Nikon D7000 Dynamic Range

[PierreVandevenne]

To futher improve DR, may we imagine ISO25 based sensors ?!

At the risk of repeating what's been said before: sensors don't have ISO values or base ISOs. They count photons. The more photons they can count (well capacity), the better it is for DR. The more the count is precise (in other words, the less noise of all source is present), the better it is for DR. That's all there is to it at the bottom with the current technology.  In that respect, the Sony sensor is clearly a winner, most probably if one reads Sony's tech papers and patent because a tremendous amount of work has been done on sensel measuring accuracy. They seem to have, at the sensel level, a processing that is more complex than what was done on a full frame basis by scientific sensors not so long ago.

The rest is just the result of some interpretation of the (true) RAW data by the camera hardware and firmware, normalized for the photographer's comfort. This is of course very important from the photographer's point of view, and there are advantages in knowing which abstract normalized setting (ISO) is best for a specific camera. But making the sensor-camera system better at an ISO value or another doesn't fundamentally change the situation.

[thierrylegros396]

OK, understood , well described !

What is really interresting in the D7000 and K5 cases is how they achieve the 3dB/stop slope !

Passive or active circuit ?!

[JR]

So the answer is, a signal level which is only 6dB above the noise floor of the camera can be representative of a deep shadow in a scene...

Finally we agree 

Seriously, I understand what you mean.. everything is relative.

Now, I know it's the case that the sun was last seen in Great Britain in the afternoon of the 23rd of August 1955 (well, let's not exaggerate. I think it might have been seen on a few occasions  since then)

Two or three occasions, that´s right.

 

[JR]

Let's now have a look at some shadows, not deep shadows, but merely areas in the shade under a tree.

Phwoar! What subtle detail and texture in that laterite stone. Can you see it? 

I certainly could at the time, when I was there.

That´s right. Everything in life is relative. DR too. And shadows 

What would be your estimate for this scene in terms of DR?

Nice pictures by the way. I like old trees.

The noise in the shadows has Canon trademarks all over it. I was using the D700 for a while earlier this year and I had to push the shadows really hard to get the same kind of noise I got from my Canon´s with no post processing in contrasty scenes.

How do you like the D7000 so far?


- John

[JR]

I think the answer depends on what in the Zone System is called your choice of "placement": at what brightness level you intend to print (or otherwise present) subject matter that is at a certain brightness level in the scene. Sometimes, there could be a shaded subject that is only 1% as bright as another, sunlit, part of the scene and you wish to place that shaded subject in the print as only a stop or so below middle gray ... at the cost of greatly reducing the relative brightness level at which the sunlit parts are placed on the print. Maybe ranges even more extreme than 100:1 can occur between, say, (a) shaded leaves or rocks that one wants to rendered light enough to show details, so not much more than one stop darker than middle gray, and (b) sunlit clouds or snow that one does not want rendered as sterile blank white, which on a print means not more than three stops lighter than middle gray. But slides and computer screens can go far brighter with the highlights than light reflected of a paper print.

BJL,

That is true. Choice of placement is important. A digression... When I work on images I use Photoshop tools like dodge&burn a lot on my B&W prints. Used carefully they can give the impression of higher DR in a print.  There is of course no real substitute for "choice of placement" and proper exposure but I find tools like d&b very useful. Knowing exactly where to place the shadows, mid-tones etc when you do the exposure is an art-form I think.

- John

 

[bjanes]

At the risk of repeating what's been said before: sensors don't have ISO values or base ISOs. They count photons. The more photons they can count (well capacity)...

I take exception to both of these statements. ISO 12232:2006 defines how the sensors in digital still cameras are rated. For a good summary, see this article on Wikipedia. The ISO saturation rating is used by DXO.

Most digital sensors do not have the resolution to count individual photons. Rather, the photons generate a charge or voltage, which is analog and is converted to a data number by the ADC (analog to digital converter). The number of photons collected is related to the data number by the camera gain (photo electrons per data number).

