NikonGear'23
Gear Talk => Camera Talk => Topic started by: Valerie S. on October 07, 2016, 06:34:31
-
Before switching to Nikon, I was used to shooting to the right - slightly overexposing (shooting RAW) and adjusting in post. I don't do it with the D500 and was wondering if anyone does this with a D5?
I've looked at many images and don't see a problem with low ISO DR. I just got the D5 today and have shoots tomorrow and Sunday and am trying to hit the ground running.
Any tips or recommendations are really appreciated. I'm still pretty new to Nikon, so everything helps.
Thanks
-
Just use the D5 as the exposure meter indicates. And I didn't have any problems with low ISO dynamic range, at least in the environnement I'm shooting (sport, action). I just hit the 10'000 clicks mark. 😊
-
The idea that you need to ETTR was invented as a selling-point for mirrorless cameras.
The basic reason to ETTR - to place your available dynamic range and as many digitised levels as possible, and minimise noise, in the important parts of the image - applies to the D5 the same as to every other digital camera. However, the better your camera's noise performance the less likely noise is to be disturbing, the higher the camera's dynamic range the less often you need to squeeze out every last stop and the higher the bit depth the less risk there is of running out of levels before you reach the shadows. (One problem ETTR is often alleged to solve but does not is low ISO noise in monochrome areas - typically blue sky).
The downside of ETTR (apart from increased post-processing load) is increased lens flare. Every stop you ETTR doubles the flare, and that applies to Nikon lenses the same as to all others. If under the conditions you are shooting in your lenses have negligible flare they will still have negligible flare if you ETTR, but if flare is a problem ETTR will make it worse. Whether your images will gain more from better shadow detail than they will lose from worse flare only you can say.
-
I dont know the D5 sensor, but I would stay away from overexposure. For Nikon, overexposure is for very brightly lit situation or low contrast scenes. The D3s and D4s didn't like recovered highlights. It is almost always better to lift shadows when using Nikon as long as you dont use high shutter speed or high ISO (Ballpark figure, <1/1000 sec or >12800 ISO). This is my experience for what its worth.
-
Underexposure is better. Usually 1/2 stop. In post-prod, if needed, it can easily be corrected. And with the usage of a lot of manual Non-Nikkor lenses, it is the way to get the correct exposure, as usual they transmit more light than the lens-description. Apart from that you gain speed, or lower ISO-value's.
-
The relationship between flare and normal image light is not dependent on the exposure as long as you don't clip any part of the image.
-
The idea that you need to ETTR was invented as a selling-point for mirrorless cameras.
The basic reason to ETTR - to place your available dynamic range and as many digitised levels as possible, and minimise noise, in the important parts of the image - applies to the D5 the same as to every other digital camera. However, the better your camera's noise performance the less likely noise is to be disturbing, the higher the camera's dynamic range the less often you need to squeeze out every last stop and the higher the bit depth the less risk there is of running out of levels before you reach the shadows. (One problem ETTR is often alleged to solve but does not is low ISO noise in monochrome areas - typically blue sky).
The downside of ETTR (apart from increased post-processing load) is increased lens flare. Every stop you ETTR doubles the flare, and that applies to Nikon lenses the same as to all others. If under the conditions you are shooting in your lenses have negligible flare they will still have negligible flare if you ETTR, but if flare is a problem ETTR will make it worse. Whether your images will gain more from better shadow detail than they will lose from worse flare only you can say.
ETTR was a mantra well before the first mirrorless cameras came about.
If I understand correctly ETTR is much less important with modern "ISO invariant" cameras.
-
No camera that I'm aware of is truly ISO invariant. In terms of tonal quality and colour, the best quality is obtained when one gives the sensor as much light as possible. After the aperture and shutter speed has been determined, ISO should then be set at a value which is as high as possible without causing significant highlight clipping. On a true ISO invariant sensor, you can always use base ISO and get the same result as you would by any higher ISO setting, but real cameras are not really like that. The D5 is about as far from an ISO invariant camera as they come, so it is better to raise the ISO than let the image be too dark at the time of exposure. Increasing ISO instead of lifting the image in post-processing reduces the impact of read noise on the image quality.
