DxOMark's new resolution magnitude: P-Mpix

[Ray]

Ray, can you expain what lead to this "histogram three stops too far to the left" problem?

I agree that one cannot alway use manual setting of shutter speed and aperture, so I suppose you were in a mode like aperture or shutter priority, but then how did you end up with those very low raw levels? Was it shutter priority plus ISO 200 leading to hitting the maximum aperture available?

BJL,
I was in full manual mode regarding aperture and shutter speed. I had adopted the principle of 'F8 and be there', because I was there. The shutter speed happened to be set at 1/200th because my earlier recent shots, with 24-120/F4 zoom attached, had been taken with the lens fully extended at 120 mm, which is 180 mm in 35 mm terms. From experience I've found that a shutter speed of at least 1/FL(35mm), in combination with Image Stabilization, is necessary to get good sharpness from very high-resolution sensors such as the D800 and D7100, when the cameras are hand-held. Some folks might disagree with this. I suppose it depends on how steadily one can hold the camera. I have no signs of Parkinson's yet.

Whilst I can't remember my precise thoughts when I took this particular shot (one of thousands taken during a period of a few weeks), I always pay particular attention to the exposure indicator stretched along the bottom of the viewfinder, whenever I shoot in manual mode. I would have noticed that a 200th at F8 was resulting in significant underexposure. I would have been reluctant to reduce exposure even further by using a shutter speed of 1/400th, and reluctant to reduce DoF by adjusting the aperture to F5.6.

Anyway, you surely realize that we are not yet dealing with completely "ISO-less" cameras, and with recent Nikons, there is something to be gained by extra amplification up to about ISO speed 400 or 800.

I can't agree that none of the recent Nikon DSLRs are completely ISO-less in practical terms. If you check out the results for the D7000 at DXOMark and compare noise levels at ISO 100 and ISO 3200, which represents a 5 stop difference, you'll find that the differences in DR measurements between ISO 100 and ISO 3200 is exactly 5 EV, or 5 stops. If I've interpreted this correctly, it means that a shot with an underexposure of 5 stops at ISO 100 will have exactly the same degree of shadow noise as that same exposure used at ISO 3200.

As regards SNR at 18% grey, the results are very close. At ISO 100 we have 41.1 dB. At ISO 3200, 26.7 dB. That's a difference of 14.4 dB. On the basis that 3 dB is equivalent to a difference of 1 EV or 1 stop, an underexposure of 5 stops at ISO 100 should result in a fall of 15 dB in SNR. One gains a mere 0.6 dB in SNR at 18% by using ISO 3200. Wow!  I bet you that difference would be invisible on any size print, even on the creamiest skin of the most beautiful model. 

The DXO results for the Nikon D800E are similar, with the exception of DR between ISO 100 and 200. It's less than it should be for an ISO-less camera. However, SNR at 18% is exactly 8 stops down at ISO 25,600, or 24 dB down, which is what 8 stops of underexposure at ISO 100 should produce. The D800E can be described as a  true ISO-less camera from ISO 200 to ISO 25,600. The loss in DR in those 7 stops of underexposure is a mere 1/3rd of a stop, and there's no loss at all in SNR at 18%. That's not bad.

If we compare the DXO measurements for the D7100, we find that the ISO-less nature is not quite as good as that of the D7000 and is best within a narrower range of ISOs, specifically from ISO 200 to ISO 1600, which represents the 3 stops underexposure that I used in my above shot of the mule.

According to DXOmark, I've sacrificed 1/3rd of a stop of DR, and 0.3 dB of SNR at 18%, by underexposing 3 stops at ISO 200 instead of raising ISO to 1600. DXOMark in their articles on such matters claim that a difference of 0.5 EV in DR should be noticeable, implying that any differences less than 0.5 EV might not be noticeable, or not significant.

In summary, the D7000 is very close to the true ISO-less camera, closely followed by the D800E. The D7100 is clearly a backward step in this regard, particularly in view of the banding which is not apparent in either the D7000 or D800E.

[BJL]

Are you suggesting that digital amplification is better than analog amplification since you suggest limit the ISO to 400 or 800? (assuming no highlight clipping).