Regards,

Bill

[PierreVandevenne]

Yes, there is a "rating" in ISO for sensors. But that doesn't mean engineers sit down and design a 50 or 400 ISO sensor or that building a "25 ISO sensor" is the way to go to improve quality. It is just a _rating_, a bit like a credit rating, that is used to describe a certain behaviour in a way that is palatable for photographers. A far as photon counting is concerned, we can of course nitpick and, usualy, I think in terms of ADU (essentially because I used CCDs before they appeared in cameras) so I am well aware that we are not necessarily counting individual photons (but we aren't too far either) and that we aren't counting them perfectly either: in fact, we are counting them  with a coarse resolution.

This being said while all these notions of stops/ISO/photographic DR are certainly useful they do obfuscate the issues. That probably explains why we have seen outlandish claims in terms of DR which are simply impossible given a specific well capacity, even if the system was a perfect noiseless photon counter. I've been toying with a small sensor simulator that I wrote to try to reproduce the behaviour of camera sensors and I am under the impression that a lot of massaging is going on in the background. One possibility to extend the DR, at least in a photographic context, would be to have sensors with variable pixel gain.

At some point, I can see that converging with software HDR and changing our expectations about how images should look.

[thierrylegros396]

One possibility to extend the DR, at least in a photographic context, would be to have sensors with variable pixel gain.

Could it be possible that K5 and D7000 are working with such a sensor ?!

[bjanes]

Yes, there is a "rating" in ISO for sensors. But that doesn't mean engineers sit down and design a 50 or 400 ISO sensor or that building a "25 ISO sensor" is the way to go to improve quality. It is just a _rating_, a bit like a credit rating, that is used to describe a certain behaviour in a way that is palatable for photographers. A far as photon counting is concerned, we can of course nitpick and, usualy, I think in terms of ADU (essentially because I used CCDs before they appeared in cameras) so I am well aware that we are not necessarily counting individual photons (but we aren't too far either) and that we aren't counting them perfectly either: in fact, we are counting them  with a coarse resolution.

No one said that an engineer sits down and designs an ISO 40 or 400 sensor. The ISO rating merely allows one to determine how the sensor will respond to a given exposure, measured in lux*seconds. The analogy of a credit score is a poor one, since the credit score is largely determined by past behavior. If the creditor loses his job or incurs large medical bills, his behavior may change, but the behavior of the sensor is determined by the laws of physics. If you expose a uniformly lit surface according to a TTL light meter reading, the sensor will receive a predetermined exposure (lumens/square meter * seconds). Roger Clark gives equations for converting to incident photons. How many of those photons are captured as photoelectrons depends on the performance of the microlenses, absorption by the Bayer filters, quantum efficiency and other factors. I agree we are counting photons in a general way.

The ISO rating of the sensor is a very useful entity. If we expose according to the meter reading with a camera using the ISO saturation rating for the sensor and standard ISO calibration for the meter, the sensor will attain approximately 12.5% saturation. If we increase the exposure by 3 EV, the sensor will be at saturation. This relationship is helpful in exposing the highlights in ETTR.

Many camera manufacturers do not use the ISO saturation standard, but one can obtain the saturation ISO from DXO. The actual ISO is usually not the nominal (stated ISO). I would imagine that an engineer would design a sensor for optimal overall performance rather than to a predetermined ISO rating. The ISO rating would then be rounded to the nearest standard ISO number.

Regards,

Bill

[PierreVandevenne]

No one said that an engineer sits down and designs an ISO 40 or 400 sensor

Well, that was essentially the message I was responding to. But I don't see this in a confrontational way.

The ISO rating of the sensor is a very useful entity.

For photography practice, I agree, just like other abstractions are.  But I don't think they are helpful as far as getting to the bottom of things.
Just take a very simple case such as the focal length issue. Thinking in terms of 35mm equivalent has led to mass confusion in the early days of digital. Think in terms of what focal length means and size of sensors and the issue becomes 6th grade level. Of course, marketing wise it was better to claim an "38-380 mm focal length lens" in a small package than to claim having a "super cropping sensor". The same happens here, but there are so many factors that it is even more confusing.