If you are not limited in the amount of light you can use, e.g. if your camera is on tripod and your subject is not moving, then obviously by all means use base ISO and a camera with a large dynamic range at base ISO (D810). The D5 isn't really the camera for such situations, it excels in action, AF, and low light.
-
The relationship between flare and normal image light is not dependent on the exposure as long as you don't clip any part of the image.
Flare is light that enters the lens but is lost by reflection at air-glass interfaces. The lost light bounces around the inside of the lens: some goes back out the way it came, some is absorbed by the internal surface of the lens, and some hits the film/sensor, lowering contrast, dynamic range and colour saturation (flare light is more-or-less white). For any given lens, the proportion of the entering light that is lost is constant, so if you double the amount of light that enters - ie, increase the exposure one stop - you double the amount lost.
Sure, the ratio of flare to image-forming light is also constant if you have no photosites at full well capacity. But flare is the same over the whole sensor and image light is bright in some areas and dark in others. If there are parts of the sensor receiving no image light, increasing exposure increases flare but not image light, and that reduces dynamic range. Because the darkest possible pixels are less dark because of flare, and the brightest possible pixels are no brighter, the scene dynamic range you can record is reduced. If you happen not to have a scene with less dynamic range than your sensor, that is good, but you often won't, and in any case the whole point of ETTR is that photographically unimportant highlights should be clipped to optimise mid-tone and shadow exposure.
-
One problem ETTR is often alleged to solve but does not is low ISO noise in monochrome areas - typically blue sky
Photon shot noise is one area where ETTR will help in a very predictable fashion. Could you elaborate why you think that this is not the case?
-
ETTR was a mantra well before the first mirrorless cameras came about.
If I understand correctly ETTR is much less important with modern "ISO invariant" cameras.
No, the "ISO invariant" does nothing at all to change the importance of ETTR. It only matters in the question of whether to set ISO in-camera or in post.
-
Flare is light that enters the lens but is lost by reflection at air-glass interfaces. The lost light bounces around the inside of the lens: some goes back out the way it came, some is absorbed by the internal surface of the lens, and some hits the film/sensor, lowering contrast, dynamic range and colour saturation (flare light is more-or-less white). For any given lens, the proportion of the entering light that is lost is constant, so if you double the amount of light that enters - ie, increase the exposure one stop - you double the amount lost.
Sure, the ratio of flare to image-forming light is also constant if you have no photosites at full well capacity. But flare is the same over the whole sensor and image light is bright in some areas and dark in others. If there are parts of the sensor receiving no image light, increasing exposure increases flare but not image light, and that reduces dynamic range. Because the darkest possible pixels are less dark because of flare, and the brightest possible pixels are no brighter, the scene dynamic range you can record is reduced. If you happen not to have a scene with less dynamic range than your sensor, that is good, but you often won't, and in any case the whole point of ETTR is that photographically unimportant highlights should be clipped to optimise mid-tone and shadow exposure.
This explanation does not quite make sense to me.
If you first set your exposure such that your flare is 7 stops below full well, and your brightest image pixel is 1 stop below full well, you may use ETTR to expose one more stop such that your brightest pixel is just at full well. Your flare is now 6 stops below full well but you have changed nothing about the ratio between image and flare. You have increased the signal to noise ratio in the midtones, let's say 3-5 stops below full well and that is good. You can now reset your black point in post to 6 stops below full well and you don't see the flare anymore.
If your exposure already clips some highlights, you normally do not apply ETTR unless you think that further clipping those highlights is not detrimental to the picture. If that is the case, let's say you have one more stop of highlights that you can clip (let's say, specular highlights). Your flare is at -7 stops as before. After increasing exposure, your flare is at -6 stops and your brightest image-relevant pixel is just at full well. Again, you have maximized signal to noise ratio and you may reset your black point in post.