Not universally, but in one somewhat common situation: once the analog gain is enough to raise all noise in its input comfortably above any noise that comes later (like ADC quantization noise), further analog gain has no further benefit in S/N ratios, and instead using perfectly noise-less digital "bit shifting" for any further level/brightness adjustments has several potential advantages over the inherent imperfections and noise sources of additional analog gain:

1. avoiding clipping of highlights by the amplifier or ADC from over-amplifying the signal from a "bright" but not overfull photosite.

2. simplifying the amplifier and/or chip design. The extreme case is if an ADC is good enough that a "base-ISO speed" amplification that sends full wells to a voltage just within the ADC's range then has the ADC quantization noise floor well below the noise floor of this input signal: then it is probably best to used a fixed gain circuit design rather than adding the imperfections of a variable gain amplifier. Hasselblad seems to do this in some backs with their combination of 12-13 stop DR sensors and good 16-bit off-board ADCs.
Another example is a method of applying analog gain in CMOS sensors described in a Canon research paper: using multiple sense capacitors at the edge of the sensor, and reading to a larger or smaller one in order to apply more or less charge gain. Here needing fewer gain levels and thus fewer sense capacitors and less switching circuitry probably has some design advantages.


The limits of 400 and 800 I mention are roughly what I see with recent Nikon and Olympus system cameras, where analog gain beyond those levels raises total noise levels almost exactly in proportion to signal levels, so offering no gain in SNR or DR or any other IQ metric I can think of; only a potential loss in DR through highlight clipping.

[BJL]

Ray,
    I see only two possibilities:

1. The banding can be completely fixed with appropriate post-processing, so choosing a higher ISO speed setting would only have been a matter of convenience, not IQ.

2. Despite the ISO-independence of the DR and SNR measurements you cite, there are other issues such as pattern noise which sometimes lead to problems like the banding you have seen when the signal from the far less than fully exposed sensor is amplified too little. That is, there can be more to the ISO-less ideal that "the ISO setting is irrelevant to IQ except to avoid amplifier clipping" than revealed by DR and SNR specs alone.

[Hans Kruse]

Not universally, but in one somewhat common situation: once the analog gain is enough to raise all noise in its input comfortably above any noise that comes later (like ADC quantization noise), further analog gain has no further benefit in S/N ratios, and instead using perfectly noise-less digital "bit shifting" for any further level/brightness adjustments has several potential advantages over the inherent imperfections and noise sources of additional analog gain:

1. avoiding clipping of highlights by the amplifier or ADC from over-amplifying the signal from a "bright" but not overfull photosite.

2. simplifying the amplifier and/or chip design. The extreme case is if an ADC is good enough that a "base-ISO speed" amplification that sends full wells to a voltage just within the ADC's range then has the ADC quantization noise floor well below the noise floor of this input signal: then it is probably best to used a fixed gain circuit design rather than adding the imperfections of a variable gain amplifier. Hasselblad seems to do this in some backs with their combination of 12-13 stop DR sensors and good 16-bit off-board ADCs.
Another example is a method of applying analog gain in CMOS sensors described in a Canon research paper: using multiple sense capacitors at the edge of the sensor, and reading to a larger or smaller one in order to apply more or less charge gain. Here needing fewer gain levels and thus fewer sense capacitors and less switching circuitry probably has some design advantages.


The limits of 400 and 800 I mention are roughly what I see with recent Nikon and Olympus system cameras, where analog gain beyond those levels raises total noise levels almost exactly in proportion to signal levels, so offering no gain in SNR or DR or any other IQ metric I can think of; only a potential loss in DR through highlight clipping.

So you are basically saying that there is no advantage in underexposing in capture and raise exposure in pp relative to raising ISO in the camera (still assuming no highlight clipping).

Given that I simply don't understand why anybody would underexposure many stops and make it impossible to use the LCD to review or even see pictures and also have to adjust exposure in e.g. Lightroom to even review the pictures initially.

I see no practical purpose in that method and also it as unncessary. However it is a benefit to use a technique whereby it is not neccessary to constantly review histograms and adjust exposure compensation. I would much prefer to have an ETTR setting on my camera so that the camera automatically would exposure to the right without having to work against the metering. This would be really beneficial on both Canons and Nikons etc. for low ISO work as well as higher ISO work. For Canons one would still need to bracket more than Nikons, but still for Nikons (or similar behavior) bracketing is needed for optimum IQ. I shot both cameras and on the Nikon D800E I see noise coming in lifting exposure by two stops where optimal exposure mostly leaves a smooth noise free tone in e.g. clouds shooting at ISO 100. Raising more than a few stops ruins IQ in the details. What my experience tells me is that at ISO 100 optimum IQ also needs optimum exposure.

