Pages: 1 2 3 [4] 5 6 ... 10   Go Down

Author Topic: Nikon D7000 Dynamic Range  (Read 126731 times)

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
Re: Nikon D7000 Dynamic Range
« Reply #60 on: November 19, 2010, 11:38:55 pm »

Correct -- for example, on many modern SLRs if you spot meter an area and take the picture at the recommended exposure (i.e., with the meter needle in the middle), that area will be about 3 to 3.5 stops below sensor saturation. In other words, you could increase the exposure time by ~3 stops before you start clipping.

I tried this technique with my 5D a few years ago. I would have the camera in 'spot meter' mode and search the scene I was about to photograph for the brightest patch, often a white cloud, but sometimes simply a white napkin in a restaurant. I would take an exposure reading of that brightest patch, then increase exposure by 3 stops. It seemed to work without fail, giving me a good ETTR exposure.

Trouble is, a lot of my photography is capturing the moment, which I find difficult enough as it is. Introducung additional concerns such as searching for the brightest patch in a scene, then calculating a 3 stop exposure increase from the spot meter reading of that brightest patch, caused me to miss the moment too often.  :D
Logged

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: Nikon D7000 Dynamic Range
« Reply #61 on: November 20, 2010, 09:48:03 am »

Correct -- for example, on many modern SLRs if you spot meter an area and take the picture at the recommended exposure (i.e., with the meter needle in the middle), that area will be about 3 to 3.5 stops below sensor saturation. In other words, you could increase the exposure time by ~3 stops before you start clipping.

Eric,

That is useful information and is consistent with what was reported by the late Bruce Fraser, but one should probably do tests for his own camera. The results would vary with the calibration of the camera's light meter and the ISO rating of the sensor. As far as I know, most camera light meters use the ISO 2721 standard, which gives similar results to ISO 2720 for hand held meters. Otherwise, one would get different results for hand held and built in metering (disregarding lens transmission and lens extension).

However, there is considerable variation in the ISO ratings for sensors. The ISO saturation rating (which is used by DXO) would give 12.7% saturation when exposure is made according to the metered reading, which corresponds to a pixel value of 99.8 in an 8 bit gamma 2.2 space. One can test his own camera by exposing a uniformly illuminated target such as a white wall or gray card and determine sensor saturation. The best way to do this would be to use a program that looks directly at the raw data such as DCRaw, Iris, or Rawnalize.

I did this for my Nikon D3 and got 12.1% saturation, which is 3 stops below saturation. This allows 0.5 EV of highlight headroom and is very close to the ISO saturation standard.  Not everyone uses one of these programs and it is more convenient to use Adobe Camera Raw (for those who use this program). However, there are complications. ACR uses a BaselineExposure offset to compensate for highlight headroom allowed by the camera. For the D3, this is +0.5 EV. In addition, the default tone curve boosts the mid-tones. Using ACR defaults, the AdobeRGB value for this exposure is 146, which is much higher than expected. Using -0.5 EV of exposure and a linear tone curve (sliders on the main ACR tab all set to zero and the point curve set to linear), gives an aRGB value of 97, which is very close to the calculated value of 99.8. Using the default exposure value and a linear tone curve gives a pixel value of 116, which is very close to mid-gray (117). Is this process what you would expect and is it valid for other cameras?

One can determine the ISO speed rating for other sensors from the DXO data. Nikons seem to have ratings similar to the ISO saturation standard, but slightly conservative. The Canon 1DsM3 allows more highlight headroom and the Phase One P65+ gives an entire f/stop headroom.

Regards,

Bill
Logged

madmanchan

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 2115
    • Web
Re: Nikon D7000 Dynamic Range
« Reply #62 on: November 22, 2010, 11:27:31 am »

Bill, good points, but keep in mind that metering and sensor saturation in the raw data are largely independent of the ISO standards. To put it directly, ISO 12232 (which concerns itself with the determination of exposure index, ISO speed ratings, standard output sensitivity, and recommended exposure index) says nothing about raw image data, and does not even contain the word "raw" in its well-written 24 pages.

Ray, true, one does not always have the luxury "in the field" to obtain optimal exposure. But if you are concerned about dynamic range and extracting the max SNR your sensor has to offer (which is what I gather from this thread topic), then that is precisely what needs to be done.

In reality, with today's cameras, there is a tradeoff between "getting the shot" and "getting optimal exposure" -- which is a shame because the technology exists to do both simultaneously.
Logged
Eric Chan

bjanes

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 3387
Re: Nikon D7000 Dynamic Range
« Reply #63 on: November 22, 2010, 10:01:00 pm »

Bill, good points, but keep in mind that metering and sensor saturation in the raw data are largely independent of the ISO standards. To put it directly, ISO 12232 (which concerns itself with the determination of exposure index, ISO speed ratings, standard output sensitivity, and recommended exposure index) says nothing about raw image data, and does not even contain the word "raw" in its well-written 24 pages.

Eric,

You know quite a bit more about these matters than I do, and I don't have the 24 page document, but I understand that the ISO saturation standard, at least the 1996 version, did talk about sensor saturation. See the DXO explanation of how they determine sensor sensitivity. They claim to be using the ISO Standard 12232.

Regards,

Bill



Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
Re: Nikon D7000 Dynamic Range
« Reply #64 on: November 23, 2010, 05:25:29 pm »

I decided to check this new performance of the 20D at high ISO for myself, taking two shots of the same high-SBR scene at equal exposure, the exposures being just right for an ETTR at ISO 1600. Of course the exposure when used at ISO 100 became a 4-stop underexposure.

I raised the shadows of the ISO 100 shot in ACR, converting both images without sharpening or noise reduction, and to my astonishment the ISO 1600 shot was so much better across the entire tonal range. Whilst the greatest improvement was observable in the shadows, there was a lesser, but still noticeable improvement in the midtones and upper mid-tones, in the ISO 1600 shot.

But here's the rub. Not with the D7000. There's no image quality advantage in using a higher ISO as an alternative to underexposing at base ISO.
Ray, what you seem to have shown is that the DR of the D7000 allows one to survive 4 stops of underexposure while the D20 did not. How important is this? Do you make that sort of exposure mistake often,or ever? It certainly does not refute my suggestion that once read noise levels are as low as in the D7000, read noise is irrelevant, because all you are comparing to is another, older, inferior camera whose total read noise at low ISO (I actually mean all noise other than shot noise) is far higher than in the D7000, and high enough to be problematic ... but maybe even then only problematic if one underexposes by four stops and then raises the levels by four stops, bringing severely underexposed shadow regions up to mid-tone levels.