I fail to imagine a situation where you would not want to use ETTR.
-
The idea that you need to ETTR was invented as a selling-point for mirrorless cameras.
The basic reason to ETTR - to place your available dynamic range and as many digitised levels as possible, and minimise noise, in the important parts of the image - applies to the D5 the same as to every other digital camera. However, the better your camera's noise performance the less likely noise is to be disturbing, the higher the camera's dynamic range the less often you need to squeeze out every last stop and the higher the bit depth the less risk there is of running out of levels before you reach the shadows. (One problem ETTR is often alleged to solve but does not is low ISO noise in monochrome areas - typically blue sky).
The downside of ETTR (apart from increased post-processing load) is increased lens flare. Every stop you ETTR doubles the flare, and that applies to Nikon lenses the same as to all others. If under the conditions you are shooting in your lenses have negligible flare they will still have negligible flare if you ETTR, but if flare is a problem ETTR will make it worse. Whether your images will gain more from better shadow detail than they will lose from worse flare only you can say.
My first digital ILC was the Nikon D70, and with that camera, utilizing a CCD sensor, I used, or tried to use ETTR, as often as possible. With that sensor ETTR was helpful in generating less noisy exposures. The CCD sensor provided no liveview capability, which negated the concept of "mirrorless"....
The D300 was my first liveview capable camera, but unfortunately, the mirror moved up and down for each shutter cycle thus negating much of the utility of liveview.
The S5 UV/IR forensic camera from Fujifilm as one of very few CCD based cameras gave approx. 30 seconds of liveview with a CCD sensor, since heating and noise were much worse issues in the "good old days".
Nikon DSLRs employ a colour RGB matrix light meter, which is very precise. Light metering and WB always worked very well with my Nikon cameras. A side note is that both light metering and WB are improved, if you use a lens with a CPU, rather than a non-CPU Ai-S lenses. This can be seen, if you install a CPU (from Bjørn Rørslett) in a manual Nikkor. Knowledge of not only focal length and aperture, but also exit pupil assists the light metering function of the camera. The WB will affect the histogram, so correct WB is necessary for a correct interpretation of the histogram.
-
This explanation does not quite make sense to me.
If you first set your exposure such that your flare is 7 stops below full well, and your brightest image pixel is 1 stop below full well, you may use ETTR to expose one more stop such that your brightest pixel is just at full well. Your flare is now 6 stops below full well but you have changed nothing about the ratio between image and flare. You have increased the signal to noise ratio in the midtones, let's say 3-5 stops below full well and that is good. You can now reset your black point in post to 6 stops below full well and you don't see the flare anymore.
If your exposure already clips some highlights, you normally do not apply ETTR unless you think that further clipping those highlights is not detrimental to the picture. If that is the case, let's say you have one more stop of highlights that you can clip (let's say, specular highlights). Your flare is at -7 stops as before. After increasing exposure, your flare is at -6 stops and your brightest image-relevant pixel is just at full well. Again, you have maximized signal to noise ratio and you may reset your black point in post.
I fail to imagine a situation where you would not want to use ETTR.
If you set your exposure so flare is seven stops below FWC, and then say, "I will ETTR" and increase flare to six stops below FWC the darkest pixel in the image, that in a flare free image was black, is one stop brighter. The maximum dynamic range has, therefore, been reduced by one stop. Sure, if the brightest pixel was one stop below FWC and is now just at FWC the dynamic range of your image is not affected. Photography has always been easy when the scene dynamic range is less than the recording medium. But if the scene dynamic range is the same as or greater than the dynamic range of the sensor - which is hardly uncommon - you have lost a stop of dynamic range because of the increased flare. You can make the flare-affected pixels black again, but you can't recover the lost information.