[Hans Kruse]

BJL,
I was in full manual mode regarding aperture and shutter speed. I had adopted the principle of 'F8 and be there', because I was there. The shutter speed happened to be set at 1/200th because my earlier recent shots, with 24-120/F4 zoom attached, had been taken with the lens fully extended at 120 mm, which is 180 mm in 35 mm terms. From experience I've found that a shutter speed of at least 1/FL(35mm), in combination with Image Stabilization, is necessary to get good sharpness from very high-resolution sensors such as the D800 and D7100, when the cameras are hand-held. Some folks might disagree with this. I suppose it depends on how steadily one can hold the camera. I have no signs of Parkinson's yet.

Whilst I can't remember my precise thoughts when I took this particular shot (one of thousands taken during a period of a few weeks), I always pay particular attention to the exposure indicator stretched along the bottom of the viewfinder, whenever I shoot in manual mode. I would have noticed that a 200th at F8 was resulting in significant underexposure. I would have been reluctant to reduce exposure even further by using a shutter speed of 1/400th, and reluctant to reduce DoF by adjusting the aperture to F5.6.

I can't agree that none of the recent Nikon DSLRs are completely ISO-less in practical terms. If you check out the results for the D7000 at DXOMark and compare noise levels at ISO 100 and ISO 3200, which represents a 5 stop difference, you'll find that the differences in DR measurements between ISO 100 and ISO 3200 is exactly 5 EV, or 5 stops. If I've interpreted this correctly, it means that a shot with an underexposure of 5 stops at ISO 100 will have exactly the same degree of shadow noise as that same exposure used at ISO 3200.

As regards SNR at 18% grey, the results are very close. At ISO 100 we have 41.1 dB. At ISO 3200, 26.7 dB. That's a difference of 14.4 dB. On the basis that 3 dB is equivalent to a difference of 1 EV or 1 stop, an underexposure of 5 stops at ISO 100 should result in a fall of 15 dB in SNR. One gains a mere 0.6 dB in SNR at 18% by using ISO 3200. Wow!  I bet you that difference would be invisible on any size print, even on the creamiest skin of the most beautiful model. 

The DXO results for the Nikon D800E are similar, with the exception of DR between ISO 100 and 200. It's less than it should be for an ISO-less camera. However, SNR at 18% is exactly 8 stops down at ISO 25,600, or 24 dB down, which is what 8 stops of underexposure at ISO 100 should produce. The D800E can be described as a  true ISO-less camera from ISO 200 to ISO 25,600. The loss in DR in those 7 stops of underexposure is a mere 1/3rd of a stop, and there's no loss at all in SNR at 18%. That's not bad.

If we compare the DXO measurements for the D7100, we find that the ISO-less nature is not quite as good as that of the D7000 and is best within a narrower range of ISOs, specifically from ISO 200 to ISO 1600, which represents the 3 stops underexposure that I used in my above shot of the mule.

According to DXOmark, I've sacrificed 1/3rd of a stop of DR, and 0.3 dB of SNR at 18%, by underexposing 3 stops at ISO 200 instead of raising ISO to 1600. DXOMark in their articles on such matters claim that a difference of 0.5 EV in DR should be noticeable, implying that any differences less than 0.5 EV might not be noticeable, or not significant.

In summary, the D7000 is very close to the true ISO-less camera, closely followed by the D800E. The D7100 is clearly a backward step in this regard, particularly in view of the banding which is not apparent in either the D7000 or D800E.

This can all be true but I do not understand why you choose this technique as there are so many disadvantages with it.

[ErikKaffehr]

Hi,

That seems to me like the optimum recipe for perfection.

The impression I have is that it is beneficial to raise ISO on Canons, perhaps up to 1000 ISO instead of just reducing exposure as higher ISO will reduce shadow noise.

Pushing up ISO will tell camera firmware and raw processor that it is a high ISO image. So raw processor can increase noise reduction.

Best regards
Erik

What my experience tells me is that at ISO 100 optimum IQ also needs optimum exposure.