And I have a hunch as to what is going on with the 20D,and even in more recent Canon DSLRs: at ISO 100, the noise floor is probably dominated by sources arising after ISO gain is applied (which Canon seems to do on the sensor): low ISO noise is largely produced in the subsequent signal transportation and/or in A/D conversion. What the D7000 has perhaps achieved is what MF backs have more or less been doing for some time in a different way: making variable analogue gain (ISO adjustment before ADC) rather irrelevant, by producing digital output that records the full SNR of the signal from the photosites. Exposure level can thus be chosen after the fact as a digital adjustment on RAW files. This has been easier for MF backs, because ISO gain cannot be applied anyway until the signal has been moved off the sensor, by which time almost all read noise is already present, and so all that is needed is a good enough, high bit-count, low-noise ADC.
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
Re: Nikon D7000 Dynamic Range
« Reply #65 on: November 23, 2010, 09:40:54 pm »

Ray, what you seem to have shown is that the DR of the D7000 allows one to survive 4 stops of underexposure while the D20 did not. How important is this? Do you make that sort of exposure mistake often,or ever? It certainly does not refute my suggestion that once read noise levels are as low as in the D7000, read noise is irrelevant, because all you are comparing to is another, older, inferior camera whose total read noise at low ISO (I actually mean all noise other than shot noise) is far higher than in the D7000, and high enough to be problematic ... but maybe even then only problematic if one underexposes by four stops and then raises the levels by four stops, bringing severely underexposed shadow regions up to mid-tone levels.

BJL,
I think you might have missed the significance of the point I was trying to make. Perhaps my comparison was a bit convoluted. I'll try again.

The underexposure of 4 stops at base ISO, whether in the 20D or the D7000, reduces the effective dynamic range of those cameras by 4 EV in those instances. Would you agree?

If we were to choose another scene which included the same tonal range and shadows as in the 4-stop underexposure at base ISO, but in addition included a brighter section (such as a sky with bright clouds) which extended the brightness range of the entire scene by another 4 stops, then we would have an ETTR exposure with both cameras which would likely demonstrate the DR capability of those cameras.

Of course, such an exposure with the 20D at ISO 1600 would then be overexposed by 4 stops, but the quality of those parts of the image that are not overexposed would remain the same as in the previous scenario, ie. better than the same range of tonal values at ISO 100, in the 20D.

However, since we are trying to be precise here, we should try to exclude the fact that the D7000 has double the pixel count of the 20D. Let's avoid as far as possible, distracting and extraneous issues.

DXOMark allows us to compare cameras at the pixel level. To do this in practice with 'real-world' images, in order to verify the relevance and accuracy of the DXO results, we should treat the D7000 as though it is a 'cropped format of a cropped format', with an additional crop factor of 1.4x.

In other words, we should use a focal length of lens that is appropriately wider (than the lens on the 20D), by a factor of 1.4, and an aperture that is appropriately wider by a factor of 1.4, when using the D7000.

In order not to put the D7000 at a disadvantage regarding lens performance, we should use F stops like F8 with the D7000 and F11 with the 20D (or F11 with the D7000 and F16 with the 20D), because the D7000 with its greater pixel density needs a higher resolution lens for the same perceived sharpness. Let's not introduce further extraneous factors of differing resolution, since it is only DR quality we are concerned about here.

Okay! So we've sorted out the practicalities of a fair, objective and meaningful, 'real world' comparison.

I predict the outcome would be:  Image quality in the region of 4 stops below full-well saturation (from both cameras at base ISO), would be better in the D7000 image, than in the 20D image, in accordance with DXOMark test results.

As I've mentioned before, when using all 16mp of the D7000, there's no contest between the D7000 and the 20D. This is not because the 20D is old technology, but because the D7000 is better technology. The 20D pixel is basically as good as the 1Ds3 pixel. The 1Ds3 sensor could be considered as 2.6 20D sensors stitched together and reorganised into a 3:2 aspect ratio. This fact is quite illuminating.

There are certain areas of performance where the 1Ds3 pixel is very marginally better than the 20D pixel, and other areas where the 20D pixel is marginally better than the 1Ds3 pixel, but such differences are of no consequence in practice, at the pixel level.

The main point is that the 1Ds3 pixel is the same size as the 20D pixel and its qualities have not seemed to have improved since the 20D was released. The 20D camera may be old technology, but its pixel is still the equal of any other in the Canon repertoire, with the possible exception the most recent 1DMk4 pixel which appears to be marginally improved in some respects.

Now, for all those who find this very convoluted, I'm afraid I have no special de-convolution algorithm available which you can download to help you clarify the issue. You'll have to search within for the appropriate algorithm.  ;D


Quote
And I have a hunch as to what is going on with the 20D,and even in more recent Canon DSLRs: at ISO 100, the noise floor is probably dominated by sources arising after ISO gain is applied (which Canon seems to do on the sensor): low ISO noise is largely produced in the subsequent signal transportation and/or in A/D conversion. What the D7000 has perhaps achieved is what MF backs have more or less been doing for some time in a different way: making variable analogue gain (ISO adjustment before ADC) rather irrelevant, by producing digital output that records the full SNR of the signal from the photosites. Exposure level can thus be chosen after the fact as a digital adjustment on RAW files. This has been easier for MF backs, because ISO gain cannot be applied anyway until the signal has been moved off the sensor, by which time almost all read noise is already present, and so all that is needed is a good enough, high bit-count, low-noise ADC.

I'm speculating something similar. Put simply, Canon amplify the low-level (underexposed) analog signal at high ISO in order to improve SNR and the general image quality.

Nikon, with the D7000, amplifies the analog signal at base ISO to achieve an extended DR and better image quality in the shadows, which is carried through to higher ISOs when underexposure at base ISO takes place and exposure compensation is subsequently applied in software.

Logged

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
Re: Nikon D7000 Dynamic Range
« Reply #66 on: November 23, 2010, 10:13:02 pm »

Ray, we are going in circles. My point is that to see the effects of the 14th stop or so of dynamic range, you have to lighten underexposed shadow regions by about for stops. Your example is based on having to use exposure four stops less than usual in order to avoid blown highlights, so that instead of a normal midtone placement about three stops below maximum, the midtones about seven stops below maximum! That means about 1/128 of full well, so about 200e- in the D7000. A midtone exposure that low will cause distinct noise problems in the important midtone areas, mostly due to photon shot noise, so reducing the badness of the greatly lightened deep shadows will probably be the least of your problems for noise effects on overall IQ.  To repeat my point from a previous post, good midtone quality requires getting more exposure there  of about 500 to 1000e- or more, and the only solution to doing that while avoiding blowing of the extreme highlights in your hypothetical example, 7 stops brighter than midtones, is increasing DR through increasing well capacity for highlights, not reducing read noise in the deep shadows.