Flare is a form of noise, but like dark noise rather than photon noise, which is how you are treating it: it is added to every pixel, and its effect is greater in darker pixels. If you increase exposure one stop you double the amount of light. But the image light is not evenly distributed across the sensor. Some parts of the image are bright and some are black: some parts of the sensor get a lot of light and some get none. But the flare light is evenly distributed across the sensor. So although the ratio of flare light to image light across the sensor as a whole is not affected by increasing exposure, the ratio is greater in the darkest parts of the image and increases further when you increase exposure.
Image post-processing is not free - obviously, in terms of time, but also in terms of information loss. The situation you would not want to use ETTR is any situation you don't need to.
-
One reason not to use ETTR is if you want the file to be usable with minimal or no adjustments in post.
-
Photon shot noise is one area where ETTR will help in a very predictable fashion. Could you elaborate why you think that this is not the case?
Low ISO noise in monochrome areas - say, a blue sky - is not photon noise. It arises from other channels - in the blue sky case, mostly the red channel - which are under-exposed and noisy. Increasing the exposure has minimal effect because there is still no red in the sky so the red channel is still under-exposed and noisy.
-
Thanks for all of the replies and information. I'll shoot as is unless the situation calls for it.
-
One reason not to use ETTR is if you want the file to be usable with minimal or no adjustments in post.
Yes, I agree but this has nothing to do with flare.
-
Low ISO noise in monochrome areas - say, a blue sky - is not photon noise. It arises from other channels - in the blue sky case, mostly the red channel - which are under-exposed and noisy. Increasing the exposure has minimal effect because there is still no red in the sky so the red channel is still under-exposed and noisy.
So if the red channel noise is NOT photon noise how would you describe it ?
-
If you set your exposure so flare is seven stops below FWC, and then say, "I will ETTR" and increase flare to six stops below FWC the darkest pixel in the image, that in a flare free image was black, is one stop brighter. The maximum dynamic range has, therefore, been reduced by one stop.
Yes, it is clear that increasing exposure has the potential of clipping any highlights that are already near full well. But that does not change in the presence of flare.
So, I still don't see how flare can affect the decision when to use ETTR or not.
If you are not using ETTR in the scenario you describe, it is because you do not want to clip highlights and not because there is any flare.
ETTR is really meant to push the data that the photographer finds relevant to the picture as far to the right as possible. If you do not want to clip more than what is already clipped, there is no point in further increasing exposure. Flare or not, I do not see how that matters.
-
Not only can ETTR push data out the top it squishes data together in the shoulder of the image. Instead try shooting with the Neutral or Flat PC. Clip only specular highlights. Put punch back into the image using an "S" curve using the Master Lightness feature of LHC.
Dave
If not using Nikon software use a preset that gives you an NEF conversion without putting and "S" curve into the conversion. Do that under your control in LCH (Lightness, Chroma, Hue).
-
Friday night physicists chat room!
-
Not only can ETTR push data out the top it squishes data together in the shoulder of the image. Instead try shooting with the Neutral or Flat PC. Clip only specular highlights. Put punch back into the image using an "S" curve using the Master Lightness feature of LHC.
Dave
If not using Nikon software use a preset that gives you an NEF conversion without putting and "S" curve into the conversion. Do that under your control in LCH (Lightness, Chroma, Hue).
I try to think of the raw conversion as something that is separate from the collection of data. When using ETTR in the field, I usually also make an image with the "right exposure" i.e. that would yield the desired brightness with standard raw conversion settings. Then I can import both the ETTR and the regularly exposed image into, say, ACR and adjust the ETTR image to match the brightness of the regularly exposed one.* I will use the exposure slider to do so, and I have found the scale on the exposure slider to match the actual exposure pretty closely, i.e. if the ETTR was two stops above the regularly exposed image, I can set the exposure slider at -2 stops in the ETTR image and have a pretty close match in overall brightness, but also the histograms are pretty similar. The ETTR image will have an extra two stops better shadow details. One can also do the same with curves, but I find the curve interface to be quite difficult to master and it will take me more time to find a suitable curve. The difficulty arises because of the nonlinear transformation from linear raw levels to RGB levels. Applying the transformations on raw levels would make this task much easier. I am not sure what exactly the exposure slider in ACR does, but it is not far from a linear scaling on the raw levels from what I can infer.