[hjulenissen]

I am a big fan of graphical representations of complex matters.

Even though my physics may be wrong, do you think that the fundamental idea behind the plot included could be used as a means to convey and discuss DR/noise/linearity aspects of a sensor and camera?

[dosdan]

A generic comment on DR. I really think that the discussion around DR is overblown. Much of the discussion at all is overblown.

My experience is that DR is essentially plentiful when shooting with present days DSLRs at base ISO correctly exposing to the right. In real life we always have lens flare that reduces the achievable DR anyway.

There has been some discussion that a 13-stop or 14-stop DR is meaningless in the real world because of lens flare.  So instead consider the DxOMark SNR 18% figures both within and between formats. To my way of thinking, this is a reasonable indicator of how well the sensors in the cameras are performing for their size.

SNR 18% comparison within and between formats


Dan.

[bluekorn]

The DR of my ability to comprehend the technical delights involved in this discussion moved far beyond the right hand edge of my cerebral histogram during my reading in the first three of four pages here. Without wading through the remaining pages I'm going to put my IQ at risk and interject a beginners question.

I frequently find myself shooting Lake Superior scenes that include 2 to 2 1/2 stops of soft fog twilight that can be moved left to right, or right to left, presumably capturing the full dynamic range of the scene at many points along the histogram. The perceived optimal image on the LCD often appears "correct" to my eye at the histogram midpoint or even to the left of midpoint. When the appearance of a darker image with limited DR is desired for aesthetic reasons do we still shoot to the right ensuring well saturation and then somehow find our way back to apparent lower light levels in post processing?

Peter

[Ray]

The DR of my ability to comprehend the technical delights involved in this discussion moved far beyond the right hand edge of my cerebral histogram during my reading in the first three of four pages here. Without wading through the remaining pages I'm going to put my IQ at risk and interject a beginners question.

I frequently find myself shooting Lake Superior scenes that include 2 to 2 1/2 stops of soft fog twilight that can be moved left to right, or right to left, presumably capturing the full dynamic range of the scene at many points along the histogram. The perceived optimal image on the LCD often appears "correct" to my eye at the histogram midpoint or even to the left of midpoint. When the appearance of a darker image with limited DR is desired for aesthetic reasons do we still shoot to the right ensuring well saturation and then somehow find our way back to apparent lower light levels in post processing?

Peter

This topic of ETTR seems inexaustible.

The problem of attempting an accurate ETTR, in my experience, is the problem of inadvertantly blowing at least one channel in the brightest parts of the image.

If you have the luxury of time on your side, you can adopt the technique of using the spot meter mode on the brightest part of the image, then in manual mode increase exposure by an experimentally determined number of stops. For the Canon 5D, as I recall, that was 3 stops slower than the spot meter reading directed at the brightest part of the scene.

I gave up this technicque because it was too time-consuming and I missed a few moments, but I was impressed with its accuracy.

To avoid missing the moment, I now prefer to auto-bracket exposure when the scene has a high range of brightness levels.

Clearly, if the scene does not have a wide range of brightness levels, and you are using a camera that does have a wide dynamic range, such as the D7000, then it's not worth stressing yourself about getting an ETTR  (the purpose of which is to lower noise in the shadows and lower midtones) unless you are using a high ISO for the need of a fast shutter speed.

At high ISO, say 800 or 1600 and above, the DR of the camera is significantly reduced. For clean midtones, an ETTR is essential in those circumstances.

If in doubt, experiment with auto-bracketing +/- 1 stop.