But perhaps you have real examples where this extreme lightening is needed even with optimal "ETTR" exposure, and is done while also achieving good midtone IQ, rather than severe underexposure for no good reason? "Shadow peeping" examples that merely crop to deep shadow regions and dramatically lighten them while ignoring midtones are not of practical interest.
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
Re: Nikon D7000 Dynamic Range
« Reply #67 on: November 23, 2010, 11:19:00 pm »

Ray, we are going in circles. My point is that to see the effects of the 14th stop or so of dynamic range, you have to lighten underexposed shadow regions by about for stops. Your example is based on having to use exposure four stops less than usual in order to avoid blown highlights, so that instead of a normal midtone placement about three stops below maximum, the midtones about seven stops below maximum! That means about 1/128 of full well, so about 200e- in the D7000. A midtone exposure that low will cause distinct noise problems in the important midtone areas, mostly due to photon shot noise, so reducing the badness of the greatly lightened deep shadows will probably be the least of your problems for noise effects on overall IQ.  To repeat my point from a previous post, good midtone quality requires getting more exposure there  of about 500 to 1000e- or more, and the only solution to doing that while avoiding blowing of the extreme highlights in your hypothetical example, 7 stops brighter than midtones, is increasing DR through increasing well capacity for highlights, not reducing read noise in the deep shadows.

BJL,
You might be going around in circles, but I'm not. My point is as straight and as linear as the exposure of a sensor.

Four stops underexposure at base ISO on the D7000 results in a maximum (engineering) DR of 9.35 EV, at the pixel level, according to DXOMark.

Four stops underexposure on the 20D at base ISO results in a maximum engineering DR of just 6.95 EV.

Those 6.95 EV of DR have better image quality on the D7000, period. I'm not even considering the extra 3 stops of DR that the D7000 is capable of. Those extra 3 stops might well be crap, of no artistic value, I agree. I'm concerned with the image quality of the 9-11 stops of DR which take the 20D to the engineering limit (and the 1Ds3 also to the engineering limit at the pixel level).

Quote
But perhaps you have real examples where this extreme lightening is needed even with optimal "ETTR" exposure..

Of course I do. I'm a practical man. Did you miss my examples of the St Isaac's Cathedral in St Petersburg, on page 2 of this thread?

The sun was near setting at about 10.30 pm, and almost visible behind the clouds. Avoiding blown highlights in the clouds surrounding the sun was impractical. If I'd reduced exposure even more to achieve this, the midtones would have been even noisier.

These are not deep shadows that are noisy, but the whole foreground.


Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Nikon D7000 Dynamic Range
« Reply #68 on: November 24, 2010, 12:38:25 am »

Hi,

I have made a few experiments with HDR in the past year. The experience has mostly been mixed, but more negative than positive.


So, what I try to do is to expand dynamic range. Problem is that we still have a limited dynamic range for display, something like 1:500 on screen and 1:120 in print. What I have found that the real world density range on my cameras (Sony Alpha) is good enough that I can extract shadow detail from an -2EV exposure.  

Messing with HDR doesn't really make the image better, even if I get less noise and better resolution in the shadows. On the other hand it's easy to get unrealistic color with HDR.

HDR: http://echophoto.smugmug.com/Special-methods/HDR/HDR/13306153_DcZHj#966794997_wt4h6

Non HDR: http://echophoto.smugmug.com/Special-methods/HDR/HDR/13306153_DcZHj#1002864735_dkeci

My Sony Alpha is not a DR champ by any means, and it has been my experience that there is a lot of good shadow detail in properly exposed DSLR images.

The screen dump below show both images. HDR on left and DNG processed in Lightroom on the right. The DNG image is parametrically processed but with a healthy amount of local manipulation. Tonal separation in the clouds is achieved by generous local application of clarity.

Best regards
Erik

But perhaps you have real examples where this extreme lightening is needed even with optimal "ETTR" exposure, and is done while also achieving good midtone IQ, rather than severe underexposure for no good reason? "Shadow peeping" examples that merely crop to deep shadow regions and dramatically lighten them while ignoring midtones are not of practical interest.
« Last Edit: November 24, 2010, 02:05:43 am by ErikKaffehr »
Logged
Erik Kaffehr
 

BJL

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 6600
Re: Nikon D7000 Dynamic Range
« Reply #69 on: November 24, 2010, 04:54:14 pm »

Four stops underexposure at base ISO on the D7000 results in a maximum (engineering) DR of 9.35 EV, at the pixel level, according to DXOMark.
Four stops underexposure on the 20D at base ISO results in a maximum engineering DR of just 6.95 EV.
...
Those 6.95 EV of DR have better image quality on the D7000, period.
Please read again what I said in post #48 in this thread: my skepticism is about whether improving beyond 12 stops of DR at base ISO is significant. So proving an advantage over an older camera offering only about 11 stops of DR is irrelevant.
Quote
Did you miss my examples of the St Isaac's Cathedral in St Petersburg, on page 2 of this thread?
... These are not deep shadows that are noisy, but the whole foreground.
I cannot find any such photo in you earlier posts in this thread; can you given me the link again?
And if it is with that "old, noisy" 20D, it is irrelevant for the reasons stated above. Only if done with a camera giving 12 stops of more of DR at the ISO in use is it relevant to my claim about the lack of benefit in improving beyond 12 stops of DR.  And if the noise is visible "in the whole foreground", that sounds like it visible in the midtones or just a bit below, and then for reasons I have already stated, it is likely to be due to shot noise, not read noise and the root of the problem is too low a maximum photon count, not too much read noise. (Perhaps for the sake of science, you need to get a D7000 and revisit that Cathedral!)
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
Re: Nikon D7000 Dynamic Range
« Reply #70 on: November 25, 2010, 09:04:06 pm »

I cannot find any such photo in you earlier posts in this thread; can you given me the link again?


Have you tried logging on before searching, BJL? It seems to be a peculiarity of the new forum software, if you are not logged on you don't see any pictures.

The images demonstrating noisy shadows are in reply #23 and reply#38.


Quote
And if it is with that "old, noisy" 20D, it is irrelevant for the reasons stated above.

BJL, why don't you be clear and state that you don't give much credence to the DXO tests, so we all know where we stand? All the tests that I've carried out, that are relevant to the DXO results, such as testing the DR limits of my 5D, and comparing the high ISO performance of the 5D with the Nikon D3, seem to correspond very closely with the DXOMark results in relation to those cameras. I have no reason to doubt the validity of their testing procedures and the accuracy of their results.

If the 20D is noisy, then so is the 1Ds3. If you were to shoot identical scenes with both cameras, using the same lens at the same aperture, then crop the 1Ds3 image to the same FoV as the 20D image, you would have two indistinguishable photos of the same file size (according to DXO's test results).