* The added benefit of taking a normal exposure is that if I botched and blew some important highlights in using ETTR (which is quite likely given the fact that the camera does not show you what exactly has been clipped), I can always recover them from the normal exposure or just use the normal exposure and suck up the inferior shadows.
-
So if the red channel noise is NOT photon noise how would you describe it ?
Sorry for the delay - I have been away.
Yes, it starts out as ordinary photon noise, but the question is why it is obvious in the final image. That is because (1) the red channel is not silent, as ideally it would be when blue light falls on the sensor, because the spectral responses of the channels are not perfectly separated, and (2) noise is imported from the red channel during raw processing. The effect of the second point is that channel noise is different for different raw converters. http://www.libraw.org/articles/channel-noise-and-raw-converters.html has examples.
-
Did ETTR ever really work or were people fooling themselves with the standard curves applied by the camera maker. For example the it may look like one has blown the highlights with a Nikon when using the Standard Picture Control but one may well find there is more useful highlight information when using the Neutral or Flat PC. Some negative exposure compensation may help. Then one puts a bit of punch back into the mid tones with an "S" curve in LCH's master brightness.
I've always considered ETTR to be dicey at best and frequently a detriment to the photo to gain a bit lower noise in the shadows. With my D2H I could loose the shadows and blow the highlights and all at the same time. The D300s was a huge relief for me.
With negative film one would expose for the shadows (Zone I) and develop from the highlights. With roll films one would develop the roll at N-1. With digital one should expose for the highlights and develop RAW files for the shadows.
Maybe I'm missing something. Maybe it was a Canon thing. I don't know. I don't see ETTR as useful with either my D300s or D800.
Dave
-
Did ETTR ever really work or were people fooling themselves with the standard curves applied by the camera maker.
....
Maybe I'm missing something. Maybe it was a Canon thing. I don't know. I don't see ETTR as useful with either my D300s or D800.
Dave
David, I think you are on to something by bringing in the historical angle. Digital raw imaging systems used to be much more limited in tonal range than they are now. I recently fooled around processing (in ACR/LR) a Canon raw shot from an APSC Canon Digital Rebel camera, circa 2005. The tonal capture range in the raw file was extremely narrow by today's standards. No matter how much you played with the raw file, it still looked like a jpeg from a budget point and shoot JPEG camera. Coincidentally, the photographer who, in my early days of digital(2006), preached ETTR was a Canon user.
-
Did ETTR ever really work or were people fooling themselves with the standard curves applied by the camera maker. For example the it may look like one has blown the highlights with a Nikon when using the Standard Picture Control but one may well find there is more useful highlight information when using the Neutral or Flat PC. Some negative exposure compensation may help. Then one puts a bit of punch back into the mid tones with an "S" curve in LCH's master brightness.
When you can recover anything, you haven't fully blown it. You might have blown one channel which allows context-sensitive highlight recovery by guessing what the color was and filling in the missing information. You think you have blown something because the clipping warnings are based on the color space you use to convert your file. You will observe than upon changing color space, the clipping areas change. Your clipping also depends on white balance, to confuse matters further.
If you want to experiment with ETTR, the first thing IMO is to get a program that shows you raw levels and histograms so you can see how far you can go until genuinely blowing channels. You will find that in many normal exposures of low dynamic range scenes your top 2 or more stops are really left unused by the camera (if going by the exposure meter). Of course this is the correct way of exposing if developing the file with standard settings should produce the correct brightness, but it is certainly a very sub-optimal way of collecting data about the scene. By pushing the data to the right (not by increasing ISO of course, but by having more exposure!), you can definitely get a much cleaner picture, particularly if your processing of the image involves expanding the data to get high contrast.