[Hans Kruse]

Hi,

That seems to me like the optimum recipe for perfection.

The impression I have is that it is beneficial to raise ISO on Canons, perhaps up to 1000 ISO instead of just reducing exposure as higher ISO will reduce shadow noise.

Pushing up ISO will tell camera firmware and raw processor that it is a high ISO image. So raw processor can increase noise reduction.

Best regards
Erik

At ISO 100 for both Canon and Nikon you will get the best IQ with the longest exposure without burning out essential detail (=highlights) and judged in Lightroom.

[BJL]

So you are basically saying that there is no advantage in underexposing in capture and raise exposure in pp relative to raising ISO in the camera (still assuming no highlight clipping).

Firstly, I am talking about "limited light" situations where using base ISO speed, say 100, would be significant underexposure, and in this case, your talk of using ISO 100 and optimum exposure is irrelevant. For example, in Ray's Nepalese photo, using ISO 100 and "properly exposure" would have given images been blurred due to excessively long exposure times.

Secondly, I am not talking about reducing _exposure_, in the sense of increasing either shutter speed or aperture ratio; I am talking solely about what ISO setting to use when one is already using the lowest acceptable shutter speed and the largest acceptable aperture, and so the sensor is already getting as much light as is possible from the scene, and this is still leaving all the photosites well short of full. (That is, "exposing to the right" in the original sense is not an option.)

With that clarified, what I am saying is that sometimes, when things are happening too quickly to allow careful metering decisions, it can be wise to be conservative in ISO setting (meaning maybe 200 to 800) and thus use a lowish level of analog gain, in order to avoid clipping of highlights --- meaning highlights where the photosite was short of full, so they are not blown in the photosite, but where a too high analog gain can push the value beyond the maximum of the amplifier and ADC.

However, I only suggest this as a last resort, when simpler options like bracketing are ruled out, for example because the scene is changing too quickly to allow more than one frame.

And only with "near ISO-less" sensors! Ray has in years past posted evidence of the IQ problems that this underamplification caused with some Canon DSLRs.

P. S. Another clarification: I am not suggesting that one always limit the ISO setting to 400 or 800 or whatever on cameras like the Nikon 7000 or 7100; when the metering is clear cut, using a higher ISO setting to get suitable levels in standard JPEG conversions seems simpler and safe. (But with some cameras, the effects of a setting like ISO 1600 might best be implemented with the same analog gain as ISO 100, plus a flag in the raw file for 16x digital shift.)

[ErikKaffehr]

Hans,

I have a small issue with 'judged in Lightroom'. Lightroom with PV2012 does some highlight recovery and often shows a nice histogram even with clipped highlights. I often use RawDigger to look at 'real' histogram in the raw file.

Best regards
Erik

At ISO 100 for both Canon and Nikon you will get the best IQ with the longest exposure without burning out essential detail (=highlights) and judged in Lightroom.

[Ray]

This can all be true but I do not understand why you choose this technique as there are so many disadvantages with it.

Hans,
I'm simply not aware of the many disadvantages you refer to. The main disadvantage that I've experienced is that one cannot show off to another person the interesting shot one has just taken, because it appears so underexposed on the camera's LCD screen.

As regards reviewing the image for the purpose of retaking the shot if it doesn't look satisfactory, that should not be a concern in the circumstances. If the shot is not satisfactory, the moment will likely have passed by the time one has enlarged the review to inspect the image.

The advantage of shooting in full manual mode at base ISO in circumstance where one knows that underexposure in relation to full-well capacity is unavoidable, is that one has the greatest certainty that  the ISO setting,  and any possible overexposure in terms of blown highlights, will not be an issue.

Have you never taken a shot with a 'minus-1-stop' exposure compensation, that has turned out to have blown highlights? I have. Shoot (in a hurry) some interesting activity that's taking place in the shadows in, say, the lower right corner of the frame, against a background of an interesting, brightly lit sky in the upper left corner, and you might find that a 'minus one stop EC' is not sufficient to bring out the detail in the sky in the top left corner of the frame.

Your approach of using auto ISO with a minus one stop exposure compensation might work just as well most of the time, but would seem to me to provide less certainty for all situations that one might encounter. Perhaps you would then recommend a minus-2-stop EC.