Below ISO 800, the 1Ds3 begins to have a very slight DR advantage which reaches a peak at base ISO, where it's 0.3EV better than the 20D. On the other hand the 20D above ISO 800 edges ahead of the 1Ds3 by a similar degree. As regards SNR at 18% grey, tonal range and color sensitivity, the pixels of both cameras appear to be identical, and that also includes the more recent 1DMkIV.

However, the 1D4 appears to have a noticeably better DR than both the 1D3 and 20D at higher ISOs. At ISO 3200 it's a whole stop better than the 1Ds3, so it's clear that Canon have made some improvement at the pixel level in the 1D4.


 
Quote
Only if done with a camera giving 12 stops of more of DR at the ISO in use is it relevant to my claim about the lack of benefit in improving beyond 12 stops of DR.

The benefit of the greater-than-12 stops of DR (of the D7000) is to be found in the 10th and 11th stops. I have never argued that the last two stops of an engineering-type specification of a DR, such as the ones that DXO seem to provide, are useable on a print.

What I do know from my own tests with the 5D and the 20D is that at base ISO detail in the 10th and 11th stops from those two cameras is too degraded to be of practical use. But at ISO 1600, such degraded detail is quite noticebly improved by a worthwhile degree, and not only in the 11th and 10th stops, but the 9th and 8th and so on.

If we accept that the D7000 has a  performance at high ISO at least equal to its competitors, such as the Canon 7D and 60D, but can achieve this image quality without the use of the higher ISO settings, presumably because image quality at base ISO is so darned good, then it follows logically that those 11th, 10th, 9th and 8th stops of DR in the D7000 at base ISO, will be less degraded and more detailed than the same stops in the 60D, 7D, 20D, 1Ds3 and so on, at base ISO.

I'm surprised you seem to be having trouble grasping this point, BJL. I always got the impression you were a very rational, logical and objective sort of guy.

You certainly seem to find access to lots of technical documents detailing sensor specification at the manufacturers'web sites. Perhaps you could do me a favour and explain to me the significance of the 'quantum efficiency' of a photodiode and the relationship between the 'electron charge' and the number of photons required to produce such charge.

For example, we seem to agree that a full-well capacity for the D7000 pixel could be around 32,768 e-.  But a pertinent question might be, 'How many photons were required to be 'absorbed' by the pixel in order to achieve such charge?.

Another pertinent question would be, ''What proportion of photons arriving at the outer surface of the sensor is blocked or absorbed by the various filters before the light even reaches the photodiode?

We know that the Color Array Filter of the Bayer-type sensor absorbs at least half of the light reaching it. The green filter has to exclude both red and blue frequencies, otherwise the pixel wouldn't be green. Furthermore, it is reasonable to assume that other filters such as the AA filter, IR filter and even the slight degree of opacity of the microlenses and filters will absorb yet more light.

Such facts seem significant to me because shot noise is caused by the slightly random and unpredictable arrival of individual photons during any given exposure of the sensor. Such shot noise exists before the photons pass through the various absorbing filters and microlenses. It therefore seems reasonable that the calculations for shot noise (the square root of the mean average of photons collected by the photodiodes) should in fact be applied to the number of photons arriving at the sensor's outer surface. Would you agree or not?

If this is the case (and I'm certainly not asserting that it is), we cannot use the 'electron charge' figure for calculations regarding shot noise.

Consider a situation where the full-well capacity of the D7000 is, for argument's sake, 32,768 e-. The 11th stop down from (and including) full-well capacity should then have a charge of just 32 e-. But that does not necessarily mean that the shot noise is sqrt of 32, or 5.65 photons which represents about 17% of the signal, which is quite high, but in fact may be the sqrt of 100 photons which represents a shot noise of just 10%.

Would you agree? If not, please enlighten me.
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
Re: Nikon D7000 Dynamic Range
« Reply #71 on: November 26, 2010, 11:31:59 pm »

Hi,

I have made a few experiments with HDR in the past year. The experience has mostly been mixed, but more negative than positive.


Erik,
All the more reason to have a camera with a high DR, to reduce the number of occasions when you might see a need to bracket exposures in order to merge to HDR.

Being aware of a need to bracket exposures in order to increase DR is one thing. Being able to do that successfully is another thing. First, the scene really does have to be stationary. I've bracketed exposures of the Himalayas with camera on tripod, only to discover back home that the very slight breeze causing movement of the tall grass in the foreground, and the slight movement of the distant clouds in the background, has made the merging to HDR process too difficult to bother with.

In such circumstances I choose the exposure which is closest to an ETTR and do my best to cover up any noise which might be apparent in the shadows.

Having successfully merged different exposures to HDR without any 'misalignment' of detail, there's sometimes a further problem of making the right 'tone-mapping' adjustments to get the image looking natural.

In my view, this is partly because one is faced with a whole new system of tonal adjustments which are different from the regular adjustments in ACR that one might be familiar with.

Another issue is the number of auto-bracked shots the camera is capable of. I've often found with Canon cameras that the maximum of +/- 2EV with just 3 shots is not ideal for merging to HDR. The extension of 4 stops of DR may be sufficient, but the gap of 2EV between each exposure may not produce in the best result. 5 frames with a 1EV difference between shots would be better.


Quote
So, what I try to do is to expand dynamic range. Problem is that we still have a limited dynamic range for display, something like 1:500 on screen and 1:120 in print. What I have found that the real world density range on my cameras (Sony Alpha) is good enough that I can extract shadow detail from an -2EV exposure.

I've noticed, Erik, that you've mentioned this factor before, quite frequently, implying that in some way the DR limitations of the print medium makes a high DR capability in the camera irrelevant or less relevant.

This is simply not true if one knows how to process an image in Photoshop. It's possible to display, in a natural way, the full dynamic range on the print of any correctly exposed high-SBR scene, even if the scene encompasses a whopping 20 stops of DR.

How is this possible? First, let me explain how the eye perceives a scene. (In simple terms so I can understand it myself).

The eye has a surprisingly narrow field of view as regards discernment of detail. Peripheral vision tends to give us the impression that the eye's FoV is quite wide, but peripheral vision is mainly good for the detection of movement. It's totally hopeless for the discernment of detail. To see all the detail in a scene that may be captured even by a moderately wide-angle lens, we have to shift our gaze from side to side, the equivalent of image stitching in a camera.

A similar situation applies to the eye's DR capability. When perusing a scene, we find in one moment we can see lots of detail in the white, fluffy clouds in the sky, then a fraction of a second later, as we shift our gaze to some shadows at the foot of a nearby tree, we can see leaf litter and lots of fine detail in the dark shade which really doesn't look all that dark anyway, but would look very dark if we were to photograph the scene with a single shot, exposing for the clouds.

It's easy to get the impression that the eye has a very wide DR capability. In a sense, it does, but only because it has the capability of bracketing exposures. The pupil of the eye, which is similar to the aperture of a camera lens, can dilate and contract in a fraction of a second in order to accommodate changing brightness levels.