This is not something that applies only to Canon and not to Nikon. Sure, sensors are different and if you can get away without using ETTR on your camera -- good for you! But the fundamental principles do not depend on the brand of sensor.
-
If you want to experiment with ETTR, the first thing IMO is to get a program that shows you raw levels and histograms so you can see how far you can go until genuinely blowing channels.
OK, I'll bite: what program for Nikon's NEF(s)?
Dave
-
I believe RAWDigger is what you should be looking for. Not freeware, though. Andrea B. uses this program a lot, you might PM her about it.
-
Thanks!
-
Yes, RawDigger is what I use.
-
Sorry for the delay - I have been away.
Yes, it starts out as ordinary photon noise, but the question is why it is obvious in the final image. That is because (1) the red channel is not silent, as ideally it would be when blue light falls on the sensor, because the spectral responses of the channels are not perfectly separated, and (2) noise is imported from the red channel during raw processing. The effect of the second point is that channel noise is different for different raw converters. http://www.libraw.org/articles/channel-noise-and-raw-converters.html has examples.
Thanks for replying.
Have you looked at the RGB composition of the "blue colour" of a "blue-sky" ? I just looked at one of my pictures and found, for the "blue sky", Blue=200, Green=100, Red=50. The numbers are rounded a bit and obviously one expects some variation depending upon "conditions". However the approximate 4:2:1 ratio is seen in a number of reports when searching for "blue sky spectrum" ( https://www.google.co.uk/search?q=blue+sky+spectrum&client=firefox-b&source=lnms&tbm=isch&sa=X&ved=0ahUKEwjm8Jqqgt3PAhVfFMAKHarcDvYQ_AUICCgB&biw=1920&bih=1091#imgrc=uwXQaJSamV9nqM%3A ).
So the "blue sky" contains 4 times as much blue light as red light. Looking at photon (shot) noise alone, this means that the signal-to noise (SNR) ratio of the blue channel is twice that of the red channel. We describe the red channel as being more "noisy" - because what is noticeable is the SNR and not the absolute amount of noise.
I do not say that demosaicing does not contribute to noise, but I'm not sure that it's the most important factor here.
Are you suggesting that perfect spectral separation between channels is desirable ? How would a sensor which achieved this recognize yellow ?
-
Thanks for replying.
Have you looked at the RGB composition of the "blue colour" of a "blue-sky" ? I just looked at one of my pictures and found, for the "blue sky", Blue=200, Green=100, Red=50. The numbers are rounded a bit and obviously one expects some variation depending upon "conditions". However the approximate 4:2:1 ratio is seen in a number of reports when searching for "blue sky spectrum" ( https://www.google.co.uk/search?q=blue+sky+spectrum&client=firefox-b&source=lnms&tbm=isch&sa=X&ved=0ahUKEwjm8Jqqgt3PAhVfFMAKHarcDvYQ_AUICCgB&biw=1920&bih=1091#imgrc=uwXQaJSamV9nqM%3A ).
So the "blue sky" contains 4 times as much blue light as red light. Looking at photon (shot) noise alone, this means that the signal-to noise (SNR) ratio of the blue channel is twice that of the red channel. We describe the red channel as being more "noisy" - because what is noticeable is the SNR and not the absolute amount of noise.
I do not say that demosaicing does not contribute to noise, but I'm not sure that it's the most important factor here.
Are you suggesting that perfect spectral separation between channels is desirable ? How would a sensor which achieved this recognize yellow ?
You don't see low-ISO monochromatic noise in B&W sensors. If the issue is noise in colour images, I am not sure it makes sense to separate out the effect of demosaicing, because without it there is no colour image. To get a three-colour image from a Bayer mosaic you have to work out the green and red channel values for the blue pixels by interpolating from neighbouring pixels, and then the noise in the red channel "infects" the blue channel.
Perfect spectral separation of channels would not affect the underlying Bayer principle.