I'll give you an example of what would appear to me to be an unsatisfactory outcome that could result from your method of using auto-ISO in conjunction with exposure compensation. But I'm no expert with this method. There might be settings in the menu of the D7100 I am unfamiliar with.

Let's consider the situation of using a 'minus 2 stop' exposure compensation in auto-ISO mode. In the circumstances of continuously changing lighting conditions that occur when walking along a mountain track, you might inadvertently underexposed at ISO 100. The best choice might have been, for example, ISO 200, which is the setting that I would use with the D7100.  As a result of the 'minus-2 stops' exposure compensation in conjunction with auto-ISO, the camera has lowered ISO to 100 and underexposed a further stop.

In such circumstances, using the D7100, your shot would have more than a full stop worse noise and DR than mine. Is this not the case? If not, please advise what settings you would use on the D7100 to avoid this situation.

In fact, having just now experimented with your recommended method on the D7100 as I write this, I'm rather puzzled that I get no indication of underexposure on the indicator at the foot of the viewfinder. It's either zero or in the overexposed region, yet I can take shots that look underexposed on the LCD review. What's going on here?

[Hans Kruse]

Firstly, I am talking about "limited light" situations where using base ISO speed, say 100, would be significant underexposure, and in this case, your talk of using ISO 100 and optimum exposure is irrelevant. For example, in Ray's Nepalese photo, using ISO 100 and "properly exposure" would have given images been blurred due to excessively long exposure times.

Secondly, I am not talking about reducing _exposure_, in the sense of increasing either shutter speed or aperture ratio; I am talking solely about what ISO setting to use when one is already using the lowest acceptable shutter speed and the largest acceptable aperture, and so the sensor is already getting as much light as is possible from the scene, and this is still leaving all the photosites well short of full. (That is, "exposing to the right" in the original sense is not an option.)

With that clarified, what I am saying is that sometimes, when things are happening too quickly to allow careful metering decisions, it can be wise to be conservative in ISO setting (meaning maybe 200 to 800) and thus use a lowish level of analog gain, in order to avoid clipping of highlights --- meaning highlights where the photosite was short of full, so they are not blown in the photosite, but where a too high analog gain can push the value beyond the maximum of the amplifier and ADC.

However, I only suggest this as a last resort, when simpler options like bracketing are ruled out, for example because the scene is changing too quickly to allow more than one frame.

And only with "near ISO-less" sensors! Ray has in years past posted evidence of the IQ problems that this underamplification caused with some Canon DSLRs.

P. S. Another clarification: I am not suggesting that one always limit the ISO setting to 400 or 800 or whatever on cameras like the Nikon 7000 or 7100; when the metering is clear cut, using a higher ISO setting to get suitable levels in standard JPEG conversions seems simpler and safe. (But with some cameras, the effects of a setting like ISO 1600 might best be implemented with the same analog gain as ISO 100, plus a flag in the raw file for 16x digital shift.)

Probably I didn't say that clearly enough, but my question to you was rather specific and not about Ray's situation. It was simply: Is there a difference in IQ of the resulting image between the two situations where you shoot 1) one picture at ISO 100 and underexposed by zero to e.g. 6 stops and 2) the same picture shot at ISO settings 100, 200, 400 and 6400? Again assuming no highlight clipping.

I assume the amplification in the camera for lifting ISO is done in the analogue domain and not in the digital domain in cameras like the Nikon D7100, D800 etc. If the sensor is truly ISO-less the two situations above should be the same except for the resolution, i.e. the number of bits representing the tones where the analogue amplification and then digitizing will represent the picture with a full 14bit resolution where digital amplification is simply shifting the bits x-number of positions (where is the number of stops). The other difference could be the noise characteristics with a noise floor that gets amplified with digital amplification but not with analogue (noise floor in the ADC). With a truly ISO-less camera the noise floor of the ADC should be zero. Are there other factors?

Btw. the 16x shift, I assume you mean 4x shift (shifting bits 4 positions) going from ISO 100 to ISO 1600.

[Hans Kruse]

P. S. Another clarification: I am not suggesting that one always limit the ISO setting to 400 or 800 or whatever on cameras like the Nikon 7000 or 7100; when the metering is clear cut, using a higher ISO setting to get suitable levels in standard JPEG conversions seems simpler and safe. (But with some cameras, the effects of a setting like ISO 1600 might best be implemented with the same analog gain as ISO 100, plus a flag in the raw file for 16x digital shift.)