Sometimes, when the change in brightness is extreme, it may take several seconds for the eye to adjust, such as coming out of a darkened cinema into the bright sunlight. But normally, a shift in brightness of just a few stops is almost instantly accommodated by a change in the aperture of the pupil.

So basically, if we compare the eye with a camera, using camera terminology, the eye is continually performing image stitching to widen its FoV, and continually bracketing exposures in order to increase its dynamic range.

A single shot from a camera, any camera, even the D7000, cannot compete with the eye as it's normally used with its continual gaze-shifting and changes of aperture.

Okay! All that must be crystal clear to you all, so let's move on to the print. How do we portray on a print, with its very limited DR, the full DR of a 20 EV scene?

Well, I'm no expert on the potential of the myriad of processes and techniques in Photoshop. I'm sure Jeff Schewe could make me look like a complete novice in that respect.

But there are a few very simple techniques that work quite well for me. First, the Shadows/Highlight tool. You need to be in advanced mode of course, but this is a tremendously useful feature in Photoshop for quick adjustments of balance between shadows and highlights. If you need more extreme adjustments, you can repeat the process.

Another simple technique (you must have noticed, I like simplicity), is to select with the lasso tool any area on the image you want to treat separately, feather appropriately (30, 50 or 100 pixels), then apply whatever adjustments you think appropriate.

By such processes, you can achieve on the print, the full detail of a bright cloud in the sky, and the full detail in the shadows at the foot of a tree in the foreground, if your camera has sufficient DR or, if you bracket exposures and merged to HDR.



Quote
Messing with HDR doesn't really make the image better, even if I get less noise and better resolution in the shadows. On the other hand it's easy to get unrealistic color with HDR
.

You are confusing the 'subjective' with the 'objective' here. The term, 'better', without a qualifying description, is about as subjective as one can possibly get. However, 'less noise' is a comparative matter which is either true or false. If it is determined to be true, objectively, through the sorts of measurements that DXO make, but subjectively one cannot see it, then one can reasonably surmise a number of possible causes.

(1) One doesn't know what 'visual' noise is and therefore cannot recognise it.

(2) The differences in noise are so small that they are irrelevant in practice, or beyond the threshold of normal discernment.

(3) The display medium, monitor or print, is not of sufficient quality, or not sufficiently well-calibrated, to enable the observation of such differences.

(4) The observer has a medical or genetic problem with his eyesight.

Of course, there may be other reasons of a more psychological nature, which perhaps we should not dwell upon. People can have 'blind spots'. Sometimes seeing the truth is too painful, so we bury our head in the sand.

Sometimes people identify so strongly with their camera equipment that it becomes part of their personality, so any criticism of their camera equipment is viewed as a criticism of themselves, their taste and their judgement.

So, if camera A has better high-ISO performance than camera B, then someone who owns camera B, of which he is very proud and which symbolizes his good taste and discernment, may first try to dispute the validity of the tests that imply that camera A has better ISO performance.

Having failed to convincingly win the argument that the tests are invalid or inaccurate, because of mounting photographic evidence that camera A really does have better high-ISO performance than camera B, the owner of camera B may then resort to other tactics to preserve his self-esteem.

A typical response might be, "I never use high ISO, so it's irrelevant." Or, in the case of camera A having a wider DR than camera B, "I'm not interested in pixel-peeping shadows. The DR of my current camera B is quite sufficient."

One sees many examples of such attitudes on DPreview.

By the way, Erik, your example images of HDR do not appear in IE when I click on them. There's something wrong.

Cheers!
Logged

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Nikon D7000 Dynamic Range
« Reply #72 on: November 27, 2010, 04:16:24 am »

Hello Ray,

Here is another link to the HDR stuff. It works for me: http://echophoto.smugmug.com/Special-methods/HDR/HDR/

I have not tested with Internet Explorer as I don't have it on my Macs.

Regarding HDR I planned an article on "doing things with multiple images" discussing stitching, HDR, and infinite DOF. Unfortunately I got stuck on HDR. The reason for that is that I failed to come up with any good example where HDR is of significant benefit, and the main reason is the cameras having enough DR.

There are some points you may miss in your lengthy response. The major one, as I see it, is that an HDR image is not in a normal representation but using some more exotic data representation, like floating point or 32bits/channel. Before you can apply any photoshop manipulation you need to do tone mapping. Photoshop CS5 has four major options for that:

- Local adaptation
- Equalize histogram
- Exposure and Gamma
- Highlight Compression

My favorite is local adaption.  When this has been applied the image will be converted to a 16-bit representation. I have tested other programs, like Photomatix and Autopano Pro (which can also handle HDR) but found that Photoshop HDR works best for me. The other programs also have issues color space. In HDR you really need to have raw conversion done in the HDR program and I prefer to use a raw converter I know.

After tone-mapping I usually go back to Lightroom as I'm much more comfortable with LR controls than with Photoshop's. That's just a matter of taste or preferences.

The image below is an example of HDR I have done. What I really like is the nice separation in clouds. This is actually one of the HDR images I'm quite satisfied with. My experience is that I could achieve about the same effect by using an image correctly exposed for highlights and use "fill light" in Lightroom to improve shadow detail and apply clarity locally to the clouds.

Regarding DR, it is not my opinion that 7-8 stops of DR is ample, more that 7-8 stops of DR is available in print. In order to fit a large DR in a print you either need to make local manipulations or tone mapping. When you work with shadow/highlight you essentially do tone mapping.

I normally use +/- 2 EV brackets with three exposures. That would give a DR (based on DxO-mark) of around 16 stops on my camera (Sony Alpha). More exposures may be better, of course, but this is what I can achieve comfortably without touching the camera

Best regards
Erik

Erik,
All the more reason to have a camera with a high DR, to reduce the number of occasions when you might see a need to bracket exposures in order to merge to HDR.

Being aware of a need to bracket exposures in order to increase DR is one thing. Being able to do that successfully is another thing. First, the scene really does have to be stationary. I've bracketed exposures of the Himalayas with camera on tripod, only to discover back home that the very slight breeze causing movement of the tall grass in the foreground, and the slight movement of the distant clouds in the background, has made the merging to HDR process too difficult to bother with.

In such circumstances I choose the exposure which is closest to an ETTR and do my best to cover up any noise which might be apparent in the shadows.

Having successfully merged different exposures to HDR without any 'misalignment' of detail, there's sometimes a further problem of making the right 'tone-mapping' adjustments to get the image looking natural.

In my view, this is partly because one is faced with a whole new system of tonal adjustments which are different from the regular adjustments in ACR that one might be familiar with.

Another issue is the number of auto-bracked shots the camera is capable of. I've often found with Canon cameras that the maximum of +/- 2EV with just 3 shots is not ideal for merging to HDR. The extension of 4 stops of DR may be sufficient, but the gap of 2EV between each exposure may not produce in the best result. 5 frames with a 1EV difference between shots would be better.


