articles can be viewed here.