I did a small test with my D800E at ISO 6400 compared with ISO 100. For the ISO 100 I lifted the exposure in Lightroom.

Here is a 1:1 crop from the two pictures as shown in Lightroom. I would say a huge difference where the ISO6400 looks rather good and the ISO 100 lifted is really poor. The histogram for the ISO 100 has a lot posterization where the ISO 6400 looks smooth.

[Hans Kruse]

Hans,
I'm simply not aware of the many disadvantages you refer to. The main disadvantage that I've experienced is that one cannot show off to another person the interesting shot one has just taken, because it appears so underexposed on the camera's LCD screen.

As regards reviewing the image for the purpose of retaking the shot if it doesn't look satisfactory, that should not be a concern in the circumstances. If the shot is not satisfactory, the moment will likely have passed by the time one has enlarged the review to inspect the image.

The advantage of shooting in full manual mode at base ISO in circumstance where one knows that underexposure in relation to full-well capacity is unavoidable, is that one has the greatest certainty that  the ISO setting,  and any possible overexposure in terms of blown highlights, will not be an issue.

Have you never taken a shot with a 'minus-1-stop' exposure compensation, that has turned out to have blown highlights? I have. Shoot (in a hurry) some interesting activity that's taking place in the shadows in, say, the lower right corner of the frame, against a background of an interesting, brightly lit sky in the upper left corner, and you might find that a 'minus one stop EC' is not sufficient to bring out the detail in the sky in the top left corner of the frame.

Your approach of using auto ISO with a minus one stop exposure compensation might work just as well most of the time, but would seem to me to provide less certainty for all situations that one might encounter. Perhaps you would then recommend a minus-2-stop EC.

I'll give you an example of what would appear to me to be an unsatisfactory outcome that could result from your method of using auto-ISO in conjunction with exposure compensation. But I'm no expert with this method. There might be settings in the menu of the D7100 I am unfamiliar with.

Let's consider the situation of using a 'minus 2 stop' exposure compensation in auto-ISO mode. In the circumstances of continuously changing lighting conditions that occur when walking along a mountain track, you might inadvertently underexposed at ISO 100. The best choice might have been, for example, ISO 200, which is the setting that I would use with the D7100.  As a result of the 'minus-2 stops' exposure compensation in conjunction with auto-ISO, the camera has lowered ISO to 100 and underexposed a further stop.

In such circumstances, using the D7100, your shot would have more than a full stop worse noise and DR than mine. Is this not the case? If not, please advise what settings you would use on the D7100 to avoid this situation.

In fact, having just now experimented with your recommended method on the D7100 as I write this, I'm rather puzzled that I get no indication of underexposure on the indicator at the foot of the viewfinder. It's either zero or in the overexposed region, yet I can take shots that look underexposed on the LCD review. What's going on here?

There are at least two major disadvantages to the method of shooting at ISO 100 and allow underexposure 1) You can't even see the pictures on the LCD, you can't review focus or show it to anyone 2) you have to adjust exposure when importing into Lightroom or any other RAW converter before you can review the pictures for selection or rejection.

Check the examples of my ISO 6400 test I have just posted. Try it with your D7100 and it will likely be much worse.

I'm not saying you would never get clipped highlights by exposure compensation -1 on matrix metering, but what I say is that it is relatively rare and rare enough as a good general setting.

My experience from shooting landscapes, where I need at adjust the lighting over the image area, is that even with the D800 lifting exposure more than 2-3 stops starts to make colors look less good and noise comes and washes out details. Surely the Canon 5D III which I also shoot is not as good. In both cases I always bracket and make sure that there is always one overexposed and one underexposed in the bracket sequence and then I choose the most exposed without clipping in Lightroom. Lightroom has since version 4 an automatic highlight recovery which recovers from clipping one or two channels and does it in 99% of the time without visual defects. I do miss a true RAW clipping indicator in Lightroom to allow me to see where the clipping occurs. The clipping can be checked in a RAW analyzer like RAW digger which I use in rare cases. This is mostly needed with the Canon but the Nikon also gets better IQ in the difficult shots by this method. I never need to look at histograms, I can just check the blinking indicators for the bracket sequence if I'm in doubt and this is quick to do and does not interrupt the flow of shooting.