I've noticed, Erik, that you've mentioned this factor before, quite frequently, implying that in some way the DR limitations of the print medium makes a high DR capability in the camera irrelevant or less relevant.

This is simply not true if one knows how to process an image in Photoshop. It's possible to display, in a natural way, the full dynamic range on the print of any correctly exposed high-SBR scene, even if the scene encompasses a whopping 20 stops of DR.

How is this possible? First, let me explain how the eye perceives a scene. (In simple terms so I can understand it myself).

The eye has a surprisingly narrow field of view as regards discernment of detail. Peripheral vision tends to give us the impression that the eye's FoV is quite wide, but peripheral vision is mainly good for the detection of movement. It's totally hopeless for the discernment of detail. To see all the detail in a scene that may be captured even by a moderately wide-angle lens, we have to shift our gaze from side to side, the equivalent of image stitching in a camera.

A similar situation applies to the eye's DR capability. When perusing a scene, we find in one moment we can see lots of detail in the white, fluffy clouds in the sky, then a fraction of a second later, as we shift our gaze to some shadows at the foot of a nearby tree, we can see leaf litter and lots of fine detail in the dark shade which really doesn't look all that dark anyway, but would look very dark if we were to photograph the scene with a single shot, exposing for the clouds.

It's easy to get the impression that the eye has a very wide DR capability. In a sense, it does, but only because it has the capability of bracketing exposures. The pupil of the eye, which is similar to the aperture of a camera lens, can dilate and contract in a fraction of a second in order to accommodate changing brightness levels.

Sometimes, when the change in brightness is extreme, it may take several seconds for the eye to adjust, such as coming out of a darkened cinema into the bright sunlight. But normally, a shift in brightness of just a few stops is almost instantly accommodated by a change in the aperture of the pupil.

So basically, if we compare the eye with a camera, using camera terminology, the eye is continually performing image stitching to widen its FoV, and continually bracketing exposures in order to increase its dynamic range.

A single shot from a camera, any camera, even the D7000, cannot compete with the eye as it's normally used with its continual gaze-shifting and changes of aperture.

Okay! All that must be crystal clear to you all, so let's move on to the print. How do we portray on a print, with its very limited DR, the full DR of a 20 EV scene?

Well, I'm no expert on the potential of the myriad of processes and techniques in Photoshop. I'm sure Jeff Schewe could make me look like a complete novice in that respect.

But there are a few very simple techniques that work quite well for me. First, the Shadows/Highlight tool. You need to be in advanced mode of course, but this is a tremendously useful feature in Photoshop for quick adjustments of balance between shadows and highlights. If you need more extreme adjustments, you can repeat the process.

Another simple technique (you must have noticed, I like simplicity), is to select with the lasso tool any area on the image you want to treat separately, feather appropriately (30, 50 or 100 pixels), then apply whatever adjustments you think appropriate.

By such processes, you can achieve on the print, the full detail of a bright cloud in the sky, and the full detail in the shadows at the foot of a tree in the foreground, if your camera has sufficient DR or, if you bracket exposures and merged to HDR.


.

You are confusing the 'subjective' with the 'objective' here. The term, 'better', without a qualifying description, is about as subjective as one can possibly get. However, 'less noise' is a comparative matter which is either true or false. If it is determined to be true, objectively, through the sorts of measurements that DXO make, but subjectively one cannot see it, then one can reasonably surmise a number of possible causes.

(1) One doesn't know what 'visual' noise is and therefore cannot recognise it.

(2) The differences in noise are so small that they are irrelevant in practice, or beyond the threshold of normal discernment.

(3) The display medium, monitor or print, is not of sufficient quality, or not sufficiently well-calibrated, to enable the observation of such differences.

(4) The observer has a medical or genetic problem with his eyesight.

Of course, there may be other reasons of a more psychological nature, which perhaps we should not dwell upon. People can have 'blind spots'. Sometimes seeing the truth is too painful, so we bury our head in the sand.

Sometimes people identify so strongly with their camera equipment that it becomes part of their personality, so any criticism of their camera equipment is viewed as a criticism of themselves, their taste and their judgement.

So, if camera A has better high-ISO performance than camera B, then someone who owns camera B, of which he is very proud and which symbolizes his good taste and discernment, may first try to dispute the validity of the tests that imply that camera A has better ISO performance.

Having failed to convincingly win the argument that the tests are invalid or inaccurate, because of mounting photographic evidence that camera A really does have better high-ISO performance than camera B, the owner of camera B may then resort to other tactics to preserve his self-esteem.

A typical response might be, "I never use high ISO, so it's irrelevant." Or, in the case of camera A having a wider DR than camera B, "I'm not interested in pixel-peeping shadows. The DR of my current camera B is quite sufficient."

One sees many examples of such attitudes on DPreview.

By the way, Erik, your example images of HDR do not appear in IE when I click on them. There's something wrong.

Cheers!
Logged
Erik Kaffehr
 

Bart_van_der_Wolf

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 8913
Re: Nikon D7000 Dynamic Range
« Reply #73 on: November 27, 2010, 06:24:39 am »

The image below is an example of HDR I have done. What I really like is the nice separation in clouds. This is actually one of the HDR images I'm quite satisfied with. My experience is that I could achieve about the same effect by using an image correctly exposed for highlights and use "fill light" in Lightroom to improve shadow detail and apply clarity locally to the clouds.

Hi Erik,

That's a nice image. The whole issue with HDR is more to do with the inability of a complete camera system to capture the scene dynamic range in a single shot. The sensor may be able to capture, say, 11 stops of range, but the lower exposed regions do not carry enough useful data to exploit in postprocessing. What's more, the lens will further severely limit the useful dynamic range of the sensor, maybe to 9 stops.

HDR allows to break free from these physical limitations. By offering better S/N ratio throughout the dynamic range, we get an enormous potential of tonality manipulation, without the noise compromises. As an example of what creative potential lies hidden in even a JPEG like the one you just posted (admittedly already tonemapped), I've attached a modified version of it (If you don't like me doing that without your permission, I''ll remove it, let me know). It's not necessarily better, because that depends on your preferences. All I wanted to show is that with the proper tools, one can exploit the potential in an image file better, and the more potential, the more options we have.

Cheers,
Bart

P.S. I used SNS-HDR for the tonemapping with the JPEG file as a source. It's just one of a myrad of possible renditions.
« Last Edit: November 27, 2010, 06:49:14 am by BartvanderWolf »
Logged
== If you do what you did, you'll get what you got. ==

ErikKaffehr

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 11311
    • Echophoto
Re: Nikon D7000 Dynamic Range
« Reply #74 on: November 27, 2010, 08:13:09 am »

Hi Bart,

I don't disagree with you view. It's just that my experience is that DSLRs have ample range. Might be that my brain is still influenced me shooting Velvia on MF.