[bluekorn]

Ray, thank you. This is very helpful. The behavior of digital cameras from capture through processing to card storage is very complex to my way of thinking. I find it challenging to ferret out the rudimentary understanding, which is my need and want, from the body of scientific discussions which quite naturally diverge into seemingly endless channels of nuance beyond my interest and ability. And I agree with you that some notion of how it all works and the ability to keep from making the same mistakes twice, the improved efficiency in process, adds considerably to the pleasure in making photographs. Thanks again. Peter 

[bjanes]

Clearly, if the scene does not have a wide range of brightness levels, and you are using a camera that does have a wide dynamic range, such as the D7000, then it's not worth stressing yourself about getting an ETTR  (the purpose of which is to lower noise in the shadows and lower midtones) unless you are using a high ISO for the need of a fast shutter speed.

At high ISO, say 800 or 1600 and above, the DR of the camera is significantly reduced. For clean midtones, an ETTR is essential in those circumstances.

I agree that if one has a camera with low read noise and a high DR (such as the Nikon D7000), obtaining a perfect ETTR is not necessary since one might blow the highlights and lose data. However, the concept of ETTR implies that a histogram to the right is the critical factor, whereas in fact the exposure (number of photons collected) is the critical factor. Older cameras such as the Nikon D3 have higher read noise at low ISO, and when f/stop and shutter speed considerations limit exposure, it is advantageous to increase the ISO to take advantage of the lower read noise.

However, read noise approaches a minimum asymptotically at higher ISO, and raising the ISO beyond this point only serves to limit head room without decreasing read noise. As Emil Martinec has pointed out, this point of diminishing returns can be determined by looking at the DR plotted vs the ISO on DXO. In a camera whose DR is limited by shot noise, the plot becomes linear and DR drops by one stop for each doubling of ISO.  When read noise is significant, the curve bows downward towards the left, as shown for the D3 plot below ISO 800.

As the plot shown below indicates, the curve for the D3 becomes linear slightly below ISO 800. Increasing ISO above 800 with this camera limits highlight headroom and does nothing for the shadows; one can leave the camera setting at 800 and merely increase exposure in the raw converter. If the histogram is to the left at ISO 1600 and to the right at ISO 800, the DR will be similar, but with ISO 1600 there is a danger of blowing the highlights. In this sense, ETTR judged by the histogram is really a misnomer.

The D7000 is linear from the start, and one really does not have to increase the ISO to obtain a better read noise. One could leave the ISO setting at base and give as much exposure as f/stop and shutter speed concerns allow, and then increase exposure in the raw converter. Of course, the image will appear dark on the camera LCD, but highlight head room will be preserved.

Regards,

Bill

[Ray]

The D7000 is linear from the start, and one really does not have to increase the ISO to obtain a better read noise. One could leave the ISO setting at base and give as much exposure as f/stop and shutter speed concerns allow, and then increase exposure in the raw converter. Of course, the image will appear dark on the camera LCD, but highlight head room will be preserved.

Regards,

Bill

That's quite true, Bill. I was really speaking in general terms in my previous post. It's not clear that Bluekorn uses a D7000.

The D7000 is in a special category that makes the concept of ETTR quite redundant when the shutter speed at base ISO, required for an ETTR, is too slow for the subject, or too slow to prevent camera shake.

Having determined that the shutter speed at base ISO necessary for a reasonably accurate ETTR is too slow for the circumstances, one can simply switch to manual mode (if one isn't already in manual mode) and increase the shutter speed to whatever one assesses is appropriate, without bothering to change ISO.  ETTR is then no longer a consideration and one doesn't have to worry about the possibility of blown highlights.

[bluekorn]

I presently shoot with a Sony R1 but in in order to avoid further graduate studies I'm considering buying the D7000 as it "is in a special category that makes the concept of ETTR quite redundant" (thank you Ray).

The "number of photons collected" is the critical factor. (In fact, I feel an intuitive resonance with this claim although I couldn't support it logically). I recall reading somewhere that the further right one goes on the histogram the greater the number of photons one will collect. The author divided the histogram into thirds and stated in support of ETTR that something like sixty percent of photons collected are available to the right hand third and only ten percent in the left hand third. I drew from this that even though the DR of many of the subjects I shoot is two to two and a half stops, I would benefit from ETTR because I collected more photons. I'm certainly over my head in this discussion but I sincerely want to capture photons to my best advantage in making prints.