Now with casual shooting like what you refer to here I use auto ISO on either manual or aperture priority and on the Canon 5D III I bracket always with the same rule as above. With the Nikon I usually also bracket but could also use the exposure compensation -1 (or -2) to get nearly as good results without the bracketing overhead. Yes, there may still be exceptions in rare cases but usually these are the cases where I can reshoot.

I do follow your wish to have a no worry shooting setup so you can concentrate 100% on capturing the moment The reason for my landscape and casual shooting technique is the same. I almost always come back with good technical quality shots using this method on both the Canon and the Nikon.

[Hans Kruse]

I have a small issue with 'judged in Lightroom'. Lightroom with PV2012 does some highlight recovery and often shows a nice histogram even with clipped highlights. I often use RawDigger to look at 'real' histogram in the raw file.

I agree to some degree on this and I use bracketing always for my landscape work not only to recover from the rare occassion where Lightroom shows no clipping for a clipped RAW file, but also to achieve to highest possible exposure to have the maximum data to work with which gives the minimum noise and the maximum flexibility to adjust. Sometimes my adjustments are huge and the quality of the RAW file is crucial to get the best result. Still HDR merging is in my opinion a plan B solution. I also use RawDigger to check when I suspect there is an issue in e.g. clouds. I wish Adobe would include an optional RAW clipping overlay on the image to see where the RAW file has any clipping to make it easier to see if the is an issue. In my experience it is very rare that there is an issue if Lightroom does not show any clipping on a default conversion and AWB from the camera. Even a wrong WB could cause clipping to occur even if the there is no clipping in the RAW file.... Rather than checking with RawDigger what I usually do is that I synchronize the editing of a picture to one that is exposed one stop less and adjust the exposure in Lightroom +1 and then check 1:1 in areas where I have a doubt if certain structures in e.g. clouds is caused by clipping or not. Sometimes a strongly lit cloud by the sun can show an area with no detail where there is still no detail when using an image with no clipping at all as shown in RawDigger. So all in all I'm happy with the results of the automatic highlight recovery in Lightroom and very seldomly I see artifacts created by this.

[Bart_van_der_Wolf]

So all in all I'm happy with the results of the automatic highlight recovery in Lightroom and very seldomly I see artifacts created by this.

Hi Hans,

But it (PV2012) also causes major highlight compression at Default settings, which tends to take a lot of life out of images with important light tones. This is something many users don't realize because it is not made clear by the default settings, but a significantly reduced Highlights slider control is required to get IMHO more natural looking images.

Cheers,
Bart

[Hans Kruse]

Hi Hans,

But it (PV2012) also causes major highlight compression at Default settings, which tends to take a lot of life out of images with important light tones. This is something many users don't realize because it is not made clear by the default settings, but a significantly reduced Highlights slider control is required to get IMHO more natural looking images.

Cheers,
Bart

Hi Bart,

I agree and I almost always use the highlight slider to bring out detail in the highlights, no matter the exposure. Same with the shadows slider. Both sliders are used in combination with the whites and blacks sliders (of course) and contrast/tone curve.

Update: However if you compare the histogram of an exposure with the compressed highlights taken down by one stop and compare with another exposure without the compression you will see an almost identical histogram in the highlights. So adding to the above I very often take the exposure down in Lightroom even when there is no compression visible of highlights and then use the highlights and shadows sliders as mentioned.

Attached histograms that shows this. The EV+1 is a EV+1 from a bracket sequence. This one is taken down by one stop in Lightroom and the resulting histogram is shown and can be compared with a EV0 histogram.

[BJL]

There are at least two major disadvantages to the method of shooting at ISO 100 and allow underexposure 1) You can't even see the pictures on the LCD, you can't review focus or show it to anyone 2) you have to adjust exposure when importing into Lightroom or any other RAW converter before you can review the pictures for selection or rejection.

I quite agree, and this is part of the reason that I would rather not have to use this technique ... but they do not directly affect Ray's energetic pursuit of maximum IQ, so I suppose that it why he puts up with those inconveniences.

But the bigger part of why I have not felt much need for this practice is that the problems that Ray faces can be mitigated or avoided by different firmware than what Nikon offers, and my EM5 offers what is for me an adequate approximation.