I appreciate your effort rendering my image. The original DNG files are here: http://echophoto.dnsalias.net/ekr/images/ToBart/

Best regards
Erik

Hi Erik,

That's a nice image. The whole issue with HDR is more to do with the inability of a complete camera system to capture the scene dynamic range in a single shot. The sensor may be able to capture, say, 11 stops of range, but the lower exposed regions do not carry enough useful data to exploit in postprocessing. What's more, the lens will further severely limit the useful dynamic range of the sensor, maybe to 9 stops.

HDR allows to break free from these physical limitations. By offering better S/N ratio throughout the dynamic range, we get an enormous potential of tonality manipulation, without the noise compromises. As an example of what creative potential lies hidden in even a JPEG like the one you just posted (admittedly already tonemapped), I've attached a modified version of it (If you don't like me doing that without your permission, I''ll remove it, let me know). It's not necessarily better, because that depends on your preferences. All I wanted to show is that with the proper tools, one can exploit the potential in an image file better, and the more potential, the more options we have.

Cheers,
Bart

P.S. I used SNS-HDR for the tonemapping with the JPEG file as a source. It's just one of a myrad of possible renditions.
Logged
Erik Kaffehr
 

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
Re: Nikon D7000 Dynamic Range
« Reply #75 on: November 28, 2010, 04:53:30 am »

I don't disagree with you view. It's just that my experience is that DSLRs have ample range. Might be that my brain is still influenced me shooting Velvia on MF.

That's a nice image, Erik, that you can be proud of. When I first saw it, I attempted to brighten the tonality of the large rock on the right, similar to what Bart has done, but it was late at night and I needed some sleep.

Which DSLR are you referring to, Erik?

If you are referring to the 24mp Sony A900 which has a very respectable DR at base ISO, more than a stop better than my 12mp Canon 5D at normalised print sizes, then I can understand why you may be quite satisfied with the relative lack of noise in the shadows, especially if you print no larger than A2.

My situation is quite different. My full frame cameras (the 5D and D700) are only 12mp, but my printer is a 24" wide format. If I wish to make a 24"x 36"print from a 5D or D700 file, then the detail and noise I see at approximately 80% on my 24" 1920x1080 LCD monitor, is what I get on a 24"x36"print at a ppi of 240.

However, if you are making an  A2 size print, say 400mm wide in portrait mode, then you should be viewing 'print' size on an HD monitor at approximately 36% enlargement (assuming you're viewing a file of 256ppi on a 24" HD monitor. At 240ppi you would get a 17" wide print, portrait mode.).

At such reduced enlargement, noise and lack of detail in the shadows is less of a problem.

I've taken more images with my 5D, since I bought it in November 2005, than with any other camera in my entire life.  Because I frequently use the camera without a tripod, often with little time to take accurate spot meter readings for ETTR purposes, I often bracket exposure, but not for the purpose of merging to HDR, but to make sure I get an exposure which is reasonably close to an ETTR.

Such bracketing may be +/- 2/3rds of a stop, and sometimes +/- 1 stop. Sometimes, I've bracketed hand held shots for merging to HDR. Then the difference in exposure between the lightest and the darkest is usually 4 stops. But hand-held shots tend to be 'hit & miss' when merging to HDR. I'd prefer to use a tripod.

The fact is, I have hundreds of pairs of shots that differ in exposure between the lightest and darkest by 2 stops, and where the darkest shot is an ETTR. The difference in noise and detail in between the lightest and darkest, not only in the deep shadows, but also in the lower mid-tones, is very obvious at 80% on my monitor, and even more obvious at 100% of course.

I tried doubling the files size of a noisy 5D shot, then viewing it at 36% on my monitor. The shadow noise was noticeably reduced. I can well imagine if the 5D had the same DR specs as the A900 at base ISO, then at A2 print size, I would be less concerned with DR issues.

However, the D7000 at base ISO and normalised print sizes, say up to 14"x21"at 240ppi, has a significant 1 & 1/2 stops higher DR rating than the A900, although it should be said that the DR advantage gradually reduces at higher ISOs so that at ISO 1600 the DR of the D7000 is only 2/3rds of a stop better, and at ISO 3200 only 1/2 a stop better.

All the other parameters, SNR, tonality and color sensitivity seem to be about the same at all ISOs, for the A900 and D7000.

I've done such a fantastic job of promoting this camera to myself, I'm unable to resist placing an order for one. I expect to receive it some time next week. I've also placed an order for the Nikkor 24-120/F4 VR which effectively doubles the cost.

The question is, '"Will I, or will I not, see the following 2-stop difference in shadows and lower mid-tones, when I compare the new camera with my old Canon 5D (old but full frame)?"
One of these 100% cropss is from an ETTR at ISO 100, F11 and 1/200th sec. The other is at the same ISO and F stop, but overexposed by 2 stops, ie 50th sec. No prizes for guessing which is which.  ;D
Logged

ejmartin

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 575
Re: Nikon D7000 Dynamic Range
« Reply #76 on: November 28, 2010, 12:51:18 pm »

Honestly, I would enjoy having an effectively "ISO-less" camera such as the D7000.  The ISO control is for me a nuisance, a workaround for the suboptimal capture chain electronics of my Canon DSLRs.  It is of course not strictly speaking an "exposure" control, it is a "rendering" control; so why not leave it to the time when I want to render the image rather than capture it?  The camera's meter should be able to suggest a rendering intent and use it for jpeg creation, but leave the amplification setting as metadata rather than a hardware amplification that throws away image data.  Being able to capture at close to base ISO and set exposure compensation in post would enable me to concentrate on the actual exposure parameters of aperture and shutter speed and not have to worry about whether the camera electronics is getting in the way of optimal capture for those exposure settings.
Logged
emil

Roy

  • Full Member
  • ***
  • Offline Offline
  • Posts: 196
    • http://
Re: Nikon D7000 Dynamic Range
« Reply #77 on: November 28, 2010, 06:21:54 pm »


Another pertinent question would be, ''What proportion of photons arriving at the outer surface of the sensor is blocked or absorbed by the various filters before the light even reaches the photodiode?

We know that the Color Array Filter of the Bayer-type sensor absorbs at least half of the light reaching it. The green filter has to exclude both red and blue frequencies, otherwise the pixel wouldn't be green. Furthermore, it is reasonable to assume that other filters such as the AA filter, IR filter and even the slight degree of opacity of the microlenses and filters will absorb yet more light.