 

[Ray]

The "number of photons collected" is the critical factor. (In fact, I feel an intuitive resonance with this claim although I couldn't support it logically). I recall reading somewhere that the further right one goes on the histogram the greater the number of photons one will collect. The author divided the histogram into thirds and stated in support of ETTR that something like sixty percent of photons collected are available to the right hand third and only ten percent in the left hand third. I drew from this that even though the DR of many of the subjects I shoot is two to two and a half stops, I would benefit from ETTR because I collected more photons. I'm certainly over my head in this discussion but I sincerely want to capture photons to my best advantage in making prints.

For the smoothest and cleanest image the camera is capable of, one should ideally strive to capture as many photons as possible, consistent with an oppropriate shutter speed and aperture, and without blowing any of the channels.

However, there are lots of trade-offs and compromises on the technical side of photography that one has to deal with when the camera is not used in fully automatic mode, so one should try to  be aware of the significance of the effects of any settings which may not be ideal as a result of the requirements for the circumstances and the lighting conditions.

That is why I recommend experimenting with different settings, different exposures through autobracketing for example, to see for oneself what effect an underexposure of 1/2 a stop, or 1 stop, or 2 stops or more, may have on the quality of certain parts of an image using a particular model of camera.

Because I'm a peripatetic type of photographer, I don't always have the time to prepare each shot for a perfect ETTR, nor the opportunity to retake the same scene if I got it wrong, so I'm very attracted to the high-DR characteristics of the D7000 which allow me to underexpose an image at base ISO by at least one stop, and maybe as much as 2 stops, whether by design or accident, and still achieve image quality in the shadows on a par with an ETTR shot of the same scene using the equivalent Canon 60D.

Comparing the D7000 at 1 stop underexposure with the 60D at half the shutter speed (double the exposure), I would not expect to see any image quality advantage in the 60D image in any respect, whether in the deep shadows, moderate shadows, midtones or highlights, with regard to DR, SNR, tonal range or color sensitivity. Slight differences in resolution due to lens quality, AA filter characteristics and sensor pixel count is another issue.

A two-stop underexposure comparison would be interesting because theoretically shot noise in the shadows of the D7000 image should then be greater, and SNR in the midtones might also be noticeably worse.

On the other hand, the advantages of a 1/100th sec exposure with the D7000, as opposed to a 1/25th sec exposure with the 60D, could be of far greater benefit than the disadvantage of a barely perceptible increase in shot noise or a barely perceptible reduction in SNR in the midtones that might only be apparent at 100% on the computer monitor, representative of a huge print of around 6ftx4ft.

[hjulenissen]

On the other hand, the advantages of a 1/100th sec exposure with the D7000, as opposed to a 1/25th sec exposure with the 60D, could be of far greater benefit than the disadvantage of a barely perceptible increase in shot noise or a barely perceptible reduction in SNR in the midtones that might only be apparent at 100% on the computer monitor, representative of a huge print of around 6ftx4ft.

I agree that having a larger practical DR is a significant recent progress in DSLRs. If not for carefully planned studio-shots of sensible DR scenes, then for practical photography where you have one chance to get it right, and light can be hard to control.

But if the differences in noise are none between a D7000@1/50s and a 60D@1/25s and "barely perceptible" between a D7000@1/100s and a 60D@1/25s, would you not also expect the differences to be "barely perceptible" between a 60D@1/25s and a 60D@1/50s?

I agree that capturing a large amount of photons is important. I am sceptical about the implicit message that perceptual image quality is somehow a 1:1 correlate with the number of counted photons. Clearly it is not, as very few photons are needed to distinguish one scene from another (say, 100 photons per pixel on average), while increasing the average number of photons per pixel from 100000 to 100100 will make no perceptual difference at all. Stephens power law etc.

-h