Plan A: an ISO speed selection system that uses one ISO speed value for JPEG conversion (including the on-camera review, in-camera JPEG, and default raw->JPEG conversion) but a fixed analog gain, so that raw files are always at the numerical levels of base ISO speeds, and there is never any amplifier clipping. The difference between the gain used to produce raw and the "JPEG intent" is flagged in raw file metadata.

Plan B: same as above as far as JPEG, and again with metadata flagging, but with the gain used in producing raw files intermediate, say one or two stops less than the "JPEG intent". This could be useful when the sensor/ADC combination is not perfectly "ISO-free", so that analog gain up to about 400 or 800 helps to mitigate effects like quantization noise.

Neither strategy uses the one stop underexposure (-1 compensation) that you mention, let alone the two stop underexposure that Ray worries about in his response; they are solely choices of when and how to amplify the signal after the sensor exposure level is set.

Ironically, these strategies are similar to ones used by some camera makers (Hasselblad, Phase One, Olympus)  that Ray and others have described somewhat disparagingly with words like "over-stating the ISO speed" and "underexposure", partly because of widespread confusion (excacerbated by some confusing description of measurements by DXO) between "less exposure" and "less analog amplification of the results of an exposure".

[Fine_Art]

I did a small test with my D800E at ISO 6400 compared with ISO 100. For the ISO 100 I lifted the exposure in Lightroom.

Here is a 1:1 crop from the two pictures as shown in Lightroom. I would say a huge difference where the ISO6400 looks rather good and the ISO 100 lifted is really poor. The histogram for the ISO 100 has a lot posterization where the ISO 6400 looks smooth.

That is a limitation of lightroom. At least for now. If you use other software that does it's calculation in floating point, the resultant histogram will be smooth. RAW Therapee is now 64 bit floating point for example. So is Images Plus.

[Hans Kruse]

That is a limitation of lightroom. At least for now. If you use other software that does it's calculation in floating point, the resultant histogram will be smooth. RAW Therapee is now 64 bit floating point for example. So is Images Plus.

I thought I would check this, but Raw Therapee will not run on my Mavericks MBP. Anyway this is only of theoretical interest for me as I don't see why I should shoot like this anyway

Update: I checked Capture One 7 and lifting exposure and brightness to the same levels as Lightroom does not create the jagged histogram. I checked some of the other exposures as it appears that +4EV on exposure is the maximum in Lightroom without creating the posterization. Histograms attached.

So it's clear that using Lightroom or ACR lifting exposure more than 4 stops is not good.

[Hans Kruse]

I quite agree, and this is part of the reason that I would rather not have to use this technique ... but they do not directly affect Ray's energetic pursuit of maximum IQ, so I suppose that it why he puts up with those inconveniences.

But the bigger part of why I have not felt much need for this practice is that the problems that Ray faces can be mitigated or avoided by different firmware than what Nikon offers, and my EM5 offers what is for me an adequate approximation.

Plan A: an ISO speed selection system that uses one ISO speed value for JPEG conversion (including the on-camera review, in-camera JPEG, and default raw->JPEG conversion) but a fixed analog gain, so that raw files are always at the numerical levels of base ISO speeds, and there is never any amplifier clipping. The difference between the gain used to produce raw and the "JPEG intent" is flagged in raw file metadata.

Plan B: same as above as far as JPEG, and again with metadata flagging, but with the gain used in producing raw files intermediate, say one or two stops less than the "JPEG intent". This could be useful when the sensor/ADC combination is not perfectly "ISO-free", so that analog gain up to about 400 or 800 helps to mitigate effects like quantization noise.

Neither strategy uses the one stop underexposure (-1 compensation) that you mention, let alone the two stop underexposure that Ray worries about in his response; they are solely choices of when and how to amplify the signal after the sensor exposure level is set.

Ironically, these strategies are similar to ones used by some camera makers (Hasselblad, Phase One, Olympus)  that Ray and others have described somewhat disparagingly with words like "over-stating the ISO speed" and "underexposure", partly because of widespread confusion (excacerbated by some confusing description of measurements by DXO) between "less exposure" and "less analog amplification of the results of an exposure".

Interesting thoughts, but more for a camera engineer than a photographer I really believe Ray is wrong about his pursuit about optimum IQ....