Such facts seem significant to me because shot noise is caused by the slightly random and unpredictable arrival of individual photons during any given exposure of the sensor. Such shot noise exists before the photons pass through the various absorbing filters and microlenses. It therefore seems reasonable that the calculations for shot noise (the square root of the mean average of photons collected by the photodiodes) should in fact be applied to the number of photons arriving at the sensor's outer surface. Would you agree or not?

If this is the case (and I'm certainly not asserting that it is), we cannot use the 'electron charge' figure for calculations regarding shot noise.

Consider a situation where the full-well capacity of the D7000 is, for argument's sake, 32,768 e-. The 11th stop down from (and including) full-well capacity should then have a charge of just 32 e-. But that does not necessarily mean that the shot noise is sqrt of 32, or 5.65 photons which represents about 17% of the signal, which is quite high, but in fact may be the sqrt of 100 photons which represents a shot noise of just 10%.

Would you agree? If not, please enlighten me.

Hi Ray,

The quote above is equivalent to saying that the shot noise of a constant light source when attenuated by absorption no longer has a Poisson distribution. If it was true, the light on a dark cloudy day would record with no more noise that on a bright sunny day, and deep shadows in a room illuminated by a narrow beam of bright sunlight would be noise-free. This is about discrete quantized events which do not scale in a linear fashion. To take another example, if we toss a coin many times and look at the variation of outcomes of only a random 10% of the tosses, your argument would predict that the variation of the 10% sample would, on average, equal the variation of the total outcomes. Clearly not so.

Another observation: note that the standard deviation of photon noise is only equal to the square root of the number of photons for large numbers of photons. For large numbers the Poisson distribution approaches the normal distribution which has the simple square root relationship. 5.65 is not a large number.

Cheers
« Last Edit: November 28, 2010, 06:32:06 pm by Roy »
Logged
Roy

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
Re: Nikon D7000 Dynamic Range
« Reply #78 on: November 28, 2010, 10:17:44 pm »

Honestly, I would enjoy having an effectively "ISO-less" camera such as the D7000.  The ISO control is for me a nuisance, a workaround for the suboptimal capture chain electronics of my Canon DSLRs.  It is of course not strictly speaking an "exposure" control, it is a "rendering" control; so why not leave it to the time when I want to render the image rather than capture it?  The camera's meter should be able to suggest a rendering intent and use it for jpeg creation, but leave the amplification setting as metadata rather than a hardware amplification that throws away image data.  Being able to capture at close to base ISO and set exposure compensation in post would enable me to concentrate on the actual exposure parameters of aperture and shutter speed and not have to worry about whether the camera electronics is getting in the way of optimal capture for those exposure settings.

Emil,
I agree completely. If one is concerned about subtleties of image quality, smoothness of tonality, noise in the shadows etc, then it follows one would also be concerned about getting an ETTR exposure at the appropriate ISO. With Canon DSLRs this involves the dual concern about avoiding overexposure which will blow out the highlights, and underexposure which will increase noise in the mid-tones, lower mid-tones and shadows.

With the D7000 it seems one might have only the one concern regarding correct exposure, and that is to avoid overexposing at base ISO, (having of course already chosen an appropriate shutter speed and aperture for the circumstances).
Logged

Ray

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 10365
Re: Nikon D7000 Dynamic Range
« Reply #79 on: November 28, 2010, 10:31:14 pm »

Hi Ray,

The quote above is equivalent to saying that the shot noise of a constant light source when attenuated by absorption no longer has a Poisson distribution. If it was true, the light on a dark cloudy day would record with no more noise that on a bright sunny day, and deep shadows in a room illuminated by a narrow beam of bright sunlight would be noise-free. This is about discrete quantized events which do not scale in a linear fashion. To take another example, if we toss a coin many times and look at the variation of outcomes of only a random 10% of the tosses, your argument would predict that the variation of the 10% sample would, on average, equal the variation of the total outcomes. Clearly not so.

Another observation: note that the standard deviation of photon noise is only equal to the square root of the number of photons for large numbers of photons. For large numbers the Poisson distribution approaches the normal distribution which has the simple square root relationship. 5.65 is not a large number.

Cheers


Hi Roy,
I'm not sure I understand your explanation. I'm working from the principle that the shot noise is proportional to the sqrt of the total number of photons that pass through the shutter.
According to this principle, the shot noise in a photo taken on a cloudy day should be less, as a proportion of the total signal, than the shot noise in the same scene taken on a sunny day, assuming appropriate ETTR exposures in each case.

Whether the sun shines or not, the full-well capacity of the camera remains unchanged. The exposure should be greater for the cloudy-day shot in order to achieve a full-well charge in at least some of the pixels.

Because the contrast and brightness range in the scene on a cloudy day is less, a greater proportion of the pixels will achieve the maximum full-well charge, and a greater proportion will achieve 3/4 potential charge, and 1/2 potential charge, and so on.

The shot on the sunny day, on the other hand, will have more shadows and deeper shadows where the signal is dominated by shot noise in the parts of the sensor recording the signal..

The total number of photons passing through the shutter, in the shot on the cloudy day, will be significantly greater, and whilst the shot noise in an absolute sense will also be greater, it will be significantly less as a proportion of the over all signal.

Let me rephrase the question because this is really outside my area of expertise, and I'm not at all sure I'm right.

Consider two hypothetical cameras. Camera A has a perfect quantum efficiency at the photodiode level and amazing transparency of all its filters and microlenses which absorb no light whatsoever. It's the mythical perfect camera.

Let's suppose that the electron charge in the 11th stop down from full-well capacity is a mean average of 32e- and that this charge has been achieved by a certain number of photons exiting the lens and passing unhindered through the open shutter to the photodiodes where they knock off, on average, 32 electrons from the silicon substrate of each diode, one photon per electron.

Let's now consider Camera B with the same number of pixels with the same full-well capacity. The difference is that Camera B has a fairly low quantum efficiency and a lot of slightly opaque filters and microlenses, including one very opaque filter, the Bayer Color Filter Array.

Let's suppose that in order to achieve that average charge of 32e- in each pixel in the 11th stop in Camera B, we need to let in 8x the amount of light. Instead of an average of 32 photons per pixel, we need an average of 256 photons per pixel. Instead of an exposure of 1/200th at F11 for camera A, we need an exposure of 1/25th at F11 for Camera B.

Now, it seems reasonable to me, in my ignorance, to surmise or deduce that Camera B will exhibit less shot noise in the RAW file, than Camera A, although Camera A, being a mythically perfect camera, will possibly have lower noise in total because the only noise it will produce (or reproduce) is shot noise.

Camera B on the other hand will likely produce images with noise which is predominantly read noise, thermal noise, fixed pattern noise and the whole gamut of various noise types. Its lower shot noise (in proportion to signal) will not help much in these circumstances.

You might be right, but I'm not convinced yet  :)  .

Logged
Pages: 1 2 3 [4] 5 6 ... 10   Go Up